TRADE AND NEW EVIDENCE OF IMPACTS IN POVERTY DEVELOPING REDUCTION: COUNTRIES JOINT PUBLICATION BY THE WORLD BANK GROUP AND THE WORLD TRADE ORGANIZATION Disclaimer The designations employed in this publication and the presentation of material therein do not imply the expression of any opinion whatsoever on the part of the World Bank, its Board of Executive Directors, or the governments they represent, or the World Trade Organization concerning the legal status of any country, area or territory or of its authorities, or concerning the delimitation of its frontiers. The responsibility for opinions expressed in this publication rests solely with their authors, and publication does not constitute an endorsement by the World Bank or the World Trade Organization of the opinions expressed. Reference to names of firms and commercial products and processes does not imply their endorsement by the World Bank or the World Trade Organization, and any failure to mention a particular firm, commercial product or process is not a sign of disapproval. This volume is a co-publication of the World Bank and the World Trade Organization. Attribution—Please cite the work as follows: World Bank Group and World Trade Organization, 2018. Trade and Poverty Reduction: New Evidence of Impacts in Developing Countries. World Trade Organization: Geneva. Acknowledgements—This publication has been coordinated by Marcus Bartley Johns (WBG), Paul Brenton (WBG), Roberta Piermartini (WTO), Mustapha Sadni Jallab (WTO) and Robert Teh (WTO). The coordinators would like to thank Sandra Rossier for her assistance in the preparation of the publication. The production of the publication was managed by Anthony Martin and Edmundo Murray. Design and layout of the publication were undertaken by Corporate Visions, Inc. Copyright © 2018 World Trade Organization ISBN (paper): 978-92-870-4521-8 ISBN (electronic): 978-92-870-4522-5 Cover images: Bigstock and Flickr World Bank ii TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES CONTENTS 3 84 Foreword The Poverty Impact of Modernising Dar es 4 Salaam Port Trade and Poverty Reduction: New Evidence of Impacts in 98 Developing Countries: Agricultural Logistics in Introduction and Overview Lagging Regions: Evidence from Uganda 18 Is Tunisian Trade Policy 120 Pro-poor? Trade Openness and Vulnerability to Poverty 36 in Viet Nam under Doi Moi Gender Welfare Effects of Regional Trade Integration on 148 Households in Ghana Glass Barriers: Constraints to Women’s Small-Scale, 58 Cross-Border Trade in Cambodia and Lao PDR Exporting, Importing and Wages in Africa: Evidence from Matched Employer-Employee 174 Data Are the “Poor” Getting Globalised? 1 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 2 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Foreword In 2015, the World Bank and the World Trade Organization published a flagship report on the role of trade in the effort to end poverty by 2030. Over the past three years, the two organizations have collaborated in various ways to advance that goal, from supporting implementation of the WTO Trade Facilitation Agreement; to assisting the poor, including women and small-scale traders, to take advantage of trade opportunities; to supporting trade reforms in the world’s poorest countries. The latest report in this collaboration comes during an uncertain time for global trade. There is a trend away from trade openness, and the unstable trade policy environment has developed into one of the key risks facing the global economy. It is essential that we do not lose sight of the significant implications for the poorest and most vulnerable. Trade is a key driver of global growth and poverty reduction. An open global economy has created opportunities for hundreds of millions of people to lift themselves out of extreme poverty. This new volume brings together contributions from researchers that detail the challenges and opportunities in maximizing the impact of trade openness for the poor. It shows the need to continue to focus on reducing high trade transaction costs faced by poor workers and consumers in developing countries, and it explains how the benefits of trade can vary between rural and urban families, and between women and men. The papers in this collection also demonstrate the value of different research methods to understand links between trade and poverty, while also highlighting areas for further research and for testing new analytical methods. The collected analysis has important implications for policies and future research. It highlights how trade openness has clear, positive impacts on poverty reduction. For example, trade can benefit the poor by reducing the price of what they consume and increasing the price of what they sell. As producers, the poor can gain by selling their output in overseas markets where they can get a better return. Trade can also benefit the poor because it allows producers of domestic goods to respond to adverse shocks to domestic supply chains by shifting sourcing abroad. Trade can also help particular groups. For example, one of the case studies in this book finds that exporting firms in Africa tend to pay women workers more than non-trading firms. For some sectors and groups, however, the research shows serious challenges. One example is lower incomes due to import competition. Another is the risk that various “behind the border” barriers – like limited competition in transport and distribution, weak infrastructure, or a lack of information on new opportunities – will negate the benefits of export opportunities or lower-cost imports. Understanding these issues is essential for designing complementary policy reforms to help maximize the overall positive gains of trade openness for the poor. This requires more policy-relevant research, as well as an effort to address data constraints in areas such as services trade or the impact of non-tariff measures. With recent World Bank forecasts indicating that the pace of poverty reduction is slowing, there is an urgent need to intensify reforms and increase investments that help the extreme poor. Maximizing the benefits of trade openness for everyone, especially the poorest and most vulnerable, will be critical to driving inclusive, sustainable economic growth and finally ending extreme poverty on the face of the earth. Jim Yong Kim Roberto Azevêdo President Director-General World Bank Group World Trade Organization 3 The Role of Trade in Ending Poverty TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Trade and Poverty Reduction: New Evidence of Impacts in Developing Countries: Introduction and Overview In recent years there has been an intensification of the The 2015 joint report emphasized that the greatest long-running debate on the effects of trade integration. impact on poverty reduction will come from a coherent This debate has focused largely on the impact of trade and multi-pronged approach that addresses these specific in advanced economies, which has risked diverting constraints. To this end, the report discussed policy actions attention away from the impact of trade on people’s lives that governments can take individually and collectively to in developing countries, and especially the extreme poor. maximize the gains from trade for the poor: This volume brings together new research, using a range of different analytical approaches, that examines how the (i) Lowering trade costs for deeper integration of extreme poor have fared following trade liberalization markets, tackling policy and infrastructure barriers in various developing countries and regions and the to goods and services trade are critical to growth challenges that poor people face in benefitting from trade. and poverty reduction; Trade has been recognized as an engine for inclusive (ii) Improving the enabling environment including economic growth and poverty reduction in the 2030 policies related to human and physical capital, Agenda for Sustainable Development. The 2015 joint access to finance, governance and institutions, WTO-World Bank publication, The Role of Trade in Ending and macroeconomic stability; Poverty strengthened the evidence that trade has played (iii) Intensifying the poverty impact of integration a critical role in poverty reduction and that the further policies by bringing a greater focus on tackling integration of developing countries into an open global remoteness from markets at the sub-national economy will be essential for achieving the goal of ending level, and facilitating the activities of poor and extreme poverty by 2030. small traders; That publication (hereafter referred to as the “2015 (iv) Managing and mitigating risks faced by the poor joint report”) has contributed to focusing the debate on that limit them from benefiting from trade the challenges that the extreme poor face given their opportunities when they arise and build poor vulnerability, as well as the strategies that are most people’s resilience to the effects of adverse effective in maximizing the positive impact of trade. One of events; and the main messages of the 2015 joint report was that trade openness or trade growth alone may not be sufficient (v) Improving data and analysis to inform policy to to end extreme poverty. The extreme poor face specific better understand the trade-related constraints constraints—due to the fact that they tend to work in rural that the poor in many countries continue to face areas, in the informal sector, live in fragile states and face including through the use of new technologies for gender inequality—that can limit their ability to benefit data generation and analysis, including big data. from wider trade-induced economic gains. 4 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The present volume builds on this, surveying subsequent women within the household. Trade can further affect research and presenting a series of country-level studies long-term development outcomes if it leads declines to complement the global approach taken in the in child malnutrition rates, higher school attendance 2015 joint report. It continues to focus on the linkages and performance etc. The effect of trade on the poor between trade and extreme poverty rather than the depends on the specific mechanism at play, the nature broader distributional effects of trade and the agenda of trade policy change (whether it increases import of shared prosperity. It seeks to build understanding of competition or market access), the specific industry the challenges surrounding each of the four constraints or firm where the poor work and household decision- that were the focus of the joint report, as well as making. In this book we focus on how the effects of trade identifying ways to overcome them, thereby maximizing depend on informality, geographical location (rural or the contribution of trade to poverty reduction. The case urban area) and gender. studies included in this volume reinforce the need to Trade can benefit the poor by spurring economic growth. move forward with the above agenda to maximize the Sustained economic growth is the most powerful tool gains from trade for the poor. One of the main objectives for poverty alleviation. A household’s income is derived of the current volume is to bring a focus to the country from sale or utilization of its resource endowments and regional level, providing insights and examples of (landholdings, capital, labour or human capital). People analytical methods that can be drawn upon for work in are poor because they have few endowments or other countries and regions.1 because the rewards received from those endowments The chapters contained in this volume highlight some new are low. Typically, the poor suffer from both afflictions. analytical approaches to understanding and addressing Rapid and sustained economic growth allows the poor the linkages between trade and poverty, while also opportunities to increase their initial endowments (save showing the ongoing challenges facing the research and to accumulate capital, get an education to increase policy community in this work. Before summarizing each human capital) and to earn better rewards for supplying chapter, this section reviews the recent literature on their resources to others, typically through the market. trade and poverty, noting where chapters in this volume Another avenue through which trade can spur economic contribute to the literature. This overview chapter growth is by increasing the pace of innovation by firms. concludes by discussing the main policy implications of First, trade liberalization increases the size of the market the chapters and priorities for future work. and the incentives to innovate. Second, to the extent that technical knowhow is embodied in products, trade Overview of the recent literature on trade and poverty, liberalization makes possible knowledge spill-overs including links with the current volume through improved access to imports. Third, an increase in the degree of openness of an economy will typically The economic literature suggests that trade openness enhance product market competition. The increase is key to poverty reduction, but must be part of a wider in productivity means more output or income can be effort. Trade influences the income of the poor through obtained by society, and therefore also by the poor, from various channels: through its effects on economic a given amount of resources. growth, relative prices, macroeconomic stability and on government revenues (Winters, 2002 and Winters Trade can benefit the poor by reducing the price of et al., 2004). The impact of trade on poverty then what the poor consume and increasing the price of what depends on decisions within the household on how the poor sell. Trade opening changes relative prices in income is allocated. Trade can itself influence how both product and factor markets. These changes affect such decisions are made, for example by empowering the members of the household as both consumers and 1 The chapters for this report were selected following a call for papers by the WTO and World Bank that provided new empirical work on the trade and poverty nexus. There was an explicit request for papers that addressed issues relating to rural poverty, fragility and conflict, gender and informality. Perhaps reflecting the challenges of data, there were few submissions that addressed trade, poverty and conflict or that had an explicit focus on informality. In selecting papers for inclusion, beyond quality the volume has sought to capture evidence from a range of countries in different regions. 5 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Trade reforms can create new opportunities, but also involve adjustment costs for the poor. Access to international markets may deliver higher average incomes to farmers who specialize in producing export crops, but may bring greater competition that reduces the demand for poor workers in import competing sectors. sellers of goods and factors of production (e.g., labour). of trade reforms may not reach the poor, if they are not As producers, the poor can gain by selling their output connected to markets. This can be the case for the rural in overseas markets where they can get a better return. poor, who tend to be low skilled and less mobile than As producers, the poor are also consumers of inputs. those living in urban areas. A study of Indian liberalization Trade can allow them to get better access to material and in the 1990s (Topalova, 2007 and 2010) showed that services inputs and technology that improve productivity although poverty in India declined, it declined less for in the production of the goods and services that the households living in rural districts more exposed to poor produce. As consumers, trade liberalization can import competition. be beneficial to the extent that it reduces the price for The country studies in this volume also highlight the imported goods. As income earners, prices can affect differentiated impact of trade reform in rural compared wages and employment. to urban areas. For instance, in Ghana, households in As an example of how poor producers can benefit from rural areas that are net producers of agricultural products trade, there is evidence that the US-Viet Nam FTA has experience greater losses in welfare than households in helped poverty reduction in Viet Nam. Families living in urban areas. The study on India (Mendoza, Nayyar and provinces that benefitted from the largest cuts in the Piermartini) shows that the products of rural households costs of exporting to the US also saw the largest decline face an average tariff in overseas markets that is 10.9 in poverty (McCaig, 2011). Benefits have also been percentage points higher than that faced by the products extended to people working in the informal sector, as of urban households. This underlines the need to do more export opportunities have promoted the reallocation to address the sources of agricultural trade costs, including of workers from micro-enterprises to the formal sector weak internal connectivity as well as market access (McCaig and Pavcnik, 2014). barriers, in order to maximize the positive impact of trade opening for the rural poor. In their study on the potential The challenge for policy, however, is to implement trade pass-through of price changes due to reforms at Dar es reforms in a way that the increases the likelihood that Salaam port in Tanzania, Depetris-Chauvin et al. highlight the benefits will reach the extreme poor. The benefits the extent to which domestic transport costs and lack of 6 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES competition among intermediaries limit the gains of trade Trade benefits the poor, if it is associated with greater from being passed on to the poor in rural areas. The study diversification and greater macroeconomic stability. by Kunaka, drawing on research in Uganda, highlights Macroeconomic volatility is usually bad for the poor the importance of information barriers for agricultural because it can reduce economic growth and adversely producers seeking to participate in trade. affect the distribution of income and generate inequalities. The poor have little access to finance to be Trade reforms can create new opportunities, but able to tackle a period of tightened liquidity, therefore also involve adjustment costs for the poor. Access to they are the most affected by macroeconomic volatility. international markets may deliver higher average incomes If domestic shocks are the major source of volatility, trade to farmers who specialize in producing export crops, but can help reduce volatility through export diversification may bring greater competition that reduces the demand (Caselli et al., 2012). For example, when a country has for poor workers in import competing sectors. Adjusting multiple trading partners, a domestic recession or a to these changes can be costly. Poor workers may require recession in any one of the trading partners translates some retraining or may need to move to another location into a smaller demand shock for its producers than when to access newly generated jobs. Also specialization into trade is more concentrated. Trade allows domestic the production of one or few crops, for example, while goods producers to respond to shocks to the domestic increasing income of the poor when prices are high, it supply chain by shifting sourcing abroad. Geographical may reduce their income when prices are low. These costs export diversification may also help reduce the impact are particularly difficult for the poor to bear given their of country-specific external shocks (Jansen, Lennon lack of resources and limited access to finance. This is and Piermartini, 2015). In fact, a recent study by the compounded by the lack of effective social safety nets in IMF (2014) finds a robust negative correlation between many poor countries that can support poor people during export diversification and output volatility in low-income periods of transition. Therefore, what should be short- countries. term adjustment costs from trade can turn out to have long-term negative outcomes for the poor. The current However, greater trade openness also implies greater volume contributes to understanding the potential for exposure to external shocks—especially in outward- increased vulnerability of households in trade-exposed oriented industries. Countries with closer trade links sectors because of trade liberalization, even when tend to have more tightly correlated business cycles trade contributes to an overall increase in incomes and (Frankel and Rosen, 2008) suggesting that trade acts as a reduction in poverty (Magrini and Montalbano). transmission mechanism to propagate a country specific shock to others. By leading to greater specialization in Existing evidence points to the importance of putting output, trade reduces diversification in production and in place appropriate policies to smooth the costs of may make a country more susceptible to idiosyncratic adjustments for the next generation, as well as in the shocks. Kose and Riezman (2001) find that because a short-term. For example, by increasing the demand for significant fraction of African countries’ exports are skills and when adequate information on this is provided, concentrated in a narrow range of primary commodities, trade can contribute to improved educational outcomes. terms of trade shocks account for 45% of the volatility Jensen (2012) shows that recruitment campaigns in in aggregate output. Moreover, adverse trade shocks can rural villages in India that provided information about cause prolonged recessions since they induce a significant job opportunities in information technology (IT) in urban decrease in aggregate investment. Koren and Tenreyro areas were associated with increased schooling of young (2007) also suggest that greater volatility in developing girls. However, primary school attendance, especially countries arises from their initial specialization in the for girls, declined in Indian regions more exposed most volatile production sectors. Economic development to international competition (Edmonds, Pavcnik and involves diversifying away from these volatile sectors. Topalova, 2009 and 2010). 7 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Opening up to trade may put the poor at risk, if it reduces The current volume includes a number of studies that government revenue. The poor benefit from the provision deepen our understanding of how the impact of trade of public goods and may be helped by government reform may not necessarily be the same for women as it transfer programmes. What happens to customs is for men. This is important since in many poor countries revenues after trade reform can therefore matter greatly women and female-headed households typically have a to the poor since the share of trade taxes in government higher propensity to be poor than men and male-headed revenues is typically high in poor countries. households. However, connecting to global markets does not The chapter on Ghana (Orkoh) finds female-headed necessarily lead to lower government revenues from households are likely to benefit more from the trade. At first glance, trade liberalization will reduce tariff implementation of the ECOWAS Common External Tariff, revenues and this will certainly occur if all trade taxes are especially in urban and coastal areas, as they are more reduced to zero. However, fostering trade may involve likely to be net consumers of products where prices are measures that do not affect tariff revenues. This is the expected to decline. The chapter on trading firms in case for reforms that reduce red tape at the border. Africa (Duda-Nyczak and Viegelahn) finds that exporting Also, liberalization typically stops short of complete tariff firms tend to pay more to women workers than non- elimination. The “Laffer curve” analysis suggests that trading firms. Combining qualitative and quantitative there is a tariff rate that maximizes customs revenues; if data on small-scale, cross-border traders in Lao PDR the initial tariff rate is above this rate, tariff liberalization and Cambodia, Seror, Clarke and Record document that can actually increase customs collection. Furthermore, to women face both visible and invisible constraints to the extent that quantitative restrictions are replaced by participating in cross-border trade, although they might tariffs, new sources of tariff revenues will be generated not be as great as those faced by women traders in Africa. by trade reform. In many countries non-discriminatory Their study underlines the importance of adequate border consumption taxes, such as a value-added tax, are also infrastructure, streamlined border procedures, and levied at the border. Hence, while tariff revenues will fall, access to finance. The chapter on India (Mendoza, Nayyar revenues from consumption taxes applied to imported and Piermartini) calculates that women consistently face goods will rise and mitigate the overall revenue loss. higher tariffs in overseas markets than men. On average, Finally, lower customs revenues will be partially or wholly women’s tariffs are 6 percentage points higher than made up for by greater collection of domestic taxes those that men face, based on the type of work in which (holding tax rates constant), as economic activity and they are employed. These studies reinforce the need to growth is stimulated by trade opening. assess the gender implications of trade reforms and to identify and address if there are particular challenges Finally, trade opening can affect the way decisions are that women face to benefit from the opportunities or made within the household, which can have an important deal with the risks that trade brings. impact on poverty. This is particularly the case if trade empowers women within the household by creating jobs In sum, the discussion above has stressed that there are for women that would not otherwise be available. For several channels through which international trade affects example, cross-border trade in Africa provides income for poverty. Evidence shows that not all the poor are affected hundreds of thousands of poor women. The emergence of equally. The effects will depend on where they live (rural the apparel sector in Bangladesh has created substantial versus urban areas), their individual characteristics (skill, jobs for women and has contributed to changing social gender), the type of trade policy change (increased attitudes towards women and girls. The empowerment import competition or export opportunities) and where of women within the household is typically associated they work (industry, firm, formal/informal sector). Since with better nutritional and educational outcomes for the effects are context specific, case study analysis of the children, which in turn leads to higher productivity in the type we present in this book is very important to better long-term. Hence, the nexus between trade and gender is understand the variety of channels through which trade a critical area for attention. can affect poverty. 8 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Description of the chapters in the current volume Evidence shows In one of a number of papers on African countries, Martínez, Baghdadi and Kruse examine the relationship that not all the poor are affected by between the reduction of most-favored nation (MFN) tariffs, the conclusion of several free trade agreements, international trade and implementation of streamlined trade procedures from the 1990s to 2005 in Tunisia and the decline in poverty. The authors estimate that the fall in the prices of consumer goods from further tariff reduction would equally. The effects will improve the welfare of low-income households by about 1% of expenditure, but would not improve welfare for depend on where they richer households (with similar distributional results across regions). This limited gain reflects a conservative live (rural versus urban calculation of the likely pass through of changes in tariffs to domestic prices; the higher pass through rate areas), their individual used in similar studies would result in significantly larger welfare improvements. Lower consumer goods prices characteristics (skill, are estimated to benefit rural households more than urban households, improve average welfare in female- gender), the type of headed and male-headed households by about the same amount, and improve welfare for self-employed, poor trade policy change workers only slightly more than for more formal workers. Household welfare also improves because lower tariffs (increased import tend to increase wages. However, wage data are available only at the industry rather than individual level, and this competition or export estimation is not reliable. opportunities) and Orkoh assesses the welfare effects of changes in the domestic prices of commodities as a result of Ghana’s where they work adoption of the common external tariff of the Economic Community of West African States (ECOWAS), with special (industry, firm, formal/ attention to gender differences. These price changes would improve the welfare of poor and female-headed informal sector). households as consumers, but reduce the welfare of both poor and rich households as producers, with male- headed households being the most affected. The net effect is around zero for poor male-headed households and negative for the rich, but positive for female-headed households in the low- and middle-income categories. While the results also depend on the geographic location of the household, among other factors, the reform is thus expected to have pro-poor and pro-female effects. The study recommends that the government introduces compensatory policies, such as income transfers that would target male-headed households that are producers 9 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES and streamline the ongoing Livelihood Empowerment Depetris-Chauvin et al. assess the likely welfare and Against Poverty Programme to help poor, male-headed poverty impact of a project to modernize the port of Dar households. It could also aim to improve infrastructure es Salaam, Tanzania. They assume that the improvements in the regions that will experience a net welfare loss, in would lead to a 5% reduction in border prices for bulk order to make them more competitive. imports and a 5% increase in the price of exports. The model allows that reductions in border prices are not Duda-Nyczak and Viegelahn study wages in exporting always fully passed through to the domestic economy and importing firms of the manufacturing sector in because of the lack of competition among importers Africa, using firm-level data for 47 countries and matched and exporters in some sectors. Besides market structure, employer-employee-level data for 16 countries. The the likely impact of the project is reduced by poor average firm-level and employee-level wage paid by infrastructure and logistics in Tanzania. Transport costs exporters is higher than that paid by non-exporting increase substantially moving away from the port, firms. It is gains from economies of scale that explain contributing to a high degree of geographical market the positive wage premium of exporters, rather than segmentation in Tanzania. To a large extent, the short- differences in skill utilization, differences in workforce term positive welfare impacts of improving the port alone characteristics, or technology transfers. In contrast, are modest for all groups of households considered. In there is no evidence that importers pay higher wages particular, the effect is very small for poor households, as than non-importers, at least when comparing firms of the incidence of international trade in the consumption the same age. The wage premium of importing at the basket and as a source of income for the poor is very employee-level is estimated to be negative. Finally, there low. In the long run, a better functioning port could is no significant differences between female and male reduce rural poverty if other policies and infrastructure workers’ wages in trading firms, but some evidences of a improvements (particularly for inland transport) are put gender wage gap between trading and non-trading firms. in place to alleviate the many constraints affecting small holders’ access to food and cash crop export markets. Market participation by small scale farmers in Uganda, particularly those in remote regions, is severely limited by inadequate transport that raises transport costs and impairs quality, and by lack of information on market conditions. The paper by Kunaka explores how infrastructure investments, cooperation between farmers and the use of information technology can help farmers access more efficient transport, reduce their dependence on intermediaries and increase their incomes. Simulations using an agent based modelling (ABM) technique indicate that strategic placement of markets, improving access to warehouse facilities, encouraging farmers associations, improving the institutional framework for contract farming, and facilitating compliance with good agricultural practices (GAP) standards could reduce transport costs, raise prices received by smallholders, and increase their participation in the market. A number of studies examine the relationship between trade opening and poverty reduction in Asia. Consistent with other studies, Magrini and Montalbano find that 10 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES the level of poverty and the risk of future poverty fell sharply in Viet Nam following the Doi Moi reforms, which included trade liberalization reforms. However, households engaged in sectors that were more exposed One reason that poor people to trade experienced increased risk, and thus a significant may not capture the full probability of falling into poverty in the near future. The greater the exposure to trade of a household’s activities, benefit from participation in the more important was risk in determining variations international markets is that the in household consumption. After Doi Moi, households goods they produce tend to be involved in farming were, on average, five times more subject to relatively high trade likely than households involved in non-farm activities to fall into poverty. Similarly, rural households faced greater barriers. risk of falling into poverty than did urban households. These findings underline the importance of helping farmers exposed to international competition to reduce their vulnerability to poverty, despite higher average incomes as a result of trade liberalization. Measures could include safeguards against excessive price volatility and Several policy improvements would expand opportunities assistance with managing risk through increasing savings for women to profit from SCCBT. Increasing transparency, (e.g., by increasing the availability of micro-financial simplifying procedures and limiting border officials’ instruments), improving access to credit, and increasing discretion would reduce the time burden women face and the availability of insurance (e.g., through community- the disadvantages from their smaller bargaining power. based risk-sharing or institutional products targeted to Improving infrastructure would also reduce the time farmers most involved in tradable cropping). required at the border. Increasing provision of childcare services would reduce household responsibilities. Easing Seror, Record and Clarke focus on small-scale cross border fees and restrictions on motorized transport would trade (SSCBT) and gender in Lao PDR and Cambodia. reduce the importance of physical strength in moving They show that such trade can provide significant goods across some border posts. Providing training in earning opportunities for women, yet women appear to gender sensitivity and formulating a Charter for Cross- be underrepresented in such activities in Cambodia and Border Traders and Brokers stating the rights and Lao PDR. Physical sexual harassment does not appear duties of all parties involved in cross-border trade would to be a significant problem. Nevertheless, women face improve women’s experience at the border. greater constraints in participating in SCCBT than do men. Inadequate border infrastructure imposes delays One reason that poor people may not capture the full in trading, while women face greater time constraints benefit from participation in international markets is that due to more household responsibilities. Access to the goods they produce tend to be subject to relatively financing for women traders is more limited due to high trade barriers. The paper by Mendoza, Nayyar and greater difficulties in obtaining loans from relatives. Piermartini analyses the average overseas tariffs faced by Women have less bargaining power vis-a-vis border different groups of Indian workers, by matching Indian officials, which increases the cost of nontransparent and household survey data collected from July 2011 to June arbitrary customs procedures and results in a higher tax 2012 with information on the industrial classification burden. Women pay higher taxes and are more likely to of products. Tariffs faced by exporters in international be controlled by quarantine services, but this is not due markets are higher, and non-tariff measures (NTMs) more to a greater likelihood of illegal behavior. numerous, on goods produced by poor workers than on goods produced by rich workers. Tariffs also are higher on 11 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES goods produced in rural and more remote areas than on much to be done to enhance the benefits for the poor of those from urban centers, on goods produced by informal increasing integration of markets across borders. enterprises than by formal ones, and on goods produced by women than by men. And the global reduction in tariffs Key policy issues from 1996 to 2012 failed to ameliorate these differences. Beyond the broad areas of complementary policies Access to international markets is more difficult for the identified in the joint report, several specific policy poor. How did we get there? Essentially, efforts to protect challenges emerge from the papers in this volume as poor workers across countries (tariff protection in China requiring more attention as part of trade policy reform and the United States is higher on goods produced by strategies in developing countries and in support under poor workers than on goods produced by rich workers) the Aid for Trade agenda: face a coordination problem. If poor workers produce the same kinds of goods, then each country’s attempts to • A focus on reducing the high trade transaction protect its own poor workers by imposing higher tariffs costs faced by poor workers and consumers and more NTMs on such goods will reduce the access of in developing countries, to realize potential all poor workers to international markets, and thus limit gains. Several sources and impacts of these high the reduction in poverty from trade liberalization. transaction costs are noted in contributions to this volume, including tariff and non-tariff barriers Conclusions on policy priorities and issues for (Mendoza, Nayyar and Piermartini) transport and deeper analysis logistics costs (Kunaka and Depetris-Chauvin), The studies presented in this book show that reducing costs associated with accessing information barriers for the goods that the poor consume, facilitating (Seror, Record and Clarke; Kunaka). The potential access to external markets for the goods that the poor benefits of reducing these costs are demonstrated produce, and connecting the poor to global markets by in several papers (for example, Orkoh; Martínez, overcoming international and domestic trade-related Baghdadi and Kruse). costs are all central to maximizing the potential benefits • Ensuring competition and efficiency in provision of trade for poverty reduction. Trade can work through of services along domestic distribution networks, a variety of channels to alleviate poverty including: which can be essential to ensure that the potential creating jobs and increasing job opportunities; making benefits from trade are realized by poor people. goods consumed by poor households cheaper; lowering Lack of access to road and rail infrastructure limits the pecuniary and non-pecuniary barriers to trade that the returns to poor exporters and the benefits fall most heavily on poorer producers; and facilitating from declining prices of imported final and the access to information and technology that can intermediate products, especially for the rural poor transform production processes or make them more (Kunaka). Lack of competition in the transportation efficient. However, this report underlines the importance and distribution sector and poor logistics services of trade policy being accompanied by appropriate and mean that the impact of trade reforms can be very sound macroeconomic and sectoral policies to ensure small for poor households (Depetris-Chauvin). that gains generated from trade are shared as widely as possible, and that trade does not exacerbate inequalities • More attention to mitigating the risks that poor or contribute to increased vulnerability. Building on producers and workers face from increased the joint report and subsequent studies, the papers import competition. Exposure to import included in this volume make substantial contributions to competition can increase the vulnerability of increasing our understanding of how trade can help drive people who are below the poverty line or have poverty reduction and steps that governments can take recently exited poverty (Magrini and Montalbano). to maximize the positive impacts from trade and mitigate There is therefore a role for policies that help adverse outcomes for the poor. Nevertheless, there is still address these risks, and make the investments 12 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES needed so that trade-driven poverty reduction A key issue that a number of papers explore is the degree results in a sustained transition from poverty to of pass-through from changes in border prices to the higher income levels. These include: (i) Improving changes in relative prices actually faced by poor people. access to finance. For the poor to invest to benefit Typically, it is assumed that there is perfect pass-through from new opportunities to export typically requires but there is increasing evidence that this does not hold in access to finance. Access to finance is essential reality, especially in regard to the key characteristics of for the poor to weather temporary periods of the poor discussed above and in the joint report: location dislocation caused by trade. However, in poor in rural areas, informality and fragility. The papers here countries this is often constrained, especially for show that assumptions about pass-through are critical women; and (ii) Expanding the social safety net and in determining estimates of the extent to which poor insurance. The poor in developing countries often people are affected by trade. lack access to formal insurance and a social safety The contributions to this volume include both ex ante net. Typically, they rely on family networks for and ex post analyses of trade reforms. These in turn these services, which may impede mobility across cover a variety of examples, including tariff liberalization, regions and occupations if there are costs involved reduction of non-tariff barriers, trade facilitation, and in changing jobs. regional trade integration. The majority of papers assess Priority issues for research and analysis the impact on poverty of potential or previous episodes of trade liberalization by modelling the change in prices One of the benefits of this volume, and of the wider set of particular sectors and linking these with the results of of submissions received in response to the call for papers household surveys. issued by the WTO and the World Bank, is that they shed further light on the value of different research methods It is clear that the papers in this volume make a significant employed to understand trade-poverty linkages, while contribution to building an understanding of the value also highlighting areas for further research and testing of existing research methods available to study trade new analytical methods. and poverty linkages, and ways in which these are being modified and improved. At the same time, the papers also In terms of the research methods used, there is no highlight the gaps in research and analytical methods for dominant approach and different methods are required understanding trade and poverty linkages. The following according to the issues being assessed and the availability priorities are suggested for future work. of data. The volume includes a number of papers which use information from household surveys to trace Data limitations remain a critical constraint to deeper through the impact of changes in relative prices due to research and analysis, especially with regard to fragile trade reforms on welfare. The papers vary in the way in and conflict states, informality and gender. There is a which they model the impact of trade reform including need for a more systematic effort to capture the lessons detailed partial equilibrium approaches and the use of a learned through the use of new methods to gather computable general equilibrium (CGE) model to explore relevant data (e.g., through data scraping, mobile phone wider inter-sectoral implications. Other papers use surveys) and through use of a wider range of data sources econometric techniques applied to detailed firm level (e.g., nutritional data; geospatial imagery). and employee data to examine the impact of exporting Understanding the potential impact of the reduction and importing on wages. Other papers use more novel of non-tariff barriers, especially for food and agro- research approaches obtaining data from detailed surveys processed products. There is very little analysis of how of small-scale traders at the border and from agent based barriers such as quantitative restrictions, differences in modelling to assess detailed issues related to the impacts regulations across countries, poor trade facilitation, and a of trade that are not possible in more aggregate sector lack of competition among key value chain players impact and economy-wide modelling approaches. 13 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES on poverty and different groups of the poor. There is by lowering many of the costs associated with trade also limited knowledge of specific policy approaches that (e.g., information costs). For example, in rural areas of can facilitate their removal. For example, more attention China the rapid growth of e-commerce for agriculture needs to be given to reducing the costs of compliance has led to significantly higher demand for value-added with standards and regulations for poor producers. Of niche agricultural goods, like organic produce.3 Access particular importance are measures related to trade in to information for farmers in countries like India is agricultural products and compliance with Sanitary and addressing information asymmetries, a key source of Phytosanitary (SPS) regulations. It remains extremely trade-related costs for the poor, and the Internet is being difficult and costly for poor producers of agricultural used to lower logistics costs in East Africa in ways that products to access markets in richer countries. There is reduce transaction costs for producers. There is also the also enormous potential for trade between developing potential for new forms of Internet-enabled trade to countries in food products, which is stymied by restrictive help firms and entrepreneurs overcome the constraints regulations and onerous inspection procedures, imposed by informality and small size. This is particularly especially for traded products. There is considerable applicable for e-commerce in goods, where there are potential for the use of new ICT technologies for rapid many examples of platforms that help connect individual testing and certification of traded commodities grown by artisans, re-sellers of goods, and farmers to markets. smallholder farmers, to reduce the costs of compliance Across these various applications of technology that could and of trading across borders. reduce trade costs facing the extreme poor, there is a need to more systematically capture the lessons learned Facilitating trade in services. Services exports provide through pilot initiatives, and apply them at greater scale new routes to deliver jobs for the poor, and in particular where feasible. for women, who are employed intensively in services sectors such as tourism, health and education, as well The relationship between informality and trade. A as improving access to critical services that can drive number of papers in this volume touch on the issue of development through increased imports.2 Sectors trade and informality but detailed analysis of the links such as tourism can be an important source of jobs and between trade and informality remain absent. None incomes for the poor in rural areas and can be a key sector of the papers submitted to the call for this volume had for fragile states as they emerge from conflict. Access to an explicit focus on informality. In part this reflects the information, logistics, education and health services can lack of easily available data and that new approaches increase productivity and allow small producers to access are required. For example, participants of the informal larger export markets. Yet there is very little analysis of economy are often very reticent to provide information the constraints to services trade that impact most heavily to what may appear as an official survey. Given very on the poor and scant guidance for governments on large numbers of the extreme poor who derive their how to address these. As with non-tariff measures, this income from the informal sector, it is essential to grow is compounded by a lack of data on services trade flows the body of country evidence on the constraints posed —although innovative, technology-enabled methods to by informality for participation in trade and how trade gather this data are helping address this, as discussed can provide opportunities for informal enterprises to below. grow and ultimately transition to the formal sector and so obtain better access to finance, information and skills. Exploiting the potential for technology to provide Understanding and addressing the risks associated with solutions that help ensure a positive impact of trade operating in the formal relative to the informal sector for the extreme poor. Access to the Internet and other is a key issue. In many cases, operating in the informal forms of digital technology holds significant promise for sector is a rational response to the risks and costs of connecting the poor to new or better trade opportunities, 2 See, for example, see Dihel and Goswami (2016). 3 http://www.chinadaily.com.cn/business/2016-12/20/content_27719089.htm and http://www.scmp.com/news/china/article/1662752/chinas-new-farmers-are-using-e-commerce-transform- agriculture 14 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES being formally registered. It may also be that the scale of operation that is feasible in small domestic economies Trade has made a limits operation to the informal sector, a constraint that can be broken by access to larger overseas markets. critical contribution Understanding and enhancing second round impacts of to poverty reduction to date, and further trade reforms. This involves taking analysis beyond the typical focus on short-term impacts on consumption and production through changes in prices brought about by trade reform, to looking at strategies and responses that integration of maximize the positive impact over time. For example, the typical approach in assessing the impact of global shocks developing countries that lead to higher food prices on poor households has been to identify whether the household is a net consumer into international or producer of food products. Given the finding that most poor households are net consumers of food, analyses markets will be using this approach typically conclude that higher food prices are likely to increase poverty in the short run (see, essential for ending for example, Ivanic and Martin (2008, 2014)). However, in the medium households adjust to higher food prices in poverty and leaving ways that can lead to lower poverty. This could be through investments, for example in fertilisers and higher yielding no one behind. seeds, that increase productivity and output. There could also be shifts within the household to activities that generate higher wage income for the household as demand for unskilled labour in rural areas increases. Finally, the impact of changing household income on living in poverty face if they are to maximize the gains poverty will depend on intra-household distributional of trade. The papers presented in this volume explore decisions which are not captured in the net food position these constraints in more detail at the country level in of the household. The latter could include policies that several regions, making a significant contribution to the further support the empowerment of women, which can literature. They showcase various research methods for have long-term development outcomes in terms of lower exploring trade and poverty linkages, while also making child malnutrition, and higher school participation and clear the areas in which research and analytical gaps attainment rates. exist, for future work to fill. In this way, it is hoped that this volume contributes not only to the growing body of Conclusion work on trade and poverty in developing countries, but Trade has made a critical contribution to poverty also contributes to further efforts among researchers reduction to date, and further integration of developing and policymakers to address the many gaps that remain countries into international markets will be essential for in this field. ending poverty and leaving no one behind. However, as the 2015 joint WTO-World Bank report emphasized, the gains from trade integration alone may not be sufficient to end poverty by 2030. Complementary efforts are required to tackle the constraints that people 15 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Caselli, F., M. Koren, M. Lisicky and S. Tenreyro (2015). “Diversification Through Trade”. CEPR Discussion Paper10775. Dihel, N. and A. Goswami (2016). The Unexplored Potential of Trade in Services in Africa, World Bank, Washington DC. Edmonds, E., N. Pavcnik and P. Topalova (2009). “Child Labor and Schooling in a Globalizing Economy: Some Evidence from Urban India.” Journal of European Economic Association, Papers and Proceedings 7 (2–3), 498–507. Edmonds, E., N. Pavcnik and P. Topalova (2010). “Trade Adjustment and Human Capital Investment: Evidence from Indian Tariff Reforms.” American Economic Journal: Applied Economics 2 (4), 42–75. Frankel, J. A. and H. Rosen (2008). “The Endogeneity of the Optimum Currency Area Criteria,”The Economic Journal 108, 1009-1025. IMF (2014). “Sustaining Long-Run Growth and Macroeconomic Stability in Low-Income Countries – The Role of Structural Transformation and Diversification – Background Notes,” Washington, DC: IMF. Ivanic, M. and W. Martin (2008). “Implications of Higher Global Food Prices for Poverty in Low-Income Countries.” Agricultural Economics 39, 405–16. Ivanic, M. and W. Martin (2014). “Short- and Long-Run Impacts of Food Price Changes on Poverty.” World Bank Policy Research Working Paper No. 7011. Kose, M.A. and R. Riezman (2001). “Trade Shocks and Macroeconomic Fluctuations in Africa,”Journal of Development Economics 65 (1), 55-80 Jansen, M., C. Lennon and R. Piermartini (2016). “ Income volatility: whom you trade with matters,” Review of World Economics 152, 127–146. Jensen, Robert. (2012). “Do Labor Market Opportunities Affect Young Women’s Work and Family Decisions? Experimental Evidence from India,” The Quarterly Journal of Economics 127(2), pp. 753–792. Koren, M. and S. Tenreyro (2007). “Volatility and Development,”Quarterly Journal of Economics 122 (1): 243-287. McCaig, B. (2011). “Exporting out of poverty: Provincial poverty in Vietnam and U.S. market access.” Journal of International Economics, 85(1), pp. 102–113. McCaig, B. and N. Pavcnik (2014). “Export Markets and Labor Reallocation in a Low-Income Country.” NBER Working Paper No. 20455. Topalova, P. (2007). “Trade Liberalization, Poverty and Inequality: Evidence from Indian Districts,” in Harrison, A. (ed). Globalization and Poverty, University of Chicago Press. Topalova, P. (2010). “Factor Immobility and Regional Impacts of Trade Liberalization: Evidence on Poverty from India.” American Economic Journal: Applied Economics 2(4), 1–41. Winters, L.A., N. McCulloch and A. McKay (2004). “Trade Liberalization and poverty: the evidence so far.” Journal of Economic Literature 42(1): 72-115. 16 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The international border– Petrapole, West Bengal, India–is a bilateral trade gateway between India and Bangladesh. 17 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Is Tunisian Trade Policy Pro-poor? 1 Inma Martínez-Zarzoso, University of Goettingen and University Jaume I Leila Baghdadi, WTO Chair Holder, Tunis Business School, University of Tunis Hendrik Wiard Kruse, University of Goettingen 1. Introduction Several other key policy changes in Tunisia’s liberalization T rade liberalization policies affect the domestic also took place in the 1990s and the 2000s. In particular, economy through their impact on prices of the reform of external trade, established by the law of goods and services. Consequently, these policies 1994, launched a first program (2000–2004) that resulted also can affect average productivity and lead to in an integrated system of electronic management of industrial restructuring. The main goal of this research external trade procedures, which reduced the time is to estimate the distributional effects of trade policy needed to complete foreign trade operations. The at the micro level using household survey data, and to second program, launched in 2005, put in place a determine whether trade liberalization affected different custom risk management system and more transparent groups of poor people differently. To our knowledge this standards and technical regulations. However, from 2005 question has not yet been addressed for Tunisia. to 2015 there were no other significant modernization steps, and trade policy remained almost unchanged, with Trade policy in Tunisia has been evolving over time only one FTA signed (with Iran in 2008—WTO, 2016). through a progressive reduction of tariff protection that Since 2015, all technical import control documents can has narrowed the gap between most favored nation be transmitted electronically, but the processing of the (MFN) tariffs and preferential tariffs and increased the numerous tax incentives still relies on paper documents. number of free trade agreements (FTAs) signed with its main trading partners. Most FTAs involve a gradual While the reduction in tariff rates and the simplification elimination of tariffs, at least for non-agricultural products of the tariff regime (there are only three tariff rates: zero, (WTO, 2005). The maximum tariff rate in 1995 was 43% 20% and 36%) have reduced distortions, trade remains for non-agricultural products and 150% for agricultural subject to extensive controls. State-owned enterprises products. Tariff reductions to bring the MFN rate close and a number of boards (Trade Board, Cereals Board to the tariff applied to preferential imports have reduced and Oil Board) exercise considerable control over the average rate from 45% in 2006 to 14% in 2016; and international trade. Imports are still subject to many the maximum rate of 150% was reduced to 36% in 2009. controls and permits, although the development plan By 2016, Tunisia had concluded trade agreements with launched in 2016 is supposed to review the role played about 60 countries. by these entities in the development process. 1 We gratefully acknowledge financial and intellectual support from the World Trade Organization Chair at Tunis Business School, University of Tunis, Tunisia and the Economic Research Forum (ERF). I. Martínez-Zarzoso is also grateful for the financial support received from Project ECO2017-83255-C3-3-P (AEI, FEDER, EU). The authors thank participants at the 22nd Annual Conference of the ERF, Caroline Freund and Chahir Zaki for their constructive comments. 18 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES According to the World Bank (2016) the overall level of poverty in Tunisia has fallen since the mid-1980s, due to increasing economic growth and several program interventions. However, inequality and social inclusion remain an issue. Trade policy in Tunisia has been evolving over time through a progressive reduction of tariff protection that has narrowed the gap between most favored nation (MFN) tariffs and preferential tariffs. 19 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The structure of the economy experienced some investment funds, energy subsidies, rural development significant changes in the 1990s and 2000s. Tunisia programs, microcredit programs, etc). However, remains an open economy, with trade in goods and inequality and social inclusion remain an issue. Given services equaling 90% of GDP in 2015. However, a loss the importance of international trade for the Tunisian of competitiveness of Tunisian firms since 2005 has been economy, it is of great interest to determine whether the reflected in a rise in the import share of GDP from 45 to decline in poverty is related to the fall in protectionism. 50% and a fall in the export share from 45 to 39%. Exports of machinery and transport equipment have increased The main question to be answered is how trade reforms sharply, while the share of agricultural products and affect domestic prices and to what extent these changes of clothing in total exports has dropped (the latter fell translate in turn into changes in household welfare. There from 30 to 16% of total exports in the second half of the are a number of channels through which households are 2000s). Participation in trade is dominated by enterprises affected by trade reform. Declines in tariffs will reduce in the coastal regions and in urban areas, with the main goods prices, and households that are net consumers exported products being olive oil, seafood, “harissa” of these goods will benefit, while net producers will and dates. Tunisia’s main trading partner remains the be hurt. In addition, changes in prices can also affect European Union (EU-28), which in 2014 received almost employment and wages. Households, as income earners, 75% of Tunisian exports and sourced 53% of its imports. may benefit as higher prices in competitive exporting However, the EU shares have fallen with imports from sectors attract more producers into a given industry China increasing from 3% of Tunisian imports in 2005 to and increase employment and subsequently also wages. 7% in 2014. Conversely, declining prices for imports will put pressure on employment and wages in import competing sectors. According to the World Bank (2016) the overall level of poverty in Tunisia has fallen since the mid-1980s (in The main result is that household welfare improves due particular, extreme and moderate poverty fell in urban to the reduction in tariffs. This effect is greater among and rural areas from 2000 to 2012), due to increasing low-income groups, since the decline in consumer goods economic growth and several program interventions, prices benefited poor more than rich households. Labor including social assistance programs (e.g., social income effects are sizable, but statistically significant Tunisia remains an open economy, with trade in goods and services equaling 90% of GDP in 2015. 20 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES only for the skilled. The rest of the paper is organized However, when taking into account the positive labor as follows. Section 2 reviews the related literature and income effect on household welfare, estimated using a presents some stylized facts. Section 3 outlines the wage equation and calculating wage-price elasticities, methodology and describes the data sources, variables a net pro-poor effect of the reduction in tariffs due to and the model specification. Section 4 presents and Mercosur is observed. The intuition behind this result discusses the results, and Section 5 concludes. is that the reduction in tariffs on traded goods raised the relative price of goods intensive in unskilled labor (increase in the price of food and beverages and decline 2. Review of the literature in price of household equipment), which reduced the Most empirical evidence at the macro level indicates wages of more-skilled workers relative to those of less- that trade openness has a positive impact on economic skilled workers. The estimated effects are small, around development in general (Doyle and Martínez-Zarzoso, 6% of initial expenditure. The main policy conclusion 2011). However, the benefits are usually unevenly is that poverty in Argentina would have been higher distributed across households. Recent literature shows without the Mercosur agreement. that the impact of tariff liberalization on households, Several recent studies, including Nicita (2009), Ural both as consumers and factory owners, is positive overall. Marchand (2012), Borraz et al. (2013) and Nicita et al. However, the distribution of gains differs significantly (2014), have applied a similar methodology, relaxing across income levels and geographic regions within some of the strong assumptions made in Porto (2006). countries. The main novelty of Nicita (2009) and Ural Marchand (2012) was to estimate the extent to which the impact of trade reforms on prices differed in rural versus urban areas. Indeed, market imperfections partially isolate households from the effects of tariff changes, and this isolation is more severe in rural areas. Nicita (2009) allow for less than full pass-through from Porto (2006) develops a method to estimate the changes in border prices of traded goods domestic prices distributional effects of trade policies using household by using an econometric model proposed by Goldberg survey data and applies it to the case of Argentina. and Knetter (1997). It is estimated that in Mexico only He finds that the average poor and middle-income 33% of the tariff reduction for agricultural products and family benefited from the Mercosur agreement. More 27% for manufactures were reflected in domestic prices. specifically, Porto (2006) assumes a unitary pass-through Contrary to Porto (2006), richer households gained more rate from tariffs to prices, and uses intra-Mercosur and from trade liberalization in Mexico than poor households. common external tariffs and import shares to compute the price changes. Next, he obtains the consumption Ural Marchand (2012) estimates similar pass-through effects by multiplying budget shares by the computed equations and finds that Indian households experienced price changes, and uses locally weighted regressions gains at all per capita expenditure levels as a result of trade (Fan, 1992) to analyze the relationship between those liberalization, while the average effect was generally changes along the distribution of per capita household greater for poor households, and varied significantly expenditure. He finds that the resulting welfare impact across the per capita expenditure spectrum. The main increases with income per capita expenditure, indicating novelty of Nicita (2009) and Ural Marchand (2012) was to a greater gain for higher-income than low-income estimate the extent to which the impact of trade reforms households. 21 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES on prices differed in rural versus urban areas. Indeed, market imperfections partially isolate households from the effects of tariff changes, and this isolation is more severe in rural areas. Ural Marchand (2012) estimates that the pass-through of tariff reductions to domestic prices was only around 40% in rural areas, compared to around 66% in urban areas. Borraz et al. (2013) also find that trade liberalization had a pro-poor effect in Brazil, as poverty fell and inequality remained unchanged. This result is mainly explained by the decrease in consumer prices after Brazil entered Mercosur, as the net impact on household welfare due to changes in wages was almost zero. Finally, Nicita et al. (2014) examines the impact of the structure of trade protection on income distribution at the household level in six Sub-Saharan African (SSA) countries. They find that trade policies in SSA tend to redistribute income from rich to poor households. The main novelty of this research is that they present a method to indirectly estimate wages, which are not available in many countries. We extend this literature to the case of Tunisia and take on board the novelties incorporated by recent studies. For instance, we take the estimated pass-through from a companion paper (Baghdadi et al., 2016) and estimate wage equations and welfare effects along the lines proposed above. We also estimate different average welfare effects by region, between rural and urban households, and by gender. We do not expect to find substantial differences among regions, since Tunisia is a small country. This study is the first investigation of the effects of trade policy on income distribution in Tunisia. Minot et al. (2010) are to our knowledge the only authors who estimate the poverty effects of trade policy in Tunisia for given scenarios, using a computable general equilibrium model (CGE) calibrated with household data for 1995. Their main results indicate that poverty will decline slightly, from 8.1% to 7.6%, if all tariffs on imports from all countries are eliminated. In contrast, we aim to estimate the effect of trade policy on the entire distribution of income and taking into account tariffs. 22 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 3. Model specification, data and number of steps to consolidate and merge the data at the variables same level of disaggregation. Moreover, concordances between the different classifications have to be manually We apply the methodology proposed in Porto (2006) constructed to match the various datasets.2 The main to recently available household-level data for Tunisia in sources for the data are national surveys, and trade and order to assess the effect of trade on income distribution. protection statistics from national and international The model is used to simulate the effects of trade sources (INS, COMTRADE and WITS). policy changes on household wellbeing along the entire distribution of expenditure per capita by extending the Expenditure shares along the distribution of income techniques used in Deaton (1989). The latter provides a are obtained from the national survey on household non-parametric, empirical methodology to explore the consumption and expenditures compiled periodically impact of small changes in prices following trade reform by the National Institute of Statistics and harmonized on household welfare. by the Economic Research Forum (ERF), which kindly provided us with the data. The 2005 survey, used in The model focuses on the effect of changes in domestic this paper, is the eighth of its kind. The survey was prices and wages that could be attributed to changes in launched in early May 2005 and lasted until the end of trade policy. A change in a tariff translates into a change April 2006, to take into account seasonal changes in in the border price of traded goods which is passed household consumption. This survey aims to identify the through to domestic prices (retail and factor prices) to current standard of living of families through an accurate a variable extent. The magnitude of the pass-through is estimation of expenditure and food consumption, and determined by country-specific or region-specific factors, to compare these findings with those of previous years. which in turn influence the extent to which trade policies The survey collects information on aspects related to the can affect domestic prices. These factors are, among expenditures and living conditions of families, such as others, domestic policies, institutions, geography, market their access to education and health services. competitiveness and infrastructures. In Baghdadi et al. (2016) we show that market concentration and market The 2005 survey includes a representative sample of power are also crucial factors affecting pass-through for 13,400 households, distributed among 1,116 counties different sectors in Tunisia. (villages, campaigns and cities) of Tunisia. The survey consists of three axes: (i) household expenditures, (ii) Based on our estimate of the pass-through, we calculate nutrition, and (iii) social and collective services. The the inverse of so-called “compensating variation,” which household expenditure axis includes the whole sample, measures how much money households would have to whereas the nutrition axis only includes half the number be given in order to be compensated for hypothetical of households present in the first axis (6,700 households). changes in prices and can be interpreted as a measure The final axis includes one-third of the household of the change in welfare. Wages are analysed using spending axis (4,450 households). Mincerian equations, which link wages to skills, age (where typically an inverse U-shaped relationship is From this survey we derive information on broad found), regions and gender. We augment this framework expenditure shares, educational attainment, literacy, to include trade policy variables interacted with the level marital status, household size, educational status of head of skill. The econometric methodology adopted in this and spouse, and industry of occupation for head and chapter is presented in detail in Appendix A. spouse, and other individual characteristics such as age, household head’s sex and geographic indicators (urban or 3.1 Data and variables rural). Information on years of schooling is not included, but the information on educational attainment reports The estimation of the impact of trade policy changes on whether respondents have completed primary or lower welfare requires the use of various sources of data and a 2 These industry conversions are available from the authors on request. 23 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES secondary, secondary, post-secondary or equivalent, 4. Main results university, or post-graduate education. We use different definitions of skilled versus unskilled labor. In fact, there 4.1. Wage estimations is a trade-off in choosing the minimum standard for The results obtained from estimating Mincerian equations skilled labor. Although the vast majority of respondents (model 3 in Appendix A) are shown in Table C.1 in the report no education at all (or no graduation), having a Appendix. Our findings show that wages increase with meaningful indicator requires that the bar should not be education (the skill dummy represents literacy) and are set too low. As a benchmark, we define a person as skilled higher for male than for female workers. Concerning the if he or she had secondary or higher education. Summary effect of trade policy, Figure 1 plots point estimates and statistics and expenditure shares are reported in Tables confidence intervals for changes in tariffs differentiated B.1 and B.2 in the Appendix. by skill level. It displays a negative correlation between Unfortunately, individual wages are not recorded in these tariffs and wages, indicating that a reduction in tariffs surveys. Instead, industry-specific wage indices were will tend to increase wages. Apparently, higher educated obtained from the INS. They were linked to the household households are affected more by changes in tariffs. data using the industry of occupation of household The plot on the right-hand side of Figure 1 uses more members. Industry is reported following the 2-digit ISIC conservative industry-clustered standard errors for the classification (Rev. 3 for 2005 and Rev. 4 for 2010). That calculation of confidence intervals. The effect of trade leaves us with 61 industries for 2005 and 67 industries policy is still significant at the 5% level for skilled workers in 2010. Using industry-level data precludes studying and at the 10% level for unskilled workers. At the same inequality within industries, which is a major drawback of time, the difference between the two effects is no longer the dataset. Tariffs are from the WITS database. Another significant. limitation of the dataset is that there is no information on These results, however, should be treated with caution. whether households depend on the sale of agricultural or As mentioned above, the variation reflects merely 61 other goods for their income. For this reason, we can only sectors for 2005 and 67 sectors for 2010 for which data study the labor income channel, and not the producer were available. Monte Carlo simulations are used as a income channel when investigating how trade policy robustness check. They support our previous results as could affect households’ income. shown in Figure C.1 in the Appendix. Figure 1: Effect of trade policy on wages. Confidence intervals Robust SE Clustered SE Average Marginal Effects of weighted average with 95% Cls Average Marginal Effects of Tariffs (weighted) with 95% Cls Skilled Skilled Unskilled Unskilled -0.9 -0.8 -0.7 -0.6 -1.5 -1.0 -0.5 0 Effects on Linear Prediction Effects on Linear Prediction Source: World Development Indicators, PovcalNet 24 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 4.2 Welfare effects elimination of tariffs leads to a 1.025% increase in wages, Figures 2–4 show the consumption, earnings and and the interaction coefficient between weighted total welfare effects along the income distribution of tariffs and skill groups is -0.335 (both from the results in households, in the scenario of full elimination of tariffs. Table C.1). Finally, the squared root of the equivalence The main assumptions used to construct the figures scale is used to calculate household size. Results using are: skill level is defined in terms of literacy; the tariff household expenditure per head and the OECD modified pass-through is 0.10 (from Baghdadi et al., 2016); the equivalence scale are very similar, as shown below. Figure 2: Consumption effects Figure 2 shows the total consumption effect due to the change in the prices of the traded goods. The solid curve shows the estimated inverse compensating variation (explained in Section 3 above), which is downward sloping, indicating a pro-poor effect of liberalization. The average consumption effect is positive and significantly different from zero. The gains extend to 1% of initial household expenditure for low-income households and are close to zero for rich households, with the exception of a few outliers that also obtain higher gains. Note: Authors’ elaboration using wage elasticities in Table C.1 and incomplete pass-through. Figure 3: Earnings effects Figure 3 shows that the earnings effect is sizable, in contrast to Borraz et al. (2013) for Brazil. It seems that poor households, again, benefit more. Results for the wage effect should in our case be treated with caution. As can be seen above there is very little variation, and it is due to intersectoral rather than inter-household variation, and should be thus mainly explained by composition effects. Note: Authors’ elaboration using wage elasticities in Table C.1 and incomplete pass-through. Figure 4: Total welfare effect Figure 4 shows the total welfare effect for all households, which has been computed by aggregating the consumption and earnings effects. The increase in the welfare of low-income households amounts to about 2.5% of initial household expenditure, while welfare increases for richer households are less. Note: Authors’ elaboration using wage elasticities in Table C.1 and incomplete pass-through. 25 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The analysis of differences across population groups share of food products in their expenditure. There is (urban versus rural, gender, regions, and employment little systematic difference in the impact of trade policy types) is based on the consumption effect, as the by gender (Figure 6), and the distributional effects estimation of the wage effect has shown to be quite across regions seem to be rather similar (Figure B.1 in noisy and lacking important interpersonal variation. the Appendix). Finally, the increase in welfare for self- Poor people in rural areas would benefit slightly more employed, poor workers (in Tunisia, a useful proxy for than their counterparts in urban areas from trade policy informality) is only slightly higher than for other types of changes in Tunisia (Figure 5), probably due to the higher workers (Figure B.2). Figure 5: Consumption effect for urban and rural areas Rural Urban 0.012 0.012 0.012 0.012 0.01 0.01 0.01 0.01 Consumption Effect Consumption Effect Consumption Effect Consumption Effect 0.008 0.008 0.008 0.008 0.006 0.006 0.006 0.006 0.004 0.004 0.004 0.004 6 7 6 8 7 9 8 10 9 10 6 7 6 8 7 9 8 10 9 10 Capita Log Adjusted per Log Household Adjusted Expenditure per Capita Household Expenditure Log Adjusted per Log Capita Household Adjusted Expenditure per Capita Household Expenditure (Square Root Equivalence Scale) (Square Root Equivalence Scale) (Square Root (Square Root Equivalence Equivalence Scale) Scale) kernel = epanechnikov, degree kernel = 0, bandwidth = epanechnikov, = 0,pwidth = 0.24, degree = 0.36 bandwidth = 0.24, pwidth = 0.36 kernel kernel = epanechnikov, = epanechnikov, degree degree = 0, bandwidth = 0,pwidth = 0.28, bandwidth = 0.28, pwidth = 0.43 = 0.43 Figure 6: Consumption effect by gender Male Female 0.014 0.014 0.014 0.014 0.012 0.012 0.012 0.012 Consumption Effect Consumption Effect Consumption Effect Consumption Effect 0.01 0.01 0.01 0.01 0.008 0.008 0.008 0.008 0.006 0.006 0.006 0.006 6 76 87 98 9 10 10 6 76 87 98 9 10 10 Log Log Adjusted per Adjusted Capita per Capita Household Household Expenditure Expenditure Log Log Adjusted per Adjusted Capita per Capita Household Household Expenditure Expenditure (Square Root (Square Root Equivalence Equivalence Scale) Scale) (Square Root (Square Root Equivalence Equivalence Scale) Scale) kernel = kernel = epanechnikov, epanechnikov, degree degree = 0, bandwidth == 0, bandwidth 0.19, = 0.19, pwidth = 0.29 pwidth = 0.29 kernel = kernel = epanechnikov, epanechnikov, degree degree = 0, bandwidth == 0, bandwidth 0.31, = 0.31, pwidth = 0.47 pwidth = 0.47 26 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 4.3 Robustness Figure 8 the total effect (the earnings effect does not The greater positive impact of trade reform for the poor change due to this alternative assumption). Similarly, in Tunisia remains if we change some of the underlying using unadjusted per capita household expenditure to assumptions. A pass-through assumption of 0.5 (in line measure income also generates a pro-poor impact of with the results in Baghdadi et al. (2016) also results in tariff reduction (the fitted curve in Figure 9 is almost the the poor benefiting more from tariff reductions than same as in Figure 4, except that it looks smoother and the rich. Figure 7 provides the consumption effect and less influenced by outliers). Figure 7: Consumption effect with alternative pass-through Figure 8: Total welfare effect with alternative tariff pass-through Figure 9: Total welfare effect using per capita household expenditure 27 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES In comparison to other studies, the impact of trade reform of income. The increase in welfare due to lower prices on welfare using the estimated 10% pass-through for on consumption goods is larger for poor than for rich Tunisia of around 1% of initial household expenditure, is households. The welfare increase due to wage effects is similar to the gains found in Borraz et al. (2013) for Brazil, also positive, and greater for poor than for rich workers. but lower than the welfare effects found in Ural Marchand However, wage effects are less accurately estimated than (2012) for India and Porto (2006) for Argentina. However, the consumption gains, because the lack of individual when using a pass-through of 50%, the benefits are very wage data means we rely on average sectoral wages. similar—around 6% of initial expenditure—to those When added to the consumption effect, we find that estimated in Porto (2006).3 the welfare of the poor increases by about 2.5% when assuming that the tariff pass-through is low (about The main policy conclusion is that trade liberalization in 10%). This is a conservative estimate, given that the Tunisia would in fact reduce poverty if it is made through pass-through could be around 50 to 60%. a reduction of tariff barriers. This result is in line with the ex-ante analysis conducted by Minot et al. (2010) A limitation of this study is that the effect of the changes (see above). Similar to our findings, they also show that in the prices of traded goods on the prices of non-traded poverty will decline more in rural than in urban areas. goods has been excluded from the analysis. Nevertheless, these effects are probably small for Tunisia, where non- 5. Conclusions traded services are highly regulated and could only weakly respond to general equilibrium effects. We leave This paper examines how Tunisian households would be this issue for further research. Also, the data required to affected by further tariff liberalization. The distributive analyse the income effect on households that sell specific impacts from the perspective of both consumers and goods—for instance agricultural—are not available. workers are considered, as well as the price transmission Another issue is that this framework only takes static mechanism. In particular, the effects of trade effects into account. Trade policy could also change the liberalization or, more generally, trade policy reform on production structure of the economy, and this in turn household wellbeing and poverty, are identified and could have an effect on welfare. compared for the analyzed sectors. The overall effect is decomposed into a consumption and income effect Summarizing, the findings suggest that trade on wages, and separate results are shown for different liberalization in Tunisia could have a net positive welfare groups of households. We distinguish between rural and effect on households and that the benefit is higher for urban households and also show the effects by region, by poorer households. However, the magnitude of these gender and by type of employment. effects is estimated to be small in economic terms. The reduction in the prices of traded goods is found to improve welfare for all households along the distribution Summarizing, the findings suggest that trade liberalization in Tunisia could have a net positive welfare effect on households and that the benefit is higher for poorer households. However, the magnitude of these effects is estimated to be small in economic terms. 3 Porto (2006) assumed complete pass-through. 28 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Baghdadi, L., H. Kruse and I. Martinez-Zarzoso, (2016). “Trade policy without trade facilitation: Lessons from tariff pass-through in Tunisia.” In: Teh, R., M. Smeets, M. Sandi Jallab and F. Chaudri (eds.), Trade cost and inclusive growth. Case studies presented by WTO chair-holders. World Trade Organization. Borraz, F., D. Ferrés and M. Rossi (2013). “Assessment of the Distributive Impact of National Trade Reforms in Brazil,” Journal of Economic Inequality 11, 215-235. Deaton, A. (1989). “Rice prices and income distribution in Thailand: a non-parametric analysis.” Economic Journal 99, 1–37. Doyle, E. and I. Martinez-Zarzoso (2011). “Productivity and Trade and Institutional Quality: A Panel Analysis.” Southern Economic Journal 77 (3), 726-752. Fan, J. (1992). “Design-adaptive nonparametric regression.” Journal of the American Statistical Association 87 (420), 998–1004. Feenstra, R. and G.H. Hanson (2003). “Global Production Sharing and Rising Inequality: A Survey on Trade and Wages”, in: E. Kwan Choi and J. Harrigan (eds.), Handbook of International Trade, Oxford: Blackwell. Goldberg, P. and M. Knetter (1997). “Good prices and exchange rates: what have we learned?”, Journal of Economic Literature 35, 1243–1272. Heckman, J.J., L.J. Lochner and P.E. Todd (2003). “Fifty Years of Mincer Earnings Regressions,” NBER Working Paper 9732. Mincer, J. (1958). “Investment in Human Capital and Personal Income Distribution,”Journal of Political Economy 66(4), 281-302. Minot, N., M.A. Chemingui, M.Thomas, R. Dewina and D. Orden (2010). “Trade Liberalization and Poverty in the Middle East and North Africa,” IFPRI Research Monograph, Washington D.C. Nicita, A. (2009). “The price effect of tariff liberalization: Measuring the impact on household welfare,” Journal of Development Economics 89: 19-27. Nicita, A., M. Olarreaga and G. Porto (2014). “Pro-poor trade policy in Sub-Saharan Africa,” Journal of International Economics 92(2), 252-265. Porto, G. G. (2006). “Using survey data to assess the distributional effects of trade policy,” Journal of International Economics 70(1): 140-160. Ural Marchand, B. P. (2012). “Tariff Pass-Through and the Effect of Trade Liberalization on Household Welfare.” Journal of Development Economics 99 (2), 265-281. World Bank Group and World Trade Organization (2015). The Role of Trade in Ending Poverty. World Trade Organization: Geneva. World Bank (2016). Tunisia Poverty Assessment 2015. World Bank, Washington, DC. © World Bank. https:// openknowledge.worldbank.org/handle/10986/24410 License: CC BY 3.0 IGO. World Trade Organization (2005). Trade Policy Review: Tunisia. WTO Secretariat, Geneva. 29 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix Appendix A: Methods A.1 Analytical Framework 4 4 Pkt can be interpreted as elements of the stacked vector P= (PT ) PN 30 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 5 5 We thank Robert Teh for suggesting this approach. 31 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix B: Data Description Table B.1: HH survey summary statistics 2005 2010 Variable Obs Mean Std. Dev. Obs Mean Std. Dev. Ln Wage 4390 5.44 0.34 3192 5.90 0.46 Weighted Average Tariff 4390 19.96 7.41 3192 26.10 6.57 Skilled 1: Literate 4390 0.54 0.50 3192 0.67 0.47 Skilled 2: Primary 3847 0.11 0.31 3153 0.11 0.31 Skilled 3: Secondary 3847 0.02 0.14 3153 0.03 0.17 Skilled 4: Post-secondary 3847 0.01 0.09 3153 0.02 0.12 Skilled 5: University 3847 0.00 0.02 3153 0.00 0.05 Age 4390 48.19 12.99 3192 48.87 12.40 Urban Dummy 4390 0.37 0.48 3192 0.42 0.49 Male Dummy 4390 0.64 0.48 3192 0.71 0.45 Table B.2: Expenditure shares 2005 2010 Expenditure shares Obs Mean Std. Dev. Obs Mean Std. Dev. Food 12315 41.93% 13.67% 11278 35.78% 12.05% Clothes and footwear 11265 7.97% 6.88% 10440 8.41% 6.79% Housing and utilities 12317 21.88% 11.94% 11281 25.53% 12.24% Transport 9028 9.77% 9.92% 8404 8.98% 8.95% Communication 9785 4.21% 3.01% 9816 5.61% 3.82% Recreation 10217 6.18% 6.30% 7102 1.46% 2.97% Education 7694 4.30% 4.24% 6502 3.40% 3.26% Personal care 12275 10.17% 8.69% 11038 8.74% 8.21% 32 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure B.1: Consumption effect by region expenditure Grand Tunis North East Grand Tunis North East 0.02 Grand Grand Tunis Tunis 0.02 North East North East Effect Effect 0.02 0.02 Consumption Effect Consumption Effect 0.015 0.02 0.015 0.02 0.02 0.02 Effect Effect Effect Effect Consumption Consumption 0.015 0.015 0.01 0.015 0.01 0.015 0.015 0.015 0.01 0.01 Consumption Consumption Consumption Consumption 0.005 0.01 0.005 0.01 0.01 0.01 0.005 0.005 0.0050 0.0050 0.005 0.005 0 0 0 0 6 7 8 9 10 11 0 0 6 7 8 9 10 11 6 7 Log Adjusted 9 8 Capita Household per 10 Expenditure 11 6 7 Log Adjusted 9 8 Capita Household per 10 Expenditure 11 6 6 7 7 8 8 9 9 10 10 11 11 6 6 7 7 8 8 9 9 10 10 11 11 per Root (Square Log Adjusted Capita Household Scale) Equivalence Expenditure (Square Log Adjusted per Root CapitaEquivalence Household Scale) Expenditure Log Adjusted Log (Squareper Adjusted per Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) Adjusted Log Adjusted Log per (Squareper Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) kernel = epanechnikov, degree kernel = epanechnikov, Root= degree = 0.29, pwidth = 0.44 0, bandwidthScale) (Square (Square Root= Root 0, bandwidthScale) Equivalence Equivalence = 0.24, pwidth = 0.36 Scale) (Square (Square Root Equivalence Equivalence Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.24, pwidth = 0.36 kernel = epanechnikov, degree = 0, bandwidth = 0.29, pwidth = 0.44 kernel kernel == epanechnikov, degree = epanechnikov, degree 0, bandwidth = 0, = 0.29, 0.29, pwidth bandwidth = = 0.44 pwidth = 0.44 = epanechnikov, degree = 0, kernel = epanechnikov, degree = 0, bandwidth = bandwidth 0.24, pwidth = 0.24, = 0.36 pwidth = 0.36 kernel North West Center East North West Center East 0.02 North West North West 0.02 Center East Center East Effect Effect 0.02 0.02 Consumption Effect Consumption Effect 0.015 0.02 0.015 0.02 0.02 0.02 Effect Effect Effect Effect Consumption Consumption 0.015 0.015 0.01 0.015 0.01 0.015 0.015 0.015 0.01 0.01 Consumption Consumption Consumption Consumption 0.005 0.01 0.005 0.01 0.01 0.01 0.005 0.005 0.0050 0.0050 0.005 0.005 0 0 0 0 6 7 8 9 10 11 0 0 6 7 8 9 10 11 6 7 Log Adjusted 8 Capita Household per 9 10 Expenditure 11 6 Log 7 Adjusted 8 Capita Household per 9 10 Expenditure 11 6 6 7 7 8 8 9 9 10 10 11 11 6 6 7 7 8 8 9 9 10 10 11 11 (Square Log Adjusted per Root Capita Household Scale) Equivalence Expenditure (Square Log Adjusted per Root Equivalence Household Scale) Capita Expenditure Log Adjusted Log Adjusted per (Squareper Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) Log Log Adjusted (Squareper Adjusted per Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) kernel = epanechnikov, degree kernel = epanechnikov, degree = 0, bandwidth = 0.3, pwidth = 0.44 (Square (Square Root= Root 0, bandwidthScale) Equivalence Equivalence = 0.39, pwidth = 0.59 Scale) (Square Root Equivalence (Square Root Equivalence Scale) Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.39, pwidth = 0.59 kernel = epanechnikov, degree = 0, bandwidth = 0.3, pwidth = 0.44 kernel = kernel = epanechnikov, epanechnikov, degree degree = = 0, bandwidth = 0, bandwidth = 0.39, 0.39, pwidth pwidth = = 0.59 0.59 kernel = kernel epanechnikov, degree = epanechnikov, degree = 0, bandwidth = 0, = 0.3, bandwidth = pwidth = 0.3, pwidth = 0.44 0.44 Center West South East Center West South East 0.02 Center Center West West 0.02 South South East East Effect Effect 0.02 0.02 Consumption Effect Consumption Effect 0.015 0.02 0.015 0.02 Effect Effect 0.02 0.02 Effect Effect Consumption Consumption 0.015 0.015 0.01 0.015 0.01 0.015 0.015 0.015 0.01 0.01 Consumption Consumption Consumption Consumption 0.005 0.01 0.005 0.01 0.01 0.01 0.005 0.005 0.0050 0.0050 0.005 0.005 0 0 0 0 6 7 8 9 10 11 0 0 6 7 8 9 10 11 6 7 8 9 10 11 6 7 8 9 10 11 6 Log Adjusted 7 per Capita 8 Household 9 Expenditure 10 11 6 Log Adjusted 7 per Capita 8 Household 9 Expenditure 10 11 6 7 8 9 10 11 6 7 8 9 10 11 per Root (Square Log Adjusted Capita Household Scale) Equivalence Expenditure (Square Log Adjusted per Root CapitaEquivalence Household Scale) Expenditure Log (Squareper Log Adjusted Adjusted per Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) Log Log Adjusted (Squareper Adjusted per Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.21, pwidth = 0.32 kernel = epanechnikov, degree = 0, bandwidth = 0.28, pwidth = 0.42 (Square Root Equivalence (Square Root Equivalence Scale) Scale) (Square (Square Root Root Equivalence Equivalence Scale) Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.21, pwidth = 0.32 kernel = epanechnikov, degree = 0, bandwidth = 0.28, pwidth = 0.42 kernel kernel = epanechnikov, degree = epanechnikov, = 0, degree = 0, bandwidth bandwidth = = 0.21, pwidth = 0.21, pwidth = 0.32 0.32 kernel = kernel = epanechnikov, degree = epanechnikov, degree 0, bandwidth = 0, = 0.28, bandwidth = 0.28, pwidth pwidth = = 0.42 0.42 South West South West 0.02 South South West West Effect 0.02 Consumption Effect 0.015 0.02 Effect 0.02 Effect Consumption 0.015 0.01 0.015 0.015 0.01 Consumption Consumption 0.005 0.01 0.01 0.005 0.0050 0.005 0 0 0 6 7 8 9 10 11 6 7 8 9 10 11 6 6 Log Adjusted 7 7 per Capita 8 8 Household 9 9 Expenditure 10 10 11 11 per Root (Square Log Adjusted Capita Household Scale) Equivalence Expenditure Log Adjusted Log (Squareper Adjusted per Capita Capita Root Household Household Equivalence Expenditure Expenditure Scale) (Square kernel = epanechnikov, degree (Square Root Root =Equivalence Scale) 0, bandwidth = Equivalence 0.31, pwidth = 0.47 Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.31, pwidth = 0.47 kernel kernel = epanechnikov, degree = epanechnikov, degree = = 0, 0, bandwidth bandwidth = 0.31, pwidth = 0.31, = 0.47 pwidth = 0.47 33 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure B.2: Consumption effect by employment type Employees Employers 0.014 0.014 0.012 0.012 Employees Employers Effect Effect 0.014 0.014 0.01 0.01 Consumption Consumption Consumption Consumption 0.012 0.012 0.008 0.008 Effect Effect 0.01 0.01 0.006 0.006 0.008 0.008 0.004 0.004 0.006 6 7 8 9 10 0.006 6 7 8 9 10 Log Adjusted per Capita Household Expenditure Log Adjusted per Capita Household Expenditure (Square Root Equivalence Scale) (Square-root Equivalence Scale) 0.004 0.004 kernel = epanechnikov, degree = 0, bandwidth = 0.23, pwidth = 0.35 kernel = epanechnikov, degree = 0, bandwidth = 0.66, pwidth = 0.99 6 7 8 9 10 6 7 8 9 10 Log Adjusted per Capita Household Expenditure Log Adjusted per Capita Household Expenditure (Square Root Equivalence Scale) (Square-root Equivalence Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.23, pwidth = 0.35 kernel = epanechnikov, degree = 0, bandwidth = 0.66, pwidth = 0.99 Self-Employed Unemployed/Not stated 0.014 0.014 0.012 0.012 Self-Employed Unemployed/Not stated Effect Effect 0.014 0.014 0.01 0.01 Consumption Consumption Consumption Consumption 0.012 0.012 0.008 0.008 Effect Effect 0.01 0.01 0.006 0.006 0.008 0.008 0.004 0.004 0.006 6 7 8 9 10 0.006 6 7 8 9 10 Log Adjusted per Capita Household Expenditure Log Adjusted per Capita Household Expenditure (Square Root Equivalence Scale) (Square-root Equivalence Scale) 0.004 0.004 kernel = epanechnikov, degree = 0, bandwidth = 0.23, pwidth = 0.35 kernel = epanechnikov, degree = 0, bandwidth = 0.28, pwidth = 0.43 6 7 8 9 10 6 7 8 9 10 Log Adjusted per Capita Household Expenditure Log Adjusted per Capita Household Expenditure (Square Root Equivalence Scale) (Square-root Equivalence Scale) kernel = epanechnikov, degree = 0, bandwidth = 0.23, pwidth = 0.35 kernel = epanechnikov, degree = 0, bandwidth = 0.28, pwidth = 0.43 Appendix C: Results for the Mincerian significant at the 10% level. However, the effect of tariffs Equation and the interaction between tariffs and skills are jointly significant. Table C.1 reports results for equation (3) using robust standard errors in column (1) and industry-clustered In Figure C.1 we report confidence intervals with robust standard errors in column (2). There is a negative standard errors (the dashed vertical lines) and point effect of tariffs on wages. The results are stronger for estimates (solid lines). In addition, we add the distribution skilled workers. Introducing standard errors clustered of point estimates in our Monte Carlo distribution. at the industry level renders our results by and large Evidently, the peak is in all cases close to the original insignificant except for the main effect which remains point estimate. 34 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table C.1: Results mincerian equation (1) (2) Variables Robust SE Clustered SE -0.686a -0.686c Weighted tariff (0.0487) (0.297) -0.136a -0.136 Weighted tariffc skill dummy (0.0404) (0.178) 0.0345b 0.0345 Skill dummy (0.0148) (0.0645) 0.00285a 0.00285 Age (0.000835) (0.00185) -2.30e-05a -2.30e-05 Age squared (7.42e-06) 1.87e-05) -0.00886c -0.00886 Urban dummy (0.00508) (0.00856) 0.0153a 0.0153 Male dummy (0.00388) (0.0234) 7.938a 7.938a Constant (0.0295) (0.0970) Observations 9,820 9,820 R-squared 0.877 0.877 Industry FE Yes Yes Time FE Yes Yes Note: standard errors in brackets. a p<0.01, b p<0.05, c p<0.1. Figure C.1: Monte Carlo simulation results Effect of Tariffs on Wages 2.5 2 1.5 Density 1 0.5 0 -1.5 -1 -0.5 0 kernel = epanechnikov, bandwidth = 0.0369 Skilled Unskilled 35 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Gender Welfare Effects of Regional Trade Integration on Households in Ghana Emmanuel Orkoh, North-West University, South Africa 1, 2 1. Introduction states, a reduction in the complexities associated with rules of origin requirements and protection of some O ver the past two to three decades, a number emerging sectors. In 2015, about 12% of ECOWAS of developing countries have pursued exports went to other member countries, 6% to other regional economic integration to harmonize African countries and about 80% outside of Africa. trade policies and increase their weight in The region ranked third in the 2016 Africa Regional global trade. Economic integration, particularly in Africa, Integration Index (AfDB, OECD, UNDP, 2017). However, has also been seen as a way to diversify the structure the potential challenge for the CET is its coherence with of African economies, boost intra-African trade and the broader objectives of Africa’s Continental Free Trade investment, build supply capacity, and sustainably reduce Area (CFTA), which seeks to harmonize or replace existing poverty (Osakwe, 2015). These integration efforts have arrangements governing trade and the movement of resulted in the creation of regional blocs such as the persons in the continent. The concern has been whether Economic Community of West African States (ECOWAS), the CFTA, (ratified by 44 out of the 55 AU member states West African Economic and Monetary Union (WAEMU), during its extraordinary summit held between 17–21 Common Market for East and Southern Africa (COMESA), March 2018 in Kigali), will add a layer of complexity or Economic Community of Central African States (ECCAS), will simplify the existing agreements enshrined in the Central African Economic and Monetary Community CETs and other Regional Economic Communities (RECs) (CEMAC), Southern African Customs Union (SACU), and (Gutowski, Knedlik, Osakwe, Ramdoo & Wohlmuth, 2016; Arab Maghreb Union (AMU). The trade-related objectives UNCTAD, 2016). of these blocs include the establishment of custom Regional trade integration through the creation of a unions,3 with a common external tariff (CET) as a major customs union with a CET has been found to have both trade policy instrument. direct and indirect implications for household poverty In January 2015, ECOWAS began the implementation of and welfare in general. Trade integration affects poverty a common external tariff (CET), a process expected to and welfare through three main channels: (a) changes in be completed by 2020. The envisioned benefits of the employment structures and wages; (b) changes in prices CET include a reduction in lost revenues that arise from and their impact on consumption and production patterns; competition in external tariff rates between the member and (c) changes in financing for social expenditure by 1 TRADE Research Focus Area, Faculty of Economic and Management Sciences, North-West University, Potchefstroom Campus, South Africa. Email: aorkoh@gmail.com 2 The author would like to thank David Zavaleta, of Universidad Católica Boliviana, Bolivia, for his very useful advice and excellent mentorship during the development of this study. Sincere appreciation also goes to the Virtual Institute team and Jane Elizabeth Casabianca, of Università Politecnica delle Marche, Italy, whose feedback and guidance have helped shape the study. This work was supported by the government of Finland [grant number 50039]. 3 A customs union is a trade agreement under which certain countries preferentially grant tariff-free market access to each other’s imports and agree to apply a common set of tariffs to imports from the rest of the world. That is, they enter into a free trade agreement and apply a common external tariff to imports from non-members (Adams, 1993). 36 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES governments (Winters et al., 2004; Harrison and Tang, decisions of households (Marchand, 2012; Nicita, 2009). 2005). Trade reforms can also affect poverty indirectly We can therefore consider the price channel as the via economic growth. Increased trade openness resulting mechanism that affects households more directly in the from trade reforms can improve access to technology short term. and hence foster productivity growth, leading to faster Between 2007 and 2015, Ghana alternated between economic growth and reduced poverty (UNCTAD, 2010). its own tariff system and the CET of the regional Conversely, increased trade restrictions can impede economic bloc of which it was a member at a given time. productivity growth and slow economic growth, leading From 2007 to 2011, the country implemented its own to increased poverty. tariffs, but in 2012 it adopted the WAEMU CET, before Among the various channels of effects, this study analyses switching back to its own tariff system in 2013 and the price channel. The focus on this channel is due to the finally adopting the ECOWAS CET starting in 2015. This fact that most often trade policies such as a CET affect followed negotiations by ECOWAS members in Dakar, import tariffs and thus domestic prices of commodities, Senegal, on the CET for the region, which concluded in which in turn affect the consumption and production October 2013 (Roquefeuil et al., 2014).4 4 According to Article 3.2(d) of the revised ECOWAS treaty, one of the aims of ECOWAS is “the establishment of a common market through: (i) the liberalization of trade by the abolition, among Member States, of customs duties levied on imports and exports, and the abolition, among Member States, of non-tariff barriers in order to establish a free trade area at the Community level; (ii) the adoption of a common external tariff and a common trade policy vis-à-vis third countries; and (iii) the removal, between Member States, of obstacles to the free movement of persons, goods, service and capital, and to the right of residence and establishment.” (see http://www.courtecowas.org/site2012/pdf_files/revised_treaty.pdf#page=4&zoom=auto,-82,12). The member countries of ECOWAS are Benin, Burkina Faso, Cape Verde, Côte d’Ivoire, the Gambia, Ghana, Guinea, Guinea-Bissau, Liberia, Mali, Niger, Nigeria, Sierra Leone, Senegal, and Togo. 37 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Ghana’s implementation of the ECOWAS CET will result The aim of this study is to measure the in significant changes in the country’s tariff structure effects that implementation of the CET for both agricultural and non-agricultural products will have on household poverty, income, (World Bank Group, 2015). Some of the tariff rates will and consumption through the price be lower and some will be higher as a result of the CET. channel, with a special focus on gender. For instance, implementation of the CET will reduce the number of commodities admitted under zero percent tariff rates from 725 to 85, but increase the number of commodities admitted under the 5% band from 375 to characteristics such as whether the household is a net 2,146. The changes in tariffs related to implementation of producer or a net consumer of goods whose prices have the CET will affect prices and, consequently, the welfare changed, urban/rural location, and economic and social of households, depending on their position as either net status. In the Ghanaian context, the extent to which this producers or consumers of these products. assertion holds remains an important policy question, especially in the wake of the country’s recent adoption Given the different roles of men and women in society and implementation of the ECOWAS CET. and the economy, trade policies such as those enshrined in the ECOWAS CET have different implications for male- The aim of this study is to measure the effects that and female-headed households. This assertion has been implementation of the CET will have on household widely supported by findings of several trade-gender poverty, income, and consumption through the price specific studies. For instance, Bird (2004) emphasizes channel, with a special focus on gender. A review of the that changes associated with trade integration may be literature reveals that even though some ex-ante studies positive or negative for women and men depending of this nature on Ghana have looked at the poverty and on their individual characteristics, including education income effects of trade liberalization (Bhasin and Annim, and skills, marital status, family size, social group 2005; Bhasin, 2012), no study has focused on the impact 38 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES of the different ECOWAS CET bands on household welfare and the gender dimensions of this impact. The aim of this study is to fill these gaps and contribute to the existing literature on the links between international trade policy and household welfare. To assess the implications of the CET, the study applies a top-down approach by combining a macro computable general equilibrium (CGE) (top-down) model and a micro (bottom-up) household model, with the latter using data at the household level (Bourguignon and Savard, 2008). Section 2 provides an overview of the evolution of Ghana’s tariffs and poverty over the years. Section 3 explains the methodology and Section 4 the data used in the study, while Section 5 describes the simulations and presents the results. The final section provides conclusions. It is hoped that the findings of this study may serve as input to policy makers and industry in formulating gender- aware policies to ensure that all population segments and household categories share appropriately in the gains and losses associated with the country’s adoption of the CET. More broadly, it is hoped that the study may help policy makers in formulating policies to enhance gender equality and promote human development. 2. Overview of Ghana’s trade reforms and poverty trends Ghana’s trade policy evolved from being fairly liberal in the 1950s to a significantly controlled regime in the 1970s, after which the country embarked on major trade liberalization and other economic reforms in the 1980s. This approach to trade policy has been greatly influenced by developments in international trade under the General 0%, 25%, and 30% the following year while some import Agreement on Trade and Tariffs (GATT) and the World controls remained in place. Further reductions were Trade Organization (WTO). It has also been shaped by made in 1986, when the higher rates were lowered to 20 trade agreements between Ghana and its major trading and 25% (Ackah and Aryeetey, 2012). partners, the country’s economic development policy, and the structural adjustment programs of the World Major trade policy reforms took place between 2007 and Bank and the International Monetary Fund, particularly 2015, when the most-favored-nation (MFN) tariff applied in the 1980s and 1990s (Ackah and Aryeetey, 2012). by the country was frequently modified. In 2012, the Significant trade liberalization in Ghana began with the country adopted the WAEMU CET, which was based on downward adjustment of tariffs in 1983, from rates of four tariff bands comprised of a zero duty on social goods 35%, 60%, and 100% to rates of 10%, 20%, 25%, and such as medicine and publications, 5% duty on imported 30%. The tariffs were further simplified and lowered to raw materials, 10% duty on intermediate goods, and 20% 39 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES duty on finished goods (Office of the United States Trade a reduction of duties on some products and an increase Representative, 2014). In 2013, Ghana switched from in the duties on others.5 As shown in Table 1, the average the WAEMU CET back to its own national tariffs. In this unweighted applied MFN tariff in 2013 was 12.8%, context, it abolished the non-ad-valorem tariffs applied compared to the 12.7% rate in 2007. The MFN rates on to petroleum products, and replaced them with ad agricultural products were generally higher than those valorem rates in January 2014. This was accompanied by on non-agricultural products. Table 1: Trends in Ghana’s most-favored-nation tariffs, 2007–2015 (percent) Ghana WAEMU Ghana ECOWAS Change Change Categories 2007 CET 2012 2013 CET 2015 2007–2013b 2013–2015b Total 12.7 12.3 12.8 12.3 1.0 -4.0 By Harmonized System categorya Agricultural products 17.5 14.9 17.3 15.6 -1.0 -10.0 Animals and products thereof 19.4 18.5 19.0 23.9 -2.0 26.0 Dairy products 20.0 14.4 20.0 16.0 0.0 -20.0 Fruit, vegetables, and plants 18.9 17.6 18.3 17.6 -3.0 -4.0 Coffee and tea 20.0 17.2 20.0 12.0 0.0 -40.0 Cereals and preparations 17.8 12.7 16.2 13.5 -9.0 -17.0 Oils seeds, fats, oil and their products 14.6 10.5 14.6 14.1 0.0 -3.0 Sugar and confectionary 11.1 13.3 11.0 13.8 -1.0 25.0 Beverages, spirits, and tobacco 19.8 19.0 19.8 17.0 0.0 -14 Other agricultural products 14.4 9.4 15.1 9.5 5.0 -37 Non-agricultural products 12.0 11.8 12.0 11.7 0.0 -3.0 Fish and fishery products 11.1 15.5 9.8 15.4 -12.0 57.0 Minerals and metals 12.2 11.8 12.5 11.7 2.0 -6.0 Chemicals and photographic supplies 11.9 7.7 12.1 8.0 2.0 -34.0 Wood, pulp, paper, and furniture 16.1 11.3 16.8 11.4 4.0 -32.0 Textiles 16.9 16.5 16.8 16.0 -1.0 -5.0 Clothing 20.0 20.0 20.0 20.8 0.0 4.0 Leather, rubber, footwear, and travel goods 14.3 14.2 15.0 12.9 5.0 -14.0 Non-electric machinery 2.8 7.3 3.1 7.0 11.0 126.0 Electric machinery 10.3 11.3 10.6 11.2 3.0 6.0 Transport equipment 6.0 11.0 5.5 10.2 -8.0 85.0 Non-agricultural products n.e.s. 15.6 14.3 15.0 14.3 -4.0 -5.0 Petroleum 9.0 7.9 4.3 7.9 -52.2 84 By ISIC sector Agriculture, hunting and fishing 15.7 13 15.1 11.5 -4.0 -24 Mining and quarrying 11.2 5.0 11.2 5.1 0.0 -54 Manufacturing 12.6 12.4 12.7 12.5 1.0 -2.0 Source: Prepared by the author based on data from WTO (2014). a: The Harmonized Commodity Description and Coding System (HS) is a multipurpose international product nomenclature developed by the World Customs Organization. b: ”Change 2007-2013” and “Change 2013-2015” are the percentage changes in tariffs for 2007–2013 (before the ECOWAS CET), and 2013–2015 (after the ECOWAS CET), respectively. WTO (2014) explains that the 2007 tariff is based on HS 2002 nomenclature consisting of 5,969 tariff lines (at the 10-digit tariff line level). The 2013 tariff is based on HS 2012 nomenclature consisting of 6,062 tariff lines (at the 10-digit tariff line level). The WAEMU tariff schedule consists of 2012 tariff rates (5,550 tariff lines at the 10-digit tariff line level) based on the HS 2007 nomenclature, while the ECOWAS tariff schedule is based on HS 2012 nomenclature consisting of 5,899 tariff lines (at the 10-digit tariff line level). According to WTO (2014), the tariff data were obtained from Ghanaian authorities. CET: common external tariff; ECOWAS: Economic Community of West African States; ISIC: International Standard Industrial Classification; n.e.s.: not elsewhere specified; WAEMU: West African Economic and Monetary Union. 5 Items for which duties were reduced from as high as 20% to duty-free included fish livers, roe and flour, seeds, clinker and bulk cement, gasoil and related products, fishing yarn and equip- ment, mosquito nets, and contact lenses. Items for which tariffs increased included mobile phones, online telephone sets, cordless handsets, rough wood, ferrous and non-ferrous metal scrap, air coolers, and battery chargers. 40 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Subsequently, as a member of ECOWAS, Ghana endorsed Figure 1: Poverty distribution in Ghana between 1991 and 2013 the ECOWAS CET, which was adopted by ECOWAS Ministers of Finance on 20 March 2014, and came into 60 55 effect on 1 January 2015. The ECOWAS CET is based on 52 50 the four tariff bands of the WAEMU CET and an additional 43 40 41 fifth band involving a 35% duty on goods in ‘sensitive’ 40 35 34.9 Percentage sectors such as poultry and rice that the government 31.9 sought to protect. The revision of Ghana’s trade policy 30 25.9 24.2 22.1 triggered by implementation of the ECOWAS CET in 2015 19.1 20 resulted in considerable changes in Ghana’s tariff structure for agricultural and non-agricultural products. Overall, 10 there was a slight reduction in the country’s average 0 unweighted applied MFN tariff from 12.8 to 12.3%. 1991/1992 1998/1999 2005/2006 2012/2013 Gender The implications of changes in import tariffs for poverty National Female Male and household welfare are important policy issues. Poverty indicators based on reports of the last four rounds of the Ghana Living Standard Survey (GLSS) show that poverty in the country declined considerably from growth was largely driven by the services and agricultural 1991 to 2013,6 although there were some variations sectors, where the shares of women’s employment are across regions and across segments of the population. higher than the shares of men’s employment. The decline in poverty since 1998/1999 was concentrated mostly in the Central, Western, Eastern, Upper East, and 3. Methodology Northern regions of Ghana. Households of farmers in general, the non-farm self-employed, and public sector This study applies a top-down approach by combining a employees enjoyed the greatest gains in their standard macro CGE (top) model and a micro (bottom) household of living, while private sector employees and households model (Bourguignon and Savard, 2008). The CGE model with unemployed heads experienced the smallest gains. used for the macrosimulation is based on the dynamic Consistent with the general reduction in the poverty (recursive) computable general equilibrium (DCGE) level, female-headed households appear to be better off model developed by Breisinger et al. (2008). The model than male-headed households and are increasingly less is an extended version of a static standard CGE model impoverished (Figure 1) (Ghana Statistical Service, 2007). developed in the early 2000s by Löfgren et al. (2002) at the International Food Policy Research Institute Poverty at the national level decreased by 52.5% (IFPRI) (Diao, 2011). The Ghana DCGE is an economy- between 1991 and 2012—with the reduction in wide, multisectoral model that simultaneously and poverty of female-headed households being slightly endogenously solves for a series of economic variables, greater than that of male-headed ones (54.3% and 52% including commodity prices. It is made up of households reductions, respectively). The poverty level remained aggregated into a small number of representative consistently lower among female-headed households household. On the other hand, the micro (bottom) model than male-headed households, which is contrary to the considers all the households in the Ghana Living Standard “feminization of poverty” hypothesis. This may be partly Survey and models their behavior. due to the fact that over these years, Ghana’s economic 6 Poverty indicators are based on reports of the last four rounds of the Ghana Living Standards Survey (GLSS). These data are subject to two caveats. First, the contribution of the various tariff reforms to this reduction in poverty remains an important policy question. This is because there have been several policy interventions, including the Livelihood Empowerment Against Pov- erty Programme and the Ghana School Feeding Programme. Their contribution to the reduction in poverty among households will be difficult to disentangle from the effects of the reform. Second, the poverty estimates of the 2012/2013 survey may not be fully comparable with the estimates of the previous four GLSS rounds because of changes in the Consumer Price Index basket and new consumer items that have been introduced onto the market, leading to changes in household consumption. Only the 2005/2006 indicators were adjusted by the Ghana Statistical Service to make them comparable to the 2012/2013 indicators (Ghana Statistical Service, 2014). 41 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The top-down approach required that the two the household to compensate for the effects of the price frameworks would be used sequentially: first, we used change. It is imperative to stress that the household can the CGE model to simulate the effect of tariff changes be both a consumer and a producer of the commodity. between 2013 and 2015 on commodity prices. Then in Assuming that the price increases and the household is a the second stage, the simulated percentage changes net producer (Si > Si*), it will benefit from this price change. in prices of goods and services were passed down to On the other hand, if the household is a net consumer the microsimulation model, taking into consideration (Si < Si*), then a price increase will make it worse off. the gender of the household head, as shown in Figure 2. In linking the parameters from the CGE to the 4. Data and description of household microsimulation model to assess the consumption and statistics poverty effects, we matched the commodities in the SAM The CGE model used in this study was built using the with the same commodities in the household survey data, 2005 SAM for Ghana, which was constructed by IFPRI and then applied the first-order approach as described based on the fifth round of the Ghana Living Standard in Deaton (1989). This approach consists of calculating Survey.7 Effort was made to update the model to 2013, the share of household consumption expenditure and but the needed data were not available. As a result, the income (where the household is also a producer in the model was used to simulate the changes in prices from case of farmers) related to the commodities for each 2007 to 2013 (before the ECOWAS CET), and from 2013 household. These shares are multiplied by the changes in to 2015, after implementation of the ECOWAS CET. The prices obtained from the CGE model, and added to obtain model is a comprehensive dataset that encapsulates all the total change in welfare. the information contained in the national income and Following Deaton (1997), the function for the net welfare product accounts and the input-output table, as well as effect of the changes in prices for each commodity can be the monetary flows between institutions in the country. specified as: The SAM estimates the structure of the Ghanaian economy in 2005 and includes detailed information on ∂X0 = Si∂lnPi–Si* ∂lnPi = (Si–Si*)∂lnPi, (1) Y 56 production sectors, six factors of production, income where Si and Si* are, respectively, the income and budget and expenditures of rural and urban households, the ∂X0 government budget, and the balance of payments shares of commodity i, and Y is the compensating variation associated with a change in the price of good (Breisinger et al., 2007). The data on tariffs (presented i. The compensating variation is the revenue that the in Table 1), obtained from WTO (2014), were based on social planner (government) would have to provide to calculations of the WTO Secretariat using data provided Figure 2: The top-down computable general equilibrium approach Output Vector of changes in prices, income, Computable general equlibrium model wage rate, interest rate and quantities (eg., output level) Input Output New income, poverty and consumption Microsimulation model level after microsimulation Output Output Female-headed households Urban and rural households Male-headed households Source: Adapted from Bourguignon and Savard (2008). 7 The SAM dataset was obtained from the IFPRI website, and the GLLS6 dataset from the Ghana Statistical Service. The SAM can be downloaded from IFPRI at http://www.ifpri.org/ publication/ghana-social-accounting-matrix-2005 and the GLSS6 from the Ghana Statistical Service at http://www.statsghana.gov.gh/nada/index.php/catalog/72. 42 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES by Ghanaian authorities. Table 2 shows the import and households were increased from 580 and 8,700 structure based on the SAM. to 1,200 and 18,000, respectively. This represents an increase of about 107% compared to the GLSS5 (Ghana In building the microsimulation model, the study relied Statistical Service, 2014). The household survey data used on the 2012/2013 round of the Ghana Living Standard for the micro-level analysis covered 16,772 household Survey (GLSS6), which provides nationally and regionally heads, with more male-headed households (69.5%) than representative indicators covering a broad range of female-headed households (30.5%) (Table 3). Most of the topics such as education, health, employment, housing female household heads lived in urban areas. conditions, migration, tourism, poverty, household agriculture, access to financial services, and asset Figure 3 compares the structure of employment and ownership. In order to address the needs of Savannah average consumption expenditure across different Accelerated Development Authority (SADA) areas and categories of households (female/male and rural/urban) also to provide nationally representative quarterly labor and shows that, in general, female-headed households force statistics, the numbers of primary sampling units spend more on food than male-headed households. Table 2: Ghana’s imports of selected commodities as a percentage of total imports Commodity Import share Commodity Import share Maize a 0.2 Clothing 4.5 Rice 3.4 Footwear 0.9 Other grainsa 0.1 Pulp and paper 0.4 Other crops a 0.2 Oils a 9.6 Chicken 1.5 Fuel 4.7 Beefa 0.7 Fertilizers 2.6 Goata 0.2 Chemicals 6.4 Other livestocka 0.4 Metals 2.7 Formal food processing 8.2 Capital goods 43.9 Dairy 0.2 Electricitya 0.1 Meat 2.8 Other servicesa 4.9 Textile 1.4 Source: Author’s calculations based on Ghana’s 2005 Social Accounting Matrix (SAM). a  Though the SAM reports imports for these commodities, no import tariffs are reported. Note: The structure of the SAM shows that there is not an import share for all commodities. Table 3: Distribution of households by gender of household head and place of residence Gender and place of residence Number of households Share (%) Male household head 11,652 69.50 Female household head 5,120 30.50 Total 16,772 100.00 Female household heads in rural areas 1,950 11.63 Female household heads in urban areas 5,532 32.98 Male household heads in rural areas 3,170 18.90 Male household heads in urban areas 6,120 36.49 Total 16,772 100.00 Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). 43 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 3: Structure of employment and average consumption by gender of household head 2500 Non-food 2195.20 2193.80 Per Capita Expenditure (Ghana cedis) Food 2000 1616.70 1550.70 1521.50 1535.10 1493.10 1500 1406.70 1409.90 1417.80 1342.90 1322.20 1319.90 1324.50 1242.20 1287.20 1000 500 0 Male Female Male Female Male Female Male Female Employed Unemployed Retired Other inactive Gender and employment status of household head Source: Prepared by the author based on the Ghana Living Standard Survey (GLSS6) data. This observation is consistent across all the categories Figure 4 shows that on average, urban households spend of employment status except households in which the more on both food and non-food items than their rural head is unemployed. Conversely average expenditure counterparts. Apart from a household headed by a on non-food items is higher in all the categories of male- retiree, all the households in the rural areas spend more headed households, except those in which the head is not on food than non-food items. As in the distribution across employed. Observe further that average expenditure and gender of the household head, the average expenditure the differences in expenditure between food and non- on non-food items and the difference between food items are relatively higher in households in which expenditure on food and non-food items are higher for the head is retired compared to the other households. households in the urban area headed by a retiree than This could be due to the fact the households with a retired the other households. Households in the rural areas that head are more likely to have more members who are in have unemployed heads spend the least on food. These the labour-force than the other category of households. distributions suggest that any changes in the prices of Figure 4: Structure of employment and average consumption by area of residence 2500 Non-food 2339.56 Food Per Capita Expenditure (Ghana cedis) 2000 1858.34 1790.36 1706.97 1636.43 1655.07 1605.75 1565.13 1456.19 1500 1156.43 1169.93 1109.48 1094.79 1000 839.83 778.45 702.84 500 0 Urban Rural Urban Rural Urban Rural Urban Rural Employed Unemployed Retired Other inactive Area of residence and employment status of household head Source: Prepared by the author based on the Ghana Living Standard Survey (GLSS6) data. 44 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES commodities due to the introduction of the CET may the change in the tariff: an increase in tariffs results in have differential effects on the income and consumption higher prices and a decrease in tariffs results in lower of households depending on the gender of the head and prices (Table 4). For all other commodities, the changes in geographical local location of the household. their prices come from indirect effects, given the general equilibrium nature of the CGE model. 5. Simulations and results 5.2 Non-parametric regression results Following the methodology described in Section 4, we use the CGE model to simulate the changes in This sub-section presents the analysis of the effects of prices of commodities after the implementation of the changes in commodity prices on household welfare. ECOWAS CET. Then we introduce the resulting changes The analysis is carried out for female- and male-headed in commodity prices in the microsimulation model to households separately and also considers regional (urban simulate the changes in welfare. This section provides and rural) and geographical disparities. The estimation of a disaggregation of the dataset based on the gender of non-parametric regressions is useful because they do not the household heads. We further disaggregate the data require specific assumptions on the distribution of the data into female-headed and male-headed households in rural or any econometric specification of the functional form of and urban areas, as well as across the 10 regions of the the relationship between the variables of interest (Deaton, country, to determine potential winners and losers from 1989; Calvo, 2014). In this analysis, the dependent variable the reform based on gender and residence. is the change in welfare due to changes in prices. The explanatory variable is the log of per capita expenditure. 5.1 Computable general equilibrium results The objective is to assess the welfare effect of the CET on households. We divide the analysis into three steps: the This sub-section presents the simulated results on prices welfare effect on households as consumers (through their from the CGE model. In simulating the changes in prices, expenditure), on households as producers (through their we introduced the changes in tariffs (Table 1) as the income), and the net welfare effect. trade shocks. The simulated prices (Table 4) involved 60 food and non-food commodities and services. The First, we calculate the welfare effects of implementation simulated results for the 2007/2013 and 2013/2015 of the CET on households as consumers by multiplying periods show that most of the commodities whose prices the budget share of each consumed item by its change decreased were non-food items. These include pulp in price8 simulated by the CGE model (Table 4). Figure 5 and paper, fertilizers, chemicals, clothing, textiles, and shows the results of the non-parametric regression. The metals. Among the 33 food items, only the price of rice downward sloping curve suggests a pro-poor effect of decreased. This reduction may have a positive impact implementation of the CET for households as consumers. on households as consumers, since rice is the second Figure 5 also shows that implementation of the CET most widely consumed cereal by Ghanaian households, favors poor female-headed households more than poor after maize. Available statistics suggest that in 2014, male-headed households. The expected improvement Ghanaians consumed 754,698 metric tons of rice and in the welfare of poor households as consumers may be imported 52% of that. This price reduction will therefore due to the reduction in the price of the commodities that benefit consumers and may further increase demand for are consumed most within these households.9 Moreover, rice, while at the same time potentially reducing local female-headed households stand to be better off than rice production if domestic producers cannot withstand male-headed households because the budget share of foreign competition. items whose price decreases is higher for female-headed households than for male-headed households. Changes in the prices of imported commodities (that reported import tariffs in the SAM—Table 2) depend on 8 This corresponds to the expression –Si* ∂lnPi in Equation 1. The negative sign indicates that an increase in price results in a decrease in welfare for households as consumers. 9 For instance, Table 4 shows that the price of rice decreases by more than 2 per cent and the budget share of rice is higher in poor households. 45 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 4: Simulated prices of commodities from 2007 to 2015 Log Sim1 Log Sim2 Change in price Commodity 2007–2013 2013–2015 (%) Maize 0.444 0.445 0.064 Yams 0.225 0.226 0.080 Groundnuts 1.024 1.026 0.215 Export vegetables 1.450 1.465 1.513 Chicken -0.001 0.035 3.637 Forest 0.001 0.001 0.000 Cocoa processing 0.001 -0.001 -0.200 Footwear 0.053 0.030 -2.303 Diesel 0.457 0.456 -0.063 Capital goods 0.625 0.642 1.698 Other nuts 0.022 0.026 0.391 Plantain 0.002 0.000 -0.200 Rice 1.320 1.299 -2.025 Cocoyam -0.313 -0.313 0.000 Other nuts -0.499 -0.496 0.329 Plantain -0.276 -0.273 0.263 Eggs 0.001 0.001 0.000 Fish 0.001 0.007 0.598 Dairy 0.108 0.099 -0.902 Wood products 0.107 0.104 -0.270 Fuel 0.001 0.001 0.000 Construction 0.003 0.002 -0.100 Transport 0.025 0.023 -0.195 Public administration 0.001 0.002 0.100 Sorghum and millet 1.206 1.206 0.030 Cowpea 1.204 1.206 0.120 Domestic fruits 0.808 0.81 0.134 Cocoa 2.342 2.339 -0.241 Beef 0.001 0.000 -0.100 Mining 0.001 0.001 0.000 Meat 0.035 0.041 0.577 Pulp and paper 0.119 0.067 -5.195 Fertilizers 0.093 0.061 -3.240 Water 0.149 0.151 0.172 Communication 0.018 0.013 -0.492 Education 0.001 0.006 0.498 Other grains 0.001 0.001 0.000 Soya beans 0.684 0.685 0.151 Export fruits -0.128 -0.121 0.68 Other crops 0.985 0.984 -0.037 Goat 0.001 0.000 -0.100 Formal food processing 0.093 0.103 0.907 Textile 0.103 0.093 -0.907 continued 46 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 4: Simulated prices of commodities from 2007 to 2015 (continued) Log Sim1 Log Sim2 Change in price Commodity 2007–2013 2013–2015 (%) Oil 0.001 0.001 0.000 Chemicals -0.088 -0.102 -1.429 Electricity 0.001 0.007 0.598 Business services 0.115 0.112 -0.268 Health 0.001 0.005 0.399 Cassava -0.728 -0.726 0.207 Palm oil 0.913 0.918 0.480 Domestic vegetables 0.903 0.902 -0.122 Other export crops 1.707 1.718 1.155 Other livestock 0.001 0.000 -0.100 Local food processing 0.001 -0.001 -0.200 Clothing 0.032 0.033 0.097 Petrol 0.491 0.490 -0.061 Metal -0.666 -0.669 -0.390 Trade 0.154 0.154 0.086 Real estate 0.002 -0.004 -0.601 Source: Prepared by the author using the computable general equilibrium model for Ghana. Note: The variables labelled Log Sim show the simulated prices of the commodities. For instance, Log Sim1 is the simulated price of the commodities in from 2007-2013. These values were used as the base values for the simulation of the prices from 2013-2015 (Log Sim2), which represents the period in which Ghana switched from its own tariff to the ECOWAS CET. The third column (change in price) is the difference between the first two columns, the log of prices in from 2007-2013 (before the ECOWAS CET) and 2013-2015 (the period of the ECOWAS CET). Figure 5: Change in welfare of households as consumers 0.1 National Female-headed Male-headed 0.05 Welfare Change (%) 0 -0.05 -0.1 5 6 7 8 9 10 Log per capita expenditure Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). Note: The consumption shares of the commodities, used as an indicator of welfare changes due to changes in expenditure, were obtained by dividing each household’s expenditure by total household expenditure and multiplying the result by the change in price obtained from the macrosimulation (computable general equilibrium). 47 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 6 presents the change in welfare of households across middle-income households. Welfare rises for as consumers based on their region of residence. The both urban and rural households at higher levels of per curves for both rural and urban areas follow the same capita expenditure. However, the increase is sharper for downward sloping shape as for the whole population. the latter than the former, possibly due to higher gains The regression curve for female-headed households lies in purchasing power from a reduction in the domestic above the one for male-headed households, indicating a prices of goods. pro-poor and pro-female effect of implementation of the We now move to the analysis of the welfare effects of CET. The only exception is for very poor urban households, the CET on households as producers. As indicated in the where female-headed households benefit less than their methodology discussion, some households are not only male counterparts. In both urban and rural areas, the consumers, but also producers who earn income from welfare gap between male-headed and female-headed producing some of the commodities analysed in this households is larger at the extremes of the expenditure study. The relationship between the change in welfare of distributions and much narrower in the middle, which may households as producers10 (Si ∂lnPi in Equation 1) and the be due to a more homogeneous consumption structure Figure 6: Change in welfare of households as consumers by area of residence Rural 0.15 National Female-headed Male-headed 0.1 Welfare Change (%) 0.05 0 -0.05 -0.1 5 6 7 8 9 10 Log per capita expenditure Urban 0.15 National Female-headed Male-headed 0.1 Welfare Change (%) 0.05 0 -0.05 -0.1 5 6 7 8 9 10 Log per capita expenditure Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). 10 In this case, the welfare effect is given by Si∂lnPi as shown in Equation 1. The expression has a positive sign, indicating that an increase in prices increases the welfare of producers. 48 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 7: Change in welfare of households as producers 0.01 0 Welfare change (%) -0.01 --0.02 --0.03 National Female-headed Male-headed -0.04 5 6 7 8 9 10 Log per capita expenditure Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). Note: The share of commodity income used as the measure of welfare due to changes in income was obtained by dividing the share of income obtained by households from the sale of commodities by total income of the household. The results were further multiplied by the change in price of the commodities from the macrosimulation. level of expenditure is positive, and the overall change in as their income declines. Poor producers stand to lose welfare at the national level is negative (Figure 7). This more than richer producers. Male-headed households means that implementation of the CET will reduce the will be the most affected, while the effect on female- welfare of both poor and rich households as producers, headed households is almost zero at all income levels. but poor households are the most disadvantaged. The This may be due to the fact that most producers are poor tariff reduction will redistribute income from producers male-headed households, for example rural farmers, for to consumers as the domestic prices of commodities whom the prices of their products have decreased (e.g. decline, and the purchasing power of producers will fall rice or cocoa). 49 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The net welfare effect for In a nutshell, the main finding of this study is that implementation of the CET will lead to a decrease in female-headed households prices of most items consumed by poor households, especially female-headed households, resulting in an is positive for those at the improvement in the welfare of these households. At lower and middle ends of the same time, it will reduce the welfare of households that are net producers. This conclusion differs from the income categories, but the findings of a similar study conducted in Nigeria, which finds that implementation of the CET produces negative for the rich. net welfare gains due to a reduction in prices of most agricultural products (Kareem, 2014). This difference in findings could be due to differences in methodological approaches, as the author used the pass-through effect Households as producers in rural areas experience approach which takes into account the combined effect greater losses in welfare than those in urban areas (Figure of wage and prices. Although the results show a pro-poor 8): the average reduction in welfare is 0.028% in rural and pro-female welfare effect, the variations in welfare areas and 0.014% in urban areas. In both rural and urban are small (less than 0.1%), perhaps because some prices areas, male-headed households are more affected than increase and some others decrease after implementation female-headed households. These differences in welfare of the CET. losses can be partly explained by the greater reliance on agriculture in rural households than in urban households, This same analysis was also performed for Ghana’s10 and in male-headed households than in female-headed regions (Western, Central, Greater Accra, Volta, Eastern, households (about 83% of households for which Ashanti, Brong Ahafo, Northern, Upper East, and Upper agriculture is the main occupation are in rural areas, West) to explore the regional dynamics of the welfare and they are largely male-headed). About 3.2 million effects of the CET. The results (Figure A.1 in the Appendix) households, representing 46% of all households, operate reveal the same structure as that for the national level in non-farm enterprises, with 52% of them in urban areas. some locations (Greater Accra, Volta, Ashanti, and Upper Almost half (49.5%) of all businesses involve trading, West). In these cases, therefore, implementation of the while the rest involve some kind of manufacturing CET is expected to have pro-poor and pro-female effects. activity. Women operate 72% of these businesses (Ghana However, the results are different for other regions. For Statistical Service, 2014). example, in the Western region, where oil exploitation has been under way for a little over four years, the We now assess the net welfare effect of the CET on effect is pro-female but not pro-poor, since the net households by adding the welfare effect on households welfare function first decreases, then increases, and then as producers and as consumers, as shown in Equation 1. decreases again as per capita expenditure increases (a Figure 9 depicts the relationship between the welfare sort of a U-shaped curve). In the Central region, the effect effect and household per capita expenditure. The curve of the CET is pro-poor, but not pro-female. In the Brong resembles the welfare effect on households as consumers Ahafo region, where agriculture is the predominant (Figure 5), because the welfare effect on households as occupation, the results indicate a net welfare loss for all producers (Figure 7) is much smaller than the one on categories of households, regardless of the income status households as consumers. The net welfare effect for or the gender of the household head. These variations male-headed households is around zero for the poor and in the net welfare effect across the 10 regions could be negative for the rich. The net welfare effect for female- explained by the heterogeneity in the production and headed households is positive for those at the lower and consumption structures of households. middle ends of the income categories, but negative for the rich. 50 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 8: Change in welfare of households as producers by area of residence Rural 0 -0.01 Welfare Change (%) -0.02 --0.03 National Female-headed Male-headed -0.04 5 6 7 8 9 10 Log per capita expenditure Urban 0.01 0.005 Welfare Change (%) 0 -0.005 --0.01 National Female-headed Male-headed -0.15 5 6 7 8 9 10 Log per capita expenditure Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). Figure 9: Net welfare effect 0.1 National Female-headed Male-headed 0.05 Welfare change (%) 0 -0.05 -0.1 5 6 7 8 9 10 Log per capita expenditure Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). 51 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES To conclude, this analysis has shown that implementation of the CET is likely to affect households in different ways, depending on their positions as either net producers or net consumers. Other determinants are the gender of the household head, geographical location, and changes in the prices of the commodities. To conclude, this analysis has shown that implementation reveal that implementation of the CET will have a positive of the CET is likely to affect households in different ways, consumption welfare effect on poor households, but a depending on their positions as either net producers or negative effect on rich households. Moreover, the CET will net consumers. Other determinants are the gender of reduce the welfare of both poor and rich households as the household head, geographical location, and changes producers, with poor households being the most affected. in the prices of the commodities. Overall, female-headed households stand to be better off than male-headed From a gender perspective, female-headed households households. Likewise, poor households will gain, while will be better off as consumers than their male rich households will lose marginally. The gain will favor counterparts. As producers, male-headed households will households in coastal regions and urban areas more than be the most affected by the reduction in welfare, while those in non-coastal regions and rural areas. Moreover, the effect on female-headed households will be almost the increase in commodity prices is expected to reduce zero. When we consider only commodities for which welfare, while the opposite holds for the commodities for prices increase, there will be a reduction in household which prices are expected to decrease. These findings are welfare. However, for commodities whose prices consistent with those of the earlier studies (see Ackah & decrease, there will be an improvement in the welfare of Aryeetey, 2012 and Quartey, Aidam, & Obeng, 2013) who households, meaning that the dominant effect is the one suggest that trade liberalization has differential effects on households as consumers. The net welfare analysis on the incidence, depth, and severity of poverty among shows that implementation of the CET will lead to a net loss for all income categories of male-headed households households in Ghana. and for rich, female-headed households. However, there will be a positive effect on female-headed households 6. Conclusions in lower- and middle-income categories. Households In 2015, ECOWAS members, including Ghana, agreed in urban areas stand to gain more than their rural to implement a common external tariff in order to counterparts. Thus, urban households tend to benefit harmonize the tariff structure and foster regional trade more from trade liberalization. On the basis of these and economic growth. The objective of this study has findings, this paper concludes that a comprehensive tariff been to assess the impact of the new tariff system on reform could be pro-poor in Ghana. prices and the resulting effect on household welfare, This study used the top-down approach. This general- with particular attention to gender differences. equilibrium analysis has the advantage of capturing the The descriptive analysis shows that female-headed direct and indirect effects of tariffs on prices. However, households spend more on average on food than male- it is important to add some caveats. First, the feedback headed households. Since poverty indicators in Ghana effects from household behavior are not taken into are based on consumption expenditure, female-headed account. Second, the CGE model uses data from Ghana’s households exhibit lower levels of poverty than male- 2005 Social Accounting Matrix. Having an updated SAM headed households. The macrosimulation analysis (done for 2013 may produce more accurate results. A further, through a CGE model) shows that implementation of useful step would be to include production factor effects the CET is likely to lead to mixed effects on commodity in the analysis, given that the CGE model also simulates prices, given that some tariffs were scheduled to increase changes in wages and capital. However, this would and others to decrease following implementation. When require additional efforts to match the survey with the we introduce the changes in prices from the CGE into SAM, an analysis that was not within the scope of the the microsimulation for the welfare analysis, the results present work. 52 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Ackah C. and E. Aryeetey (2012). Globalization, trade and poverty in Ghana. International Development Research Centre. Ottawa. Adams C. (1993). For Good and Evil: The Impact of Taxes on the Course of Civilization. Rowman and Littlefield. AfDB, OECD and UNDP (2017). African Economic Outlook 2017: entrepreneurship and industrialization. OECD Development Centre, Paris. Available at: www.africaneconomicoutlook.org Bhasin, V.K. and S. Annim (2005). Impact of elimination of trade taxes on poverty and income distribution in Ghana. Revised Report submitted to the Global Development Network (January). Bird, K. (2004). A framework to analyse linkages between trade policy, poverty reduction and sustainable development. Paper for the Africa Trade and Poverty Programme of the Department for International Development. London. Bourguignon, F. and L. Savard (2008). Distributional effects of trade reform: An integrated macro-micro model applied to the Philippines. In: The Impact of Macroeconomic Policies on Poverty and Income Distribution: Macro-micro Evaluation Techniques and Tools, F. Bourguignon, M. Bussolo, and A. Pereira da Silva, eds. Palgrave-Macmillan. Houndmills, United Kingdom. Breisinger, C., J. Thurlow and M. Duncan (2007). A 2005 social accounting matrix (SAM) for Ghana. International Food Policy Research Institute. Washington, DC. Breisinger, C., X. Diao, J. Thurlow and R.M. Al-Hassan (2008). Agriculture for development in Ghana: New opportunities and challenges. IFPRI Discussion Paper 00784. International Food Policy Research Institute. Washington, DC. Calvo, P.A. (2014). Welfare impact of wheat export restrictions in Argentina: Non-parametric analysis on urban households. In: Trade Policies, Household Welfare and Poverty Alleviation: Case Studies from the Virtual Institute Academic Network. UNCTAD/DGS/2014/3. United Nations Conference on Trade and Development. New York and Geneva. Available at: http://vi.unctad.org/tap/docs/chapters/argentina.pdf. De Melo, J., and A. Laski (2014). Reflections on adapting to the ECOWAS CET. FERDI Policy Brief B-107. Fondation pour les Études et Recherches sur le Développement International. Available at: http://archive-ouverte.unige. ch/unige: 46919. Deaton, A. (1989). Rice prices and income distribution in Thailand: a non-parametric analysis. The Economic Journal 99: 1–37. Diao, X. (2011). Economywide impact of Avian influenza in Ghana–A dynamic computable general equilibrium (DCGE) model analysis. International Journal of Livestock Production 2(10): 145–58. Ghana Statistical Service (2007). Pattern and trends of poverty in Ghana, 1991–2006. Ghana Statistical Service. Accra. Ghana Statistical Service (2014). Ghana Living Standards Survey round 6 (GLSS 6): Main report. Ghana Statistical Service. Accra. Gutowski, A., T. Knedlik, P. N. Osakwe, I. Ramdoo and K. Wohlmuth (Eds.). (2016). Africa’s Progress in Regional and Global Economic Integration-Towards Transformative Regional Integration (Vol. 18). LIT Verlag Münster. Harrison, A. and H. Tang (2005). Trade Liberalization: Why so Much Controversy in Growth in the 1990s: Learning from a Decade of Reforms. Edited by N. Roberto Zagha, World Bank, 2005. Kareem, O.I. (2014). The welfare impact in Nigeria of the ECOWAS common external tariff: A distributional effects analysis. In: Trade Policies, Household Welfare and Poverty Alleviation: Case Studies from the Virtual Institute Academic Network. UNCTAD/DGS/2014/3. United Nations Conference on Trade and Development. New York and Geneva. Available at: http://vi.unctad.org/tap/docs/chapters/nigeria.pdf. 53 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Löfgren, H., R.L. Harris and S. Robinson (2002). A standard computable general equilibrium (CGE) model in GAMS (vol. 5). International Food Policy Research Institute. Washington, DC. Marchand, B.U. (2012). Tariff pass-through and the distributional effects of trade liberalization. Journal of Development Economics 99(2): 265–81. Nicita, A. (2009). The price effect of tariff liberalization: Measuring the impact on household welfare. Journal of Development Economics 89(1): 19–27. Office of the United States Trade Representative (2014). 2014 National trade estimate report on foreign trade barriers. Available at: https://ustr.gov/sites/default/files/2014%20NTE%20Report%20on%20FTB.pdf. Osakwe, P.N. (2015). Transformative regionalism, trade and the challenge of poverty reduction in Africa. UNCTAD Trade and Poverty Paper Series No. 1. United Nations Conference on Trade and Development. New York and Geneva. Quartey, P., P. Aidam and C.K. Oben (2013). The impact of trade liberalization on poverty in Ghana. Available at: https:// www.gtap.agecon.purdue.edu/resources/download/2665.pdf. Roquefeuil, Q.D., D. Plunkett and F. Ofei (2014). The impact of closer regional economic integration on food security in West Africa: Focus on the ECOWAS common external tariff. ECDPM Discussion Paper 154. European Centre for Development Policy Management. Maastricht. UNCTAD (2010). Virtual Institute teaching material on trade and poverty. United Nations Conference on Trade and Development. New York and Geneva. UNCTAD (2012). Virtual Institute teaching material on contemporary issues in international macroeconomics, trade and finance. UNCTAD/GDS/2011/2. United Nations Conference on Trade and Development. New York and Geneva. UNCTAD (2016). African Continental Free Trade Area: Policy and negotiation options for trade in goods. United Nations Conference on Trade and Development. New York and Geneva. Available at: http://unctad.org/en/ PublicationsLibrary/webditc2016d7_en.pdf. United Nations (2008). ISIC Revision 4.0. International Standard Industrial Classification of All Economic Activities. Available at: http://unstats.un.org/unsd/cr/registry/isic-4.asp Winters, L.A., N. McCulloch and A. McKay (2004). Trade liberalization and poverty: the evidence so far. Journal of Economic Literature 42(1): 72–115. World Bank Group (2015). Assessing the Economic Impact of the ECOWAS CET and Economic Partnership Agreement on Ghana. World Bank, Washington, DC. Available at: https://openknowledge.worldbank.org/handle/10986/23242 License: CC BY 3.0 IGO.” WTO (2014). Trade policy review report by the secretariat, Ghana. Report No. WT/TPR/S/298. Geneva. World Trade Organization. 54 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix Table A.1: Summary statistics of the net welfare effect by commodity No. of Standard No Commodity observations Mean deviation Minimum Maximum 1 Cassava 5,750 0.000146 0.000828 -0.00152 0.006 2 Yams 5,750 0.000062 0.000825 -0.00206 0.008 3 Plantains 5,750 4.84E-05 0.000752 -0.0042 0.007 4 Oils 5,750 -1.8E-05 0.000406 -0.00167 0.005 5 Vegetables 5,750 -0.00016 0.000461 -0.004 0.004 6 Domestic fruits 5,750 3.61E-05 0.000522 -0.00192 0.007 7 Maize 6,643 0.001613 0.002319 -0.00252 0.006 8 Rice 6,643 -0.00021 0.002597 -0.015 0.013274 9 Cocoa beans 6,647 0.000942 0.001619 0.0000 0.004 10 Processed cocoa 16,750 -6.64E-06 2.17E-05 -0.001 0.0000 11 Sorghum 6,643 0.000143 0.000723 -0.00222 0.006 12 Groundnuts 6,643 0.00082 0.001898 -0.00144 0.007 13 Goats 7,145 0.000482 0.00153 -0.00269 0.008 14 Other livestock 7,145 0.000422 0.001489 -0.0065 0.007 15 Palm oil 5,750 -1.3E-05 0.000738 -0.009 0.009 16 Chicken 7,145 0.00215 0.009269 -0.02122 0.042 17 Fishing 7,145 6.72E-05 0.000416 -0.00098 0.003 18 Cocoyam 5,753 1.05E-05 0.000158 0.0000 0.005 19 Wood 6,647 1.29E-06 0.000068 0.0000 0.004 20 Other crops 6,647 9.09E-05 0.000661 0.0000 0.007 21 Other nuts 6,647 0.000219 0.001297 0.0000 0.01 22 Beef 16,750 -5.6E-05 0.000128 -0.00392 0.000 23 Dairy products 16,750 -0.00039 0.000383 -0.005 0.000 24 Eggs 16,750 -3.8E-05 7.14E-05 -0.0032 0.000 25 Petrol 16,750 -4.1E-05 0.00016 -0.00375 0.000 26 Transport 16,750 -0.00014 0.000221 -0.00463 0.000 27 Other services 16,750 -4.1E-05 8.77E-05 -0.004 0.000 28 Clothing 16,750 -0.00184 0.000836 -0.004 0.000 29 Electricity 16,750 -0.00038 0.000597 -0.005 0.000 30 Fuel 16,750 -3.3E-05 7.62E-05 -0.00095 0.000 31 Furniture 16,750 0.00061 0.002368 0.000 0.038022 32 Textile 16,750 0.000119 0.000172 0.000 0.004405 33 Fertilizers 16,750 0.0003 0.001678 0.000 0.024262 34 Footwear 16,750 0.000581 0.001044 0.000 0.016 35 Formal processed food 16,750 -4.44E-07 1.86E-06 -.0000793 0.000 Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). 55 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure A.1: Net income share by gender and region of residence Western Central 0.3 0.05 Welfare change (%) Welfare change (%) 0.2 0 0.1 -0.05 0 -0.1 -0.1 5 6 7 8 9 10 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed Greater Accra Volta 0.3 0.1 Welfare change (%) Welfare change (%) 0.2 0.05 0.1 0 0 -0.05 -0.1 -0.1 5 6 7 8 9 10 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed Eastern Ashanti 0.2 0.15 0.1 Welfare change (%) Welfare change (%) 0.1 0.05 0 0 -0.1 -0.05 -0.2 -0.1 5 6 7 8 9 10 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed Brong Ahafo Northern 0.1 0 Welfare change (%) Welfare change (%) 0.05 -0.05 0 -0.1 -0.05 -0.15 -0.1 5 6 7 8 9 10 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed 56 Upper East Upper West 0.2 0.2 nge ( %) 0.1 nge (%) 0.1 Eastern Ashanti 0.2 0.15 0.1 Welfare change (%) Welfare change (%) 0.1 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS 0.05 IN DEVELOPING COUNTRIES 0 0 -0.1 -0.05 -0.2 -0.1 Figure A.1:5Net income 7 gender 8 6 share by 9 of residence and region 10 (continued) 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed Brong Ahafo Northern 0.1 0 Welfare change (%) Welfare change (%) 0.05 -0.05 0 -0.1 -0.05 -0.15 -0.1 5 6 7 8 9 10 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed Upper East Upper West 0.2 0.2 Welfare change ( %) 0.1 Welfare change (%) 0.1 0 0 -0.1 -0.1 -0.2 -0.2 5 6 7 8 9 10 5 6 7 8 9 10 Log per capita expenditure Log per capita expenditure National Female-headed Male-headed National Female-headed Male-headed Source: Prepared by the author based on the 2012/2013 round of the Ghana Living Standard Survey (GLSS6). 57 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Exporting, Importing and Wages in Africa: Evidence from Matched Employer-Employee data 1 Marta Duda-Nyczak, United Nations Economic Commission for Africa (UNECA)2 Christian Viegelahn, International Labour Organization (ILO)3 1. Introduction economic and social development. This has found T expression in the rapid shift towards a more integrated he economic and social development of the African market in recent years. In particular, trade within African continent has been on the agenda of some of the Regional Economic Communities (RECs) policy makers and the international community has been liberalized continuously. Current trade policy for decades. With over a billion inhabitants focuses on connecting some of the existing free trade and the fastest growing population worldwide, the areas to create an even larger internal market, with the African market presents an enormous potential. Despite ultimate objective of a customs union that integrates remarkable economic growth rates, however, many all countries in Africa. Negotiations for the Tripartite countries on the continent struggle to translate this Free Trade Area, a free trade agreement between potential into significant improvements in socio-economic the Common Market for Eastern and Southern Africa, indicators. International trade is considered by many as the East African Community and the Southern Africa one of the main contributors to reductions in poverty and Development Community, consisting of 27 countries, the improvement of livelihoods (Dollar and Kraay, 2004; were launched in 2011. Also, negotiations are underway Le Goff and Singh, 2014). This stance has been adopted to create the Continental Free Trade Area integrating the in global policy making, with trade forming an integral trade in goods and services between 54 member states part of the 2030 Sustainable Development Agenda of the African Union. of the United Nations. The Sustainable Development These developments are likely to increase the number Goals (SDG) include the objective to double the share of African firms that are able to engage in trade. The of least developed countries’ (LDC) exports in global question arises whether this opening up to trade can exports by 2020. Thirty-four of the 48 LDCs are located benefit workers in terms of higher wages. Wages are on the African continent, implying that this endeavor is an important form of labor income in many countries. particularly relevant for Africa. The share of workers in wage and salaried employment International trade is also viewed by a large number of has been growing rapidly, even in Africa where informal policy makers in Africa as a potential driver of sustainable employment arrangements still tend to dominate. 1 All views expressed in this paper are those of the authors and do not reflect those of the institutions they are affiliated with. The authors would like to thank colleagues from the International Labour Organization, the African Development Bank, the World Bank and the World Trade Organization for their comments on earlier versions of this paper. In particular, we would like to thank Paul Brenton, Michela Esposito, Marcus Bartley Johns, Linguère Mously Mbaye, Mustapha Sadni Jallab, Bill Shaw and Robert Teh. The authors would also like to thankfully acknowledge comments from participants of the African Economic Conference in Addis Ababa in December 2017. 2 United Nations Economic Commission for Africa (UNECA). Email:duda-nyczak@un.org 3 International Labour Organization (ILO), Research Department. Email: viegelahn@ilo.org 58 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES A current trade policy focuses on connecting some of the existing free trade areas to create an even larger internal market, with the ultimate objective of a customs union that integrates all countries in Africa. 59 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES According to ILO estimates (ILO, 2018), almost one-third from Indonesia suggests that increased access to foreign of all workers in Africa were wage earners in 2017, many inputs through trade liberalization has led to higher of them employed by the private sector. The wage level wages, while the impact of a decline in output tariffs is determines these workers’ standard of living, and low less pronounced (Amiti and Davis, 2011). wage levels are often directly related to the prevalence There are various channels through which firms’ export of poverty. Indeed, labor income in Africa is often not and import status can affect wages at the firm level. For sufficient to lift workers above poverty levels, and about a firm to participate in international trade, it is important 56% of all African workers lived in either moderate or to have a high-skilled workforce, which in the presence of extreme poverty, on less than $3.10 PPP a day, in 2017. a skill premium on wages then leads to a higher average This paper uses a novel dataset that includes firm-level firm-level wage. The trading activity of a firm can also and employee-level data to explore the relationship give rise to technology upgrading, induced by technology between exporting, importing and wages in African transfers from the trading partner, which may increase manufacturing firms. This dataset forms part of the workers’ productivity and can therefore lead to higher World Bank Enterprise Survey and comprises 65 firm- wages. Moreover, the extension of a firm’s business to level surveys conducted in 47 African countries over export markets increases the scale of a firm, allowing 2006–2014, with information on firms’ export and import it to benefit from economies of scale, and some of status, as well as information on the average wage. For these benefits may be passed on to workers in terms of 16 of these firm-level surveys, matched employee data higher wages. Assuming a certain degree of rent sharing with information on individual worker wages are available between firms and workers, any standard bargaining to complement the firm-level analysis. These data are model would predict that the gains in productivity that comparable across all countries included and enable us are reaped by the firm would at least partially passed to control for individual worker characteristics, which is on to workers in terms of higher wages, depending on unique for Africa. The data also facilitate analysis of the workers’ bargaining power. relationship between firms’ export and import status, The wages that firms are able to pay are strongly related and wages, by sector and by country. to firm performance. Both exporting and importing There is a large body of literature that has looked at the involves fixed costs that only the most productive firms relationship between trade at the firm level and average can afford to pay, which implies that only firms whose wages that firms pay to their workers, with studies largely productivity exceeds a certain threshold engage in trade confirming a wage premium of firms engaged in trade. (Melitz, 2003; Kasahara and Lapham, 2013). At the same For manufacturing firms in the United States, it has time, firms can learn by exporting, as they have to satisfy been documented that both importers and exporters the needs of foreign customers which may be more pay higher wages on average than non-traders (Bernard, demanding in terms of product quality, and also face Jensen, Redding and Schott, 2007, 2012). Based on competition from foreign producers, which may force employer-employee level data for Germany and Italy, it them to become more productive (De Loecker, 2013). has been found that exporters pay higher wages than But it is also likely that firms can derive productivity non-exporters, after controlling for various firm and gains from importing once they have started to import, worker characteristics (Schank, Schnabel and Wagner, for example through learning from new technologies 2007; Macis and Schivardi, 2016). There is also evidence embedded in foreign inputs, access to a better quality of a positive wage premium of exporting for China, of inputs, or access to a larger variety of inputs (Ethier, driven by different firm characteristics such as ownership, 1982; Markusen, 1989; Grossman and Helpman, 1991). export orientation and location (Fu and Wu, 2013). Also The empirical literature confirms such a positive impact for Mexico, exporting has been found to increase wages, of increased access to foreign inputs on firm productivity especially at the upper end of the wage distribution (Amiti and Konings, 2007; Stone and Shepherd, 2011; (Frías, Kaplan and Verhoogen, 2012). Firm-level evidence Halpern, Koren and Szeidl, 2015), while restricted access 60 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Labor income in Africa is often not sufficient to lift workers above poverty levels, and about 56% of all African workers lived in either moderate or extreme poverty, on less than $3.10 PPP a day, in 2017. to foreign inputs in turn can lead to within-firm input The number of studies that look into the firm-level reallocation with a negative impact on firm performance consequences of trade in the African context is limited, (Vandenbussche and Viegelahn, 2016). given the scarcity of firm-level databases from this region. Milner and Tandrayen (2007) investigate the relationship There is a vast body of literature that confirms that between exporting and wages, using employer-employee wages that women receive are on average lower than matched data for manufacturing firms in six countries in those of men (Blau and Kahn, 2017). The question arises Sub-Saharan Africa. They find a positive overall association whether the wage premium of exporting and importing between individual earnings and the export status of the for women differs from the corresponding premium for firm. Yet, they find that the wage premium is positive men. Empirical evidence has so far been mixed. Using only when firms export to African markets, and it turns employer-employee-level data from Norway, the gender negative when exporting to more competitive markets. wage gap has been found to be larger within exporting In a study with larger country coverage, exporting is firms than within non-exporting firms, provided that found to have positive spinoffs on employment and women are perceived by employers to be less committed wages across a wide range of developing countries, workers than men (Boler, Javorcik and Ulltveit-Moe, including countries on the African continent (Brambilla et 2015). Policy measures that decrease these perceived al., 2017). There are to our knowledge no studies in the gender differences in commitment are found to narrow African context that focus on importing and its impact on differences in the gender wage gap. Other studies in the labor market. contrast find evidence for a lower gender wage gap in exporting firms and higher wages for women in exporting Other studies focus on the relationship between firms (World Bank, 2012). exporting and productivity. Based on firm-level data 61 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES from Cameroon, Ghana, Kenya and Zimbabwe, there is exporting premium on wages is confirmed when using evidence for both firm self-selection into exporting and employer-employee data, which allows us to control for learning-by-exporting (Bigsten et al., 2004). A causal individual worker characteristics. On the basis of these relationship between exporting and productivity has also data, we also do not find any positive wage premium of been found on the basis of firm-level data for Ethiopian importing, in line with the firm-level results. If anything, manufacturing firms, with strong evidence in favor of both workers employed by importers receive lower wages, the self-selection and learning-by-exporting hypotheses, when compared to their counterparts in non-importing demonstrating that exporters pay higher average wages firms. and employ more workers than non-exporters (Bigsten This paper also investigates the channels that are driving and Gebreeyesus, 2009). Mengistae and Pattillo (2004) our results. We find that neither productivity gains show an average total factor productivity premium through increased skill utilization or the employment of and a premium in productivity growth for exporting certain types of workers, nor productivity gains through manufacturers in Ethiopia, Ghana and Kenya. Some technology transfers can fully explain the positive wage evidence also has recently been provided on the impact premium of exporting. Instead, it appears that the of increased access to foreign inputs. Bigsten et al. (2016) positive wage premium of exporters is due to productivity analyze firm-level data for Ethiopian manufacturing firms gains through economies of scale. The paper also finds and show that a reduction in output tariffs has no impact indirect evidence for a weaker bargaining power of on firms’ productivity, while reductions in input tariffs workers employed by importers, when compared with increase firms’ productivity. those employed by non-importers. Finally, the paper This paper contributes to the literature in four ways. shows that there is no significant gender wage gap within First, this is among the first papers that uses employer- trading firms in the sample. This is different from non- employee level data in the African context. With these trading firms, where a statistically significant wage gap data, we are able to measure the average firm-level wage can be identified. premium of exporting and importing, controlling for a The next section describes in more detail the data that variety of firm-level characteristics. Similarly, we are able are used in this paper. Section 3 presents the underlying to determine the average wage premium of individual empirical methodology to estimate the wage premium of workers, after controlling not only for firm-level but exporting and importing, both at the firm level and at the also for individual worker characteristics. Second, this employee level. Section 4 discusses the results. The final paper considers the relationship of both exporting and section concludes. importing to wages, adding to the literature that has predominantly, and in the context of wages exclusively, focused on exporting. Third, this paper investigates the 2. Data channels that make trading firms pay wages that are 2.1 Firm-level data different from non-trading firms. Finally, this paper adds This paper uses firm-level data for manufacturing firms to the so far scarce literature on the gender wage gap from the World Bank Enterprise Surveys. The data are and its relation to firms’ exporter and importer status. cross-section data, comparable across different surveys. The results presented in this paper suggest that firm-level The database consists in total of over 15,391 observations wages paid by exporters to their workers are on average for manufacturing firms, comprising data from 65 higher than in firms not engaged in exporting, even surveys conducted in 47 African countries between 2006 after controlling for firm characteristics such as capital and 2014. For one country, the Democratic Republic intensity, electricity intensity, foreign ownership and of Congo, data from three surveys are available. For firm age. The average wages paid by importers and non- 16 countries, we have data from two surveys. For the importers are statistically not significant from each other, remaining 30 countries, data have only been collected after adding firm age as a control variable. A positive once. For different surveys, the sample size varies 62 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES between 21 observations for a survey conducted in 2009 nor import. Firms are on average 17.4 years old and 11% in Liberia, and 2,015 observations for a survey conducted of them are foreign-owned, defined as foreign investors in 2013 in Egypt. The average sample size across surveys having an ownership share that is greater than 50%. In is 237 observations. Appendix A lists all the surveys that terms of workforce characteristics, the average number are considered in this paper. of full-time permanent employees reported by firms is 82. Twenty-one percent of these workers are women The firm-level data that are included in the database are and 77% are production workers. The average share of representative of formally registered, privately owned temporary employees in total employment (temporary firms that employ at least 5 workers. On the basis of the plus full-time permanent) is almost 12%. The average information provided in the survey, firms can be assigned years of education of firms’ production workforce is 8.7 to the manufacturing sectors in which they operate. We years.4 The average repurchase value of firms’ capital distinguish between 8 manufacturing industries, namely stock is more than double its annual sales revenue. Firms food and beverages, textiles and garments, wood and pay electricity costs that on average amount to around paper, chemicals, non-metals and plastics, metals and 3% of their sales revenue. The annual average wage is just machinery, furniture and all other manufacturing not above 6170 USD (constant 2011). included in the preceding categories. To measure technological advancement, we estimate Table 1 shows basic descriptive statistics for the firm-level total factor productivity (TFP), where a Cobb Douglas database that is used in this paper. The table indicates that production function is estimated in logarithmic form, 53% of all firms in the sample are importers, while only separate for each survey. As input factors, we consider 23% are exporters. Out of all firms, 17% are both exporters the repurchase value of the capital stock, labor costs and importers, 5% are exporters but do not import, 36% and raw material expenses. The estimated residual do not export but are importers, and 42% neither export Table 1: Descriptive statistics on African manufacturing firms Mean Sd. Min Max N Exporter dummy (1=exporter) 0.23 0.42 0.00 1.00 14,972 Importer dummy (1=importer) 0.53 0.50 0.00 1.00 13,837 Firm age 17.40 15.29 0.00 190.00 9,808 Ownership (1=foreign) 0.11 0.31 0.00 1.00 15,075 Capital stock value over sales 2.16 4.47 0.00 51.37 9,244 Electricity costs over sales 0.03 0.06 0.00 0.48 12,273 Sales (million 2011 USD) 17.03 595.31 0.00 52,578.34 13,757 Average wage (000 2011 USD) 6.17 119.89 0.00 9453.91 13,270 Labor productivity (000 2011 USD) 24.84 146.25 0.00 6,460.50 11,185 Log(TFP) -0.02 0.59 -2.03 2.70 8,817 Production workers’ average education (years) 8.68 3.77 1.55 14.92 10,833 Employment (full-time permanent) 82.40 609.68 1.00 64,000.00 15,207 Female share in employment (%) 21.25 25.58 0.00 100.00 13,847 Production worker share in employment (%) 76.69 18.03 0.00 100.00 12,113 Temporary employment share (%) 11.93 19.36 0.00 99.67 14,585 Source: Monetary values are converted into USD (2011 constant), using data on exchange rates and GDP deflators from World Bank’s World Development Indicators database. 4 The average years of education of firms’ production force are reported only for less than two thirds of all firms. The remaining firms report intervals (e.g. 0-3 years, 3-6 years etc.) instead. For these firms, we transform intervals into years, by using the corresponding average value for each category that is obtained on the basis of the sample of firms that report the exact years. For example, the category from 0-3 years translates into a value of 1.55, as 1.55 is the average years of education for firms that fall into that category, based on available data. 63 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES corresponds to TFP. More details on the TFP estimation have no or only primary education, 17% took part in procedure can be found in Appendix B. vocational training and 6% have a university degree. The remaining 55% of employees have secondary education. 2.2 Matched employer-employee level data Twenty-one percent of workers are trade union members. The average worker age is 32 years. Workers Employee-level data for at least some firms are available have on average more than 8 years of work experience, in 16 of the 65 surveys. In total, we have data for 7,692 of which more than 5 years is experience with the current employees working in 1,385 firms, with between 1 and 10 employer. The average monthly wage of a worker in the employees per firm. For 353 firms, data on 10 employees database is 540 USD (constant 2011), which translates are collected. For 25 firms, only data on one employee are into an annual wage of 6480 USD (constant 2011), available. The employee data are available from surveys which is very close to the average annual firm-level in Angola, Botswana, Burundi, Democratic Republic wage reported in Table 1. The average monthly wage of Congo, Gambia, Ghana, Guinea, Guinea-Bissau, of female workers in the database is 850 USD (constant Mauritania, Namibia, Rwanda, South Africa, Swaziland, 2011), translating into an annual wage of 10200 USD Tanzania, Uganda and Zambia, which are all Sub-Saharan (constant 2011). While the average wage for women is African countries. Data are from surveys conducted in higher than the average wage for men in the sample, the 2006 and 2007. standard deviation of women’s wage is almost double Table 2 shows employee-level descriptive statistics. the standard deviation of the overall wage, indicating a We find that 21% of employees in our sample work for large wage variation among women. exporters while 56% work for importers. The respective shares of workers that work for exporters and importers 3. Methodology are hence very similar to the share of exporting and In this paper, we run two types of empirical analyses. First, importing firms in the firm-level database, reported in we use firm-level data to estimate the wage premium Table 1. Among the employees, 28% are women, 53% are of exporting and importing, controlling for a variety of married and 94% have a full-time permanent contract. firm-level characteristics. Then we take the estimation With regards to the education level, 22% of employees to the employer-employee level, which enables us to add Table 2: Descriptive statistics on employees in African manufacturing firms Mean Sd. Min Max N Exporter dummy (exporter=1) 0.21 0.41 0.00 1.00 7,682 Importer dummy (importer=1) 0.56 0.50 0.00 1.00 7,692 Employee wage (monthly, 000 2011 USD) 0.54 6.78 0.00 364.89 6,648 Female employee wage (monthly, 000 2011 USD) 0.85 12.19 0.00 364.89 1,835 Female (yes=1) 0.28 0.45 0.00 1.00 7,692 Married (yes=1) 0.53 0.50 0.00 1.00 7,649 Full-time permanent employed (yes=1) 0.94 0.24 0.00 1.00 7,667 No or primary education only (yes=1) 0.22 0.42 0.00 1.00 7,692 Vocational training (yes=1) 0.17 0.37 0.00 1.00 7,692 University degree (yes=1) 0.06 0.25 0.00 1.00 7,692 Trade union member (yes=1) 0.21 0.41 0.00 1.00 7,672 Age (years) 31.98 8.21 12.00 71.00 7,669 Experience with current employer (years) 5.23 4.86 0.00 48.00 5,880 Total experience (years) 8.11 6.62 0.00 54.00 5,838 Source: Monetary values are converted into USD (2011 constant), using data on exchange rates and GDP deflators from World Bank’s World Development Indicators database. 64 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES individual worker characteristics to our set of firm-level electricity intensity—the ratio between electricity costs control variables. Reported standard errors are always and sales—into the regression. The latter is intended to robust. control for the type of technology that is used. Electricity costs are likely to be lower if production mainly occurs At the firm-level, we estimate the following equations: through manual labor than if production is largely automated. Moreover, we include foreign ownership status and firm age as variables that might be correlated with the average wage. Finally, we include the logarithm of firm age to control for differences in wages between where equation (1) is estimated on the full sample start-ups and firms that have been longer in the market. of manufacturing firms, equation (2) is estimated by manufacturing sector m and equation (3) is estimated by When moving to the employer-employee level, the survey conducted in country c and year t. Index i stands estimated equations look as follows: for a particular firm that belongs to a manufacturing sector m and is observed in the survey conducted in country c and year t. The dependent variable logW stands for the logarithm where equation (4) is estimated on the full sample of the average wage paid by the firm to its employees, of manufacturing firms, equation (5) is estimated by calculated as total labor costs divided by the number of manufacturing sector m and equation (6) is estimated by full-time permanent employees.5 The exporter dummy survey conducted in country c and year t. The equations variable EX takes a value of one if the firm exports at least are similar to (1)–(3), but now include variables that carry some of its goods, including direct exports and exports a subscript w that stands for an individual employee. Now through an intermediary. Similarly, the importer dummy wages are employee-specific and a new vector of control variable IM takes a value of one if the firm imports at least variables accounts for individual worker characteristics, Y.7 some of its raw material inputs, including both direct imports and imports through an intermediary. β, βm and As control variables, we include dummy variables that βct are the main coefficients of interest and measure the respectively take on a value of one when the worker is overall, sector-specific and survey-specific wage premium a woman, married, full-time permanent employed, or of exporting. γ, γm and γct are the respective coefficients trade union member. We also consider dummy variables that measure the wage premium of importing. The wage that indicate workers’ education level, including no or premia of exporting and importing should be interpreted only primary education, vocational training or a university respectively as average premia across export destination degree. We include workers’ age, workers’ total work countries and import origin countries. εct is a survey fixed 6 experience, and workers’ experience with the current effect, εm is a sector fixed effect and εctmi is the error term. employer as explanatory variables. Moreover, we account for the same set of firm characteristics as in the firm-level With regards to firm-level control variables, as summarized regressions, with the exception of firm age, which is only in vector X, we control for the type of economic activity available in 3 out of the 16 surveys that have employee by including capital intensity—the ratio between the data.8 repurchase value of the capital stock and sales—and 5 This measure for the average wage is a proxy, given that labor costs in the numerator are the costs for all workers, while full-time permanent employment in the denominator does not include all workers. The use of the ratio between labor costs and the sum of full-time permanent and temporary employment as an alternative proxy does not affect any of our main results in this paper. 6 Data on import origin countries are not available. Data on export destination countries are only available for 22% of exporting firms, over half of which declare zero exports to developed countries. Most firms focus their trade on African countries. Some firms declare a mix of export destinations (with a third category of other countries). The non-existent or limited data on import origin countries or export destination countries prevent us from undertaking any analysis along these dimensions. 7 The survey fixed effects in equations (1) and (2) correspond to country-year fixed effects. The survey fixed effects in equations (4) and (5) also correspond to country-year fixed effects, which in this case are equivalent to country fixed effects, given that there is only one survey with employer-employee level data per country. 8 The employee-level data also includes information on whether a worker is foreigner, which could have an impact on the wage level. Still we decided to not use this variable as control variable, as only 5% of all workers in our sample are foreigners. We made sure that the main results of this paper also hold when excluding foreigners from our sample. 65 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES In order to investigate the differences in the gender The positive wage premium of exporters and importers is wage gap between exporting and non-exporting firms, confirmed when including both exporting and importing and between importing and non-importing firms, we status simultaneously as explanatory variables into the also estimate equation (4) after adding as explanatory regression (column 3). Also, the estimated wage premia of variables two interaction terms between the dummy exporters and importers remain positive and statistically variable that indicates whether a worker is a woman, and significant after controlling for capital stock value and firms’ exporter and importer status, respectively. electricity costs relative to sales, and foreign ownership (columns 4 and 5). The difference between exporters and 4. Results non-exporters is estimated to be larger than the difference 4.1 Firm-level results between importers and non-importers.
Results change partially, however, when adding firm age as an additional This section starts by showing evidence derived from firm- control variable. The coefficient for importer status then level data. Table 3 shows the Africa-wide coefficients for becomes insignificant, while the coefficient for exporter firms’ exporting and importing status, which respectively status remains significant and implies that exporters pay correspond to the estimated average difference in almost 18% higher wages than non-exporters (column wages between exporters and non-exporters, and 6). This implies the absence of any differences in wages between importers and non-importers. In the most basic between importers and non-importers that have the specification, in which there are no control variables same firm age. aside from sector and survey fixed effects, exporters are estimated to have a wage premium of 18% over non- Table 4 indicates that results are not driven by a particular exporters (column 1), while importers’ wage premium sector, but hold across many sectors. The table shows over non-importers is estimated at 19% (column 2). This the results obtained from estimating the specifications is in line with previous results in the literature that firms in columns (3) and (6) of Table 3 on samples that are engaged in trade pay higher wages (Bernard, Jensen, restricted to firms in particular sectors. In the more basic Redding and Schott, 2007, 2012). specification, a positive wage premium of exporting Table 3: Exporting, importing and the average wage (firm-level)—full sample Dependent variable: Log(Wage) (1) (2) (3) (4) (5) (6) 0.178a 0.210a 0.268a 0.234a 0.175a Exporter (0.038) (0.037) (0.035) (0.036) (0.062) 0.191a 0.155a 0.081a 0.064b -0.004 Importer (0.028) (0.028) (0.027) (0.027) (0.051) -0.032a -0.031a -0.034a Capital stock over sales (0.005) (0.005) (0.007) -2.351a -2.327a -2.108a Electricity costs over sales (0.401) (0.398) (0.520) 0.322a 0.344a Foreign owned (0.045) (0.088) 0.071a Log(Firm age) (0.026) Sector FE Yes Yes Yes Yes Yes Yes Survey FE Yes Yes Yes Yes Yes Yes R2 0.71 0.73 0.73 0.80 0.80 0.75 Number of observations 13137 12319 12254 8818 8787 3827 a  Indicate statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the full sample of firms. b  Indicate statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the full sample of firms. 66 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES is found for firms in the food and beverages, textiles number of observations is too small to obtain reliable and garments, chemicals, and non-metals and plastics results for most of the countries. sectors. In this specification, importing can be associated Based on the specification that only includes exporter and with a positive wage premium in the food and beverages, importer status, we estimate a positive wage premium wood and paper, and metals and machinery sectors, as of importing for 18 out of the 34 survey datasets and well as in other manufacturing that is not classified in any a negative premium for 3 datasets. When in addition of the other sectors. In the more elaborate specification, controlling for capital intensity, electricity intensity, exporting remains associated with higher wages in firms ownership and firm age, we estimate one significantly that belong to the textiles and garments, and non-metals positive and one significantly negative coefficient for and plastics sectors, and becomes associated with higher the relationship between firms’ importing status and the wages in the wood and paper sector. The wage premium wages they pay to their workers on average. of importing, in contrast, disappears for all sectors except other manufacturing. The results presented in this section so far point to only limited evidence for a wage premium of importing, Table 5 includes results obtained from estimating the with no wage premium estimated in the specification specifications in columns (3) and (6) of Table 3 on individual that includes firm age as a control variable. The results, survey datasets. We only run the regression on samples however, point strongly to the existence of a positive with at least 100 firm-level observations. We again find wage premium of exporting. To explore why exporters strong evidence for a wage premium of exporting. Out pay higher wages than non-exporters, the next section of 34 survey datasets with at least 100 observations, we investigates which channels contribute to the wage find for 18 survey datasets evidence for a positive wage premium of exporting. premium of exporting and only for one dataset evidence for a negative premium. Also in the more elaborate 4.2 What are the channels? specification, we find evidence for positive wage premia of exporting for 2 out of 10 survey datasets. With the There are numerous channels that can explain differences inclusion of firm age as a control variable, the underlying in wages between trading and non-trading firms. This Table 4: Exporting, importing and the average wage (firm-level)—by sector Dependent variable: Log(Wage) Sector Regressors: Exporter/Importer Regressors: Exporter/Importer Capital stock over sales Electricity costs over sales Foreign owned Log(Firm age) N Exporter Importer N Exporter Importer Food & beverages 3,079 0.299a 0.110b 851 0.120 -0.036 Textiles & garments 2,365 0.191a 0.045 815 0.344a -0.050 Wood & paper 1,364 0.155 0.171 b 433 0.392 b -0.102 Chemicals 752 0.255b 0.100 242 0.229 -0.063 Non-metals & plastics 1,159 0.278b 0.145 444 0.393c 0.109 Metals & machinery 1,582 0.215 0.245 a 502 0.135 0.012 Furniture 1,314 -0.130 0.006 314 -0.238 -0.281 Other manufacturing 639 0.137 0.391a 226 0.052 0.399b Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (2) with OLS on samples of firms from a  different sectors. Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (2) with OLS on samples of firms from b  different sectors. Indicates statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (2) with OLS on samples of firms from c  different sectors. 67 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 5: Exporting, importing and the average wage (firm-level)—by survey Dependent variable: Log(Wage) Survey Regressors: Exporter/Importer Regressors: Exporter/Importer Capital stock over sales Electricity costs over sales Foreign owned Log(Firm age) N Exporter Importer N Exporter Importer Angola 2006 212 0.096 -0.097 Botswana 2006 113 0.069 0.189 Burundi 2006 102 0.802a 0.688a Côte d’Ivoire 2009 152 1.036a 0.944a DRC 2006 149 0.359 0.294a DRC 2010 104 1.381 -0.209 DRC 2013 212 1.079 c 0.480c Egypt 2013 1,776 0.231a 0.053 1195 0.306a 0.038 Ethiopia 2011 191 0.792 b 0.344 Ghana 2007 292 0.169 -0.319a Ghana 2013 286 0.330 0.356* 105 0.256 -0.258 Guinea 2006 135 0.181 0.249 Kenya 2007 396 0.175c 0.308a Kenya 2013 321 0.318c 0.524a 206 0.270 0.238 Madagascar 2009 180 0.027 0.148 107 0.030 0.153 Madagascar 2013 214 0.081 -0.042 110 -0.262 0.160 Mali 2007 301 0.448a 0.190a Mauritius 2009 131 0.082 0.204 Morocco 2013 132 -0.089 0.246 Mozambique 2007 341 0.810 b 0.291a Namibia 2006 103 0.470 c 0.361c Nigeria 2007 948 0.620a -0.274a Nigeria 2014 622 -0.759 c -0.733c 256 0.551 -1.746a Senegal 2007 259 0.610 a 0.263 a Senegal 2014 179 0.617b 0.736a South Africa 2007 680 0.438 a 0.148b Tanzania 2006 272 -0.030 0.431a Tanzania 2013 173 0.113 0.690a 103 -0.014 0.646a Tunisia 2013 297 0.010 -0.142 224 0.044 -0.200 Uganda 2006 307 0.382b 0.476a Uganda 2013 221 0.223 0.106 Zambia 2007 304 0.406a 0.290b Zambia 2013 270 0.397b 0.601a 110 0.294 0.115 Zimbabwe 2011 353 0.298 c 0.027 327 0.289 c 0.035 Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (3) with OLS on samples of firms a  belonging to different surveys. Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (3) with OLS on samples of firms b  belonging to different surveys. Indicates statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (3) with OLS on samples of firms c  belonging to different surveys. 68 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES section empirically investigates the role of some of the Nyczak and Viegelahn, 2016). Also, due to the increased main channels that have been discussed in the literature need of firms for trade-related operational services, over the recent years. the share of production workers is likely to be lower in trading firms. All these workforce composition effects Discussion of the channels can have an impact on the average wage that is paid at One channel that can explain differences in wages the firm level. between trading and non-trading firms involves The trading activity of a firm can also give rise to differences in skill utilization across firms (Brambilla, technology upgrading, induced by technology transfers Lederman and Porto, 2012; Frazer, 2014). Exporters from the trading partner, which may increase workers’ often need to produce high quality products to compete productivity and can therefore lead to higher wages. successfully in foreign markets, especially if these markets Export destination has been shown to play a crucial are in developed economies (Verhoogen, 2008). To role in determining the wage premium in the case of produce high quality products, exporters are likely to have South Africa (Rankin and Schoer, 2013). Exporters pay a large share of high-skilled workers in their workforce. higher wages only when exporting to more-developed Importers are likely to require high-skilled workers to economies, whereas firms exporting to regional less- be able to absorb and work with the knowledge and developed markets are actually characterized by negative technology embedded in imported inputs (Kugler and wage premia. Also, a causal relationship between Verhoogen, 2012). Both exporters and importers rely exporting to high-income markets and paying higher on operational services related to logistics, marketing wages has been demonstrated globally (Brambilla and and finance, tasks that are typically executed by highly- Porto, 2016), indicating that firms and their workers may skilled workers (Matsuyama, 2007). Provided that skills benefit from technology upgrades induced by exports to are remunerated through higher wages, trading firms more developed economies. would be expected to pay higher average wages, simply because they have on average a higher-skilled workforce. Finally, the extension of a firm’s business to export markets increases the scale of a firm, thus lowering There are also other workforce composition effects average costs in the presence of increasing returns to that may be a channel for wage differences between scale. This will increase workers’ productivity, and hence trading and non-trading firms. Trading and non-trading potentially their wages. The relevance of international firms may differ in the share of temporary employees, scale economies for productivity within exporting firms female employees and production workers in their has empirically been shown for the case of Chinese Taipei total workforce. This will affect the average wage level, (Hwang, 2003). provided that temporary employees, female employees and production workers have a wage that differs from Empirical results the wage paid to permanent employees, male employees and non-production workers, respectively. Table 6 shows results from regressions at the firm level, where we investigate in columns 1–4 the role of the Recent evidence points to an increased use of temporary four above-described channels in explaining differences employment among trading firms (Machikita and Sato, between trading and non-trading firms. In the following, 2016). This may be driven by a preference for lower we consider four different variables as measures of dismissal costs and more flexibility which come with the four channels discussed above. To proxy for skills temporary employment contracts (Aleksynska and Berg, utilization, we use the average number of production 2016). It may also be driven by reduced incentives for workers’ years of education. To account for other firms to employ workers that acquire firm-specific skills, workforce characteristics, we respectively use the share which typically are permanent rather than temporary of women and production workers in full-time permanent employees. Similarly, there is also evidence for a more employment, and the share of temporary workers in total female workforce, especially among exporters (Duda- 69 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES employment. To proxy for the level of technology, we use on exporting remains positive and significant, the total factor productivity (TFP), which corresponds to the coefficient implying a wage premium of 18.2%, which is portion of output that is not explained by the amounts of very close to the 17.5%, obtained from the specification inputs used in production. Finally, for economies of scale, where this variable was not included. This suggests that we use firms’ total sales. skill utilization does not explain the difference in wages between exporters and non-exporters. Importing status We start by including these variables one by one as remains insignificant, even after including a measure for explanatory variables and observe how the estimated skills utilization in the regression. coefficients for firms’ export and import status react. As expected, we find firms’ skill utilization, measured Similarly, the coefficient on exporting remains positive by the average number of production workers’ years of and significant when including individual workforce education, to positively affect wages (column 1). But characteristics, including the share of temporary, female even after controlling for skills utilization, the coefficient and production worker employment, as additional explanatory variables (column 2). The difference Table 6: Exporting, importing and the average wage (firm-level)—different channels Dependent variable: Log(Wage) Skill Workforce Labour utilization characteristics Technology Scale productivity All (1) (2) (3) (4) (5) (6) 0.182a 0.206a 0.179a -0.154b 0.072 0.013 Exporter (0.064) (0.062) (0.060) (0.061) (0.057) (0.058) 0.010 0.023 0.026 -0.199a -0.086b -0.071 Importer (0.052) (0.051) (0.048) (0.049) (0.046) (0.044) -0.030a -0.037a -0.018a -0.009 -0.006 -0.009b Capital stock over sales (0.005) (0.007) (0.005) (0.008) (0.009) (0.004) -1.981a -1.892a -1.049c 0.107 0.823c 0.651 Electricity costs over sales (0.537) (0.532) (0.554) (0.456) (0.484) (0.481) 0.370a 0.450a 0.280a 0.033 0.059 0.037 Foreign owned (0.090) (0.093) (0.090) (0.081) (0.077) (0.087) 0.074a 0.056b 0.069a -0.059b 0.010 0.011 Log(Firm age) (0.026) (0.026) (0.024) (0.024) (0.023) (0.020) Log(Average production 0.106b 0.025 workers’ years of education) (0.044) (0.039) -0.004a -0.000 Female worker share (0.001) (0.001) -0.004b -0.001 Production worker share (0.002) (0.002) 0.006a 0.001 Temporary worker share (0.001) (0.001) 0.348a -0.480a Log(TFP) (0.041) (0.061) 0.290a 0.002 Log(Sales) (0.015) (0.017) 0.503a 0.723a Log(Labour productivity) (0.023) (0.042) Sector FE Yes Yes Yes Yes Yes Yes Survey FE Yes Yes Yes Yes Yes Yes R2 0.76 0.76 0.78 0.78 0.81 0.86 Number of observations 3530 3513 3507 3827 3575 2970 Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the full sample of firms. a  Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the full sample of firms. b  Indicates statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the full sample of firms. c  70 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES between exporters and non-exporters is estimated to be 20.6%. Importing status once more remains insignificant, even after including measures for different workforce characteristics in the regression. We then include TFP as a proxy for technology into the regression (column 3). Once more the coefficient on exporting remains positive and significant, with an estimated value of 17.9%. Differences in technology that may arise from technology transfers from trading partners are therefore unlikely to be responsible for the wage premium of exporters. Including TFP into the regression also does not change the statistical insignificance of the estimated importer coefficient. As proxy for the fourth channel, we include the firm’s total sales into the regression (column 4). The positive wage premium of exporting vanishes and even becomes negative. Moreover, the wage premium of importing becomes negative and statistically significant. This finding suggests that economies of scale play a major role in explaining differences between trading and non- trading firms. Achieving economies of scale through exporting hence appears to be a key channel through which exporting firms have higher average wages than for Belgium that associates increased import competition non-exporting firms. with decreased bargaining power for workers (Abraham, All channels that explain wage differences between Konings and Vanormelingen, 2009). trading and non-trading firms will work through increased Finally, we include all variables introduced in this section labor productivity. For example, technology transfer at the same time into the regression (column 6). It should in favor of exporting firms can only result in a wage be noted that TFP, sales, and labor productivity are premium if it boosts labor productivity. Labor productivity highly correlated by construction, as all include sales as hence summarizes all channels in one variable. When an ingredient, which makes it impossible to interpret the including labor productivity as an explanatory variable signs of the respective coefficients. More importantly, (calculated as the difference between sales and raw however, exporter and importer status are insignificant, material expenses per full-time permanent worker), as expected. we find no evidence for a positive wage premium of exporting (column 5). However, we find weak evidence 4.3 Employee-level results for a negative wage premium of importing. In other words, when comparing importers and non-importers The previous section did not find any evidence of a with identical labor productivity, importers pay on positive wage premium of importing, after controlling average lower wages than non-importers. This result for firm age, but strong evidence for a positive wage suggests that workers in importing firms reap a smaller premium of exporting. An empirical analysis of the share of the value added that is generated per worker, channels that are likely to drive this result suggested compared to non-importers, which could be evidence in that the positive wage premium of exporting is mainly favor of reduced bargaining power of workers in these the result of productivity gains from economies of scale. firms. This, for example, would be in line with evidence Technology transfers from the trading partner, as well 71 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES as the composition of firms’ workforce in terms of skills, differ between workers employed by trading firms and gender, type of contract, and type of task in contrast do those that are employed by non-trading firms. Table 7 not account for the wage differences observed between shows the relation of firms’ exporter and importer status exporters and non-exporters. with wages of employees, using different specifications. Without any control variables, we find for our sample of The previous section considered firm-level wages but did workers that wages of workers in exporting firms are 16% not control for individual worker characteristics such as higher than the wages of workers that work for firms not gender, marital status, age, level of education or years of engaging in export markets (column 1). This coefficient is experience, which might be driving some of the results. quantitatively similar to that estimated in the firm-level This section uses matched employer-employee data from regressions, reported in Table 3. In contrast, the wages 16 surveys, and analyzes whether employee-level wages of workers in importers are now on average around 5% Table 7: Exporting, importing and the average wage (employee-level)—full sample Dependent variable: Log(Wage) (1) (2) (3) (4) (5) (6) (7) (8) 0.163a 0.186a 0.166a 0.106a 0.125a 0.090b Exporter (0.034) (0.035) (0.036) (0.037) (0.037) (0.037) -0.053b -0.088a -0.097a -0.045c -0.070a -0.082a Importer (0.027) (0.028) (0.029) (0.026) (0.026) (0.028) -0.004 -0.006c Capital stock over sales (0.003) (0.003) -0.145 0.234 Electricity costs over sales (0.353) (0.387) -0.140a -0.085c Foreign owned (0.050) (0.048) -0.044 -0.041 -0.038 -0.018 Female (0.028) (0.028) (0.028) (0.029) 0.076a 0.074a 0.075a 0.077a Married (0.027) (0.027) (0.027) (0.027) -0.263a -0.266a -0.261a -0.260a No or primary education (0.029) (0.029) (0.029) (0.029) 0.331a 0.330a 0.330a 0.332a Vocational training (0.034) (0.035) (0.034) (0.035) 1.017a 1.033a 1.022a 1.038a University degree (0.053) (0.053) (0.053) (0.055) 0.005 0.025 0.006 0.017 Trade union member (0.031) (0.029) (0.031) (0.032) 0.007b 0.007b 0.007b 0.007b Experience with employer (0.003) (0.003) (0.003) (0.003) 0.012a 0.013a 0.012a 0.013a Total experience (0.003) (0.003) (0.003) (0.003) 0.009a 0.009a 0.009a 0.009a Worker age (0.002) (0.002) (0.002) (0.002) Sector FE Yes Yes Yes Yes Yes Yes Yes Yes Survey FE Yes Yes Yes Yes Yes Yes Yes Yes R2 0.83 0.83 0.83 0.82 0.87 0.87 0.87 0.87 Number of observations 6,641 6,648 6,641 6,286 5,067 5,074 5,067 4,855 Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the full sample of a  employees (16 surveys). Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the full sample of b  employees (16 surveys). Indicates statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the full sample of c  employees (16 surveys). 72 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES lower (column 2). Even when including firms’ exporter The employee-level regressions include the same firm- and importer status simultaneously as explanatory level control variables as the firm-level regressions. The variables in the regression, exporter status remains exception is firm age which was included in the firm-level positively associated with wages, while importer status regression, but cannot be included in the employee-level remains negatively associated with wages (column 3). regressions, as it is only available for 3 out of the 16 This also holds after including capital intensity, electricity surveys, resulting in a sample size that is too small. intensity and ownership status of the firm that employs These results are largely in line with the results obtained the worker (column 4). from the firm-level regressions. There is strong evidence When including individual worker characteristics as of a positive wage premium of exporting, and of the control variables, the wage premium of importers absence of a positive wage premium of importing, and remains negative and significant, while the wage even evidence for a negative wage premium. Worker premium of exporters remains positive and significant. characteristics are also found to only partially explain the This holds in all specifications (columns 5–8). As expected, differences in wages between trading and non-trading workers that are married and older age receive a higher firms, which also confirms the results obtained from firm- wage on average. Gender and trade union membership, level data. in contrast, do not appear to be related to individual Tables 8 and 9 show the estimations of the specification workers’ wages. The education level of workers explains in columns (3) and (6) of Table 7 by sector and by survey, to a large extent individual worker wages, with no or only respectively. The sector-specific results indicate that primary education being associated with lower wages, the negative wage premium for workers in importing vocational training being associated with higher wages, firms is particularly driven by workers in the textiles and and a university degree being associated with much garments, and metals and machinery sector. The results higher wages than workers with secondary education. for workers in exporting firms are dependent on the Total work experience also relates to wages positively, sector. There is a positive wage premium of exporting in particular work experience with the current employer. Table 8: Exporting, importing and the average wage (employee-level)—by sector Dependent variable: Log(Wage) Sector Regressors: Exporter/Importer Regressors: Exporter/Importer Individual worker characteristics Capital stock over sales Electricity costs over sales Foreign owned Individual worker characteristics N Exporter Importer N Exporter Importer Food & beverages 1,427 0.300 a 0.046 1375 0.301 a 0.038 Textiles & garments 866 -0.324a -0.339a 821 -0.146 -0.464a Wood & paper 646 0.312a -0.084 623 0.313a -0.085 Chemicals 360 -0.113 0.013 360 -0.119 -0.016 Non-metals & plastics 376 0.074 -0.088 347 0.037 -0.174 Metals & machinery 582 0.396a -0.148b 568 0.362a -0.076 Furniture 635 0.463 a 0.027 586 0.003 0.019 Other manufacturing 175 -0.572a 0.208 175 -0.541a -0.023 Indicate statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (5) with OLS on samples of employees a  from different sectors. Indicate statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (5) with OLS on samples of employees b  from different sectors. 73 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 9: Exporting, importing and the average wage (employee-level)—by survey Dependent variable: Log(Wage) Survey Regressors: Exporter/Importer Regressors: Exporter/Importer Individual worker characteristics Capital stock over sales Electricity costs over sales Foreign owned Individual worker characteristics N Exporter Importer N Exporter Importer Angola 2006 266 0.000 -0.058 246 0.000 -0.062 Botswana 2006 113 0.476 b 0.127 Burundi 2006 107 0.162 -0.061 DRC 2006 342 -0.111 0.032 342 -0.122 0.027 Ghana 2007 566 -0.000 -0.011 546 0.024 -0.020 Guinea 2006 224 -0.009 0.293a 193 -0.179 0.181 Mauritania 2006 124 -0.620 -0.140 121 0.478 b -0.043 Namibia 2006 279 0.053 0.367a 259 0.086 0.454a Rwanda 2006 171 0.051 -0.233 171 0.066 -0.250c South Africa 2007 1,087 0.134 a -0.105 b 1,073 0.136 a -0.103b Swaziland 2006 116 0.227 0.131 Tanzania 2006 336 -0.120 -0.065 326 -0.222c -0.088 Uganda 2006 323 0.010 -0.670 a 307 -0.075 -0.720a Zambia 2007 899 0.069 0.070 896 0.043 0.104 Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (6) with OLS belonging to different surveys. a  Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (6) with OLS belonging to different surveys. b  Indicates statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (6) with OLS belonging to different surveys. c  in the food and beverages, furniture, wood and paper, more or less pronounced in exporting compared with and metals and machinery sector. The wage premium non-exporting firms, or in importing compared with non- is in contrast negative for textiles and garments, and importing firms.9 other manufacturing. The results by survey vary greatly, with both significantly positive and significantly negative While we did not find any statistically significant gender coefficients being estimated. wage gap on average for the full sample of workers (see columns 5–8 of Table 7), there might still be differences 4.4 Gender and the difference in wages in the gender wage gap that depend on whether firms are exporters or importers. Therefore we re-do the This section examines whether firms’ exporting and analysis, but now include two interaction terms between importing status is related to gender wage differentials. the dummy variable that indicates whether the worker The literature provides plentiful empirical evidence that is a woman, and respectively the dummy variables that wages of female workers are on average lower than wages indicate firms’ exporter and importer status. Including paid to male workers (Blau and Kahn, 2017). Depending these interaction terms enables us to separate workers on the market, the gender wage gap has various causes, that are employed by firms engaged in exporting, ranging from discriminatory labor practices to overall importing or both, from workers that are employed by cultural attitudes. There is, however, no strong theoretical non-trading firms. As shown in Table 10, the coefficient premise that this gender wage gap would be significantly estimated for Female indicates a gender wage gap for 9 Boler, Javorcik and Ulltveit-Moe (2015) argue that exporters require a higher commitment from their employees due to more exposure to competition. If commitment is remunerated especially in exporting firms and women are perceived by employers to be less committed, this could explain a more pronounced gender wage gap within exporting firms. 74 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 10: Gender and the differences in average wages between trading and non-trading firms (employee-level)—full sample Dependent variable: Log(Wage) (1) (2) (3) (4) 0.092 b 0.115 b 0.069 Exporter (0.045) (0.045) (0.045) 0.044 0.033 0.067 Exporter* female (0.066) (0.067) (0.068) -0.058c -0.082a -0.093a Importer (0.030) (0.030) (0.031) 0.048 0.041 0.043 Importer* Female (0.053) (0.054) (0.055) -0.006c Capital stock over sales (0.003) 0.246 Electricity costs over sales (0.389) -0.085c Foreign owned (0.048) -0.055c -0.069c -0.070c -0.059 Female (0.031) (0.037) (0.038) (0.039) 0.076a 0.074a 0.075a 0.076a Married (0.027) (0.027) (0.027) (0.027) -0.262a -0.266a -0.261a -0.260a No or primary education (0.029) (0.029) (0.029) (0.030) 0.331a 0.331a 0.331a 0.333a Vocational training (0.035) (0.035) (0.035) (0.035) 1.017a 1.034a 1.022a 1.038a University degree (0.053) (0.053) (0.053) (0.055) 0.005 0.024 0.006 0.017 Trade union member (0.031) (0.029) (0.031) (0.032) 0.007b 0.007b 0.007b 0.007b Experience with employer (0.003) (0.003) (0.003) (0.003) 0.012a 0.013a 0.012a 0.013a Total experience (0.003) (0.003) (0.003) (0.003) 0.009a 0.009a 0.009a 0.009a Worker age (0.002) (0.002) (0.002) (0.002) Sector FE Yes Yes Yes Yes Survey FE Yes Yes Yes Yes R2 0.87 0.87 0.87 0.87 Number of observations 5,067 5,074 5,067 4,855 Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the full sample of a  employees (16 surveys). Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the full sample of b  employees (16 surveys). Indicates statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the full sample of c  employees (16 surveys). non-trading firms in 3 of the 4 specifications. In turn, 4.5 Robustness checks positive coefficients on interaction terms suggest that Finally, in Appendix B we present the results of a number there is no evidence for a gender wage gap in trading of robustness checks for the results presented above. firms. The results indicate that the wage premium of First, we assess whether outliers, in the form of the exporting and importing does not significantly vary with highest and lowest wage observation for each firm may the sex of the worker. 75 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES have overly influenced the results from the employee We endeavor to identify the channels that can explain our level analysis. Second, the results of firm-level regressions findings. Productivity gains through economies of scale are based on 65 surveys from 47 countries, while the explain the positive wage premium of exporters; neither employee-level regressions are based on 16 surveys productivity gains through increased skill utilization or the from 16 countries. To check the comparability of these employment of certain types of workers, nor productivity regressions we run the firm-level regressions on data gains through technology transfers, contribute to from the same 16 surveys for which also employee data exporters’ wage premium. Workers in importers are are available. Finally, we further limit the firm-level sample found to have weaker bargaining power than those to include only those firms for which employee data employed in non-importers, and are thus able to only are available, given that employee data are not available reap smaller shares of the value added that is generated. for all firms that form part of the 16 surveys that collected These results are somewhat surprising, given that the employee-level data. In all cases, these robustness checks trade literature has typically found that both exporting confirm the main findings of this analysis. and importing can be associated with higher wages. The 5. Conclusions arguments provided by this literature go beyond a mere effect of trade on firm performance through higher sales. This paper studies the relationship between exporting, The trade literature also associates exports with gains due importing and wages in Africa, using firm-level data and to increased foreign competition and skill premia, and employer-employee-level data from the World Bank imports with gains due to access to new technologies, Enterprise Surveys. On the basis of firm-level data, we and a better quality and wider variety of inputs. If these find that the average wage paid by exporters to their gains exist and are at least partially passed on to workers, workers is higher, even after controlling for such firm we would expect to find higher wages in exporting and characteristics as capital intensity, electricity intensity, importing firms. foreign ownership and firm age. The average wage paid by importers relative to non-importers is by contrast not This paper clearly indicates that there are other factors higher, after adding firm age as a control variable. On that contradict these general findings in the African the basis of employer-employee data, we can confirm context. On the one hand, African exporters are a positive exporting premium on wages, even after frequently incapable of competing in terms of product controlling for individual worker characteristics. Workers quality in more sophisticated markets outside of Africa. that are employed by importers are found to—if anything Thus, economies of scale resulting in lower prices —receive lower wages, when compared to workers remain the only viable channel to enter export markets, employed by non-importers. largely at the regional level. The strong price sensitivity 76 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The findings indicate that there is no significant gender wage gap within trading firms in the sample, while there is some evidence for a gender wage gap within non- trading firms. These results suggest that trading firms in the African context appear to contribute to gender equality, at least based on the data sample that this paper has worked with. of African customers, generally characterized by low Matched employer-employee data are only rarely personal income, reinforces this rationale. This particular available, especially for developing countries. In addition mechanism of competing through quantity as opposed to presenting results about the linkages between trade to quality can thus explain why economies of scale drives and wages, this paper also showed the value of analysis the positive premium on wages of exporters in Africa, of employer-employee data to inform policies, pointing and why other channels, such as skill utilization may play to the benefits of regularly collecting such data and a smaller role. increasing its quality. Such data are useful not only to examine the questions addressed in this paper, but also On the other hand, the non-existent or even negative questions that are related to a wider range of labour wage premium of importing is likely to be rooted in market issues. the nature of imports on the continent. The limited diversification of African economies means that some Given the ongoing regional and sub-regional integration inputs can only be obtained by importing. This reduces efforts of African countries, it is important to better the potential gains reaped from imported inputs, which understand under which conditions firms are able to are rather a source of higher costs than a way of having a benefit from the potential gains of trade, what the comparative advantage over firms sourcing domestically. potential bottlenecks to achieving these gains are, and In the absence of domestic raw material inputs, the how the gains can be translated into decent jobs for all higher material costs oblige firms to seek savings by workers. For trade liberalization to be sustainable and cutting other spending, including wages. In addition, inclusive, it is important to learn about the conditions high unemployment and large shares of informality in under which workers can reap at least some of the gains African countries shifts the bargaining power towards that are being made, especially in a continent like Africa. employers, which is confirmed by the negative regression This paper used available matched employer-employee results when controlling for labor productivity. data to provide evidence on these mechanisms and sought to contribute to a better understanding of the The paper also included an analysis of the gender gains from trade and how they are shared with workers dimension. The findings indicate that there is no in the African context. significant gender wage gap within trading firms in the sample, while there is some evidence for a gender wage gap within non-trading firms. These results suggest that trading firms in the African context appear to contribute to gender equality, at least based on the data sample that this paper has worked with. 77 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Abraham, F., J. Konings, and S. Vanormelingen (2009). “The effect on globalization on union bargaining and price-cost margins of firms,” Review of World Economics, Vol. 145, pp. 13–36. Aleksynska, M., and J. Berg (2016). “Firms’ demand for temporary labour in developing countries: necessity or strategy?”. Conditions of Work and Employment Series No. 77, INWORK, Geneva: International Labour Office. Amiti, M., and D.R. Davis (2011). “Trade, firms, and wages: theory and evidence,” Review of Economic Studies, Vol. 79, pp. 1–36. Amiti, M., and J. Konings (2007). “Trade liberalization, intermediate inputs, and productivity: Evidence from Indonesia,” American Economic Review, Vol. 97, No. 5, pp. 1611–1638. Bernard, A.B., J.B. Jensen, S.J. Redding, and P.K. Schott (2007). “Firms in international trade,” Journal of Economic Perspectives, Vol. 21, No. 3, pp. 105–130. ______(2012) “The Empirics of Firm Heterogeneity and International Trade,” Annual Review of Economics, Vol. 4, pp. 283–313. Bigsten, A., and M. Gebreeyesus (2009). “Firm Productivity and Exports: Evidence from Ethiopian Manufacturing,” Journal of Development Studies, Vol. 45, No. 10, pp. 1594–1614. Bigsten, A., P. Collier, S. Dercon, M. Fafchamps, B. Gauthier, J.W. Gunning, A. Oduro, R. Oostendorp, C. Pattillo, M. Soederbom, F. Teal, and A. Zeufack (2004). “Do African manufacturing firms learn from exporting?” Journal of Development Studies, Vol. 40, No. 3, pp. 115–141. Bigsten, A., M. Gebreeyesus, and M. Soederbom (2016). “Tariffs and Firm Performance in Ethiopia,” Journal of Development Studies, Vol. 52, No. 7, pp. 986-1001. Blau, Francine D., and Lawrence M. Kahn (2017) “The Gender-Wage Gap: Extent, Trends, and Explanations,” Journal of Economic Literature, Vol. 55, No. 3, pp. 789-865. Boler, Esther Ann, Beata Javorcik, and Karen Helene Ulltveit-Moe (2015) “Globalization: A Woman’s Best Friend? Exporters and the Gender Wage Gap.” CEP Discussion Papers 1358, London: Centre for Economic Performance. Brambilla, I. and G. Porto (2016). “High-income export destinations, quality and wages,” Journal of International Economics, Vol. 98, pp. 21–35. Brambilla, I., D. Lederman, and G. Porto (2012). “Exports, export destinations, and skills,” American Economic Review, Vol. 102, No. 7, pp. 3406–3438. Brambilla, I., N. Depetris-Chauvin, and G. Porto (2017). “Examining the export wage premium in developing countries,” Review of International Economics, Vol. 25, No. 3, pp. 447-475. De Loecker, J. (2013). “Detecting Learning by Exporting,” American Economic Journal: Microeconomics, Vol. 5, No. 3, pp. 1–21. Dollar, D. and A. Kraay (2004). “Trade, growth and poverty,” The Economic Journal, Vol. 114, pp. F22–F49. Duda-Nyczak, M. and C. Viegelahn (2016). “Exporters, importers and employment: Firm-level evidence from Africa,” Indian Journal of Labour Economics, Vol. 59, No. 3, pp. 309-341. Ethier, W. (1982). “National and international returns to scale in the modern theory of international trade,” American Economic Review, Vol. 72, pp. 389–405. Frías, J., D. Kaplan, and E. Verhoogen (2012). “Exports and Within-Plant Wage Distributions: Evidence from Mexico,” American Economic Review: Papers and Proceedings, Vol. 102, No. 3, pp. 435–440. Frazer, G. (2014). “Imports, import sources and skill utilization.” Working Paper. Fu, D. and Y. Wu (2013). “Export wage premium in China’s manufacturing sector: A firm level analysis,” China Economic Review, Vol. 26, pp. 182–196. 78 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Grossman, G., and E. Helpman (1991). Innovation and Growth in The Global Economy, Cambridge, Massachusetts: MIT Press. Halpern, L., M. Koren, and A. Szeidl (2015). “Imported inputs and productivity,” American Economic Review, Vol. 105, No. 12, pp. 3660–3703. Hwang, A.R. (2003). “Exports, Returns to Scale, and Total Factor Productivity: The Case of Taiwanese Manufacturing Industries,” Review of Development Economics, Vol. 7, No. 2, pp. 204–216. International Labour Office (2018). World Employment and Social Outlook: Trends 2018, Geneva. Kasahara, H., and B. Lapham (2013). “Productivity and the decision to import and export: theory and evidence,” Journal of International Economics, Vol. 89, pp. 297–316. Kugler, M., and E. Verhoogen (2012). “Prices, plant size and product quality,” Review of Economic Studies, Vol. 79, No. 1, pp. 307–339. Le Goff, M., and R. J. Singh (2014). “Does trade reduce poverty? A view from Africa,” Journal of African Trade, Vol. 1, No. 1, pp. 5–14. Machikita, T., and H. Sato (2016). “Temporary jobs and globalization.” IDE Discussion Paper No. 585, Chiba: Institute of Developing Economies, Japan External Trade Organization (JETRO). Macis, M., and F. Schivardi (2016). “Exports and Wages: Rent Sharing, Workforce Composition, or Returns to Skills?” Journal of Labour Economics, Vol. 34, No. 4, pp. 945–978. Markusen, J.R. (1989). “Trade in producer services and in other specialized intermediate inputs,” American Economic Review, Vol. 79, pp. 85–95. Matsuyama, K. (2007). “Beyond Icebergs: Towards a Theory of Biased Globalization,” Review of Economic Studies, Vol. 74, No. 1, pp. 237–53. Melitz, M.J. (2003). “The impact of trade on intra-industry reallocations and aggregate industry productivity,” Econometrica, Vol. 71, No. 6, pp. 1695–1725. Mengistae, T., and C. Pattillo (2004). “Export Orientation and Productivity in Sub-Saharan Africa,” IMF Staff Papers, Vol. 51, No. 2, pp. 327–353. Milner, C., and V. Tandrayen (2007). “The Impact of Exporting and Export Destination on Manufacturing Wages: Evidence for Sub-Saharan Africa,” Review of Development Economics, Vol. 11, No. 1, pp. 13–30. Rankin, N., and V. Schoer (2013). “Export Destination, Product Quality and Wages in a Middle-Income Country. The Case of South Africa,” Review of Development Economics, Vol. 17, No. 1, pp. 64–73. Saliola, Federica, and Murat Seker (2011). “Total factor productivity across the developing world.” Enterprise Surveys, Enterprise Note Series, Enterprise Note No. 23, Washington DC: World Bank. Schank, T., C. Schnabel, and J. Wagner (2007). “Do exporters really pay higher wages? First evidence from German linked employer-employee data,” Journal of International Economics, Vol. 72, No. 1, pp. 52–74. Stone, S., and B. Shepherd (2011). “Dynamic gains from trade: the role of intermediate inputs and equipment imports.” OECD Trade Policy Papers No. 110, Paris: Organization for Economic Co-operation and Development. Vandenbussche, H., and C. Viegelahn (2016). “Trade protection and input switching.” CEPR Working Paper No. 11395, London: Centre for Economic Policy Research. Verhoogen, E. (2008). “Trade, Quality Upgrading and Wage Inequality in the Mexican Manufacturing Sector,” The Review of World Economics, Vol. 123, No. 2, pp. 489–530. World Bank (2012). “Globalization’s impact on gender equality: What’s happened and what’s needed,” Chapter 6 in World Development Report 2012: Gender equality and development , Washington D.C.: World Bank. 79 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix Appendix A: Methods Table A.1: Surveys Country Years Country Years 1 Angola 2006, 2010 a 25 Madagascar 2009, 2013 2 Benin 2009 26 Malawi 2009, 2014 3 Botswana 27 Mali 2007, 2010 4 Burkina Faso 2009 28 Mauritania 2006,a 2014 5 Burundi 2006,a 2014 29 Mauritius 2009 6 Cameroon 2009 30 Morocco 2013 7 Cape Verde 2009 31 Mozambique 2007 8 Central African Republic 2011 32 Namibia 2006,a 2014 9 Chad 2009 33 Niger 2009 10 Congo, Republic of 2009 34 Nigeria 2007, 2014 11 Côte d’Ivoire 2009 35 Senegal 2007, 2014 12 DRC 2006, 2010, 2013 a 36 Rwanda 2006,a 2011 13 Djibouti 2013 37 Sierra Leone 2009 14 Egypt 2013 38 South Africa 2007,a 15 Eritrea 2009 39 South Sudan 2014 16 Ethiopia 2011 40 Sudan 2014 17 Gabon 2009 41 Swaziland 2006 18 Gambia 2006 42 Tanzania 2006,a 2013 19 Ghana 2007,a 2013 43 Togo 2009 20 Guinea 2006 a 44 Tunisia 2013 21 Guinea-Bissau 2006 45 Uganda 2006,a 2013 22 Kenya 2007, 2013 46 Zambia 2007,a 2013 23 Lesotho 2009 47 Zimbabwe 2011 24 Liberia 2009 Notes: This table lists all the surveys with firm-level data that are included in the analysis. This survey also has employee-level data available. a  Appendix B: Estimation of firm-level We estimate this production function in logarithmic total factor productivity form, where the estimated equation can be written as: This paper relies on firm-level estimates of total factor (8) productivity (TFP) as a measure of firm efficiency. To estimate TFP, we follow Saliola and Seker (2011) and use As measure of output, we use firm-level sales. The a simple Cobb-Douglas production function, that can be capital stock is measured as the replacement value of specified as follows: machinery, vehicles, equipment, land and buildings. Labor input is given by the number of fulltime permanent Y = TFP · LαKβM γ (7) employees, while material input in the data corresponds to raw material expenses. The residual of the estimated with Y denoting output, L denoting labor input, K equation corresponds to an estimate of total factor denoting capital input and M denoting material inputs. productivity, in logarithmic form. 80 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix C: Robustness checks Excluding for each firm the worker observation with the C.1 Controlling for outliers highest and lowest wage, we obtain results that are very similar to those based on the full-sample analysis. Table To validate the findings presented in this paper we checked Annex C.1. presents the estimated coefficients for the that the estimated coefficients are not exceedingly driven variables of interest in specifications (1)–(6), where the by certain outliers. The first robustness check considers control variables respectively correspond to the control whether outlier observations overly affect the results variables included in the regressions whose results were of the employee-level analysis. The highest and lowest shown in Table 7 of this paper. In the case of both importing wage observation per firm are dropped from the dataset, and exporting premia, the coefficients maintain the same and the same employee-level analysis is conducted as in sign as in the full-sample estimation, i.e., negative for section 4.3. Table C.1: Importing, exporting and the average wage (employee-level)—no outliers Dependent variable: Log(Wage) (1) (2) (3) (4) (5) (6) (7) (8) 0.135a 0.164a 0.141a 0.091b 0.116a 0.074c Exporter (0.039) (0.041) (0.041) (0.043) (0.043) (0.042) -0.076b -0.108a -0.124a -0.061b -0.087a -0.105a Importer (0.031) (0.032) (0.033) (0.030) (0.030) (0.031) -0.004 -0.009c Capital stock over sales (0.004) (0.005) -0.179 0.137 Electricity costs over sales (0.382) (0.386) -0.128b -0.080 Foreign owned (0.061) (0.058) -0.002 0.003 0.006 0.029 Female (0.029) (0.030) (0.029) (0.030) 0.031 0.027 0.030 0.034 Married (0.030) (0.030) (0.030) (0.030) -0.239a -0.243a -0.237a -0.237a No or primary education (0.032) (0.032) (0.032) (0.033) 0.326a 0.325a 0.327a 0.327a Vocational training (0.037) (0.038) (0.037) (0.038) 0.913a 0.930a 0.918a 0.945a University degree (0.062) (0.062) (0.063) (0.066) -0.006 0.013 -0.005 0.003 Trade union member (0.037) (0.034) (0.037) (0.038) 0.008b 0.008b 0.008b 0.009b Experience with employer (0.004) (0.004) (0.004) (0.004) 0.009a 0.009a 0.009a 0.009a Total experience (0.003) (0.003) (0.003) (0.003) 0.009a 0.009a 0.009a 0.009a Worker age (0.002) (0.002) (0.002) (0.002) Sector FE Yes Yes Yes Yes Yes Yes Yes Yes Survey FE Yes Yes Yes Yes Yes Yes Yes Yes R2 0.86 0.86 0.86 0.86 0.89 0.89 0.89 0.89 Number of observations 4161 4166 4161 3931 3253 3258 3253 3121 Indicate statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the sample of employees a  (16 surveys), where the employees with the highest and lowest wage per firm have been excluded. Indicate statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the sample of employees b  (16 surveys), where the employees with the highest and lowest wage per firm have been excluded. Indicate statistical significance at the 10% level. Reported standard errors are robust. Regression results are obtained from estimating equation (4) with OLS on the sample of employees c  (16 surveys), where the employees with the highest and lowest wage per firm have been excluded. 81 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES importing and positive for exporting. In addition, they are When including firm age in the regression (column 6), significant in all the estimated specifications, which is also however, both exporting and importing status become the case in the original regressions. Results are hence not insignificant. When estimated on the full-sample, this was driven by outlier observations for wages. only the case for importing status. Given, however, that firm age is only available in 3 out of the 16 surveys, the Limiting country coverage number of observations is too small to draw any reliable conclusions from that. The results of firm-level and employee-level regressions on the full sample cannot be directly compared. While the Limiting country and firm overage results of firm-level regressions are based on 65 surveys from 47 countries, the employee-level regressions are As a third robustness check, we further limit the firm-level based on 16 surveys from 16 countries. One way to sample to include only those firms for which employee make these regressions somewhat comparable is to run data exist (employee data are not available for all firms the firm-level regressions on the data from the same 16 that form part of the 16 surveys that collected employee- surveys for which also employee data are available. level data). The model specifications are identical with the regressions used in the firm-level analysis of section Table Annex C.2 shows results of the firm-level 4.1 and reported in Table 3 of this paper. regressions presented in Table 3 of this paper, but for the restricted sample. The coefficients on both exporting and Running the firm-level regressions on the sample importing status are positive and statistically significant restricted in size to firms represented in the employee in all but the last specification. This corresponds to the dataset yields similar results, as illustrated in Table Annex results obtained from the regressions run on the full C.3. Exporting and importing status are both positively sample. Estimated magnitudes of the coefficients based associated with wages. The specification that includes on the restricted sample tend to be larger than those firm age as a control variable results in a sample that is based on the full sample, reinforcing earlier findings. too small to provide any meaningful results. Table C.2: Importing, exporting and the average wage (firm-level) – only surveys represented in employee-level data Dependent variable: Log(Wage) (1) (2) (3) (4) (5) (6) 0.314a 0.257a 0.227a 0.191a -0.002 Exporter (0.043) (0.043) (0.044) (0.044) (0.153) 0.245a 0.195a 0.180a 0.157a -0.099 Importer (0.033) (0.033) (0.033) (0.033) (0.142) -0.036a -0.035a -0.063b Capital stock over sales (0.006) (0.006) (0.028) -2.867a -2.794a -7.144a Electricity costs over sales (0.640) (0.626) (2.010) 0.331a 0.388 Foreign owned (0.052) (0.461) 0.135b Log(Firm age) (0.057) Sector FE Yes Yes Yes Yes Yes Yes Survey FE Yes Yes Yes Yes Yes Yes R2 0.85 0.85 0.85 0.86 0.86 0.92 Number of observations 2961 2961 2960 2838 2837 164 Indicates statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the sample of firms that a  belong to the 16 surveys for which also employee data are available. Indicates statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the sample of firms that b  belong to the 16 surveys for which also employee data are available. 82 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table C.3: Importing, exporting and the average wage (firm-level)—only firms represented in employee-level data Dependent variable: Log(Wage) (1) (2) (3) (4) (5) (6) 0.325 a 0.292 a 0.272 a 0.254 a -0.079 Exporter (0.064) (0.065) (0.066) (0.065) (0.189) 0.176a 0.130a 0.120a 0.100b 0.005 Importer (0.046) (0.046) (0.045) (0.045) (0.184) -0.038a -0.036a -0.121b Capital stock over sales (0.007) (0.006) (0.046) -2.731a -2.802a -1.451 Electricity costs over sales (0.631) (0.636) (5.662) 0.301a -0.565b Foreign owned (0.072) (0.267) 0.015 Log(Firm age) (0.084) Sector FE Yes Yes Yes Yes Yes Yes Survey FE Yes Yes Yes Yes Yes Yes R2 0.86 0.86 0.86 0.87 0.87 0.93 Number of observations 1382 1383 1382 1311 1310 80 Indicate statistical significance at the 1% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the sample of firms that a  belong to the firms within the 16 surveys that also report employee data. Indicate statistical significance at the 5% level. Reported standard errors are robust. Regression results are obtained from estimating equation (1) with OLS on the sample of firms that b  belong to the firms within the 16 surveys that also report employee data. 83 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The Poverty Impact of Modernising Dar es Salaam Port 1 Nicolás Depetris-Chauvin, Geneva School of Business Administration, HES-SO2 Pablo Depetris-Chauvin, Universidad Torcuato Di Tella3 Francis Mulangu, Millennium Challenge Corporation4 1. Introduction particularly the impact on urban versus rural households T and on different income levels. his study assesses the likely impact of the The study is organized as follows. The next section modernization of the Port of Dar es Salaam presents basic information regarding the planned on household welfare and poverty in Tanzania infrastructure and operational improvements for the and neighboring countries. Trade volumes in Dar es Salaam Port. Section 3 introduces the trade and Tanzania increased more than 10% per year in the last poverty methodology based on the use of micro data, decade, and international trade has been one of the which enables the identification of the different channels engines of growth in the country. However, the current through which trade may affect poverty. Section 4 state of Dar es Salaam port is a severe constraint on describes trade and industrial policy and the structure further growth. Increasing the efficiency of the port is a of production and trade flows in Tanzania. Based on the key challenge; container vessels have to wait an average findings of section 4 and on complementary information of more than 10 days before berthing, and dwell times on market structure and infrastructure, section 5 average another 10 days. The costs associated with estimates the expected impact of the port project on the inefficiencies in the port are partially related to trade flows and prices. It covers a limited basket of congestion. The situation is more critical for imports goods which are a small proportion of the cargo going than for exports; the inefficiencies act as an implicit tax through the port, but make up a very large share of the on imports and to a lesser extent as a tax on exports consumption basket of poor households and are also an (Morisset, 2013). important source of their income. Section 6 presents We assume that port modernization would result in a a poverty profile for Tanzania. In particular, it describes reduction of 5% in border prices for bulk cargo, and consumption and income patterns for different levels measure the impact on the economy and on poverty in of livelihood, distinguishing between rural and urban the short run. The analysis proceeds in two-steps: first, households. Section 7 combines the information on price the extent of the transmission of border prices to retail changes (section 5) and the household data (section 6) and farm gate prices will be evaluated; second, these to provide estimations of the short-term welfare impact. estimated price changes will be used to determine Section 8 concludes with some recommendations. the welfare effects for different demographic groups, 1 This study has been funded by UK aid from the UK Government, however the views expressed do not necessarily reflect the UK Government’s opinions or official polices. We thank Tim Bushell and DFID Tanzania seminar participants for useful comments and discussion. All errors are our responsibility. 2 Geneva School of Business Administration, HES-SO. 3 Universidad Torcuato Di Tella. 4 Millennium Challenge Corporation. 84 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 85 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 2. Enhancement of the Dar es in the world. Tanzania was ranked only 86th in UNCTAD’s Salaam Port Liner Shipping Connectivity index, behind South Africa (31st), Ghana (64th) and Kenya (85th) on the African Dar es Salaam Port handles approximately 90% of the continent. The problem is further heightened by weak country’s international sea trade and serves as a transit road and rail connectivity to and from the port. The poor port for landlocked countries Burundi, Democratic performance of the port acts as an implicit tax on exports Republic of Congo, Rwanda, Uganda, Zambia and Malawi. and imports, and constrains growth in Tanzania and the The port has a rated capacity of 4.1 million (dwt) dry landlocked neighbors served by the port. cargo and 6.0 million (dwt) bulk liquid cargo. It has a total quay length of about 2,000 metres with eleven deep- Trade has been one of the engines of growth in recent water berths. years in Tanzania. Since 2000 export volumes have increased 7.74% a year and import volumes 11.65% As a consequence of inadequate investment and a a year. In 2013 Dar Port handled a total of 13.5 million deterioration in port management, the port suffers from tons of cargo (compared to 12.1 million tons in 2012) and insufficient capacity, frequent power cuts, high dwell merchandise trade amounted to more than US$17 billon, time and exceptionally high port congestion. Bulk, RoRo almost 52% of Tanzania’s GDP. Given the strategic role and Container customers generally rely on their own played by the port in the economy of the country and operations. Dar es Salaam port is one of the least efficient 86 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES the region, and the evident deterioration in its capacity The overall effect of globalization in a developing to deal with increasing trade flows, the Tanzania Ports country may depend on the provision of complementary Authority, the Ministry of Transport and Trade Mark East policies, institutions, and infrastructure, highlighting the Africa (TMEA) worked together to put forward a support importance of public policies. program. The Government later agreed to partner with Food is often the largest household expenditure for poor TMEA, the United Kingdom Department for International people, while much of their income will come from wages Development (DFID) and the World Bank to implement a and, for rural households, from sales of agricultural series of reforms to increase the efficiency and handling produce. The modernization of the port should reduce capacity of Dar Port through improvements in port delays, wastage and losses, and probably increase the infrastructure and cargo clearance procedures so that it level of competition in the logistics associated with is better able to handle future growth in trade. trade. This will reduce the prices of imported goods While there is a long-standing consensus among (though the level of pass-through will depend on other academics and policy-makers on the positive role of port complementary policies) and increase the level of infrastructure investments in fostering trade and growth, competition with local producers. It will also provide the link with poverty reduction is weaker. The trade enhanced export opportunities both for local producers literature emphasizes the gains from trade, but it also and landlocked neighboring countries. All these changes acknowledges that there are unavoidably winners and will have distributional impacts. losers from trade. This creates a potential distributional The conceptual framework in this paper is organized conflict as well as potential adverse effects on equality. around the two-step approach of the trade and poverty In what follows the research develops a methodology literature.5 The first step involves an assessment of how to estimate the welfare impact of the proposed the infrastructure project will affect trade flows and improvement in the Dar es Salaam port, identifying the how these changes in trading opportunities will affect winners and losers from increased trade. the prices of goods and production factors. This step 3. Trade and poverty methodology will require an assessment of the effect of the reform on border prices and how those changes on border There is no general framework that predicts the effect prices would be transmitted to retail and producer prices of trade on poverty. Globalization poses both risks and potentially to wages. The extent of pass-through and opportunities for developing countries and poor will depend, among many factors, on the trade and citizens of those countries. Access to new markets for production structure, the existing trade and industrial exporting firms in developing countries potentially policy, sectoral market structure (level of competition creates employment and increases the salary of workers among importers and exporters) and the degree of in those sectors. Local firms can also access better inputs market integration in Tanzania. The second step uses and technology, helping to close the productivity gap household surveys to assess the poverty impacts of observed in most developing countries. However, in those changes in trade. It will follow the standard first the presence of market failures it is not clear that the order effects approach, as in Deaton (1989, 1997). Using gains from trade will be observed. Moreover, trade microdata from the household surveys, consumption could potentially increase unemployment, poverty and income shares derived from the production and and income inequality in the short and medium term, consumption of different goods will be used to evaluate making it unsustainable socially, economically, and the consumption, income, and overall impacts of a given politically (Artuc et al., 2015; Artuc and Porto, 2016; and price change. Dix-Carneiro, 2014). This suggests that the relationship between international trade and poverty is complex. 5 See Chapter 1 of World Bank and World Trade Organization (2015) for the macroeconomic links between growth, trade, and poverty. 87 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 4. Production and trade patterns through its membership in the East African Community and policy in Tanzania Customs Union, non-tariff barriers (NTBs) remain major impediments to trade and business development in Tanzania experienced rapid economic growth over the the EAC. In addition, poor transport infrastructure and past decade and a half. GDP increased on average 6.7% the pattern of specialization of the involved countries per year from 2000 to 2016, compared to 3% during severely limit regional exchanges. Finally, sectoral 1990–2000. During the same period, the share of the policies that affect trade in key commodities can have an agricultural sector, which employs 75% of the labour important impact on poverty. These include food export force, in GDP declined. In 2016 agriculture accounted for bans and import quotas for rice and sugar, among others. 31% of Tanzania’s GDP, industry accounted for 2%, and services for 42%. The most important trade partners of the country are in Asia and Europe. Trade with neighboring countries is An analysis of the impact of trade on poverty in Tanzania growing but is still small, except perhaps for Kenya. Dar needs to take into account several trade and industrial es Salaam port provides transit services for neighboring policy issues. First, while external trade tariffs have landlocked countries, but faces competition from fallen sharply in Tanzania over the years and tariffs Mombasa (Kenya), Durban (South Africa) and to a less on trade with neighbors have been mainly eliminated extent Nacala (Mozambique) (see Figure 1). Figure 1: Major trade corridors in East Africa Major Corridors in East Africa Major Hubs and Ports Capital Cities International Boundries Main Border Crossings Nacala Corridor Central Corridor Dar es Salaam Corridor Northern Corridor Overlapping Corridor Route Lakes (Ferries) Railways Trunk Roads 88 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Understanding the agricultural sector is essential for changed. The country could potentially increase its share estimating the poverty impact of the enhancement of the in global markets for these and other cash crops, and a port of Dar es Salaam, as most families, and in particular more efficient port could contribute to that. most poor households, are involved in agriculture. Farming or livestock husbandry is the main occupation for 5. The expected impact of the port the majority of the population in all regions except for project on trade flows and prices Dar es Salaam and Mjini Magharibi. The main activity for This study assumes that the project’s rearrangement 79.4% of the families in rural areas is farming, livestock of existing port infrastructure and improvements in husbandry or fishing, compared to 30.5% in urban areas administrative procedures would reduce border prices for (Tanzania, 2012). bulk imports, and increase border prices for exports, by 5%. World Bank (2013) finds that the total cumulative In the agricultural sector, production is dominated by cost of the delays and additional monetary costs for the food crops and livestock. The main products include meat Dar es Salaam port are equivalent to a tariff of about 5% (US$783 million), bananas (US$711 million), dry beans on bulk imports. This figure was obtained by comparing (US$685 million), maize (US$668 million), milk (US$578 the performance of Dar es Salaam with the port of million), cassava (US$571 million) and rice (US$482 Mombasa. The tariff equivalent was computed as the million). Tobacco, cotton and cashew were the largest sum of the direct monetary costs, the cost of waiting at cash crops for the export market (FAOStat, 2015). anchorage and the inventory cost, based on an average Tanzania’s three main import products in terms of value of US$1,137 per ton for dry bulk imports. Interviews value are wheat, oil palm, and refined sugar. The three with several local stakeholders confirmed that this is a products are mainly imported through Dar es Salaam reasonable figure. port. However, as will be seen in the household data, Estimating the impact on exports of port improvements these products do not account for a large share of the is complicated. While an improvement in the efficiency consumption basket of households under the national of the port should increase the competitiveness of poverty line (mostly rural households). These products Tanzanian exports, producers in Tanzania face many are more important for urban households, including constraints other than the inefficiencies of the port. those below the international poverty line of US$3 per According to the KPMG/World Bank (2013) “Pulse of day. Cost reductions from port improvements may not the Tanzanian Economy” survey, the state of the port is be passed to consumers of these three products, for a a serious constraint for one third of the mid-size firms number of reasons. Imports of sugar are controlled by the surveyed. In addition, 43% of the respondents cited Sugar Board of Tanzania, and the price is kept artificially corruption, followed by tax rates and regulations (29%) high to encourage investments by local producers. and access to finance (24%) as major constraints. Imports of wheat are dominated by one firm (SSB), which Moreover, other surveys point to deficiencies in railways accounts for almost three quarters of the market. Palm and road infrastructure, as well as unreliable energy oil competes with other sources of edible oil produced supply, as important constraints on businesses. The locally (sunflower, seeds, and cotton). lack of an adequately-educated force is also a severe The main agricultural export products are coffee, constraint. Tanzania is ranked 131st out of 189 countries tobacco, and cashew. Food crop exports are very low. in Doing Business 2015 (World Bank), somewhat better Between 2000 and 2011, the configuration of Tanzania’s than position 145 in 2014. top agricultural export items has not changed. Export To study the transmission of border prices to retail and growth has been important in the case of tobacco but farmgate prices, it is important to consider the level disappointing for coffee and cashews. These are products of competition in the value chains for the different that are exported with little domestic processing, and the products that are produced and traded in a country (Horn required level of technology to produce them has not 89 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES and Levinshon, 2001; Porto et al., 2011; Swinnen and The model traces how the allocation of factors of Vandeplas, 2010 and 2014). In Tanzania, international production to various cash and food crops depends market conditions combine with domestic market on competition along the supply chain and on the configurations in shaping agriculture growth and poverty constraints faced by different types of farmers. Farmers reduction. While the farming sector is composed mostly choose how to allocate their resources (land, labor) of smallholders, the lower layers of the value chains are to the production of subsistence food for their own usually dominated by a small number of firms. Farmers consumption, production of marketable food surplus, or may suffer from the non-competitive behavior of other production of cash crops for exports.6 These activities agents along the chain, or be constrained from selling offer different prices and entail different production output in markets because transport and other services costs. In equilibrium, because of risk and food security are not available or are too costly. issues, all farmers produce for their own consumption. However, depending on farm-gate prices, costs, and other This study uses a model developed by Depetris-Chauvin constraints (such as infrastructure, transport costs, risks), et al. (2017) to estimate the impact of the enhancement some farmers specialize in food production altogether, of the port of Dar es Salaam on poverty. This model while others devote some resources to export products. explains the allocation of factors of production to various We develop a game-theory model of supply chains in cash cash and food crops and how this allocation depends on crop agriculture, where market structure is characterized competition along the supply chain and on the constraints by many smallholders and a few exporters. These faced by different types of farmers. The model describes exporters buy raw inputs from the farmers and sell them the behavior of farms, exporters and importers in a simple (perhaps after some processing) in international markets partial equilibrium setting. There are different versions of at given prices. Firms enjoy oligopsony power internally the model to deal with the three basic scenarios that we and set farm-gate prices. face in our empirical work. That is, a first version of the model explores the case of cash crop production (mostly The oligopsony game delivers the equilibrium farm- for exports). In the case of exported cash crops, farmers gate prices that the firms offer to farmers. In setting sell products to oligopsonies, which then undertake the these prices, each exporter takes into account its own international trading. This version can be used to study characteristics, the characteristics of other exporters, crops such as cotton, coffee, tea, tobacco, cacao, vanilla, and the endogenous responses of the farmers (which etc. This model is then adapted to deal with the case of a in turn affect food and cash crop production). Once the country that is a net exporter of a food crop. In this case, equilibrium of the model is found, it is possible to perform there are oligopsonies in charge of exports, but there is comparative static exercises to study how farm-gate also a domestic residual market of net-consumers of that prices depend on various parameters of the economy, crop. Food crop exports can include any relevant crop in a including the degree of imperfect competition and particular country, namely maize, rice, fish, livestock, etc. competition policies, household costs and constraints, Finally, a different version of the model is developed for and so on. In this study, the model is used to study how the case of a country that is a net importer of a food crop changes in border prices from a reduction in the cost of whereby excess demand is met via international trade, using the port affect domestic prices for a given level of and net-consumers must purchase these agricultural market structure, based on key parameters that capture goods from oligopolies. The three versions of the model various household constraints and institutional access. share common elements, such as the structure of utility, A 5% reduction in the implicit tax imposed by the port the constraints on production and the market structure, inefficiencies will increase the price of exportable crops but differ in the way the models are solved to account for (Table 1, first half). Farmers producing cotton receive exportable and importable prices. 9.41% more for their cash crop in the current scenario 6 The agricultural household model is based on the well-known models of Barnum and Squire (1979), Singh, Squire and Strauss (1986), de Janvry, Fafchamps, and Sadoulet (1991), Benjamin (1992), and Taylor and Adelman (2003). 90 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 1: Effect of increased efficiency of Dar es Salaam Port on farm gate and retail prices Baseline Perfect Competition Increase of 5% in: Border Price for cotton 9.41 20.45 Border Price for cassava 5.79 6.89 Reduction of 5% in: Border Price for rice -4.42 -4.12 Border Price for maize -4.21 -4.33 Border Price for dairy -3.80 -2.62 Border Price for wheat -2.07 -16.39 Source: Model simulations. Note: Figures denote percentage change. but could receive up to 20.45% more if there were These results are more relevant for Dar es Salaam more competition in the supply chains for cotton. This and neighboring regions than for the whole country. estimation of the pass through is very high and takes Infrastructure and logistics in Tanzania are poor, even by into account the supply responses of the farmers in African standards. While Dar es Salaam has connections equilibrium. However, the supply response to an increase to all inland regions and neighboring countries through in the price of cotton and most other agricultural a series of trunk and regional roads, and in some cases products is limited because of the many constraints railways and lake ferries, the state of the infrastructure affecting farmers in Tanzania. The impact on cassava is deficient. Most regions have very low densities of farmgate prices is smaller than for cotton. This is because roads, and unpaved roads are not exploitable during foreign markets play a very small role in cassava with the rainy season (Iimi et al., 2015). Thus, the degree of most of the production consumed locally, often as own geographical market segmentation is likely very high in consumption or informally traded. Processors in cassava Tanzania. Moreover, according to the National Sample are very competitive, so the results for the simulation Census of Agriculture 2007/2008, the prices of otherwise assuming perfect competition are similar to the baseline. homogenous goods differ by up to 50% between regions. For perishable goods the differences can be even higher, The price reductions for four imported food products are 400% in the case of milk. This confirms a high level of not fully transmitted to the domestic economy (second market segmentation within Tanzania, so that port part of Table 1). For wheat, where imports are concentrated modernization is likely to have little short-term significant among a few importers and local production is a very impact on regions distant from Dar es Salaam, including small fraction of domestic consumption, more than the neighboring land-locked countries. half of the price reduction is captured by the importers. However, in the extreme case of perfect competition, the The methodological framework presented above does local price of wheat could fall more than 16% following a not take into account the impact of changes in prices 5% reduction in border prices. The pass through to local caused by trade policy and trade facilitation on wages, consumers for dairy products, rice, and maize is less than for example in expanding sectors vis-à-vis contracting 100%, but nevertheless significant. In these sectors local sectors. Also, the impact on the wages of skilled workers production satisfies a large share of the local demand, may differ from that on the wages of unskilled workers. with imports being in most cases residual. In the three Unfortunately, data required to estimate the wage sectors there is already a healthy level of competition, so impact were not available for this study. As the poverty the results for simulations assuming perfect competition profile in next section will show, wages actually make up are not significantly different from the results for the only 16.2% of Tanzanian households total income, much status quo. lower in rural areas (8.9%) than in urban areas (47.7%). 91 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 6. Poverty profile of Tanzania national level, 65.4% of households’ budgets is allocated to food. This share is larger for rural households (69.6%) While poverty rates have fallen over the last 15 years, than for urban households (50.9%), because incomes are poverty remains high in Tanzania. Roughly 90% of the higher among urban households, who thus spend more population live on less than three US$3 a day at 2005 on other goods and services than on food. Maize accounts PPP (2008/2009 Tanzania National Panel Survey). Poverty for the largest share of household food consumption. is more common in rural than in urban areas: using the On average, maize represents 15.7% of Tanzania’s national poverty line, one third of the households in rural household expenditure (17.7% of rural expenditure and areas live in poverty and approximately 90% of Tanzania’s 8.6% of urban expenditure). Rice accounts for 4.8% of poor people live in rural areas.7 the budget, with slightly higher shares among urban The household survey provides detailed information households. Cassava accounts for 4.8% of expenditures on income and consumption patterns, which are used in rural areas and for only 1.1% in urban areas. in identifying the potential welfare effect of trade. For Rural households’ income comes mostly from own rural households, cash expenditures account for 59.2% consumption, while cash income makes up only 32.4% of of the total budget, while own consumption accounted income (of which sales of agricultural products account for the remaining 40.8% (Table 2). By contrast, 93.9% of for 16.9% points and wages 8.9% points) (Table 3). By urban households’ expenditures is cash spending. At the Table 2: Budget shares Tanzania Total Rural Urban Total consumption per capita 100.0 100.0 100.0 Expenditures 66.9 59.2 93.9 Food 32.3 28.8 44.8 Manufactures 15.2 14.7 16.8 Services 19.4 15.7 32.3 Other 0.0 0.0 0.0 Own-consumption 33.1 40.8 6.1 Own-consumption food 33.1 40.8 6.1 Own-consumption other 0.0 0.0 0.0 Total food consumption 65.4 69.6 50.9 Total crops 39.4 43.2 26.2 Maize 15.7 17.7 8.6 Rice 4.8 4.4 6.4 Livestock 5.9 6.1 5.1 Cassava 3.9 4.8 1.1 Cowpea 4.4 4.9 2.7 Yam 0.3 0.3 0.1 Wheat 1.0 1.0 1.2 Groundnut 1.5 1.8 0.4 Sweet potato 1.9 2.3 0.5 Milk 1.9 2.1 1.1 Source: Tanzania National Panel Survey (2008/2009). 7 The household data for our analysis come from the 2008/2009 Tanzania National Panel Survey. The dataset contains information on 3280 households. The sample is representative at the national level but not for each region. Thus, the study can only distinguish impacts between urban and rural areas but not among regions. 92 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES contrast, cash accounts for 78.4% of total income of urban 7. The poverty impact of the port households. Of that 78.4%, wage income accounts for improvement 47.7% age points and enterprise (mostly informal) This section presents a discussion of the poverty impacts income 20.2% age points. Maize represents 20.7% of rural of the simulation results from section 5. It distinguishes household income and 7.4% of urban household income. three different levels of livelihood. Households below the Rice (4.5%), livestock and milk (5.6 and 3.9%) and cassava poverty line in the National Panel Survey (NPS) are defined (6.4%) are also important sources of income in rural areas, as poor, households above the NPS poverty line but below but not so much in urban areas. While important as a source the US$3 a day (PPP) as vulnerable, and households above of export revenue, cash crops such as coffee, tea, cotton, the US$3 a day line as non-poor. The analysis is done using tobacco, and groundnuts do not on average generate the first order approximation analysis of Deaton (1989, a large share of income for Tanzanian households. This 1997), which implies that the impact of a price change is because in general, smallholder farmers prefer food can be approximated using income shares and budget crops and when they do produce cash crops, they do not shares as measures of exposure. specialize. The data also shows that except for cotton, cash crops for the export market are mostly produced by The welfare impacts of the price changes from the port farmers in households with incomes above the national project are reported in Table 4 for six important cash poverty line. and food crops in Tanzania. The Table reports average Table 3: Income shares Tanzania Total Rural Urban Total income per capita 100.0 100.0 100.0 Incomes 41.1 32.4 78.4 Food (agriculture) 15.1 16.9 7.2 Wage 16.2 8.9 47.7 Enterprises 8.0 5.2 20.2 Transfers 1.8 1.4 3.3 Own-consumption 58.9 67.6 21.6 Own-consumption food 58.9 67.6 21.6 Own-consumption other 0.0 0.0 0.0 Total food income and AC 74.0 84.5 28.8 Total crops 49.2 56.6 17.0 Maize 18.2 20.7 7.4 Rice 4.0 4.5 1.6 Livestock 5.0 5.6 2.4 Cassava 5.4 6.4 1.1 Cowpea 4.4 5.2 1.2 Yam 0.4 0.5 0.0 Wheat 0.8 1.0 0.0 Groundnut 2.6 3.0 1.2 Sweet potato 2.8 3.3 0.2 Cotton 1.4 1.7 0.2 Tobacco 0.7 0.9 0.0 Milk 3.5 3.9 1.6 Source: Tanzania National Panel Survey (2008). 93 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES results for the total population, the poor, the vulnerable, The results also point to the distributional impact of and the non-poor, for rural households and for those the port improvements. First, urban households benefit households that are producers of the crop for which the more than rural households do from cheaper imported price changes. The tables show, for each demographic food crops. This is expected, as urban households are group, changes in the monetary income, expenditure and mostly consumers of food while rural households are welfare attributed to each crop. both consumers and producers. Moreover, monetary expenditures on food are typically higher for urban The income effects of price changes as a result of port households, as an important share of the food consumed modernization vary by commodity. The largest change in rural areas is home produced, particularly among the in income is generated by the rise in the price of cotton, poorest households. While the study does not take into as total household income increases by 1.4% and the account explicitly the reduction in cost of containerized income of producer households by 17.1%. One reason imports, it is likely that this will benefit urban households is that raw cotton is only produced and not consumed more than rural ones, as urban households’ propensity directly by households, so affected households can only to consume imported manufactured goods is larger. The benefit from the price rise. By contrast, all income groups poor and the vulnerable benefit from the decrease in are on average net consumers of rice, maize, cassava, the prices of maize, rice, dairy and wheat and from the and wheat, so they will benefit from price reductions increase in the price of cotton, and are only hurt by the (cheaper imports) and will be hurt by price increases increase in the price of cassava. Overall, the poor and (more expensive exportable food crops). The opposite the vulnerable as a whole are likely to benefit from the is true for households that are net producers. While on improvements in the port. From the simulations it is hard average households are net consumers of milk, in rural to predict whether urban or rural poor will benefit most. areas households are net producers. However, given the high degree of market segmentation For most crops, shocks, and affected populations, the within Tanzania, it is likely that the poor and vulnerable welfare impacts of the simulations are less than 1% of households around Dar es Salaam will benefit the most. total household expenditures. The only exception is the Finally, it is not clear whether poor, vulnerable or non-poor impact on cotton producers. These results are expected, households would benefit more. However, the structure given the nature of the exercise considered here, and of Tanzania’s imports and the structure of consumption of they are comparable to other results in the literature on poor and non-poor households suggests that the reform the topic (Lederman and Porto, 2013). will favor non-poor or vulnerable households more than those with incomes below the national poverty line. Given the high degree of market segmentation within Tanzania, it is likely that the poor and vulnerable households around Dar es Salaam will benefit the most from port improvements. 94 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table 4: Welfare impact of port improvement (Baseline Scenario) Income Expenditure Welfare Income Expenditure Welfare Maize Wheat Total 0.82 6.31 0.23 Total 0.02 0.49 0.01 Poor 0.69 4.29 0.15 Poor 0.02 0.57 0.01 Vulnerable 1.28 10.13 0.37 Vulnerable 0.02 0.42 0.01 Non poor 0.80 8.20 0.31 Non poor 0.01 0.32 0.01 Rural 0.98 6.01 0.21 Rural 0.03 0.31 0.01 Rural poor 0.91 3.79 0.12 Rural poor 0.03 0.31 0.01 Rural vulnerable 1.10 9.78 0.37 Rural vulnerable 0.03 0.31 0.01 Rural non poor 1.02 8.11 0.30 Rural non poor 0.02 0.31 0.01 Producers 4.99 3.66 -0.06 Producers 4.25 0.37 -0.08 Rice Cotton Total 0.79 3.30 0.11 Total 1.40 0.00 0.13 Poor 0.66 4.00 0.15 Poor 1.65 0.00 0.16 Vulnerable 0.82 3.42 0.11 Vulnerable 1.33 0.00 0.12 Non poor 1.10 2.11 0.04 Non poor 0.23 0.00 0.02 Rural 0.93 2.45 0.07 Rural 1.71 0.00 0.16 Rural poor 0.83 2.90 0.09 Rural poor 1.89 0.00 0.18 Rural vulnerable 0.92 3.02 0.09 Rural vulnerable 1.66 0.00 0.16 Rural non poor 1.09 1.71 0.03 Rural non poor 0.45 0.00 0.04 Producers 11.45 1.07 -0.46 Producers 17.09 0.00 1.61 Cassava Dairy Total 0.02 1.01 -0.06 Total 0.49 0.59 0.00 Poor 0.01 0.55 -0.03 Poor 0.55 0.70 0.01 Vulnerable 0.03 1.14 -0.06 Vulnerable 0.47 0.57 0.00 Non poor 0.05 1.65 -0.09 Non poor 0.36 0.39 0.00 Rural 0.03 1.04 -0.06 Rural 0.54 0.48 0.00 Rural poor 0.01 0.49 -0.03 Rural poor 0.63 0.54 0.00 Rural vulnerable 0.04 1.21 -0.07 Rural vulnerable 0.58 0.52 0.00 Rural non poor 0.05 1.50 -0.08 Rural non poor 0.40 0.42 0.00 Producers 6.88 0.05 0.40 Producers 12.79 0.23 -0.48 Note: Figures denote percentage change. One argument of why the poor could benefit from the the fertilizer adoption rate among poor farmers is port reform and trade in general (World Bank, 2013) almost zero, as fertilizer is in general used by well off, is through improved access to cheaper inputs, tools, medium-size farmers producing cash crops for the export and materials. For instance, rural poor would benefit market. In the case of construction material, besides from cheaper fertilizers and urban poor from cheaper the protection of the local cement industry, imports of construction materials. However, these effects are likely clinker are also concentrated with one company (Maweni to be limited in Tanzania. Fertilizer is at the moment Limestone) accounting for 68.8% of the market. Most only imported, and in 2013/2014 the two largest of this imported construction material is used in urban companies (Yara and Premium Agro) accounted for 56% areas and therefore will not benefit much the households of the market (the five largest companies had 98.1 %). under the poverty line (92% are based in rural areas) It is unlikely that they would pass much of the savings but may benefit vulnerable urban households (incomes from lower fertilizer prices to farmers, as the recently- above the poverty line but less than US$3 a day). cancelled fertilizer subsidy program showed. Moreover, 95 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 8. Conclusions The study does not consider the effect of the port project on wages. In general, in the short run, increased trade This contribution assesses the poverty impact of reduces wages in import-competing sectors and increases proposed improvements to the Dar es Salaam Port. wages in export sectors. The port improvements are Currently, the port suffers from several infrastructure likely to increase imports but have little effect on exports and operational deficiencies that increase the cost both in the short run. In the long run, cheaper manufactured for imported and exported goods. These costs partially imports may negatively affect the development of some reflect port congestion, where the situation is more labor-intensive manufactures, and this sector can have critical for imports than for exports. These inefficiencies a positive impact on poverty reduction. However, more act as an implicit tax on imports and to a lesser extent as a work is needed to understand this effect, especially tax on exports. An improvement in the operations of the given the very fast process of urbanization taking place port could reduce border prices up to 5% for bulk cargo. in Tanzania. Besides the concrete impact on poverty The simulations of the short-term effect on selected bulk and inequality, it also would be important to estimate goods show that the pass through from border prices the efficiency gains from the port improvements, and to retail and farm gate prices will be less than a 100%, develop metrics that allow us to quantify any possible limiting the potential impact of cheaper imports. This is trade-off between efficiency gains and inequality or due partially to importers’ market power, which enables poverty. them to capture some of the cost saving. The effect of the Finally, beyond the concrete findings of this study, port improvement would be geographically concentrated it is important to note that international trade has in Dar es Salaam, as inadequate roads and railways result undoubtedly greatly contributed to the growth in a high level of market segmentation. performance of Tanzania in the last ten years. While this The short-term impact on poverty of the port high growth has not translated in drastic reductions in improvements is positive, albeit small. However, the poverty for many reasons (Atkinson and Lugo, 2010), reduction in poverty may be accompanied by an increase trade has contributed to the steady reduction in the in inequality, as non-poor households are expected poverty headcount and the poverty gap. The impressive to benefit proportionally more than the poor and growth in trade and the inadequate investments to keep vulnerable ones. In the long run, a better functioning up with this growth are why the port is currently suffering port could more significantly reduce rural poverty and from severe congestion. Failure to significantly improve raise agricultural productivity if complementary policies port operations could severely constrain Tanzania’s and infrastructure improvements (particularly for inland exports and imports, jeopardizing one of the engines of transport) were adopted to improve access of the poor growth and poverty reduction of the country. to export opportunities and to cheaper imported food, tools and inputs. 96 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Artuc, E., D. Lederman, and G. Porto (2015). “A Mapping of Labor Mobility Costs in the Developing World,” Journal of International Economics 95(1), pp. 28-41. Artuc, E., and G. Porto (2016). Global Labor Market Frictions and The Dynamics of the Gains from Trade. Work in Progress. Atkinson, A., and M. Lugo (2010). Growth, poverty and distribution in Tanzania. Working Paper 10/0831. International Growth Centre. Barnum, H., and L. Squire (1979). “A model of an Agricultural Household. Theory and Evidence,” World Bank Occasional Papers No 27. Benjamin, D. (1992). “Household Composition, Labor Markets, and Labor Demand: Testing for Separation in Agricultural Household Models,” Econometrica 60, No.2, pp. 287–322. de Janvry, A., M. Fafchamps, and E. Sadoulet (1991). “Peasant Household Behaviour with Missing Markets: Some Paradoxes Explained,” Economic Journal 101, pp. 1400–1417. Deaton, A. (1989). Rice Prices and Income Distribution in Thailand: a Non-parametric Analysis. Economic Journal 99 (Supplement), pp. 1-37. Deaton, A. (1997). The Analysis of Household Surveys - A Microeconometric Approach to Development Policy. Baltimore: Johns Hopkins Press. Dix-Carneiro, R. (2014). “Trade Liberalization and Labor Market Dynamics,” Econometrics. Depetris-Chauvin, N., G. Porto, and F. Mulangu (2017). Agricultural Supply Chains, Growth and Poverty in Sub-Saharan Africa: Market Structure, Farm Constraints and Grass-root Institutions. Springer. Food and Agriculture Organization of the United Nations (2015). FAOSTAT statistics database. Horn, H. and J. Levinsohn (2001). “Merger Policies and Trade Liberalisation,” Economic Journal 111, pp. 244–277. Iimi, A., R. Humphreys and S. Melibaeva (2015) Firms? Locational Choice and Infrastructure Development in Tanzania: Instrumental Variable Spatial Autoregressive Model. World Bank Policy Research Working Paper No. 7305. KPMG/World Bank (2013): “Pulse of the Tanzanian Economy.” Survey. Lederman, D. and G. Porto (2014). The Price is Not Always Right: On the Impacts of (Commodity) Prices on Households (and Countries). World Bank Policy Research Working Paper No. 6858. Morisset, J. (2013). Tanzania economic update: opening the gates : how the port of Dar es Salaam can transform Tanzania. Tanzania economic update; issue no. 3. Washington DC ; World Bank. Porto, G., N. Depetris-Chauvin, and M. Olarreaga (2011). Supply Chains in Export Agriculture, Competition, and Poverty in Sub-Saharan Africa. The World Bank and Centre for Economic Policy Research. Singh, I., L. Squire, and J. Strauss, eds. (1986). Agricultural Household Models: Extensions, Applications and Policy. Baltimore, Johns Hopkins Press for the World Bank. Swinnen, J., and A. Vandeplas (2010). “Market Power and Rents in Global Supply Chains,” Agricultural Economics 41, pp. 109–120. Swinnen, J., and A. Vandeplas (2014). “Price Transmission and Market Power in Modern Agricultural Value Chains,” LICOS Discussion Paper #347/2014. Tanzania (2012). Population and Housing Census of 2012. National Bureau of Statistics. Taylor, E. and H. Adelman (2003). “Agricultural Household Models: Genesis, Evolution and Extensions,” Review of Economics of the Household 1(1). World Bank (2013). Tanzania Economic Update: Opening the Gates - How the Port of Dar es Salaam Can Transform. Tanzania economic update, Issue no. 3. Washington DC. World Bank. World Bank and World Trade Organization (2015). “The Role of Trade in Ending Poverty”. Washington DC: World Bank. 97 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Agricultural Logistics in Lagging Regions: Evidence from Uganda Charles Kunaka, Macroeconomics, Trade and Investment Global Practice, World Bank 1. Introduction foods, and regional cooperation in trade and transport S infrastructure development. mall scale farmers face many hurdles when attempting to connect to global markets. The economies of lagging regions are typically dominated While infrastructure and trade facilitation by trade in agricultural products, which have a high improvements are helping to reduce overall elasticity to logistics costs. Rural connectivity is defined trade costs, the challenges faced by such farmers are as the “degree to which rural dwellers can take advantage most acute at the local level. Efforts to eradicate poverty of their access to infrastructure, services, and markets therefore need to start with constraints at the farmgate. to pursue economic opportunities” (World Bank, 2014). The problems faced are compounded by the general It is important to identify options for governments and lack access to proper agricultural inputs, technology and the private sector to contribute to reductions in trade intermediate services. costs, by implementing interventions to broaden access to logistics services at the sub-national or regional level, However, in the research on market access there has not in order to expand the participation of remote regions in been much attention paid to the needs of poor areas domestic and global supply chains. within countries and the constraints faced by small producers in isolated regions to reach domestic and Stifel and Minten (2008) claim that the positive international markets. A few studies, such as Raballand et correlation between poverty and isolation is driven by al. (2010) and Kunaka (2010), applied micro-level analysis high transportation costs. In turn, high costs have typically by looking at the relationship between road infrastructure been linked either to a lack of passable roads and/or lack and road transport and the access of small producers of transportation services. Casaburi et al. (2013) find that to markets. The present paper aims to contribute to improvements to rural roads reduce the market prices of this growing knowledge base by exploring measures to local crops; the effects are stronger in markets that are lower costs to market for producers in areas that can further from major urban centers and in less productive be characterized as lagging.1 It underscores policies and areas. In addition, Gollin and Rogerson (2010) look at factors that favor the development of logistic services the link between transport costs and the size of the and enable small scale producers in remote regions to quasi-subsistence sector in Uganda and conclude that improve their access to markets in support of strategies the high dispersion of prices across geographic space for inclusive growth, cross-border regional trade in staple reflects underlying transportation costs that prevent any 1 The approach should be applicable generally to logistics analysis at a sub-national scale in both leading and lagging regions. 98 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Small scale farmers face many hurdles when attempting to connect to global markets. While infrastructure and trade facilitation improvements are helping to reduce overall trade costs, the challenges faced by such farmers are most acute at the local level. 99 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES arbitrage between regions. Also in Uganda, Raballand et Earlier work by Kunaka (2010), Raballand et al. (2010) and al. (2009) find that after a certain threshold of access to Farole (2013) point to the importance of consolidating rural road infrastructure is met, the effects of additional volumes to reducing unit logistics costs in lagging regions, investments are not only empirically unclear, but also which would extend the distances farmers are able to probably economically unsustainable considering the low trade. Typically, costs are highest at the local level (first volumes generated by smallholder farming production mile), as that is where volumes are smallest and costs systems. Moreover, in order to have affordable road per unit highest. As such, supply chains in thin markets freight services—where competition is achieved— tend to have a cascading pattern of consolidation and consolidation and the use of ICT are probably needed. value addition to reduce the ratio of costs to volume and cost to value. Consequently, products pass through Proximity to markets is seemingly a key factor in farming several logistics nodes, which can add to costs and transit outcomes. For instance, agricultural production and times, as well as affect overall system reliability. Analysis proximity to urban markets in Sub-Saharan Africa of logistics services therefore should consider: (i) the are highly correlated (Dorosh, Wang and You, 2008). location and availability of core logistics infrastructure Moreover, about 60% of total crop production in Sub- such as transport, handling and storage facilities that Saharan Africa, but only about 40% of the population, influence the direction and concentration of trade flows; is between 2 and 9 hours away from a market. Also, the (ii) the availability and cost of services linking different adoption of high-input technology, such as modern seeds points of production to domestic, regional and global and fertilisers, is negatively correlated with travel time to markets; and (iii) the interactions between logistics urban markets. players who intermediate the physical movement of Minten et al. (2013) find that in Ethiopia transaction and products. transportation costs together add between 20% (for Based on the evidence, it can be argued that the the least remote farmers) and more than 50% (for the performance of logistics systems in lagging regions is a most remote farmers) to the fertilizer prices charged at function of the interaction of different systems, rooted the input distribution center. The authors also suggest in various types of networks: physical, social, financial, that farmers who live about 10 km from the distribution and information, among others. However, there is center face per unit transaction and transportation costs great coincidence in the social, financial and information as large as the costs needed to bring the fertilizer from networks such that they can be conveniently combined the international port to the input distribution center into one sub-system of supply chain organization. So, (about 1,000 km). These effects might partly explain the ultimately, we can view the system as comprised of two low modern input adoption rates in parts of Ethiopia. main models: one that represents the physical movement Raballand et al. (2009) find that transport of small loads by of goods through physical networks defined by physical truck in Uganda is 10 times more expensive per kilometer transport and logistics infrastructure and a second one than movement by bicycle or motorcycle. The typical that represents the relationships between shippers, small-scale farmer in the study produces on average logistics actors, processors and exporters of goods. This between 40 kg to 3 tons of agricultural goods per year paper aims to explore the interactions between the two and thus are able to transport their product mainly by and propose measures that may be taken to reduce the bicycle or motorcycle. Transport by truck only becomes isolation from markets of farmers in lagging regions. profitable for loads of at least 500 kg of product per trip that are transported no fewer than 50 km (DFID, 2005). 2. Organization of agricultural At the current production levels, a consolidation center supply chains in thin markets (warehouse) would be needed to collect the production To properly account for the distribution of rents in of at least 600 farmers. agricultural supply chains in lagging regions, it is important to identify the main actors in the chains (Figure 100 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 1: Generalized structure of agricultural supply chains Processor 3rd Party Distribution Seed Equipment Smallholding Processor Manufacturer Agro-chemical Infrastructure Medium, Large Farm Exporter Retailer Irrigation Technical support Plantation Manufacturer Food Distribution Transport Collection Consolidation Storage Transport Transport Storage Storage Supplier Grower Processor Distributor Procure Procure input Procure s inputs Sorting, Packing Market,Prepare Market, Prepar e Prepare Plant, Cultivate, Plant, Cultivate,Harvest Harvest Milling, Polishing Distribute Distributeto retail to retail Store Post-harvest Post-harvestprocessing processing Heating, Mixing Outlet Outlet Source: Kunaka, Saslavsky and Watanuki, 2015. 1). A distinguishing characteristic of agricultural supply The early steps in the process of selling produce, chains in lagging regions is that the products are handled which consist of a combination of on-farm and multiple times between point of production and the local transport, tend to be the more costly and market. This poses several challenges, including lengthy difficult part of the journey, often conducted on time to market which leads to product losses; mixing of infrastructure with no clear ownership structure.2 the produce of different farmers, which compromises Rural freight transport ends for the farmer at the quality and reduces opportunities to sell to more buying point near the road side or at rural hubs, selective and premium markets; and inability to achieve both serving as nodes for the intermodal transfer economies of scale, which leads to high costs. Tracing of cargos. Rural hubs may have a special function, the path of movement of product must start at the farm- crucial for the development of the whole area. gate and include the primary stages of marketing as well • Traders: In areas with small volumes of production as midstream segments of the supply chains. the main choice available to producers in thin The main categories of actors along the agricultural markets is to sell through traders. Fafchamps and supply chains in lagging regions are: Hill (2005) find that in Uganda the likelihood of selling to the market increases with quantity sold • Growers: They range in size from smallholders and proximity to the market. They also find that to plantations. Each has different linkages to the number of itinerant traders who tour the the supply chain. The smallholders deliver their countryside in search of coffee rises when the price crops to collection points where the crops are increases. Their purchase prices do not increase aggregated prior to processing. The medium-to- proportionately to their sale prices to established large farms deliver directly to the processor, while trader-millers, suggesting that they take advantage the largest farms and plantations often process of farmer’s ignorance about price movements. their own crops. 2 Much of the infrastructure tends to be rudimentary, consisting of pedestrian paths or tracks used by animal drawn carts. 101 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES • Processors: They add value to the crop supplied by as long a period as possible to reduce peak demand for the grower. There is usually a series of processes processing capacity. On the other hand, if the crop has a involved in converting the crop into final products relatively long shelf life and the inputs for the processor for sale to the consumer. These can be categorized can be supplied from inventory, then there is less demand according to the extent of processing, as shown in for integration between the grower and processor. Table 1. Two ways that market integration can be enhanced, with • Distributors and Retailers: The structure of the important implications for farmer outcomes, is through food distribution chain is increasingly determined by cooperatives and contract farming arrangements. modern food retailing in the form of supermarkets and large retail chains for food and food services. 2.1 Cooperatives and producer associations They establish a demand-driven relationship with Cooperatives have a long history in the farming sector. the distributors, and often also the processors. They typically play an important role in how farmers Supermarket chains have dominated food retailing are organized to pursue common interests. Generally, in developed countries for about 50 years but only cooperatives have twin objectives, to minimize transaction began to appear in developing countries in the costs and increase value for members by managing some 1990s, largely as a result of foreign investment of the downstream processes such as packaging, storage from Europe, Japan and the United States, and in and marketing. When producers own and control the Africa from South Africa. In general, their influence chain, they are in a better position to maximize returns in Uganda is still nascent though Kenyan and South for farmers and to avoid being restricted to their closest African supermarket chains have been making markets. inroads. While supermarkets in Kenya account for about 20% of urban food sales, this percentage is These organizations offer several benefits to estimated at approximately 1% for Uganda (Evers smallholders, including: et al., 2014; Neven and Reardon, 2006a; Tschirley, 2010). • Obtaining inputs, services and/or assisting in marketing outputs by offering significant scale The level of integration between the grower and economies; processor in the food supply chain depends on physical • Helping farmers manage post-harvest processing, characteristics of the crop, the type of processing and the by achieving scale economies in the required market for the goods produced. The two utilize different capital investment; logistics service providers who may provide transport, storage and other services. If the crop has a short shelf • Assisting farmers to access finance and applying life, then the supply chain must be tightly integrated peer pressure as a mechanism for enforcement of and the harvesting of the crop must be spread out over financial obligations; and Table 1: Food processing activities Package Preserve Transform Chilled Frozen Freeze dried Washed Pressed Distilled Separate, Sort Pasteurized Sterilized/Cultured Dried Heated/Smoked Roasted Gutted, Scaled Sliced/Ground Cooked Pack Bottled Canned, Sealed Hulled/Polished Source: Kunaka, Saslavsky and Watanuki, 2015. 102 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES • Facilitating logistics for distribution of inputs and collection of outputs. 2.2 Contract farming Contract farming arrangements for smallholders have attracted considerable interest in the last few years throughout Africa, in part due to concerns that farmers are not capturing enough value from exports or domestic sales (Hazell et al.. 2007). Contract farming is often seen as an alternative means of tackling suboptimal investment and other market failures, since this mechanism fosters economies of scale, thus reducing transaction costs (Dorward et al., 2004). Some authors find robust evidence directly linking contract farming with increased farmer income in developing countries (Kirsten and Sartorius, 2002). Nonetheless, others doubt that contract farming schemes generate sustainable income benefits allowing foreign investment in the sector. Improvements for participants or improve rural inequality (Kirsten and in infrastructure are important for the efficiency and Sartorius, 2002; Raynolds, 2002). competitiveness of the agricultural sector. Agricultural Despite the seemingly positive relationship between research is necessary to support the introduction of contract farming participation and crop income, new varieties and types of crops. All of these measures, contractual problems have appeared between combined with dissemination of information through outgrowers 3 and large agribusiness firms, resulting in improved communications infrastructure, would help market exit of smallholders. The contractual problems growers to become more competitive. are probably often magnified by inadequate quality and These policy improvements would benefit the enforcement of legal frameworks for contracts (Wiegratz agricultural sector as a whole, not just contract farming. et al., 2007). Specific areas in which government can benefit contract The type of contract used depends on the type or variety farming include support for outgrower schemes, of crop, the method of cultivation, and the availability of similar arrangements for supporting the organization appropriate inputs including seed, agrochemicals, credit of smallholdings, and strengthening financial and legal and technical assistance. regulations for contracts. The latter should be relatively light since the strength of contract farming lies in the use Policies towards agricultural development, trade of self-enforcing agreements. and competitiveness, infrastructure and research, and macroeconomics have a significant impact on Agricultural supply chains in lagging regions therefore the potential for contract farming. Government can involve several actors. Their level of integration, either support contract farming by encouraging direct linkages forced by market requirements, through voluntary between farmers and the organized food supply chain association or contractual relationships can have processors. Government also can create a favorable significant impacts on economic outcomes for farmers. environment for the development of food supply chains The remainder of this paper explores these issues in the through liberalization of trade in agricultural products, specific case of Uganda and runs some experiments to restricting government involvement in this trade and test outcomes under different scenarios. 3 In this context an outgrower is farmer who is in a contractual partnership with a larger farmer or company to produce specific products. 103 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 3. Case Study: Trade in fresh Due to its tropical location, Uganda is endowed with a produce in Uganda wide variety of fruits, which account for a large part of the country’s non-traditional agricultural exports. The Agriculture is the most important sector of the Ugandan major fruits produced in Uganda include passion fruit, economy, contributing up to 26% of GDP and over papaya, jackfruit, citrus, pineapple, mango, avocado, 70% of exports.4 Downstream industries are especially apple, banana, cavendish, watermelon, guava, grape, dependent on agriculture: food processing alone strawberry, melon and tree. There is a lot of potential accounts for up to 40% of total manufacturing. While the to grow exports in these products and to diversify the economy of Uganda has been growing strongly over the export basket. past two decades, at an average of over 7% p.a., growth has been uneven. Partly driven by the uneven reduction Between 2001 and 2011, production of fresh vegetables in poverty and persistent inequality, the authorities in Uganda almost doubled, from 585,000 tons to almost have introduced new programs, including the “Rural one million tons. Meanwhile, exports multiplied by a Development Strategy,” and “Prosperity for All”5 with the factor of 40 in volume, but only by 13 times in terms of aim of raising rural incomes and reducing the income gap. value. On the other hand, production and exports of fresh fruits remained stable between 2001 and 2011 According to Bolwig (2012), the Ugandan agricultural in both weight and value. In both cases the percentage market structure is divided into three major groups: of production exported is negligible, close to or smaller subsistence and small scale, medium, and large than 1%. The most notable change in trade patterns is in farmers. Agricultural production is dominated by small- the geographical composition of exports: while African scale farmers, consisting of approximately 2.5 million markets represented less than 1% of both fruits and households (or 90% of the farming community), the vegetables exports in 2001, by 2011 they represented majority of which owns less than 2 acres of land each. between 65 and 75% of total external sales. 2001-2011 The major fruits produced in Uganda include passion fruit, papaya, jackfruit, citrus, pineapple, mango, avocado, apple, banana, cavendish, fresh vegetable production watermelon, guava, grape, strawberry and melon. in Uganda almost doubled 585,000 tons to almost 2.5M Agricultural production is dominated by small-scale farmers, consisting of 1 million tons approximately 2.5 million households There is a growing volume of trade in fresh produce between Uganda and neighboring countries, especially South Sudan. 4 From World Bank World Development Indicators. 5 The “Prosperity for All” program started as a post-2006 election plan to implement the new government’s manifesto. The program aspires to see that every household earns sustainable income to transform itself into an economically viable entity. With particular emphasis on agriculture, the plan envisages raising agricultural productivity through zoning, providing start-up capital and inputs, supporting mechanization and value addition (agro-processing). It has run parallel to the PEAP, which is GOU’s PRSP. 104 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Transport of Ugandan produce between regional and the market for sale, even though doing so would reduce global markets differs: regional markets are reached everyone’s costs. However, fieldwork for this paper found almost exclusively by road, while overseas markets are that this has become quite common, facilitated through connected by a combination of land and air transport the use of mobile phones. services. Government interventions are necessary to establish 3.1 Regional markets logistics infrastructure in rural areas. The core infrastructure has traditionally been roads and some There is a growing volume of trade in fresh produce storage facilities. While Uganda has a dense road between Uganda and neighboring countries, especially network and high traffic volume compared with its East South Sudan. The volume of traffic between the two African neighbours, the district roads are often in poor countries has increased rapidly, particularly since South condition, making for low accessibility to rural areas Sudan gained independence in 2011. There are no (Ranganathan and Foster, 2012). As a result, transport established suppliers for food (maize, pulses, sugar, oil or costs in Uganda at all levels are very high in relation meals ready to eat) produced in South Sudan. Nearly all to neighboring economies. The high transport costs the states experience cereal deficits, which are balanced impair the development of the agricultural sector in the through imports, mostly from across the border with country (Natamba et al., 2013; Gollin and Rogerson, 2010; Uganda, DRC and Ethiopia, but also from overseas. The Collinson et al., 2005). The burden of transportation costs thriving trade in food products in the border region on export prices erodes the returns to capital, reducing of Uganda and South Sudan is driven by the significant investment and hindering economic growth. differences in the prevailing prices in South Sudan and northern Uganda (Table 2). To reduce such costs, the government of Uganda has invested significantly in road construction and upgrading Traders purchase produce in Uganda at the farmgate or over the last decade. Nonetheless, Uganda’s RAI (rural in market towns. A typical shipment starts at the farm access index6) is still low at only 53% compared to other level, is consolidated or is transported to a buying point countries such as Bangladesh which has an index of 86% or to local and regional markets. Consolidation is the only (Iimi et al., 2016). Given that Ugandan agriculture is mainly way to develop synergy with higher volume-lower cost composed of low density subsistence small farming, transport services to the nearest export market. Though reaching an RAI of 100% might not be sustainable there is some empirical evidence of this (e.g., Raballand, considering the considerable investment that would be 2010), how such consolidation takes place is often not required (Raballand et al., 2009). Carruthers et al. (2008) fully explored. It cannot be assumed that cooperation to estimate that the Ugandan government would need to trade will necessarily take place on its own accord (Gibbon, invest 3.6% of GDP for a period of 10 years to reach an 2001). Fafchamps and Hill (2005) find that farmers in RAI level of 75%. Given the unavailability of resources Uganda used not to trust each other sufficiently to for that magnitude of investment, other measures are enable one of them to carry all neighborhood produce to needed to reduce overall logistics costs. The design of Table 2: Maize and beans retail and wholesale prices, April 2010 Maize Beans Logistic Centers Retail price Wholesale price Retail price Wholesale price Juba—South(ern) Sudan 17,12.5 1,182.5 2,700 2141 Gulu—Northern Uganda 342.5 400 1,350 1,150 Source: Ephrem Asebe, 2012. 6 The rural access index measures the “share of the population who live within 2 km of the nearest road in good condition in rural areas” (Iimi et al (2016). 105 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES such measures should take into account how farmers farmers then individual costs can be high, as shipment actually reach domestic and regional markets. volumes will be small resulting in high unit costs. It is in this context that the World Bank (2012) found ICT effective In addition to roads, there is an important nexus in reducing transport and logistics costs with just-in-time between transport and storage infrastructure which is information on consolidation and market trends. often ignored in rural projects. The location of storage facilities, which serve also as consolidation points, is Ready access to information has three primary effects on critical in enabling small scale producers to connect to producers: regional and global supply chains. As argued above, • It reduces information asymmetry and enhances a certain level of infrastructure development is often farmers’ ability to negotiate higher prices, even with necessary to provide the foundation for improvements market intermediaries. in the quality of logistics services. When storage is not available, produce must be harvested a few hours before • Farmers can decide to delay selling their produce it is taken to market, which limits the distance the produce until market conditions improve, although storage can be transported and its shelf life. Storage facilities in costs can be high if the right facilities are not Uganda are typically provided by the private sector, but in available. a patchy manner. • Farmers can switch to alternative markets, if they Grabowksi (1999) emphasizes that without an external have the necessary transport and are able to change coordination mechanism to bring buyers and sellers their destination market. together, farmers may lack the assurance required to undertake production while customers may be unable to In addition to facilitating information flows, ICT has find suppliers. In this regard, technological innovations contributed to agricultural efficiency through the are rapidly transforming marketing practices across East proliferation of payment systems based on mobile Africa, including in Uganda. Electronic intermediation phones. There are a large number of developing and through ubiquitous coverage by mobile telephony in developed countries that now have such systems in place. particular has disrupted rural marketing and led to Kunaka (2010) finds also that ICT helps manage payment virtual integration of supply chains. One of the impacts processes in lagging regions by enabling all players to has been though improved price discovery. Farmers be prepared for market transactions. Otherwise some and traders often have imperfect or no knowledge of studies (e.g., Gong et al., 2007) find that farmers who sell prevailing market conditions, which may result in them directly to market without advance information can suffer delivering produce when conditions are not conducive to payment delays unless they sell to an intermediary but at high returns. As the products are often bulky and of low a lower price. East Africa has several successful mobile value the farmers can suffer significant losses. Mobile payment systems, which emulate the first successful7 phones enable farmers to check prices and general branchless banking platform, M-Pesa, introduced in market conditions before deciding when to sell their Kenya in 2007.8 These systems make it easier and faster produce thereby reducing their exposure to risk of losses. for farmers to receive payment without having to travel long distances. However, while the use of IT in agro-markets is widespread in Uganda, the impact on logistics costs is not clear. Each Connectivity to regional markets by small scale farmers farmer is put in a position where they can decide which in Uganda utilizes elements of all of the above. Farmers market to access or when to take their product to the either sell produce to traders who take products into market. If this is done without coordinating with other neighboring markets or they get together and organize a 7 Morawczynski and Pickens (2009) says that in August 2009 M-Pesa had 7 million customers transferring 150 million KSh (US$2 million) a day in small transactions averaging about 1,500 KSh (US$20). The system had handled over 130 billion KSh (US$1.7 billion) since 2007. 8 It was initially launched in 2003 to allow borrowers to re-pay loans with airtime using the network of Safaricom airtime resellers. Mobile phone customers put it to various other users, and it was re-launched as a branchless bank service with limits on transactions and credit holdings. The second launch was supported by DFID and Vodaphone (40% shareholder of Safaricom) and the telecom consultancy firm Sagentia. 106 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES truck to ferry fresh produce of several farmers to markets in neighboring countries. Only a few farmers travel, while In 2013, fresh the rest keep track of sales and prices through mobile telephony. The regional markets, especially South produce was the Sudan and the eastern part of the Democratic Republic of Congo, hold great potential for increased trade from third largest export Uganda. They also have the advantage, compared to markets in the Middle East or Europe, in that they do not by air in Uganda (22% impose strict standards on product quality and handling. As argued below, strict product standards either increase of export tonnage), costs for farmers or can exclude small scale farmers completely from some markets should they be unable to after fish (46%) and comply with the set standards. flowers(26%). 3.2 Overseas markets Due to its landlocked nature and long distances to overseas markets, air transport is important for global Europe among them. In fact, Europe is the largest connectivity of Uganda producers. In fact, the fresh market for Uganda’s exports by air. About 19% of the produce sector in Uganda is concentrated within 40 km cargo is transported to African airports, where Southern of Entebbe Airport. Proximity to the airport is quite Africa captures a larger share than does the East African important to defining a zone of production of perishables Community (EAC). Within East Africa, some freight can like flowers, fresh fruits and vegetables, which tend to be transported by road instead of air transport. have a short shelf life. In 2013, fresh produce was the The distribution of Uganda produce in the overseas third largest export by air in Uganda (22% of export markets, particularly in the Middle East, is usually through tonnage), after fish (46%) and flowers (26%). However, supermarkets. As mentioned above, supermarkets the volume of production is generally not sufficient to increasingly play an oversized role in the marketing of support the regular supply of airlift capacity. In general, produce in both the East Africa region and internationally. for exports, cargo is nearly equally distributed between However, the supermarkets have strict requirements in passenger and freighter flights. Still, the belly freight terms of packaging, quality and other standards as well is less than the worldwide average of 70%. To support as the traceability of products. Compliance with these the market some private cargo handlers have developed requirements often has to start at the farms in countries facilities at the airport for the temporary storage of the like Uganda, through some central processing facilities produce of many farmers and to enable the airlines to where the produce is washed, weighed, bagged and then cater to this market. The airlines provide a cut-off tagged. As a result, the ability of producers in Uganda time of a few hours before a flight for all produce to be to meet the pressures from supermarkets and to comply delivered and processed ready for export. This enables with global standards is influenced in part by, and in turn farmers to plan the ideal moments to harvest produce is affecting, the topology of fresh fruit and vegetable and deliver it while it is still fresh. supply chains. Whereas in the past the produce of Most of the flights carrying freight out of Entebbe airport different farmers might have been mixed in the same are destined for the Middle East or Europe. Nearly 75% packaging, there is increasing pressure to keep separate of air exports leave on flights to these two regions. A the produce from different farmers so that it can be easily significant proportion of the cargo on flights to the identified. This can work contrary to the requirements to Middle East is transferred to flights to other destinations, reduce unit costs of accessing some of the markets. 107 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 3.3 Typology of supply chains to regional or own refrigerated trucks. Some of the large farmers global markets work with outgrowers who possess 1–2 ha each. In turn, each outgrower can have a relationship with Consistent with the framework in Figure 1, four main another 100–200 smallholder farmers. types of fresh produce export supply chains have emerged in Uganda, which determine whether the exports are 2. Farmer associations. Most of Uganda’s fresh fruits destined for the regional or overseas markets: and vegetables are produced by smallholders, mainly for self-consumption or local markets. There 1. Large farmer-led supply chains. There is only a small are 2–3 million smallholder producers of fresh fruits number of large fresh produce farmers (at least 30 and vegetables, but only a few hundred supply hectares of land and 300–350 workers) in Uganda. produce to regional supermarkets through producer These farmers package and label their own brand marketing organizations (PMOs), cooperatives or products, control their own cool chain, and face little outgrower schemes. They own on average less difficulty in meeting the standards or quantities than 2 acres of land and are spread in the Southern, required by Uganda-based regional supermarkets Central and Eastern regions of the country. Producer or supermarkets in neighboring countries (Kenya, associations such as Hortexa,9 cooperatives, and Rwanda, Burundi). Only a few possess GlobalGAP marketing organizations, which have about 20–30 certification and export to overseas markets. They farmers each, consolidate volumes and export them. can transport produce to supermarkets using their Typically, production relies on family and unskilled Table 3: Summary of characteristics of supply chains Contract farming Transaction Large farmer Association Trader 4a 4b Intermediaries Farm gate to Farmers bring to brings to collection PMO collects from PMO collects from collection centers Process at farm collection center centers farm gate farm gate Sort and grade at The team leader Exporter sort/grade own processing inspects and at his collection Done by PMO at Done by PMO at Sorting/grading location receives center collection center collection center Sort and grade at The team leader Exporter sort/grade own processing inspects and at his collection Done by PMO at Done by PMO at Packaging location receives center collection center collection center Agent accompanies Transport to airport/ goods to regional Transport to airport regional market Transport to airport markets or regional market Done by PMO Done by PMO FOB/CIF CIF/FOB CIF CIF CIF FOB Supermarket/ Supermarket/ Supermarket/ Final market Wholesaler Wholesaler/retailer Wholesaler/retailer Wholesaler Wholesaler Traceability Risks No/Yes No Yes Yes Post harvest losses Yes Significant Significant Average Average Timely delivery of cargo to destination Normal Significant Significant Average Average Subject to change Access to reliable regional market, High risks due to Subject to changes Subject to changes market Average politics lack of traceability in the world market in the world market Transport price fluctuation Low risk Very high High High High Source: World Bank. 9 HORTEXA is a legal entity that brings together horticultural producers and exporters. It was formed in 1990 to help exporters deal with their common problems collectively. In 1998 it was decided that farmers could also be members. Among other activities, the organization works to promote awareness of the high quality and availability of Ugandan fresh produce, with special focus on international markets. It had played a major role in helping growers upgrade their production and postharvest handling methods to meet international export requirements. 108 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES labor, with limited access to specialized advisory structured surveys and modeling of supply chain actor services. Production is also characterized by low use interactions. The modeling of the relationships between of pest and disease control, planting materials or soil supply chain actors employed a relatively new technique, fertility enhancement measures. Most companies or called agent based modeling (ABM). ABM techniques are producer associations have very limited investment a class of computational models and simulations based on capacity to increase production or explore market a large number of acting and interacting agents (Axelrod opportunities. Only a very small number sells directly and Tesfatsion, 2005; Tesfatsion and Judd, 2006). ABM to European wholesalers (Evers et al., 2014). models are robust and less demanding than econometric models with regard to availability of aggregate data, 3. Traders. It is common for small scale farmers to sell making them especially attractive for policy analysis to traders who then consolidate and export product. in studies of developing countries. ABM models also Itinerant traders purchase product from farmers and differ from other conventional simulation models that either export it or act as local agents for regional are grounded in mathematical programming, because buyers. Generally, farmers who have large quantities they can capture explicitly the interactions between are more likely to take their produce to the market, actors (farmers and intermediaries, for example), thus often using their own transport, while those who sell allowing for the study of transaction and information small quantities are likely to sell at the farmgate, to costs. These models can fully account for the spatial traders or agents. dimension in agricultural activities, and thus the role of 4. Contract firms. These are firms that may or may internal transport costs and the physical features of land. not grow any product but instead consolidate Consequently, the explanatory power of the models is product from farmers, carry out quality control and suited to policy questions of interest. packaging, and sell to domestic markets, regional The outcome of a simulation using ABM is largely markets (South Sudan, Kenya, DRC, Rwanda, and determined by the agents at the micro level, who are Tanzania), and international markets. Such firms acting between themselves and the environment in source product when there is demand, consolidate which they are set. Hence, ABM provides a way to and market it, typically to national or international study the behavior of agents as a result of a set of wholesalers and supermarkets. different scenarios, and then to test the impact of policy changes. Compared to conventional modeling, ABM 4. Data collection and analysis makes it much easier to model heterogeneous and not Given these characteristics of the fresh produce sector in fully rational agents, provides more flexibility to model Uganda, we carried out experiments to assess the role of farmers’ decision-making processes, and facilitates the different dimensions of the logistics systems for access inclusion of non-economic factors, which leads to a to market. The experiments were conducted to test four better representation of the regional-spatial variations in policy issues: agriculture. ABM can capture not only individual decision making, but also negotiation processes between supply • What is the impact of further improving transport infrastructure? chain actors, and incorporates time and geographic information system (GIS) elements for geographical • Where should storage facilities be provided? placement. The agents’ interactions result in a macro • What is the impact of associations on logistics costs equilibrium as a consequence of the model’s bottom-up and farmer participation in product markets? approach, generating better results than in top-down modeling (Axelrod and Tesfatsion, 2012). Moreover, ABM • What is the impact on farmers of complying with enables constructing and solving problems that are non- national and global product standards? tractable by usual analytical models (Billari et al., 2006; The experiments were based on intensive consultations Romero-Calcerrada et al., 2008). Lastly, ABM is also useful with farmer associations, field data collection through in the representation of changing behavior, starting with 109 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES individual actions and leading to of individuals sharing conditions change or when pricing mechanisms are more information, adapting, and learning. readily available. The ABM approach is particularly suited to these The modelling was implemented by the Quantitative and questions in the Uganda context, as the supply chains Applied Spatial Economic Research (QASER) laboratory have a wide range of organizational models (labor vs. at University College London. Data for the model were capital intensive, production location in periphery vs. core, collected from districts in the Central and Eastern regions time-sensitive vs. non-time sensitive goods, fragmented of Uganda, and used to define the spatial, geographical vs. integrated chains) and international transport means and demographic environment of the agent-based (shipping, air) requiring a varied scope for analysis. The system. The spatial and geographical environment is agricultural supply chains are characterized by a wide captured using a network structure. The nodes of the spectrum of agents who operate in highly fragmented network are represented by the GIS locations of villages in networks (farmers, traders, wholesalers, and exporters). Uganda. The links between the village nodes correspond These diverse agents operate within the same context to the available road infrastructure, which is mostly but have diverse goals, resources and available unpaved. A field survey was conducted in districts near information, and operate under different constraints. Kampala, including Luwero, Mayangayanga and Kikyusa The ABM approach may enable us to reveal unanticipated (Figure 2). A standardized questionnaire was designed underlying dynamics, for instance how specific types and administered during December 2013 to farmers, with of farmers/agents behave in terms of output and sales the objective of covering different aspects of agricultural decisions (location of sale, contracting, etc.) when market production, marketing and logistics practices, for five Figure 2: Study areas depicting ABM-generated locations of farmers (red), warehouses (green) and traders (blue). Source: Kunaka, Saslavsky and Watanuki, 2015. 110 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES representative products: hot pepper, chillies, matooke, conditions” network, which leaves road infrastructure okra, and sweet potatoes. Farmers were selected using quality unchanged. Based on this finding, a direct a stratified sampling method. Interviews were held also intervention in Entebbe Airport’s hinterland roads with producer cooperatives and associations. appears to be more effective than upgrading all roads in the Central and Eastern regions of Uganda in terms of The results of the experiments are summarized increasing participation rates. Although the ABM model below. is generally not well suited for identifying individual 4.1 Impacts of road infrastructure factors causing a particular outcome, it was noted that investments a considerable amount of smallholder agricultural production takes place around the Entebbe area. This A typical intervention to try and improve logistics in rural finding is consistent with other work, especially by areas is to upgrade all roads. As mentioned above, this Raballand, et al. (2009). is implied in the definition of the RAI. The experiment tested the impact of roads investment in Central and Improving access to markets can lead to a shift away from Eastern regions of Uganda, in the districts of Luwero, a supply-based situation of the agriculture sector where Mpigi, Masaka, Iganga, Mitiyana, Kamuli, and Mukono, farmers merely sell whatever crop surplus they have, to to upgrade all roads to all-weather standards (Figure 2). a demand-driven agriculture where farmers produce for Specifically, two possible intervention scenarios were the markets. However, this shift is complex, and there explored. First, rather than assuming that all roads are numerous constraints that still prevent Ugandan are upgraded to achieve 100% RAI in Uganda, road farmers from taking full advantage of participation improvement was targeted to agricultural-oriented in the market. The non-traditional export agriculture districts. On the second intervention, an improvement goods market remains underdeveloped, and depending to 100% RAI of all roads around the airport in Entebbe on the product can fluctuate significantly in terms of within a radius of 30 km was considered. pricing, production volume, export volume, and export values (Ugandan Bureau of Statistics, 2013), in part due The simulation showed that improving roads in the to erratic weather conditions or to changes in demand. agricultural districts would produce a 60% decrease in Penetrating international markets requires concerted travel time. Improving the infrastructure would affect efforts to increase production to a sufficient level, to the micro-economic behavior of farmers, as assumed develop efficient supply chains, and to establish market in the ABM construct. Agents/farmers are affected quality and standards. by road quality upgrading through “speed to market” improvements, which reduces post-harvest losses, 4.2 Consolidation of agricultural produce among other things. Moreover, the increase in market In an effort to look into logistics markets more closely, accessibility enables farmers/agents to augment their we ran a second experiment where we assumed that capital because of more frequent trade exchanges. the agriculture trader selects the best shipping route Greater connectivity will mean that farmers will be able between an origin and a destination to minimize their to sell their surplus production, and thus will be less transport costs while also satisfying demand. To keep likely to rely solely on their agriculture product for their transport costs low, the trader will offer to buy the subsistence. farmer’s crop at the farmgate (origin) and then transport Over the course of one simulated year, we do not observe it to a warehouse (destination). The trader can reduce long-lasting benefits in terms of trade growth or an costs by consolidating the produce of several farmers increase in overall productivity and welfare as a result and lowering unit delivery costs. The product price to of these transport improvements. Nonetheless, the be used in these transactions varies from point to point number of farmers participating in market exchanges and is determined by the trader in advance. It is assumed is significantly increased as compared to the “existing that the trader must determine prices at each warehouse 111 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES destination point for his offer to be acceptable to the impact that associating with other farmers would have on farmer. Moreover, the trader chooses the most suitable farmers’ market participation, transport cost, crop prices, price and tries to maximize his profit. However, the trader and volumes sold to traders. Association between two offers a price that also includes shipment costs, so the farmers implies that they share information, produce and price offered to the farmer will depend on the location sell together at the same price, consolidate and transport of the warehouses. together in the same warehouse and at the same logistics cost, and share revenues in proportion to the amount In the simulation approximately 40 control warehouse produced. locations were preselected in accordance with the geographical location of the 7 districts of study. In the Under the base conditions scenario of the model, it is ABM simulations, a trader decides whether the farmer assumed that only minimal coordination is taking place. must pay for transport (by offering a price that reflects Farmers are bounded rational agents, and they make transport costs) in relation to his (the trader’s) own decisions as a function of the information available to budget. Our simulation suggest that the outcomes them. Through association the farmers can share the depend on the number of farmers who decide to sell resources for paying the shipping cost, pass information to a trader and that the traders are better off selling at to one another and can gain some market power and specific consolidation centers. In addition, the cost of influence prices of their products. It is assumed that transport will significantly affect the farmer’s decision on association is realized between farmers no farther than whether or not to sell his crop. 30 kilometers (half a day walking) from each other and thus benefit from some agglomeration economies within Based on average probabilities of warehouse use their local area. Each farmer may start with a network of computed in the simulations, 5 warehouses emerged farmers known to him, and this network is then shared as potential consolidation centers (see Figure 2). The every time a new multiplier attachment/coordination centers, or rural hubs, are those that would minimize with a new farmer/outgrower is developed. The model transport costs for all the market participants, be they is designed so that the sharing of the network of known farmers or traders. Two of the selected sites are in the farmers occurs within one calendar year, so the spread Jinja area, and the others in Kabasanga (West of Kampala) of information is thus across one year to allow for and Luwero (North of Kampala, near the Kampala-Gulu adjustment and clearing of noise. Since the association Highway). In terms of characteristics the rural hubs are process is very time-consuming from the point of view those that have high population density, and agriculture of computing, we consider that the coordination process is the main economic activity. Their main attraction is takes place every 3 months. that farmers can sell their products quickly and minimize losses due to produce deterioration. This result confirms The results of the simulation are striking. In the case the existence of a hierarchy of markets and the fact that of no association, approximately half of the farmers the more central locations especially those with relatively (agents) cease to participate in the market (Figure 3a). large demand can play a role in how rural logistics are However, when the farmers associate, participation organized. As volumes increase, it would be important rates rise sharply. This result suggests that association for rural hubs to have improved facilities for storing among farmers may be highly significant over the longer- produce and to minimize post-harvest losses. term. Association results in lower transport costs: the curve related to the association among farmers lies 4.3 Farmer coordination strategies conspicuously below the curve representing the case of no association (Figure 3b). The third experiment explored the effects of farmer coordination strategies. The ABM enabled us to model The farmgate prices under associations of the crops coordination initiatives/association among farmers considered (hot peppers, chillies, matooke, okra, and through the spread of information that can be of any type sweet potatoes) are always higher than prices of crops (formal and informal). The model is used to explore the without association. For any crop, the supply chain is 112 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES composed of many intermediaries, each taking a margin a fall in the number of products sold to traders (Figure at every stage of the chain. When farmers form an 3c). A plausible explanation for this pattern is farmers, association, they can coordinate their actions and avoid by associating with each other and reducing their costs some of the intermediary stages by selling directly to while also ensuring more reliable supply, can sell directly the major wholesaler/trader. Generally, through the crop to final markets often at higher prices. price mechanism, the formation of associations leads to Figure 3: Effects of farmer association on market participation, transport costs, and volume of trading Number of bankrupt agents as function of time 300 250 Bankrupt agents 200 150 100 50 Association No Association 0 0 50 100 150 200 250 300 350 Days Transport costs 12 Association Total transport costs Ugsh (in billions) No Association 10 8 6 4 2 0 0 50 100 150 200 250 300 350 Days Number of product sold to traders 30 Association No Association 25 Product sold to traders 20 15 10 5 0 0 50 100 150 200 250 300 350 Days Source: Kunaka, Saslavsky and Watanuki, 2015 113 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 4.4 Adoption of food safety and quality and otherwise be compliant with buyer and/or regulatory standards requirements (Diaz Rios et al., 2009). However, due to little exposure to mainstream global buyer chains – linked Uganda’s export market has increased steadily over to their focus on ethnic and wholesale trade - Ugandan the past two decades; however, its non-traditional crop shippers have not faced much pressure from buyers to market is still poorly structured. Given its high transaction modify their practices and that of farmer suppliers (Diaz costs, the non-traditional crop market also has not Rios et al., 2009). been able to implement standards and quality control mechanisms adequately. This is a major hindrance to the The ABM was used to explore the impact of implementing development of trade, and particularly impedes sales to good agricultural practices (GAP) at the exporter level. It the major international importers such as the European is assumed that only some farmers apply GAP procedures. Union. The results of the simulations were mixed. On the one hand, the adoption of GAP procedures would in theory Increasingly therefore, the Government of Uganda has ensure access to the specific markets where set standards embraced compliance with private standards such as have to be complied with. In general, the introduction GlobalGAP. Good Agricultural Practices (GAP) “address of standards and controls was found to be conducive environmental, economic and social sustainability for on- to the dissemination of information on the benefits farm processes, and result in safe and quality food and associated with GAP. This will in turn impact how produce non-food agricultural products” (FAO COAG, 2003). In is handled all along the chain. However, the results also Uganda adopting GAP was deemed necessary to maintain suggest that there would be increased fragmentation the country’s presence in key developed markets and as of volumes as not all farmers will be able to meet the a tool for differentiation. Different interventions were standards. Fragmentation of volumes will increase unit geared to increase awareness of the standards and to costs of delivery of the products to markets. Specifically, facilitate their adoption, mainly through capacity building, products that are typically consolidated by traders, who technical and financial assistance—especially to cover tend to buy and mix produce from different farmers, certification costs. The assistance has predominantly may no longer meet the traceability requirements that been channelled through “lead firms” that are supposed are a feature of most GAP standards. Consequently, to provide support to smallholder farmers to adopt GAP shipment sizes would reduce and unit costs would go While there are legitimate reasons for specifying how products should be handled all along the supply chain the standards can also have the effect of locking out some of the small and more marginalized producers, typically those in lagging regions. 114 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES up. At the same time, the simulations suggested that However, as the results of the simulations suggest, the GAP implementation would not have a dramatic impact net effect of adopting high product handling standards on trader prices, although they increase at a slightly are double edged. While there are legitimate reasons faster rate than without GAP implementation. This latter for specifying how products should be handled all along phenomenon could be because the reduced volume the supply chain the standards can also have the effect is comprised of higher quality produce that meets the of locking out some of the small and more marginalized standards and therefore fetches higher prices in the producers, typically those in lagging regions. Such market. producers are often not able to meet any high standards at least without suffering the effects of high transport 5. Conclusions and logistics costs. The farmers can therefore end up relying on only the local markets and to a limited extent A thread running through the results of the experiments regional markets. This can have significant implications run in Uganda is that small scale farmers benefit most on poverty eradication outcomes. from coordinating among themselves in accessing information, searching for markets, deciding when to go A second consolidation strategy is virtual, through to market and when adopting new technologies. the use of ICT. There are several actors in agricultural logistics flows in rural regions in low income countries. One consolidation strategy is physical. The typical Among them are various types of intermediaries who shipment originating in rural areas in developing countries often may retain a large share of the rents from farming starts at the farm level, is transported to a storage compared to the farmers who do the actual production. facility and from there to buying points or to local and Providing farmers with information on prices or other regional markets. The development of road networks market conditions can influence where and when product can have a significant influence over the direction of flow is delivered to the market. In markets in lagging regions of agricultural shipments. As products move from the there is great uncertainty on how much or when product farmgate to market in lagging regions volumes can be should be ready for market. In some agro-supply chains increased and costs lowered by consolidating flows into there have been attempts to minimize this uncertainty by larger and larger vehicles. The larger vehicles require entering into contract farming arrangements between better roads than are available at the local level and producers and downstream processing or marketing which lead to central places where markets are larger. In enterprises. A more recent approach has been to exploit fact, there are discernible rural freight hubs, serving as information technology, to facilitate price discovery storage and logistics transfer points. In that regard the and assessment of market conditions and to coordinate placement of rural hubs is quite important to reduce between farmers on when to go to market. The appeal of costs in lagging regions. these features is reflected in a surge in the development of A complement of rural hubs is their link with urban areas. farmer applications for mobile phones, especially in Africa. Increasingly modern food retailing is dependent on A third consolidation strategy is one that is several supermarkets and large retail chains for food and food decades old, in the form of social organizations. Farmer services. Supermarkets are playing a transformative role organizations help individual producers overcome their in reorganizing and upgrading supply chains for fresh immediate disadvantage of small scale production. produce. They require produce of specific standards, Through collaboration farmers can combine their produce sizes, packaging, freshness, etc., which forces suppliers into large volumes, and gain some market power and to also update their systems if they want access to high retained incomes. However, as social formations, growing urban markets. The reconfiguration of supply many farmer associations fail to transform over time, and chains extends all the way to the farmgate. If small focus on issues that are removed from raison-d’etre and scale producers want to benefit from urbanization and are not sustainable. Nonetheless, farmer organizations globalization then they have to comply and meet these have a relevant role to play in logistics in lagging regions. requirements. 115 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Asebe, E. (2012). field survey report for World Bank. Axelrod, R. (1997). The complexity of cooperation. Princeton University Press (Princeton). Axelrod, R., and L. Tesfatsion (2005). “A Guide for Newcomers to Agent-Based Modeling in the Social Sciences.” In: Leigh Tesfatsion and Kenneth L. Judd (Eds.), Handbook of Computational Economics, Vol. 2: Agent-Based Computational Economics. Beckmann, M. J. (1976). “Spatial equilibrium in the dispersed city: In: Environment, regional science and interregional modeling (pp. 132-141). Springer Berlin Heidelberg. Billari, F. C., T. Fent, A. Prskawetz and J. Scheffran (2006). “Agent-based computational modelling: an introduction.” In: Agent-Based Computational Modelling (pp. 1-16). Physica-Verlag HD. Bolwig, S. (2012). Poverty and Gender Effects of Smallholder Organic Contract Farming in Uganda. USSP WORKING PAPER. IFPRI. Carruthers, R., and R. R. Krishnamani (2008). “Improving Connectivity: Investing in Transport Infrastructure in Sub- Saharan Africa.” Background Paper 7 for the Africa Infrastructure Country Diagnostic, World Bank, Washington, DC. Casaburi, L., R. Glennerster, and T. Suri (2013). Rural Roads and Intermediated Trade: Regression Discontinuity Evidence from Sierra Leone. Available at SSRN 2161643. Collinson, C.D., U. Kleih, D.G. Burnett, A. Muganga, J. Jagwe, and R.S.B. Ferris (2005). Transaction cost analysis for selected crops with export potential in Uganda, prepared for the Plan for the Modernisation of Agriculture by the Natural Resources Institute, UK, and the International Institute of Tropical Agriculture, Nigeria. ASARECA/ IITA Monograph 6, IITA, Ibadan, Nigeria. 168 pp. DFID (2005). “The Value of Time in Least Developed Countries: The African Studies.” Final report for contract R8307, Department for International Development, London. Dorosh, P., H. G. Wang, L.You, and E. Schmidt (2010). Crop production and road connectivity in Sub-Saharan Africa: a spatial analysis. World Bank Policy Research Working Paper Series, Vol. Dorward, Andrew, Shenggen Fan, Jonathan Kydd, Hans Lofgren, Jamie Morrison, Colin Poulton, Neetha Rao, Laurence Smith, Hardwick Tchale, Sukhadeo Thorat, Ian Urey and Peter Wobst (2004). “Institutions and Policies for Pro-poor Agricultural Growth,” Development Policy Review 22, Issue 6, November 2004, pp. 611-622. Evers, Barbara, Maggie Opondo, Stephanie Barrientos, Aarti Krishnan, Flavia Amoding, and Lindani Ndlovu (2014). Global and regional supermarkets: implications for producers and workers in Kenyan and Ugandan horticulture, Capturing the Gains: economic and social upgrading in global production networks. University of Manchester. Working Paper 39. Fafchamps, M., and R.V. Hill (2005). “Selling at the farmgate or traveling to market.” American Journal of Agricultural Economics 87(3), 717-734. 116 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES FAO Committee on Agriculture (2003). Development of a Framework for Good Agricultural Practices, Seventeenth Session. Rome, March 31 – April 4, 2003. Food and Agricultural Organization. Farole, T. (ed) (2013). The Internal Geography of Trade: Lagging Regions and Global Markets. World Bank. Washington, DC. Furlong, W. J., and G. A. Slotsve (1983). “ Will That Be Pickup or Delivery?”: An Alternative Spatial Pricing Strategy. The Bell Journal of Economics, 271-274. Gibbon, P. (2001). “Upgrading primary production: A GCC approach.” World Development 29 (2), pp. 345–363. Gollin, D., and R. Rogerson (2010). Agriculture, roads, and economic development in Uganda (No. w15863). National Bureau of Economic Research. Gong, W., K. Parton, R. J. Cox, and Z. Zhou (2007). “Transaction costs and cattle farmers’ choice of marketing channels in China: A Tobit analysis.” Management Research News 30(1), 47-56. Grabowski, R. (1999). “Coordination Problems, Risk Diversification, and Chance Economic Development and the Traditional Sector.” Journal of Asian Economics 10 (3): 475–87. Hazell, P., C. Poulton, S. Wiggins and A. Dorwood (2006). “The Future of Small Farms: Synthesis Paper” Based on Research Workshop organised by IFPRI, ODI and Imperial College, Wye. Iimi, A., F. Ahmed, E. C. Anderson, A. S. Diehl, L. Maiyo, T. Peralta-Quirós and K. S. Rao (2016) New Rural Access Index: Main Determinants and Correlation to Poverty. Washington, DC. World Bank. Policy Research Working Paper 7876. Kirsten, J. and K. Sartorius (2010). Linking agribusiness and small-scale farmers in developing countries: Is there a new role for contract farming? Development Southern Africa 19:4, 503-529, Kunaka, C. (2010). Logistics in Lagging Regions: Overcoming Local Barriers to Global Connectivity. World Bank. Washington, DC. Kunaka, C., D. Saslavsky and M. Watanuki (2015). Small Payers in Big World, Unpublished paper, World Bank. Minten, B. and C. Barrett (2008). “Agricultural Technology, Productivity, and Poverty in Madagascar,” World Development 36, issue 5, p. 797-822. Minten, B., B. Koro and D. Stifel (2013). The last mile (s) in modern input distribution: Evidence from northwestern Ethiopia. Ethiopia Strategy Support Program-2 Working Paper, 51. Neven, D. and T. Reardon (2004). “The Rise of Kenyan Supermarkets and the Evolution of their Horticultural Product Procurement Systems,” Development Policy Review Vol 22 (6), pp.669-699. Raballand, G., P. Macchi and C. Petracco (2010). Increasing roads investment efficiency in rural Areas. World Bank, Washington DC. Raballand, G., P. Macchi, D. Merotto, and C. Petracco (2009). Revising the roads investment strategy in rural areas: an application for Uganda. World Bank Policy Research Working Paper Series, No. WPS5036. World Bank. Ranganathan, Rupa and Vivien Foster (2012). Uganda’s infrastructure : a continental perspective (English). Policy Research working paper; no. WPS 5963. Washington, DC: World Bank. 117 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Diaz Rios, L., S.Jaffee, S.Henson and J. Mugisha (2009). “Not Quite Up to Standard: The Legacy of Two Decades of Government and Donor Assistance to Uganda’s Horticultural Export Sector.” Agriculture and Rural Development Discussion Paper, World Bank, Washington, DC. Romero-Calcerrada, R. and J. Millington (2008). “An Agent-Based Model of Mediterranean Agricultural Land-Use/Cover Change for Examining Wildfire Risk.” Journal of Artificial Societies and Social Simulation 11(4): 4. Ronner, E., and K. E. Giller (2013). Background information on agronomy, farming systems and ongoing projects on grain legumes in Uganda. Stifel, D. and B. Minten (2008). Isolation and agricultural productivity. Agricultural Economics 39(1), 1-15. Wiegratz, J., P. Nyabuntu and C. Omagor (2007). “Case Studies of Lead Firm Governance Systems in The Context of Commercialization of Smallholder Agriculture in Uganda.” Uganda Programme for Trade Opportunities and Policy, Kampala, Uganda. World Bank (2009). Not yet up to standard: the legacy of two decades of private, governmental, and donor efforts to promote Ugandan horticultural exports. World Bank (2012). The Transformational Use of Information and Communication Technologies in Africa. Washington D.C: The World Bank. 2012. World Bank (2014). Review of Smallholder Participation in Transforming Agri-Food Supply Chains in East, Working Paper. 118 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 119 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Trade Openness and Vulnerability to Poverty in Viet Nam under Doi Moi 1 Emiliano Magrini, Food and Agriculture Organization of the United Nations2 Pierluigi Montalbano, Sapienza University (IT) and University of Sussex (UK)3 1. Introduction Doanh, 2009; Coello et al., 2010). Empirical analyses consistently highlight the increased importance of F ollowing the so-called “Asian option” of transition, international trade in the Vietnamese economy as well from the early 1990s Viet Nam adopted the Doi as the positive correlation between trade liberalization, Moi (renovation) process, a combination of growth and poverty reduction. liberalization, stabilization and structural reforms. However, these studies focus mainly on the first sub- This included two main waves of trade liberalization, period, when the process of liberalization was still one in the 1990s and a second in the 2000s (Coello at restricted and subject to trade licences. Moreover, al., 2010). The first wave lasted from the initial opening the studies do not examine the relationship between of the country until approximately 2001 and foresaw openness and vulnerability to poverty. This is because the total abolition of trade licences and the removal they generally overlook the possible impact of the of most quantitative restrictions (Thanh and Duong, opening process on households’ exposure to risk as 2009). The second wave—still ongoing—includes the well the role of trade openness as one of the possible full involvement of the country in the global network channels of risk. of reciprocal trade agreements (both multilateral, WTO accession in January 2007, and bilateral, such as This work aims at addressing this gap, assessing agreements signed with the United States in 2001 as well differences in households’ vulnerability according to as FTA negotiations with the EU concluded in 2016). specific features such as the typology of economic activities (farm versus non-farm), gender, and trade Extensive empirical investigation of trade liberalization exposure. The value added of this analysis lies in taking and poverty dynamics in Viet Nam has been carried out advantage of a full set of available rounds of household (Irvin, 1997; Fritzen, 2002; Jenkins, 2004; Nadvi et al., surveys in Viet Nam to give a careful interpretation of 2004; van de Walle and Cratty, 2004; Jensen and Tarp, the cross-sectional evidence of risk-induced household 2005; Nguyen and Ezaki, 2005; Fujii and Roland-Holst, vulnerability, its determinants, and its heterogeneity 2008; Niimi et al., 2007; Abbott et al., 2009; Heo and across “trade-related” industries.4 1 We are very grateful to L. Alan Winters, Andy McKay, Chris Elbers and Julie Litchfield for helpful comments on earlier drafts of this work and to all the participants in the Seminars and Conferences where previous versions of this paper have been presented. 2 Agricultural Development Economics Division, Food and Agriculture Organization of the United Nations, Rome, Italy. 3 Department of Economics and Social Sciences, Sapienza University of Rome (IT), Department of Economics, University of Sussex (UK). 4 Because of the lack of panel data, our analysis is not able to directly control for cross-sectional household heterogeneity or for measurement error and their evolution over time. The main problem is that the cross-sectional variation in vulnerability to poverty across the various trade-exposed sectors can actually be driven by a number of factors other than risk (e.g., differences in household characteristics across sectors due to self-selection) that are unobservable to the researcher. By providing sound empirical techniques and taking advantage of the full set of available household and community controls, we are confident to be able to minimise the relevance of unobservables and provide useful upper bound of the phenomena under analysis. 120 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Empirical analyses consistently highlight the increased importance of international trade in the Vietnamese economy as well as the positive correlation between trade liberalization, growth and poverty reduction. 121 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The share of Moreover, the risk of future poverty for households engaged in activities directly affected by trade the vulnerable liberalization was driven by high volatility, not from expected mean consumption below the poverty line. The population in the above results are key for policymaking. They highlight a link between trade openness and risk-induced relatively more vulnerability, underlining the need to address vulnerability to poverty, even in the context of trade liberalization trade-exposed policies that result in a net reduction in poverty. sectors fell more The paper is organized as follows: section 2 reviews the literature and presents the conceptual framework; slowly than in section 3 presents the data; section 4 shows the empirical results; and section 5 concludes and provides key policy non-traded implications. sectors. 2. Trade openness and vulnerability to poverty: the conceptual framework The literature on trade liberalization and poverty The contribution of this paper is twofold: using six Living dynamics in Viet Nam has reached consensus on the Standards Measurement Surveys in Viet Nam (covering following issues: price liberalization has had a great the period 1992-2008), we first assess the level and impact on agricultural households since 1986 (Niimi et al., changes over time in the shares of vulnerable people 2007), with a substantial poverty reduction for rice net across economic sectors, organized according to their producers that exceeds that for rice net consumers (Heo relative degree of trade exposure; second, we measure and Doanh, 2009); trade liberalization has been beneficial how much of households’ consumption variation (which to the poor thanks to the highly labor intensive structure is at the core of vulnerability analysis) can be explained by of Vietnamese exports;5 the negative effects of trade its stochastic ex-ante component, namely the variance of liberalization occurred mainly in coffee production after income within trade-exposed groups, as well as by actual 1998 (Ha and Shively, 2008). income shocks, defined as the component of income However, a key issue remains unanswered: has trade variation unexplained by observables. openness magnified households’ exposure to risk and Our main results are the following. Vulnerability to raised their vulnerability to poverty? The topic is currently poverty fell in Viet Nam during the Doi Moi period, debated by practitioners, whereas it is largely ignored together with an increased share of its stochastic (risk) by the trade literature (Montalbano, 2011). In principle, determinant. The share of the vulnerable population trade can change the level of risk faced by households in the relatively more trade-exposed sectors fell more in two ways: by changing the riskiness of existing slowly than in non-traded sectors. Even after Doi Moi, activities, for instance, by altering the weight of foreign farming households engaged in the production of export compared with domestic shocks faced by the economy, crops and import-competing crops faced higher levels of or by shifting the composition of household activities, for vulnerability than those engaged in the production of example switching from subsistence food crops to cash non-traded crops or in non-farm activities. crops (McCulloch et al., 2001). 5 Abbott et al. (2009) claim that the poverty impacts of trade reforms in Viet Nam are even larger than those anticipated by existing model predictions, because of the intrinsic limitations of the most common applied methods and because they generally overlook the fact that institutional rather than tariff reforms have been the main driving factor behind recent development in Viet Nam. 122 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The poor face particularly severe challenges if trade management strategies (as well as different risk coping reform increases risk. Their ability to insure themselves ones when shocks occur), and that households adopting against adverse impacts tends to be limited, while their the second option (changing behaviour to get benefit of traditional coping mechanisms may be ineffective in trade liberalization) do not have appropriate risk sharing dealing with the greater exposure to foreign shocks and strategies, we would register different welfare impacts changes in incentives generated by trade liberalization ex-post. (Dercon, 2001; 2005). Moreover, the poor may lack information on the risks associated with the new 3. Data activities induced by openness (Winters et al., 2004). Our empirical analysis uses the standard measure of Trade openness can also affect government ability to vulnerability to expected poverty (VEP)6 (explained in adopt price stabilization policies or contribute to the detail in Appendix B), drawing on cross-sectional data for elimination of institutions or policies aimed at smoothing the following years: 1992, 1998, 2002, 2004, 2006 and domestic prices (Winters, 2002; Winters et al., 2004). 2008. Data come from two different sets of Vietnamese In all the above cases, trade openness can have an household surveys: the Viet Nam Living Standards Survey impact on households’ optimal economic activities and, (VLSS) and the Viet Nam Household Living Standards eventually, lead to net welfare effects that are less Survey (VHLSS).7 The variable used for consumption is positive than expected in the long run (Winters, 2002; the real per capita food and non-food expenditure in the Winters and al., 2004; Calvo and Dercon, 2007). This, past 12 months, re-adjusted by price indexes of regions together with the presence of risky assets (Elbers et al., and months. Poverty lines for computing vulnerability are 2007), may explain ex-ante their unwillingness to pursue expressed in Vietnamese Dong as follows: 1,160,000 for high average returns linked to the different activities 1992; 1,790,000 for 1998; 1,915,000 for 2002; 2,070,000 opened up by trade reforms and eventually the possibility for 2004; 2,559,000 for 2006; 3,360,000 for 2008. to fall into poverty traps (Carter and Barret, 2006; Dercon and Christiaensen, 2011; Barrientos, 2013). For instance, in the Vietnamese context, poor farmers in the midst of trade reform have two options. The first one is to rely on conservative choices (for example, subsistence farming) as their main risk management strategy, thus insulating themselves from trade-related risks. This leaves them still vulnerable to shocks that existed before liberalization (for example, natural ones), and fails to improve their income. The second option is to make changes in production in response to the new incentives generated by trade liberalization (for example, moving to an export crop such as coffee), with an expected increase in mean income as well as an increase in its volatility. With this choice they could climb out of poverty, but remain vulnerable to risks that existed before liberalization as well as the new ones relating to openness. Assuming that different risks (domestic and foreign) call for different risk 6 For a taxonomy of the main methods applied in vulnerability analysis, see Montalbano (2011). 7 The VLSS was undertaken in the period 1992/93 using a sample of 4,800 households, of which 4,000 were re-interviewed in 1997/98, out of a sample of 6,000 households in total. The VHLSS collected information from a new sample of 29,530 households in 2002; 9,188 in 2004; 9,189 in 2006 and 2008. Unfortunately, as reported by Pham and Reilly (2007) and Le and Booth (2010), the sampling frame for VHLSS differs substantially from that of VLSS: whereas VLSS used the 1989 Population Census, the VHLSS 2002 exploited the Population and Housing Census from 1999. As a result, while there are short panel samples from the last waves, no household was re-interviewed between the VLSS and the VHLSS and, generally speaking, a comparison between VLSS and VHLSS rounds is not possible. 123 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The variable used for household real per capita income roads, water pipelines, public transports, urban/rural has been derived by aggregating income into six major environment). categories: income from crops, income from agricultural Since VLSS and VHLSS surveys do not relate production sidelines, household business income, wage income, gifts and external trade, we group households according to and remittances, and other residual sources of income. the trade openness of their sector of specialization, as While we acknowledge possible measurement errors, in Coello et al. (2010). This requires matching the ISIC when errors are random errors with a mean of zero, and code of any sector with the SITC classification used in the variable with errors is used as a dependent variable, trade data and classifying sectors as follows: exported as in our case, it is well known that those errors will not manufactures; import-competing manufactures; non- cause estimation bias. Furthermore, as suggested by traded services; and agricultural goods. A further Nakata et al. (2009), measurement errors in retrospective breakdown of the agricultural sector is also provided, expenditure reports seem to be systematically related as follows: rice (considered separately because of its to household size. This suggests that the inclusion of special status for the Vietnamese economy: it acts as the household size as one of the control variables in our main staple food as well as the main cash crop); the main regressions contributes to mitigating biases arising from agricultural export products, other agricultural export measurement errors in consumption. products, import-competing crops and subsistence crops. The set of covariates used for our consumption Thus, we come up with eight trade-related production estimates includes household characteristics (such as sectors articulated into traded and non-traded, farm characteristics of the head of household, i.e., linear and and non-farm activities (see Table A.1 for details on the quadratic age, marital status, sex, linear and quadratic surveyed industries included in each sector). terms of family size and number of children); education achievements (primary, secondary, upper secondary, Figure 1 reports the average levels of mean real per technical/vocational, university) as well as village-level capita consumption for each trading group across time infrastructure characteristics (such as the presence of (Table A.2 in Appendix A provides additional statistics Figure 1: Real per capita consumption (average levels by trade categories in VN Dong) 8000 7000 6000 Per capita consumption 5000 4000 3000 2000 1000 0 1992 1998 2002 2004 2006 2008 Years Exporting industries Import-competing industries Non-traded non farm Rice Main export crops Other export crops Import-competing crops Non-traded food Source: Authors’ calculations 124 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES on real per capita consumption, real per capita income, overall VEP and its risk-induced sub-component for the current values of assets/durables, and the number each household in the sample.8 Table 1 reports the new of surveyed households by each category). The figure statistics alongside the poverty rates for each round of shows that, although both farm and non-farm activities household data.9 actually increased their consumption levels over time, Both poverty and vulnerability fell during the Doi Moi people involved in non-farm activities are on average reforms. The share of Vietnamese households under the characterized by higher consumption levels than farmers poverty threshold fell from more than 50% at the eve of (the highest consumption is registered by households the liberalization process to about 16% in 2008, while involved in non-traded non farm), followed by export the share of vulnerable households fell from around 56% industries and import-competing manufacturing (and, in 1992 (68% in the case of rural households) to 8.3% in more recently, by export crops). Conversely, households 2008 (10.2% of rural households). The decline in poverty involved in rice production (actually the vast majority of was greatest at the start of the liberalization process sampled ones, see Table A.2) show, on average, the lowest (between 1992 and 1998) and more relevant for rural level of real per capita consumption. This is consistent households than for urban households: vulnerable urban with the fact that incidence of poverty is lower in non- households were about 7% of the total at the beginning farm sectors than in farm sectors (with the exception of of the openness process, falling to about 0.5% already in farm main-exports and non-traded crops) and fell sharply 1998. The same pattern is confirmed when we disentangle in households engaged in non-traded farm activities. farm and non-farm households’ activities, although 4. The empirical analysis 25% of households involved in non-farm activities were vulnerable in 1992. Vulnerability was higher among male- Our empirical analysis adds new pieces of information to headed households (9.4% in 2008) than in female-headed the standard picture of poverty and trade liberalization households (4%). in Viet Nam under Doi Moi, by computing both the Table 1: Vulnerability and poverty in Viet Nam (1992-2008) 1992 1998 2002 2004 2006 2008 Poverty Rate in the Survey 55.2 29.9 28.0 19.4 15.3 16.4 VEP Rate (%) 56.1 21.5 18.3 10.8 7.1 8.3 Non-Farm 25.1 6.5 6.7 3.4 2.0 2.0 Farm 69.0 30.9 27.9 17.5 12.0 11.7 Rural 68.2 29.8 23.6 14.0 9.3 10.2 Urban 7.1 0.5 0.5 0.1 0.2 0.2 Female 43.5 13.5 8.4 4.4 2.7 4.0 Male 60.7 24.4 21.1 12.6 8.3 9.4 Risk-induced VEP (% vulnerable) 18.7 33.7 31.0 31.2 32.6 31.1 Non-Farm 30.3 47.8 45.9 46.9 46.3 61.7 Farm 17.0 31.9 28.0 28.5 30.4 28.2 Rural 17.7 33.4 30.9 31.2 32.6 31.0 Urban 61.0 87.5 39.4 100.0 33.3 50.0 Female 22.6 40.5 39.5 50.6 46.0 53.7 Male 17.7 32.4 30.0 29.3 31.4 28.6 Source: Authors’ calculations. Note: VEP rates = shares of vulnerable households on total sampled households. 8 As is common practice, we consider households as vulnerable if they show a probability higher than 0.50 to fall into poverty at least once in the following two years. To this end, we compute vulnerability as one minus the probability of no episodes of poverty, as follows: Vht+k=1-[P(lncht>lnz)]2, given the information set at t. 9 Ex-ante vulnerability and ex-post poverty should be viewed as different statistics: while we can compare their evolution over time, we cannot draw any cross comparisons between them. For those who are interested in this, Imai et al., (2011) suggest a method of making such a comparison by means of a multinomial logit model, adding VEPh,t−1 as one of the arguments. 125 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES A different picture comes out if we look at the share higher incidence of risk-induced vulnerability among of the risk-induced component of vulnerability (i.e., urban than rural households, as well as in households the component of vulnerability associated with a high involved in non-farm than in farm activities (even if in both estimated variance of consumption, but expected cases the former categories show very low percentages consumption above the poverty line). In this case, after a of vulnerable households overall). In other words, our common drop moving from VLSS to VHLSS (between 1998 analysis shows that the nature of vulnerability changed and 2002), probably due to the substantial difference in over time (from poverty-induced to risk-induced). the sampling frame between the two surveys, the risk- Table 2 reports the breakdown of the vulnerability induced vulnerability never fell below the threshold of statistics by trading sector between farm and non-farm 31% of the overall VEP. Moreover, differently than in activities.10 For each trading sector and surveyed year, the overall measure, a higher share of female-headed it shows the total percentage of vulnerable households households than male-headed households are vulnerable and the percentage of vulnerable households that are by the risk-induced VEP measure, and the former share considered as risk-induced. The percentage of vulnerable rises, with more than 50% of vulnerable female-headed people decreased steadily in all trade-related sectors households risk-induced in 2008. Also remarkable is the Table 2: Overall and risk-induced vulnerability by farm and non-farm activities and trade-related sectors 1992 1998 2002 2004 2006 2008 Non-farm activities Export manufactured goods 22.4 10.0 10.8 5.3 3.8 2.3 Import manufactured goods 43.6 6.1 8.0 4.1 2.9 3.2 Non-traded non farm 18.9 5.8 5.5 2.8 1.4 1.4 Farm activities Main export agricultural products 54.5 14.9 25.9 11.0 3.0 3.4 VEP rate (%) Other export agricultural products 51.1 26.3 25.3 16.8 7.3 9.3 Import-competing crops 58.3 39.5 36.8 26.8 13.2 19.3 Non-traded crops 43.8 22.0 10.8 2.8 1.1 1.9 Rice 71.6 32.1 27.8 17.8 13.4 12.3 Net consumer 45.1 16.4 13.3 7.8 4.3 5.4 Net producer 68.2 27.5 20.5 14.4 10.3 11.1 Non-farm activities Export manufactured goods 31.8 48.5 39.4 41.9 40.9 55.6 Import manufactured goods 26.4 43.8 45.1 45.5 37.5 60.0 Non-traded non farm 33.3 48.3 49.0 49.4 52.4 66.7 Farm activities Risk-induced VEP Main export agricultural products 23.6 56.8 32.1 52.4 55.6 50.0 (% vulnerable) Other export agricultural products 20.9 17.1 31.8 31.3 62.5 45.7 Import-competing crops 14.3 25.5 20.2 25.4 40.0 25.3 Non-traded crops 42.9 62.5 50.9 75.0 60.0 50.0 Rice 16.5 32.4 28.5 26.5 28.2 26.4 Net consumer 20.5 27.5 30.4 34.6 37.7 35.4 Net producer 17.5 37.9 34.0 28.8 29.9 29.3 Source: Authors’ calculations. 10 Both the F-statistics of the one-way ANOVA and the Levene’s T-test reject in each round of observations the null hypotheses that the means and the variances of the estimated income residuals are the same across trade-related production groups. We are thus confronting heterogeneity in unexplained stochastic components when households are gathered by trade-related sectors. 126 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES (with the usual jumps moving from VLSS to VHLSS). As in the case of the risk-induce component of vulnerability. a result, in 2008 (our last year of observation), all trade- Notwithstanding the fact that the average income/ related sectors register, without exception, a lower consumption of households involved in main-export percentage of vulnerable households than in 1992. crops is similar to that of households involved in non- Nevertheless, farm activities show higher percentages traded non farm activities (see Table A.2), the share of than non-farm ones, with the relevant exception of vulnerable people in the former is higher than in the households producing non-traded crops. latter for all years. This is noteworthy if we consider the low incidence of poor households involved in export According to our VEP estimates, the sectors with the crops and the roughly equal distribution of income across lowest percentage of vulnerable households are non- deciles within that sector. traded non farm and non-traded crops (in both cases, the percentage of vulnerable households is below Hence, we can argue that the hypothesis of heterogeneity 2% in 2008). Among farm activities, the production in vulnerability by trade sector is not rejected by the sector with the highest percentage of vulnerable empirical data in Viet Nam. Furthermore, all non-farm households is import-competing sectors, followed by activities register in 2008, generally speaking, a higher rice. Acknowledging the peculiar nature of the rice sector share of risk-induced vulnerability than farm ones, where which is, at the same time, the main production sector import-competing crops and rice seem to be the least and the main source of food for Vietnamese households, exposed. Although the share of risk-induced vulnerable the last two rows of Table 2 show the decomposition of households is computed on a smaller total number vulnerability patterns between rice net producers and of vulnerable households, this is a relevant issue for net consumer households; although the shares of the policymaking. At the same time, we should acknowledge vulnerable are higher among net rice producers than the inherent weaknesses of VEP of measuring risk among net rice consumers, the opposite pattern holds appropriately (see Appendix B). 127 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES To shed light on the black box of the risk-induced VEP Figure 3 shows the evolution of the net contribution of component, we further disentangle the relative weight the ex-ante component of income innovation in reducing of its various determinants by calculating the so- households’ consumption by clustering households called dispersion importance (Achen, 1982), 11 i.e., the across groups of industries classified as traded, not traded proportion of the variance in consumption explained by and rice. The picture highlights a higher average of the the different covariates in the vector X. 12 Figure 2 plots ex-ante stochastic component in the case of the trading the average values over the six surveys (the estimated sectors compared with non-traded ones, especially in the coefficients for each round of the observations are most recent rounds, net of the usual jump between VLSS reported in Table A.3 in Appendix A). It shows that all and VHLSS. the non-stochastic covariates are statistically significant Even if our exercise cannot be considered a proper test and show the expected signs.13 The striking feature of of consumption behavior under risk—because of its static our empirical outcomes is that both our ex-post and ex- nature—this last result confirms that we are confronting ante stochastic components of income14 are the most heterogeneity in the variance of income innovation important determinants of household consumption which is correlated with the degree of trade openness fluctuations.15 Figure 2: Dispersion importance of the determinants of household consumption (estimated beta coefficient of per capita consumption, period 1992-2008) 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 ris k ck s ck ba n Ag e ge siz e siz e ren rrie d se x on on on tio n tio n ad s icit y ter ort te ho ho Ur .A ild ati ati ati ca ca Ro ctr Wa sp an ss gs Sq HH .H H ch Ma ad uc uc uc du du Ele Tra n Ex Po Ne Sq No He ed ed r ed he ye ary ary pe Te c rs i t Pr im nd Up ive Upper bound co Un Se Point est. (Av.) Lower bound Source: Authors’ calculations 11 Standardized coefficients are the regression coefficients when all variables have been standardized to mean zero and variance one (z scores). For more details, see Achen (1982). 12 See eq. B.7 in Appendix B and the estimated coefficients reported in Table A.3 in Appendix A. 13 The signs of age and its square coefficients confirm, in principle, the well-known concave age-consumption profile, even if the decreasing rate is in this case meaningless. Not surprisingly, having children reduces household per capita consumption while being married increases it. The significance of the parameter associated with the household dimension also mitigates pos- sible measurement error bias. Whether the head of the household is male or female is correlated with consumption too. The education variables also behave as expected, that is, higher levels of education correspond to higher levels of consumption. Lastly, the presence of a set of village characteristics (urban status and availability of paved roads, electricity, tap water and public transport) are associated with a higher level of consumption as well. 14 The outcomes of the income equation (eq. 1) which are used to separate the ex-ante and ex-post components of risk are reported in Table A.4 in Appendix A. 15 For sensitivity purposes additional estimates of eq. B.7 (see Appendix B) were carried out, including dummies for trade categories. On the one hand, this helps us capture possible unobserv- able income effects other than those already controlled for by the observable characteristics, neutralizing differences in average income between groups (i.e., households in different trade categories show heterogeneous consumption because of heterogeneous income). On the other hand, while the risk term is supposed to capture both within and between group effects, the inclusion of trade categories acknowledges that some risks can be common to households in the same trade group and allows us to isolate the risk effect within groups (i.e., risks are identified within the groups) better than in the estimates without trade categories. While the overall fit of the model with the trade dummies slightly improves, the coefficients of the risk terms do not change significantly. The above evidence suggests that the trade dummies mainly capture differences in mean income that do not influence the risk channel depicted above. 128 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES of production sectors. Again, if we are willing to assume that this could be caused only by unobserved it is the upper bound of a proper measure of the trade- heterogeneity other than risk, especially if we take into induced risk component, the plain conclusion is that, account that it is negatively correlated with consumption not only is risk increasing over time in Viet Nam under behaviour. If that were the case, it would be a very Doi Moi, but that its relevance (in terms of net relevant issue for policymaking anyway since it would contribution to the variance of household consumption) also imply a revision of the assumed trade benefits for is proportionally higher the higher the trade exposure the welfare of Vietnamese households working in the of the sector the household is involved in. It is unlikely most exposed trading sectors. Figure 3: Evolution of the net contribution of the risk component on average household consumption (1992-2008) in traded, rice and not traded sectors. 1992 1998 0 0.1 -0.002 0.05 0 -0.004 -0.05 -0.006 -0.1 -0.008 -0.15 -0.01 -0.2 -0.012 -0.25 -0.014 -0.3 -0.016 -0.35 Traded Rice Non-traded Traded Rice Non-traded Upper bound Upper bound Point est. (Av.) Point est. (Av.) Lower bound Lower bound 2002 2004 0.02 0.01 0.015 0.01 0 0.005 -0.01 0 -0.005 -0.02 -0.01 -0.03 -0.015 -0.02 -0.04 -0.025 -0.03 -0.05 Traded Rice Non-traded Traded Rice Non-traded Upper bound Upper bound Point est. (Av.) Point est. (Av.) Lower bound Lower bound 2006 2008 0 0.05 0 -0.05 -0.05 -0.1 -0.1 -0.15 -0.15 -0.2 -0.25 -0.2 -0.3 -0.25 -0.35 Traded Rice Non-traded Traded Rice Non-traded Upper bound Upper bound Point est. (Av.) Point est. (Av.) Lower bound Lower bound 129 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 5. Conclusions process, the risk of future poverty is mainly driven by the risk-induced component. This implies that the threat This paper presents a comprehensive analysis of of falling into poverty does not come from an expected vulnerability to poverty in Viet Nam during Doi Moi. The mean consumption below the poverty line, but from its results show a decreasing trend in vulnerability to poverty high volatility. By further investigating the determinants along with a decreasing trend in poverty, confirming the of consumption volatility we finally highlight the role of well-known positive impact of the reforms—including risk heterogeneity across households according to their trade liberalization—on the overall performance of the degree of risk exposure. country. By these measures, the liberalization process reduced both the observed poverty as well as the risk of These results provide some useful insights to future poverty. policymakers. First of all, they show that “risk-induced” vulnerability is relevant and significant even in absence However, a more disaggregated picture on the distribution of ex-post shocks. Second, they demonstrate that the of these benefits reveals that the encouraging results liberalization process needs to be accompanied by shown at the aggregate level hide the presence of some additional support to households engaged in those farm subsets of the population who face increased risk and activities more exposed to international competition, thus a high probability of falling back into poverty in the since trade openness can magnify risk. This is because near future. Our analysis tests if this risk depends on the liberalization changes the riskiness of existing activities, relative position of a household with respect to some altering the weight of foreign relative to domestic specific features such as the typology of its economic shocks faced by the economy and, as a consequence, activities (non-farm versus farm), trade exposure, and the households’ optimal economic activities. This is gender. Despite the fall in the vulnerability level from especially true for the smallholder because of their poor 56% to 8% over the sampled period, we still observe ability to take advantage of the positive opportunities that after Doi Moi those employed in farm activities are, created by trade reforms, their weak capabilities to on average, five times more likely to fall into poverty insure themselves against adverse impacts and, possibly, compared to households engaged in non-farm activities. the lack of information about the risks associated with The same is true when we look at the distinction the new activities induced by openness Interventions between rural and urban areas, making evident that to address these issues should primarily target trade- farmers in rural areas still deserve special attention by induced vulnerable households. First, we need to better policymakers interested in limiting an increase of poverty protect them from excessive price volatility, in the spirit in the near future. Finally, when we look specifically at of the global trade negotiations on special safeguard the risk-induced components of vulnerability to poverty, mechanisms. Second, we also need to help them to carry we detect a relatively higher incidence of vulnerable out progressive choices and take full benefit of trade households in non-farm activities and in female-headed reforms. This means fostering their ability to take risks households. consciously. This can be done by supporting self-insurance via savings (through micro-financial instruments), assisting Our estimates also show that vulnerability to poverty income risk management by providing access to credit, varies systematically according to trade exposure of sustaining community-based risk-sharing and pushing the surveyed households, especially for those involved in public and private institutions to develop new insurance farm activities. In particular, farmers engaged in the products targeted to farmers most involved in tradable production of export crops and import-competing crops cropping. still registered higher levels of vulnerability after Doi Moi than those engaged in non-traded crops or non-farm activities and, in some cases, also a new increase in recent years. More interestingly, for the categories exposed to international trade and, therefore, the liberalization 130 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Abbott, P., J. Bentzen and F. Tarp (2009). Trade and Development: Lessons from Vietnam’s Past Trade Agreements. World Development 37(2), pp. 341-353. Achen, H. (1982). Interpreting and using regression. SAGE Publications Ltd., United Kingdom. Amemiya, T. (1977). The maximum likelihood estimator and the nonlinear three stage least squares estimator in the general nonlinear simultaneous Equation model. Econometrica 45, pp. 995-968. Banks, J., R. Blundell and A. Brugiavini (2001). Risk Pooling, Precautionary Saving and Consumption Growth. Review of Economic Studies, Wiley Blackwell, vol. 68(4), pages 757-79, October. Barrientos, A. (2013). Does vulnerability create poverty traps? In Chronic Poverty (pp. 85-111). Palgrave Macmillan UK. Blundell, R., and T. Stoker (1999). Consumption and the timing of in-come risk. European Economic Review, Elsevier, vol. 43(3), pages 475-507, March. Calvo, C., and S. Dercon (2007). Vulnerability to poverty. CSAE WPS 2007-03, Oxford: University of Oxford. Carroll, C., and A. Samwick (1997). The nature of precautionary wealth. Journal of Monetary Economics 40(1), pp. 41-71. Carroll, C. and A. Samwick (1998). How Important Is Precautionary Saving? The Review of Economics and Statistics 80(3), pp. 410-419. Carter, M. R. and C. B. Barrett (2006). The economics of poverty traps and persistent poverty: An asset based approach. Journal of Development Studies 42, 178-199. Chaudhuri, S. (2003) Assessing vulnerability to poverty: concepts, empirical methods and illustrative examples. New York: Mimeo, Department of Economics, Columbia University. Chaudhuri, S. and G. Datt (2001). Assessing Household Vulnerability to Poverty: a Methodology and Estimates for the Philippines. World Bank Draft. Chaudhuri, S., J. Jalan and A. Suryahadi (2002). Assessing Household Vulnerability to Poverty from Cross-sectional Data: A Methodology and Estimates from Indonesia, (mimeo), Department of Economics, Columbia University: New York. Christiaensen, L. and K. Subbarao (2005). Towards an Understanding of Household Vulnerability in Rural Kenya. Journal of African Economies 14(4), pages 520-558. Coello, B., M. Fall and A. Suwa-Eisenmann (2010). Trade liberalization and poverty dynamics in Vietnam 2002-2006. PSE Working Papers, HAL. Dercon, S. (2001). Assessing Vulnerability to Poverty. Oxford: Mimeo, Jesus College and CSAE, Department of Economics, Oxford University. Dercon, S. (2005). Risk, Insurance, and Poverty: A Review. In: Dercon, S. (ed.), Insurance against Poverty, Oxford University Press. Dercon, S., and L. Christiaensen (2011). Consumption risk, technology adoption and poverty traps: Evidence from Ethiopia. Journal of Development Economics 96 (2), 159-173. 131 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Elbers, C., J. W. Gunning and B. Kinsey (2007). Growth and Risk: Methodology and Micro Evidence, World Bank Economic Review, Oxford University Press, 21(1), pp. 1-20. Foster, J., J. Greer and E. Thorbecke (1984). A class of decomposable poverty measures. Econometrica 52(3), 761-766. Fritzen, S. (2002). Growth, inequality and the future of poverty reduction in Vietnam. Journal of Asian Economics 13, pp. 653-657. Fujii, T., and D. Roland-Holst (2008). How Does Vietnam’s Accession to the World Trade Organization Change the Spatial Incidence of Poverty? In: Globalization and the Poor in Asia (pp. 47-89). Palgrave Macmillan UK. Giles, J. and K. Yoo (2007). Precautionary Behavior, Migrant Networks, and Household Consumption Decisions: An Empirical Analysis Using Household Panel Data from Rural China. The Review of Economics and Statistics 89 (3), pp. 534-551. Gourinchas, P. and J. Parker (2002). Consumption Over the Life Cycle. Econometrica 70(1), pp. 47-89. Guiso, L., T. Jappelli and D. Terlizzese (1992). Earnings uncertainty and precautionary saving. Journal of Monetary Economics 30, 307-337. Gunther, I., and K. Harttgen (2009). Estimating Households Vulnerability to Idiosyncratic and Covariate Shocks: A Novel Method Applied in Madagascar. World Development 37(7), pp. 1222-1234. Ha, D., and G. Shively (2008). Coffee Boom, Coffee Bust and Smallholder Response in Vietnam’s Central Highlands. Review of Development Economics 12(2), pages 312-326. Heo, Y. and N.K. Doanh (2009). Trade Liberalisation and Poverty Reduction in Vietnam. The World Economy 32(6), pp. 934-964. Hubbard, R., J. Skinner and S. P. Zeldes (1994) Expanding the Life-Cycle Model: Precautionary Saving and Public Policy, American Eco¬nomic Review, 84(2), pp. 174-79. Imai K., R. Gaiha and W. Kang (2011). Vulnerability and poverty dynamics in Vietnam. Applied Economics 43(25), pp. 3603-3618. Irvin, G. (1997). Vietnam: Adjustment, Growth and Poverty. Journal of International Development 9 (September), pp. 783-801. Jalan, J., and M. Ravallion (2001). Behavioral responses to risk in rural China. Journal of Development Economics, 66(1), pp. 23-49. Jenkins, R. (2004). Vietnam in the Global Economy: Trade, Employment and Poverty. Journal of International Development, 1613-28. Jensen H.T. and F. Tarp (2005). Trade Liberalization and Spatial Inequality: a Methodological Innovation in a Vietnamese Perspective. Review of Development Economics 9(1), pp. 69-86. Kamanou, G. and J. Morduch (2004). Measuring Vulnerability to Poverty. In Dercon, S. (ed.), Insurance against Poverty, Oxford University Press. Le, H. and A. Booth (2010). Urban-Rural Living Standard Inequality in Vietnam, 1993-2006. Mimeo. 132 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES McCulloch, N., L. A. Winters and X. Cirera (2001). Trade liberalization and poverty: a handbook. London, Centre of Economic Policy Research Meghir, C. and L. Pistaferri (2004). Income Variance Dynamics and Heterogeneity. Econometrica, Econometric Society, 72(1), pages 1-32. Montalbano, P. (2011). Trade Openness and Developing Countries’ Vulnerability: Concepts, Misconceptions, and Directions for Research. World Development 39(9), pp. 1489-1502. Nakata, H., Y. Sawada and M. Tanaka (2009). Asking retrospective questions in household surveys: evidence from Vietnam. Discussion Paper Series 10-E-008. Tokyo: RIETI. Nadvi, K., J.T. Thoburn, B.T. Thang, N. Ha, N. Hoa, D. Le and E. De Armas (2004). Vietnam in the global garment and textile value chain: impacts on firms and workers. Journal of International Development 16(1), pages 111-123. Nguyen, T.D. and M. Ezaki (2005). Regional Economic Integration and its impacts on growth, poverty and income distribution: the case of Vietnam. Review of Urban & Regional Development Studies 17: 197-215. Niimi, Y, P.V. Dutta and L. A. Winters (2007). Trade liberalisation and poverty dynamics in Vietnam. Journal of Economic Integration 22(4), 819-851. Pham, T. H., and B. Reilly (2007). The gender pay gap in Vietnam, 1993–2002: A quantile regression approach. Journal of Asian Economics 18(5), 775-808. Pritchett, L., S. Sumarto and A. Suryahadi (2000). Quantifying Vulnerability to Poverty: A Proposed Measure with Application to Indonesia, World Bank Policy Research Department Working Paper No.2437, World Bank: Washington DC. Skinner, J. (1988). Risky income, life-cycle consumption and precautionary saving. Journal of Monetary Economics 22, 237-255. Storesletten, K., C. Telmer and Y. Amir (2004). Consumption and risk sharing over the life cycle. Journal of Monetary Economics Elsevier, vol. 51(3), pages 609-633, April. Thanh, V. T., and N. A. Duong (2009). Vietnam after two years of WTO accession: What lessons can be learnt?. ASEAN Economic Bulletin 26(1), 115-135. van de Walle, D. and D. Cratty (2004). Is the emerging non-farm market economy the route out of poverty in Vietnam? The Economics of Transition, The European Bank for Reconstruction and Development, 12(2), pp. 237-274. Winters, L. A. (2002). Trade liberalisation and poverty: what are the links?. The World Economy 25(9), 1339-1367. Winters, L. A. (2004). Trade liberalisation and economic performance: an overview. The Economic Journal 114, F4-F21. Winters, L. A., N. McCulloch and A. McKay (2004). Trade liberalization and poverty: the evidence so far. Journal of Economic Literature 42(1), 72-115. 133 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix Appendix A: Methods Table A.1: Industries classification by trade-related sectors Exports Non-Farm Non-Traded Non-Farm Fishing, aquaculture Recycling Mining of coal and lignite: extraction of peat Electricity, gas, steam and hot water supply Extraction of crude petroleum and natural gas Collection, purification and distribution of water Wearing apparel: dressing and dyeing of fur Construction Footwear Sale, maintenance and repair of motor vehicles Wood and of products of wood and cork Wholesale trade and commission trade Office, accounting and computing machinery Retail trade, repair Hotels and restaurants Import-Competing Non-Farm Land transport; transport via pipelines Forestry, logging and related service activities Water transport Mining of uranium and thorium ores Air transport Food products and beverages Supporting and auxiliary transport activities Tobacco products Post and telecommunications Textiles Financial intermediation Tanning and dressing of leather: luggage Insurance and pension funding Paper and paper products Activities auxiliary to financial intermediation Coke, refined petroleum products and nuclear fuel Real estate activities Chemicals and chemical products Renting of machinery and equipment Rubber and plastic products Computer and related activities Other non-metallic mineral products Research and development Basic metals Other business activities Fabricated metal products Public administration and defense Machinery and equipment Education Electrical machinery and apparatus Health and social work Radio, television and communication equipment Sewage and refuse disposal, sanitation Medical, precision and optical instruments Activities of membership organizations n.e.c. Motor vehicles, trailers Recreational, cultural and sporting activities Furniture; manufacturing n.e.c. Other service activities Private households as employers Main Export Farm Extraterritorial organizations and bodies Black pepper Exports Cashew, coffee Import-Competing Farm Rubber, tea Apples, grapes Fresh vegetables Other Export Farm Indian Corn Bananas Jackfruit, durian Cassava manioc Jute, ramie Coconut Mulberry Cotton Oranges, limes Cabbage, cauliflower Other leafy greens Mango, Papaya Plums, potatoes Peanuts Sugar cane Pineapple Tobacco Sesame seeds Tomatoes Soy beans Specialty rice Non-Traded Farm Sweet potatoes Custard apple (subsistence) Litchi, logan, rambutan Rice Sapodilla Water morning glory Source: Coello et al., (2010). 134 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: Main descriptive statistics of sampled households by farm and non-farm activities and trade-related sectors (all monetary values are in VN Dong) Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 1992 Non-farm activities Mean 2192.451 4411.844 50768.46 Std Dev. 1561.628 4115.385 60503.03 Exporting industries Min 659.5261 702.151 770 Max 9416.787 28112.09 254030 Obs. 90 90 90 Mean 1703.968 4010.86 40157.48 Std Dev. 1096.584 4141.78 69949.56 Import-competing industries Min 644.5936 685.3751 420 Max 6964.31 32100.77 557640.00 Obs. 248 248 248 Mean 2141.634 4513.755 54588.53 Std Dev. 1344.579 4356.647 110483.2 Non-traded non farm Min 632.6236 588.8931 325 Max 13302.89 31179.41 1856910 Obs. 764 764 764 Farm activities Mean 1205.835 2228.961 11760.01 Std Dev. 588.1142 2181.356 14503.86 Rice Min 632.6989 581.9226 250 Max 9823.781 32836.96 200165 Obs. 1984 1984 1984 Mean 1415.444 3392.292 14058.95 Std Dev. 763.8515 3268.074 13966.92 Main export crops Min 655.5554 595.1584 700 Max 5502.093 20253.97 65835 Obs. 79 79 79 Mean 1422.605 2839.797 12518.11 Std Dev. 685.3914 3739.296 16657.67 Other export crops Min 641.1921 601.6185 145 Max 4300.459 25073.38 125210 Obs. 115 115 115 Mean 1434.692 2303.38 16439.59 Std Dev. 785.5209 1808.142 52780.9 Import-competing crops Min 638.0425 583.5345 310 Max 4542.778 11332.75 429000 Obs. 68 68 68 Mean 1713.105 3394.557 9093.655 Std Dev. 821.2369 2458.819 8167.273 Non-traded food Min 766.6361 707.5646 1240 Max 3904.79 9992.304 38770 Obs. 29 29 29 (continued) 135 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: continued Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 1992 Mean 1501.673 2994.423 24725.37 Std Dev. 985.8579 3260.65 61269.39 Total Min 632.6236 581.9226 145 Max 13302.89 32836.96 1856910 Obs. 3377 3377 3377 1998 Non-farm activities Mean 3412.447 5272.34 36800.08 Std Dev. 2260.809 4494.307 40726.13 Exporting industries Min 781.2977 580.001 2891 Max 13071.95 31198.08 320369 Obs. 313 313 313 Mean 4128.725 6742.565 37319.39 Std Dev. 2521.305 4987.061 41747.78 Import-competing industries Min 1000.463 725.8027 1789 Max 15113.75 28302.5 339667 Obs. 246 246 246 Mean 4575.739 7008.84 39891.78 Std Dev. 2869.457 5516.867 44393.46 Non-traded non-farm Min 672.0535 607.9286 1606 Max 18447.21 33397.65 569448 Obs. 1444 1444 1444 Farm activities Mean 2188.854 3272.888 29498.69 Std Dev. 1134.081 2615.77 15094.29 Rice Min 641.6957 580.1642 4395 Max 17954.53 32352.02 187352 Obs. 2233 2233 2233 Mean 2913.869 6095.626 50035.58 Std Dev. 1396.513 5332.938 22506.29 Main export crops Min 668.3075 641.0767 13251 Max 7743.051 31930.03 161200 Obs. 243 243 243 Mean 2371.039 3299.853 30405.18 Std Dev. 1352.125 2531.955 16198.06 Other export crops Min 642.0324 616.2089 6555 Max 12183.87 16451.14 162416 Obs. 257 257 257 Mean 2223.277 4110.429 30030.66 Std Dev. 1124.119 3511.499 14803.79 Import-competing crops Min 763.335 687.3796 5162 Max 7330.38 23243.84 101753 Obs. 369 369 369 136 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: continued Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 1998 Mean 2944.834 5428.627 36216.86 Std Dev. 1560.53 4595.308 18145.84 Non-traded food Min 1133.982 650.2026 4147 Max 12939.04 27087.97 119059 Obs. 107 107 107 Mean 3075.931 4828.271 34363.51 Std Dev. 2202.777 4409.839 30268.27 Total Min 641.6957 580.001 1606 Max 18447.21 33397.65 569448 Obs. 5212 5212 5212 2002 Non-farm activities Mean 3581.795 6192.238 90606.57 Std Dev. 2319.123 4590.791 135475.1 Exporting industries Min 666.2547 908.5842 780 Max 18474.96 32929.32 1612400 Obs. 1882 1882 1882 Mean 3993.802 6906.522 99507.89 Std Dev. 2495.643 4745.421 139213.7 Import-competing industries Min 774.4517 877.4553 800 Max 17656.49 32483.17 1128750 Obs. 1715 1715 1715 Mean 4610.45 7149.777 122618.8 Std Dev. 2846.255 4654.243 182739.7 Non-traded non farm Min 776.3353 600.4697 330 Max 18206.18 32900.31 2690650 Obs. 8192 8192 8192 Farm activities Mean 2370.043 3881.12 37594.71 Std Dev. 1262.312 2637.034 50991.64 Rice Min 636.3497 592.4973 400 Max 16062.52 32126.7 1653200 Obs. 9992 9992 9992 Mean 2865.149 4745.447 87781.05 Std Dev. 1681.762 3278.303 108277.1 Main export crops Min 661.9562 697.8714 1100 Max 15316.62 32263.17 936000 Obs. 1181 1181 1181 137 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: continued Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 2002 Mean 2641.501 4309.072 42263.15 Std Dev. 1633.2 3023.609 58158.55 Other export crops Min 678.9702 683.3324 810 Max 16090.9 30092.4 1260690 Obs. 1129 1129 1129 Mean 2438.909 4152.161 41147.29 Std Dev. 1495.762 3064.595 68697.35 Import-competing crops Min 632.3506 766.8577 500 Max 12432.19 30170.09 1301850 Obs. 1712 1712 1712 Mean 3212.053 5584.73 80363.81 Std Dev. 1635.302 3774.486 88996.52 Non-traded food Min 781.5004 1139.245 1250 Max 10509.04 33041.25 759300 Obs. 501 501 501 Mean 3314.752 5368.969 75403.3 Std Dev. 2309.087 4038.796 128504.5 Total Min 632.3506 592.4973 330 Max 18474.96 33041.25 2690650 Obs. 26304 26304 26304 2004 Non-farm activities Mean 4194.41 7098.479 166570.5 Std Dev. 2373.103 4872.359 243364.5 Exporting industries Min 659.4932 1068.277 1300 Max 18009.55 31422.96 1600000 Obs. 567 567 567 Mean 4751.13 7275.56 209212.8 Std Dev. 2883.759 4727.285 289705.9 Import-competing industries Min 804.9464 1373.189 2000 Max 17426.08 31739.94 2048380 Obs. 506 506 506 Mean 5442.173 7799.166 240375 Std Dev. 3058.865 4758.865 309790.3 Non-traded non farm Min 762.8577 742.0001 600 Max 18538.53 32610.54 3400000 Obs. 2548 2548 2548 Farm activities Mean 2963.063 4482.403 67849.33 Std Dev. 1632.501 3113.673 114108 Rice Min 636.2792 662.9399 500 Max 15168.72 32610.57 2250000 Obs. 2891 2891 2891 138 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: continued Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 2004 Mean 3897.313 6512.622 152681.3 Std Dev. 2201.884 4347.045 174252 Main export crops Min 660.0689 723.0797 10000 Max 15519.49 30273.98 1980700 Obs. 379 379 379 Mean 3343.519 5054.932 90692.55 Std Dev. 2152.001 3978.047 143519 Other export crops Min 649.9424 618.9159 1500 Max 15193.62 29642.49 1262000 Obs. 372 372 372 Mean 2900.893 4627.119 69519.02 Std Dev. 1667.725 3135.846 108253.4 Import-competing crops Min 671.829 993.0854 2000 Max 10585.72 21311.11 1020000 Obs. 417 417 417 Mean 4114.86 5867.191 162244.5 Std Dev. 2018.181 4038.008 180554 Non-traded food Min 1184.327 878.6608 2000 Max 12254.45 26558.29 1039000 Obs. 140 140 140 Mean 4056.495 6091.663 147345.6 Std Dev. 2617.109 4341.264 235251 Total Min 636.2792 618.9159 500 Max 18538.53 32610.57 3400000 Obs. 7820 7820 7820 2006 Non-farm activities Mean 5484.575 8104.154 178169.8 Std Dev. 2910.144 4973.715 243101.4 Exporting industries Min 1267.986 1358.74 1800 Max 17637.29 31921.04 2014000 Obs. 561 561 561 Mean 5830.48 8496.445 225846 Std Dev. 2881.169 4785.573 280211.4 Import-competing industries Min 1176.05 1406.036 2800 Max 17756.82 32552.32 1643450 Obs. 519 519 519 Mean 6827.813 8997.766 264558 Std Dev. 3450.417 4902.625 319453.3 Non-traded non farm Min 930.5538 1295.668 417 Max 18586.1 33385.17 2400000 Obs. 2664 2664 2664 139 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: continued Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 2006 Farm activities Mean 3909.159 7039.552 83972.15 Std Dev. 2107.268 4862.857 116821.2 Rice Min 672.7744 1287.076 1500 Max 18482.16 33404.89 2400000 Obs. 3242 3242 3242 Mean 5693.544 8731.128 257178.7 Std Dev. 2959.028 5593.56 283022.7 Main export crops Min 1234.334 1582.989 8000 Max 17913.47 31283.97 2090000 Obs. 290 290 290 Mean 4698.161 6307.187 131835.8 Std Dev. 2419.178 3901.14 155912.8 Other export crops Min 779.4249 1259.98 2500 Max 13095.32 25989.33 1230200 Obs. 215 215 215 Mean 4671.11 6601.615 118328.4 Std Dev. 2424.836 4061.37 206149.1 Import-competing crops Min 1258.001 1334.528 5000 Max 17009.25 24843.68 2000000 Obs. 220 220 220 Mean 5629.526 7690.066 180524.9 Std Dev. 3147.224 5826.151 170270.9 Non-traded food Min 1548.176 1482.076 4400 Max 17925.98 31339.33 916500 Obs. 90 90 90 Mean 5276.399 7919.612 171695.1 Std Dev. 3079.31 4964.403 248729.6 Total Min 672.7744 1259.98 417 Max 18586.1 33404.89 2400000 Obs. 7801 7801 7801 2008 Non-farm activities Mean 7431.721 8807.368 305554.5 Std Dev. 3326.838 5019.76 392335.1 Exporting industries Min 1890.487 1399.764 3000 Max 18603.3 32504.56 3200000 Obs. 357 357 357 140 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.2: continued Real pc Current value of Trade sectors Statistics consumption Real pc income Assets/Durables 2008 Mean 7213.916 8786.651 306531.6 Std Dev. 3409.095 5256.447 362044.7 Import-competing industries Min 1202.683 831.3502 6000 Max 18455.94 32561.08 3006300 Obs. 584 584 584 Mean 7998.437 9280.396 362571.2 Std Dev. 3773.926 5262.812 449682.1 Non-traded non farm Min 1290.584 909.3856 2400 Max 18620.55 33084.13 3023950 Obs. 1151 1151 1151 Farm activities Mean 5315.16 7805.027 125349.8 Std Dev. 2800.208 5300.841 161537.2 Rice Min 682.2064 857.7307 1199 Max 18584.97 33315.67 2065000 Obs. 3032 3032 3032 Mean 7612.386 8280.179 389977.8 Std Dev. 3490.674 5379.078 376865.4 Main export crops Min 1485.559 1027.04 3000 Max 18552.71 28746.93 2118500 Obs. 328 328 328 Mean 6193.856 6795.463 162927.4 Std Dev. 3323.963 4636.765 190772.5 Other export crops Min 1199.062 954.6352 3388 Max 17675.93 31684.09 1530000 Obs. 369 369 369 Mean 5374.714 5834.104 152408.7 Std Dev. 3098.08 3888.351 223396.5 Import-competing crops Min 1300.961 1141.184 2200 Max 18198.8 29505.73 1803800 Obs. 384 384 384 Mean 6875.863 7992.48 283518.5 Std Dev. 3200.153 5353.364 343254.9 Non-traded food Min 1828.302 1525.087 4000 Max 16404.09 32152.04 1724500 Obs. 102 102 102 Mean 6300.213 8070.583 215784.7 Std Dev. 3373.366 5239.045 309481.4 Total Min 682.2064 831.3502 1199 Max 18620.55 33315.67 3200000 Obs. 6307 6307 6307 Note: All monetary values are in VN dongs. 141 Table A.3: Consumption estimates (1992-2008) 1992 1998 2002 2004 2006 2008 beta se beta se beta se beta se beta se beta s.e. Exante risk -0.601 0.022 -1.302 0.027 -0.900 0.011 -0.970 0.022 -1.586 0.028 -1.627 0.030 Pos shocks 3.078 0.103 4.710 0.094 5.715 0.058 5.480 0.102 6.113 0.108 5.869 0.121 Neg shock 2.382 0.097 4.039 0.104 5.881 0.052 5.574 0.109 6.175 0.108 5.776 0.116 Urban 0.377 0.042 0.358 0.021 0.309 0.005 0.278 0.013 0.182 0.017 0.111 0.015 Age 0.016 0.003 0.012 0.002 0.010 0.001 0.003 0.002 0.010 0.002 0.015 0.002 Sq. age 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 HH size -0.057 0.008 -0.080 0.008 -0.088 0.003 -0.079 0.007 -0.087 0.007 -0.095 0.008 Sq. HH size 0.003 0.001 0.004 0.001 0.005 0.000 0.004 0.001 0.004 0.001 0.005 0.001 No children -0.078 0.005 -0.099 0.005 -0.105 0.002 -0.113 0.004 -0.111 0.004 -0.115 0.005 Married 0.072 0.018 0.118 0.014 0.094 0.006 0.094 0.012 0.078 0.012 0.043 0.014 Head sex -0.057 0.016 -0.041 0.012 -0.070 0.005 -0.105 0.010 -0.052 0.010 -0.014 0.011 Primary education 0.152 0.014 0.118 0.016 0.157 0.004 0.180 0.009 0.172 0.009 0.205 0.010 Secondary education 0.263 0.016 0.297 0.018 0.281 0.005 0.329 0.010 0.300 0.010 0.339 0.011 142 Upper education 0.364 0.025 0.458 0.020 0.449 0.007 0.473 0.014 0.451 0.014 0.453 0.016 Tech education 0.339 0.021 0.473 0.025 0.576 0.008 0.596 0.013 0.588 0.013 0.564 0.015 Univsity education 0.591 0.047 0.763 0.028 0.792 0.009 0.865 0.016 0.806 0.015 0.854 0.018 Roads -0.013 0.023 0.092 0.018 -0.024 0.006 0.081 0.011 0.023 0.012 0.022 0.013 Electricity 0.029 0.028 0.229 0.021 0.143 0.006 0.133 0.020 0.323 0.021 0.185 0.030 Water 0.126 0.039 0.171 0.022 0.134 0.005 0.048 0.012 0.069 0.016 0.041 0.014 Transport -0.017 0.014 0.032 0.011 0.051 0.004 0.055 0.008 0.057 0.008 0.052 0.008 Constant 6.199 0.086 6.082 0.075 6.847 0.028 7.176 0.059 6.254 0.066 6.670 0.073 No Obs. 4222 5446 27140 8117 8162 6702 Province Dummies yes yes yes yes yes yes F 29376.991 53429.174 334587.8804 87373.78528 94107.34282 77907.02981 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Prob > F 0.000 0.000 0.000 0.000 0.000 0.000 R-squared 0.998 0.999 0.999 0.999 0.999 0.999 Adj R-squared 0.998 0.999 0.999 0.999 0.999 0.999 Root MSE 1.898 1.917 1.915 1.879 1.914 1.893 Note: Feasible Generalized Least Squares (FGLS) coefficients. TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.4: Income regressions (1992–2008) dep.variable: log of real per capita income 1992 1998 2002 2004 2006 2008 Demographic characteristics 0.000682 0.0173a 0.0103a 0.00204 0.0128a 0.0176a Age of the household head (0.908) (0.001) (0.000) (0.574) (0.000) (0.000) 0.0000117 -0.000136a -0.0000836a -0.0000341 -0.000114a -0.000167a Age2 of the household head (0.841) (0.006) (0.000) (0.325) (0.000) (0.000) -0.0147 -0.0373b -0.0747a -0.0256b -0.0619a -0.0668a Household Size (0.496) (0.048) (0.000) (0.037) (0.000) (0.000) 0.000731 0.00204 0.00422a 0.000954 0.00336a 0.00380b Household Size2 (0.637) (0.181) (0.000) (0.344) (0.003) (0.029) -0.0872a -0.113a -0.118a -0.118a -0.110a -0.117a No. of Children (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 0.0347 0.134a 0.100a 0.0782a 0.113a 0.117a Married Head (0.326) (0.000) (0.000) (0.001) (0.000) (0.000) 0.00275 -0.00864 -0.0396a -0.0598a -0.0459b -0.0401+ Head sex (0.937) (0.757) (0.000) (0.004) (0.013) (0.106) Education 0.119a 0.103a 0.131a 0.125a 0.141a 0.161a Primary education (0.002) (0.005) (0.005) (0.000) (0.000) (0.000) 0.206a 0.280a 0.228a 0.238a 0.244a 0.287a Lower secondary education (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 0.282a 0.424a 0.355a 0.288a 0.310a 0.384a Upper secondary education (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 0.213a 0.349a 0.437a 0.381a 0.423a 0.450a Tech/voc education (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 0.305a 0.559a 0.569a 0.550a 0.517a 0.640a University (0.001) (0.000) (0.000) (0.000) (0.000) (0.000) Occupation 0.0898 0.223a 0.0914a 0.103a 0.112a 0.133a White-collar (0.156) (0.000) (0.000) (0.001) (0.000) (0.000) 0.267a 0.182a 0.110a 0.0343 0.0618b 0.00921 Personal services (0.000) (0.000) (0.000) (0.228) (0.016) (0.757) 0.106c 0.0286 0.0286b -0.0165 -0.0240 -0.00320 Production (0.052) (0.436) (0.024) (0.478) (0.269) (0.930) -0.00468 -0.0462 -0.0173 -0.0830a -0.0604b -0.0179 None (0.913) (0.193) (0.230) (0.003) (0.035) (0.587) Village characteristics -0.0574 -0.0840 0.0785a 0.0114 -0.0538c -0.0987a Urban (0.655) (0.240) (0.000) (0.695) (0.083) (0.007) -0.0853 -0.0326 -0.0452b 0.0606b -0.0611b 0.00751 Roads (0.258) (0.613) (0.038) (0.014) (0.026) (0.804) 0.0533 0.315a 0.110a 0.0843 0.302a 0.201a Electricity (0.445) (0.000) (0.000) (0.150) (0.000) (0.006) 0.117 0.105 0.0719a 0.0162 0.0341 0.0616c Water (0.314) (0.185) (0.000) (0.518) (0.231) (0.051) Transport 0.00384 0.0308 0.0408a 0.0269c 0.0373b 0.0416b 7.627a 6.959a 8.114a 8.743a 8.121a 8.420a Constant (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) 143 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Table A.4: Income regressions (1992–2008) continued dep.variable: log of real per capita income 1992 1998 2002 2004 2006 2008 Dummies for trade categories (0.933) (0.437) (0.003) (0.094) (0.016) (0.015) 0.0144 -0.0550 -0.0337c 0.0231 -0.00539 -0.00468 Exporting industries (0.857) (0.380) (0.081) (0.435) (0.845) (0.887) 0.0457 0.101c 0.0471b 0.0361 0.0661b 0.0506c Import-competing industries (0.463) (0.054) (0.011) (0.197) (0.013) (0.072) -0.265a -0.293a -0.260a -0.273a -0.0573b -0.0000530 Rice (0.000) (0.000) (0.000) (0.000) (0.013) (0.998) 0.182 0.136 -0.103a -0.0208 0.0811c -0.0187 Main export crops (0.264) (0.243) (0.000) (0.636) (0.086) (0.706) -0.162c -0.275a -0.198a -0.214a -0.231a -0.202a Other export crops (0.070) (0.000) (0.000) (0.000) (0.000) (0.000) -0.223b -0.0979 -0.186a -0.227a -0.125a -0.204a Import-competing crops (0.016) (0.016) (0.000) (0.000) (0.005) (0.000) 0.0609 0.0715 -0.0839b -0.193a -0.188a -0.134b Non-traded food (0.660) (0.536) (0.019) (0.001) (0.009) (0.041) Province Dummies Yes Yes Yes Yes Yes Yes Adjusted R2 0.263 0.357 0.427 0.353 0.296 0.305 Obs 3377 5212 26304 7820 7801 6307 a p<0.1 b p<.05 c p<0.1 144 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix B: Methodology for empirical analysis a Further details on the computation of this measure for different time horizons will be provided later on. For additional details see Pritchett et al. (2000); Christiaensen and Subbarao (2005); Chaudhuri and Datt (2001); Chaudhuri et al. (2002) and Chaudhuri (2003); Kamanou and Morduch (2004); Gunther and Harttgen (2009). b In practice, the stochastic nature of consumption is acknowledged by assuming that there is heterogeneity in consumption volatility around the mean. Thus, it addresses the issue of heter- oskedasticity by using a 3-steps Feasible Generalized Least Squares (FGLS) econometric procedure suggested by Amemiya (1977). 145 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES c Note that the lack of panel data prevents us from exploiting the time dimension. Hence, we are assuming the unexplained component of income in cross-section data in Eq. B.3 to proximate stochastic innovation. This is not unreasonable: while it is true that the unexplained component also contains non-stochastic unobservables as well as measurement error, it is not necessarily true for the variances of income innovations within sub-samples of households grouped according to their trade openness position. d For identification purposes, the occupation characteristics are assumed to influence consumption behavior only through income. e According to Skinner (1988) and Guiso et al (1992), the exponent of the scaling factor measures the sensitivity to the level of expected wealth exhibited by the reaction to uncertainty. If the exponent is more than zero, the effect of risk on consumption declines with the household’s resources and the decline is faster the higher the value. Usually, the adopted value is two and this is why we use the square of that ratio. f The current value in thousand dong of the households’ fixed assets and durable goods has been used as a proxy for wealth in the denominator of the scaling factor. Robustness checks using alternative proxies for wealth such as the linear combination of the principal component factors or observed consumption have been implemented. They show the same pattern, suggesting that the negative relationship between ex-ante risk and consumption volatility seems to be robust to alternative empirical proxies for wealth. 146 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 147 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Glass Barriers: Constraints to Women’s Small-Scale, Cross-Border Trade in Cambodia and Lao PDR 1 Marlon Seror, Richard Record and Julian Clarke, World Bank Group 1. Introduction a positive welfare impact on poor households beyond those directly involved in this activity (World Bank, 2011). B order checkpoints in developing countries Third, trade offers a way for women to earn money often teem with traders transporting small outside the household, which may foster empowerment. quantities on foot or pushing carts alongside In this context, trade facilitation projects are traditionally trucks that sport the insignia of formal built on the expectation that the automation, streamlining companies. Those small-scale, cross-border traders may and simplification of procedures2 will foster economic eventually be superseded by larger import-export firms. activity and eventually reduce poverty. Small-scale cross- But during the process of development, their trade may border traders, including informal, female and other be a valuable avenue for poverty alleviation and women’s categories of potentially vulnerable traders, may benefit empowerment. This chapter focuses on the latter in the at the margins of such projects, e.g., from improvements context of small-scale, cross-border trade in Cambodia in transparency. However, they carry small quantities and Lao People’s Democratic Republic (Lao PDR). It and may fall under customs declaration thresholds. analyzes recent survey research undertaken by the They are poorly educated and thus cannot cope with the World Bank and draws conclusions about the key policy administrative tasks demanded of formal firms, and their implications for facilitating the poverty-reducing impact profit margins may be so thin that compliance with the of women’s participation in small-scale, cross-border same customs duties and other border procedures facing trade. firms would prevent them from trading at all (Lesser Small-scale, cross-border trade (SSCBT) is thought and Moisé-Leeman, 2009). Trade policy in developing to provide several benefits to developing countries. countries thus tends to focus on large, formal firms and First, the literature emphasizes its importance as a firms that might consider going formal, even though source of employment and financial resources for poor many traders are unlikely to formalize in the medium run. smallholders and landless households, particularly on In Cambodia and Lao PDR, the two countries on which a country’s geographical (and often socioeconomic) this chapter focuses, women tend to be overrepresented fringes. Second, SSCBT plays an important role in in unpaid family labor, while wage-earning jobs are mostly reducing price differences and volatility, thus having 1 The research undertaken for this chapter was supported by the Trade Development Support Program in Cambodia, the Second Trade Development Facility Multi Donor Trust Fund in Lao PDR, and the Umbrella Facility for Gender Equality. 2 See World Bank (2012a) for Cambodia, and EMC (2012) and World Bank (2014a) for Lao PDR. 148 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Women face specific challenges in SSCBT. Besides the “crushing weight of family responsibilities,” women are more likely to face capital constraints, market smaller quantities and have difficulties accessing information on market opportunities. 149 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Our findings highlight the Cambodian-Thai border (Kusakabe et al. 2008) and long-distance traders between Lao PDR and Thailand that although they (Walker 1999). In Africa, small-scale, cross-border trade is largely carried out by women (World Bank 2011). do not interpret it Yet women face specific challenges in SSCBT. Besides the as gender-based “crushing weight of family responsibilities” (UN Women 2012), women are more likely to face capital constraints, discrimination, market smaller quantities and have difficulties accessing information on market opportunities (World Bank 2012c). women are found to Women are also more likely to be illiterate, which restricts their access to, and knowledge of, trade policies and suffer from a higher procedures (USAID 2012) and thus further limits business development. Women often have to hire brokers, which tax rate and a tax eats into their profit margins, or seek assistance from officials, who are predominantly male and not trained schedule that deters to work in gender-sensitive environments (World Bank 2012b). This may fuel extortion and even harassment, as them from upgrading shown in East Africa (World Bank 2011, UN Women 2012). to more profitable Whereas this study does not find women reporting the dramatic level of abuse highlighted in the East African cross-border trade context, women may face binding constraints in their activity as small-scale, cross-border traders. These activities. challenges may be “visible” and acknowledged by (at least some of) the actors in the border economy, e.g., discriminatory tariffs or gender-based violence. The barriers to SSCBT that women face may also be “invisible,” i.e., not recognized by those actors as related to gender, or indirectly—through regulations, norms, infrastructure, etc., that adversely affect women—constraining women’s participation in cross-border trade. taken up by men (UNIFEM, WB, ADB, UNDP and DFID/UK The contribution of this study is thus to shed light on 2004, World Bank 2012b). Since women’s employment the obstacles, both visible and invisible (“glass barriers” opportunities are often limited by cultural norms, to trade), that prevent women from making the most of restrictions on mobility for safety reasons and household SSCBT for income generation and empowerment. To this responsibilities, the fact that trade is considered an end, we rely on an innovative mix of original qualitative acceptable occupation for women in the Mekong sub- and quantitative data to both voice the concerns of the region (ibid.) makes cross-border trade a valuable avenue actors in the border economy and econometrically detect for women’s empowerment. Lao female traders, for constraints that they fail to perceive or would not express. instance, were found in an early study to often earn more Our combination of qualitative and quantitative data than their husbands (Walker 1999), which may allow further allows us to infer the constraints faced by women them to gain financial independence. Prior to this study, who selected out of SSCBT. Following the metaphor of women were known to dominate some subcategories of Hausmann, Klinger and Wagner (2008), we shall strive traders in the Mekong sub-region, e.g., fish traders across to voice the concerns of both “camels” (women actually 150 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES participating in cross-border trade—or the “Sahara tax schedule that deters them from upgrading to more desert”) and “hippopotami” (who would, but cannot, profitable cross-border trade activities. Along with capital engage in this activity—are absent from the “Sahara”— constraints, this may explain the lower share of women in and thus do not appear in our quantitative data). small-scale, cross-border trade than among own-account workers and the self-employed. Our study first documents that in contrast to other parts of the world female traders in the Mekong sub- The structure of this chapter is as follows. In the next region seldom report abuse and gender-based violence section, we present the study design and methodology or discrimination; yet women are underrepresented in for data collection. In Section 3, we provide an overview of small-scale, cross-border trade despite a potential for the border economy in Cambodia and Lao PDR. Section 4 expansion and their dominance in trade and services then investigates gender-related constraints to women’s away from border checkpoints. We next establish that small-scale, cross-border trade. Section 5 discusses the poor infrastructure is a key challenge for traders and results and delineates policy implications. acts as an “invisible” source of discrimination, women being more time constrained and thus disproportionately 2. Study Design and Methodology affected. Our findings further highlight that although they do not interpret it as gender-based discrimination, The profiles of and challenges faced by the women women are found to suffer from a higher tax rate and a and men who deal with border authorities for a living 151 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES cannot be easily described, given the dearth of data on • In all checkpoints visited for preliminary the topic in Cambodia and Lao PDR. Neither country observations, stakeholder interviews were carried holds a register of small-scale, cross-border traders, as out with border agency (customs, immigration, they usually operate only with the documents necessary Camcontrol, etc.) staff and management; both to enter the neighboring country. This may entail small- and large-scale, formal and informal traders registration with a local government agency, but these and brokers; transporters who do not act as brokers; traders are seldom registered as foreign traders with a and various border users and local dwellers. Based central ministry. Besides, informality often carries stigma, on the preliminary observations and stakeholder which means that they may be reluctant to acknowledge interviews, three checkpoints were selected their line of business. for further study: Bavet (Svay Rieng province, Cambodia), Poipet (Banteay Meanchey province, Given the lack of a list of small-scale, cross-border traders Cambodia) and Vangtao (Champasak province, Lao and brokers, an innovative mix of survey strategies was PDR), on the Cambodian-Vietnamese, Cambodian- implemented. Three major challenges were involved in Thai and Lao-Thai borders, respectively (Map 1). collecting data on the population of interest: (i) making sure that interviewees are indeed involved in small- In each of the selected checkpoints—Bavet, Poipet scale, cross-border trade; (ii) establishing a list of border and Vangtao—more detailed qualitative data crossers to get an accurate picture of trade patterns were gathered through focus group discussions and the population; and (iii) drawing a sample of border (FGDs). They consisted of open questions about crossers from that list to gather representative data. small-scale, cross-border trade patterns and the people involved in them. FGDs are helpful to The following approaches were implemented for data understand the overall picture of small-scale, collection: cross-border trade through traders’ and brokers’ • Preliminary observations were made at various experiences, as well as through the eyes of those checkpoints in Cambodia and Lao PDR, on the discouraged from engaging in SSCBT. Focus groups borders with Thailand and Viet Nam—see Table 1. separated women and men to build trust and elicit Preliminary observations were meant to: (i) select truthful information about gender-specific issues. economically important, diverse and typical border Representativeness was an essential aspect of checkpoints to include in the study; (ii) identify each group. The information obtained was used to research questions for further investigation; and determine the data collection strategy and refine (iii) define the survey methodology and inform the questionnaire for the quantitative part of the survey instruments. study. Table 1: Checkpoints visited for preliminary observations and stakeholder interviews Cambodia Lao PDR Poipet international checkpointa Vangtao International checkpointa Small, bilateral checkpoints in Poipet Paktaphan Border with Thailand Malai Daung International Port Bavet International checkpointa Dansavanh International checkpointb Border with Viet Nam Small, bilateral checkpoints in Bavet Small bilateral checkpoint near Dansavanh Srmo checkpoint a  Checkpoints selected for IDIs and FGDs. The study was piloted in Dansavanh, but this checkpoint was not retained because all small-scale cross-border traders and brokers there are Vietnamese and reside in Vietnam. b  This raised difficulties in terms of logistics and legitimacy since the counterparts for this study are the Cambodian and Lao governments. Moreover, Vietnamese crossers were reluctant to cooperate with the survey team, presumably because many of them are brokers although brokers should be Lao nationals or permanent residents (Financial ministry of the Lao PDR 2005). 152 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Map 1: Checkpoints selected for focus group discussions and in-depth interviews reports may be biased, as involvement in SSCBT can be sensitive. Therefore, we decided to implement a census of all border crossings corresponding to our definition of SSCBT—see Section 3—for 2–3 days. Basic information about the crosser, her role in the crossing and the shipment were recorded in a sampling frame used to randomly sample respondents for in-depth interviews. Based on our qualitative data, we define our population of interest as follows: brokers or traders who deal with authorities themselves and are involved in small- scale trade, i.e., in the trade of goods that cross the border in human-powered vehicles or vehicles with fewer than four wheels. The rationale for this definition is made explicit in Section 3. • The second stage consisted of in-depth interviews (IDIs), which were first piloted at all shortlisted checkpoints. The IDIs provided detailed information on both border crossers (demographics, education, past experiences as a trader/broker, perception of challenges, etc.) and crossings (goods transported, purchase value, selling price, etc.).4 Sample sizes were 55 for Bavet, 55 for Poipet and 48 for Source: Map data © 2015 Google. Text and lines in red added by the authors. Vangtao. Respondents for IDIs were selected from • A two-stage quantitative data collection approach the sampling frame through stratified random was adopted. First, a sampling frame was sampling to ensure sufficient sample sizes for cross- established—at the border gate proper3—to get group comparison, in particular across gender.5 a clear and accurate picture of SSCBT patterns. Since the sampling frame is an exhaustive list of Randomly sampling households in villages near shipment crossings, the sample was representative the checkpoints was ruled out based on qualitative of crossings and populations (as defined in Section information, as some traders travel long distances. 3) at the selected checkpoints.6 Moreover, as highlighted in the literature, self- 3 Border checkpoints are the natural place to conduct surveys of small-scale traders: all goods traded across the border, wherever they are produced, bought or sold, must cross the border at some point. Qualitative data indeed made it clear that a negligible share of small-scale cross-border trade is carried out outside checkpoints, as goods must then be carried on foot, which inflates transportation costs. There are however informal routes within checkpoint zones, small by-roads that are less thoroughly monitored by border officials. Our sampling design captures those routes. 4 For the sake of comparability, some of the IDI questions were inspired by the surveys carried out by World Bank (2011) and UN Women (2012). Our survey instruments are available upon request. 5 The variables used for stratification, carried out at the checkpoint level, include the role of the crosser in the shipment (i.e., trader and broker), gender and nationality. Sampling weights were computed based on the stratum-specific probability of being sampled. The results presented in the chapter are weighted to restore representativeness. 6 One important caveat is seasonality. First, we are confident that the days on which the census was carried out were typical, i.e., did not coincide with or fall near any holiday. Since virtually all traders and brokers work at the border every day, they are not seasonally selected. It is however possible that seasonal traders work during holidays, and our conclusions do not apply to them. Second, it is possible that some goods are seasonal, e.g., agricultural produce. Traders however specialize in one type of goods and trade in those goods all year. Traders who sell vegetables will thus sell different vegetables (e.g., cabbage vs. carrots) around the year but not different types of goods. 153 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 3. Overview of the border economy Declaration Document (ACDD). They instead fill out a simplified customs form, called “Customs Declaration 3.1 Definition of the population of interest Form for Retailed Declarants” (or “Customs Regime Form Small-scale cross-border trade is an elusive concept. 44”) in Lao PDR.7 “Informality” thus rather comes from the Different definitions have been used in the literature, way duties are applied by officials: For “small” shipments, different rules apply depending on the country, the value duties are more likely to be negotiated or estimated by and quantity of goods traded per crossing may vary from rules of thumb than dutifully calculated, as is usually one checkpoint to the next, and SSCBT includes a variety the case for “large” shipments. Field observations and of actors. Thus, we need to develop an alternative, unified stakeholder interviews revealed that officials determine definition of SSCBT. to which category a shipment belongs (“large” or “small”) based on the means of transportation. The rationale The literature proposes a variety of definitions that behind this is probably that means of transportation is revolve around shipment value or the degree of formality readily observable and a good predictor of shipment size of trade activities. UN Women (2012) considers that “all and value. revenue-generating cross-border commercial activities with a daily transaction value of less than 100 U.S. dollars Based on field observations and stakeholder interviews, (USD) per trader” qualify as “small-scale” and that traders the population of interest is thus defined by two criteria: are “informal” if they are not registered and pay no • People who deal with authorities. These comprise: income taxes, although they might pay export or import (i) traders who do not hire brokers and thus pay taxes, and pass through official border crossings with taxes and fees and interact with border authorities appropriate travel documentation. World Bank (2011) in general themselves; and (ii) brokers who do that defines informal trade as “unorganized small-scale on behalf of traders.8 Transporters who do not act trade which does not appear in the customs record.” It as brokers and traders who do not interact with may however be “official” in the sense that “traders go authorities are not included in the population of through official border posts, pay a crossing fee to the interest. This criterion is meant to capture those who immigration office, and if processed appropriately pay a are effectively involved in cross-border trade and duty on imports” (ibid.). thus the most directly affected by border conditions. Such definitions are impracticable in our context for two It is grounded in the qualitative interviews and reasons: (i) small-scale cross-border trade is not “informal” observations carried out in the first phase of the in the case of Cambodia and Lao PDR by any of the usual study, which revealed a clear divide between definitions; and (ii) we had to create a sampling frame absentee traders and people present at the border, by implementing a census of crossings, and the constant while there is some overlap between own-account flow of crossers prevented us from asking detailed traders and brokers—see Figure 1 in the next section; questions (about the value of the shipment and formality and of the transactions) to determine eligibility. • People involved in “small-scale” goods trade9 All traders and brokers in Cambodia and Lao PDR indeed (as determined by the first criterion) and who cross fill customs forms and are subject to duties. Small- the border in human-powered vehicles or vehicles scale traders and brokers, legally defined in terms of with fewer than four wheels. All the checkpoints registration with the relevant ministries or shipment selected for this study indeed have clear (informal) value (typically, below USD 100) are exempted from rules to distinguish between “small” and “large” trade a full customs declaration using the ASEAN Customs based on the type of vehicle used. Trucks are always 7 The use of a Customs declaration shows that small-scale trade is not synonymous with an evasion of duties or legal requirements. 8 Brokers and traders who are exempted from taxes and fees but would interact with authorities if controlled are part of our study population. 9 Preliminary observations and stakeholder interviews revealed that the trade of services (hairdressers, housekeeping, etc.) constitutes a very marginal activity at the studied checkpoints. We therefore exclude such traders from the population of interest. 154 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES considered as large, carts or motorbikes as small.10 contractual forms.12 Qualitative evidence highlights a Whereas this criterion might not be relevant outside watershed in the SSCBT population between traders and Cambodia and Lao PDR, it is suitable for our setting transporters-brokers on the one hand and “absentee” as it (i) facilitated the establishment of a sampling traders on the other. frame, (ii) is in line with officials’ rules of thumbs and Traders rarely engage in brokering, i.e., dealing with (iii) is predictive of the “informality” of taxes. authorities on behalf of other traders, and never work In what follows, the population of interest is referred to only as transporters, while brokers rarely trade on their as “traders and brokers” or “SSCBTers.” own accounts but are often hired as transporters, i.e., carrying goods across the border but not dealing with 3.2 Structure of the small-scale, authorities. There is no overlap between “absentee” cross-border trade population traders and either cross-border traders or brokers. Figure 1 illustrates this with a simple Venn diagram. In 6% of cases This definition reflects a striking feature of border brokers also trade on their own accounts. No trader who economies in the Mekong sub-region: the structural deals with authorities and was thus eligible for sampling divide between “absentee” traders (who are not was found to hire brokers. This implies that the traders considered cross-border traders),11 own-account traders who hire own-account traders as occasional brokers are and transporters-brokers, and the differences in this “absentee” traders who never cross the border. structure across checkpoints. An arrangement between a trader and broker can also assume one of several Qualitative evidence hints at the importance of overall Figure 1: SSCBT Is divided in three distinct activities despite some overlap between brokers and traders 6% (5%) of brokers (traders) also act as traders (brokers) Own-account traders Brokers “Absentee” traders 10 Private cars, tractors, pick-up trucks or minivans are in a “grey zone”: They cannot transport goods across the border without going through specific procedures or with a fee that SSCBTers would seldom accept to pay, preferring other (smaller) means of transportation.. 11 Whereas eliciting information from “absentee” traders would yield interesting information about trade patterns, the determinants of informality and the choice of hiring a broker, qualitative information and pilot experience made it clear that they could not be contacted through their brokers. Brokers themselves often deal with intermediaries, typically cart owners in Poipet, who do not deal with authorities and do not handle the goods at any time but rent carts to several brokers and are contacted by traders. 12 Transporters-brokers are mostly (93% of crossings) remunerated based on how much they have to transport, which is determined either in kilograms or by the number of items; the rest are paid a lump sum. Different contractual arrangements are available and were observed in the field: (i) brokers may keep whatever they can save on taxes and fees (brokers are remunerated in this manner for 91% of crossings) or give back all savings to the trader (8%); and (ii) they may be responsible for the goods in case of confiscation (76%) or not (15%), or share the responsibility (10%) according to idiosyncratic agreements, e.g., depending on whether forbidden goods are concealed in the shipment. 155 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES traffic as an explanatory factor for a predominance of that are sold to wholesalers and retailers, whereas own- brokers at a checkpoint. Busier checkpoints can indeed account traders typically sell their goods directly to final lead to delays, and traders therefore incur significant consumers, which implies thicker margins.13 losses. This fuels demand for specialized transporters- Brokers’ lack of knowledge of the local market is partly brokers who know how to get heavy carts across the explained by their higher probability of being migrant border faster, where to stop for dealing with authorities workers from other provinces, which also means that and how to minimize taxes and fees. The volume of trade, they are more vulnerable to changes in the local legal both by large and small firms, is much larger in Poipet than environment.14 In the checkpoints visited, SSCBTers were in the other two checkpoints. Therefore, intermediaries always Cambodian and Lao on the border with Thailand specialized in getting relatively large quantities of goods whereas at least a significant minority of SSCBTers was through a congested checkpoint are much needed in Vietnamese at checkpoints on the border with Vietnam, Poipet, and most SSCBTers there are brokers (Figure preponderantly from border regions. Only 5% of Vangtao 2). Conversely, brokers are almost absent in Bavet and traders were born and live in different places, and none completely absent in Vangtao. of them was born in a different province. About one-fifth Being a broker is less desirable than being a trader. Besides (19%) of SSCBTers in Bavet and a majority (86%) in Poipet being more physically demanding, brokering implies were born in a different province than the one they interacting with border authorities, negotiating taxes and currently live in. sometimes smuggling illegal or high-tax goods (illegal or undeclared goods may constitute part of the shipment). 3.3 Gender composition of the small-scale, Depending on the contractual arrangement, brokers may cross-border trade population be responsible for confiscated goods. As shown by our Checkpoints differ widely in their shares of female qualitative data, brokers would become traders, had they SSCBTers. Overall, 41% of crossings are performed by better access to capital and knowledge of local demand female SSCBTers. Figure 3 shows that this share varies and supply. The quantitative survey indeed confirms by checkpoint. It is higher in Vangtao, where 60% of that brokers are more likely to take care of shipments Figure 2: The composition of SSCBT by activity differs Figure 3: Share of crossings by gender significantly across checkpoints and gender 100 80 70 80 Share of crossings (%) Share of crossings (%) 60 60 50 40 40 30 20 20 10 0 0 Bavet Poipet Vangtao Men Women Bavet Poipet Vangtao Own-account traders Male SSCBTers Brokers Female SSCBTers Source: Authors’ calculations. Source: Authors’ calculations. 13 There are also stark differences across checkpoints. Goods are mostly sold to final consumers in Bavet (74% of crossings) and Vangtao (71%), followed by retailers, 25% and 22%, respectively. In Poipet, goods are sold to wholesalers in 38% of crossings, to retailers in 35% and to final consumers in 22% of crossings. 14 In Poipet, fees for the necessary “immigration card” increased dramatically for non-local residents of Banteay Meanchey province six months prior to the study. The process had also become stricter, as birth and registration certificates were required. The regulatory change was too recent to assess whether it was generating informal arrangements or whether transporters-brokers just accepted the hike. We expect little room for negotiation, as the card is issued by the Thai police and the relationship between Cambodian border users and Thai officials is notoriously poor. Moreover, the interviewees never mentioned this as an issue unless we specifically asked about travel documents. 156 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES crossings are done by women, and lower in Bavet (37%) Figure 4: The share of women in SSCBT is lower than in comparable job categories and Poipet (29%). Reasonable assumptions 15 about sampling suggest that 68% of the SSCBT population is female in Vangtao, 41% in Bavet and 32% in Poipet. The 100 discrepancy between the shares of women in crossings 80 Share of women (%) and in the population of border crossers is an indication 60 of a lower crossing frequency among female SSCBTers. 40 Those differences partly reflect the structure of the 20 SSCBT population. Women are indeed underrepresented 0 among brokers. Figure 3 shows that 79% of female Estimated Estimated Own-account/ Estimated Self-employed SSCBT SSCBT Self-employed, SSCBT in trade, SSCBTers are own-account traders, as against only 56% of population population nationwide population Champasak province Bavet Poipet their male counterparts.16 Only 25% of brokers’ crossings CSES (2013) Vangtao LFS (2010) Male Female are carried out by women, as against 50% of own-account Source: Authors’ calculations. traders’ crossings. Female traders however resemble brokers more than male traders in one important respect: They are much more likely to sell goods to wholesalers and of physical sexual harassment in the Cambodian and retailers than to final consumers, which hints at thinner Lao checkpoints surveyed. Great care was taken to elicit profit margins and may signal fewer trade opportunities. truthful answers about such a sensitive topic as gender- The share of women among small-scale, cross-border based verbal and physical abuse. Female interviewers traders and brokers is lower than among own-account were recruited to carry out IDIs and moderate FGDs. All workers and the self-employed in the country. The interviewers were trained to ask gender-sensitive survey nationally representative 2013 Cambodia Socio-Economic questions in a non-judgmental manner, minimize report Survey (CSES) shows that in Cambodia 54% of the own- bias and write down comments for field supervisors account or self-employed workers are women (National when they suspected reticence. We find no report of Institute of Statistics, Ministry of Planning, Kingdom of physical sexual harassment in the IDIs, which is consistent Cambodia, 2014). This rough comparison suggests that with FGDs and stakeholder interviews. SSCBTers the proportion of women in SSCBT is lower than we would however mention verbal harassment targeting women, expect from looking at jobs in the same broad category. in particular authorities’ insistent questions “for the Similarly, Lao PDR’s 2010 Labor Force Survey (LFS) can be purpose of flirting” and gender-specific insults, either used to compare the gender composition in SSCBT with discriminatory or with sexual innuendos.18 The relatively that among the self-employed in wholesale and retail safe situation of female SSCBTers at Cambodian and Lao (but not necessarily small-scale, cross-border) trade in the checkpoints is reassuring. Nevertheless, the lower share same province. We find as well that the share of women is of women in SSCBT than in other trade-related self- lower among Vangtao traders—see Figure 4. 17 employment remains a puzzle and may hint at constraints preventing women from entering SSCBT. The object of Contrary to previous studies, e.g., World Bank (2011) and Section 4 is to shed light on the constraints that women UN Women (2011) in East Africa, there are no reports face in this activity. 15 These assumptions include: (i) Crossers who cannot be uniquely identified (e.g., because of a missing phone number) are different crossers. This is reasonable given that qualitative and quan- titative evidence suggest most SSCBTers cross daily. (ii) The SSCBTers active during the survey period are similar to the general SSCBT population. Qualitative information and pilots suggest that most SSCBTers are active all year round. Seasonal crossers are not captured by the study design. Sampling was carried out and IDIs fielded in early September 2014 in Cambodia and early November 2014 in Lao PDR. (iii) No trade occurs outside sampling hours, which were set to avoid missing any crosser. We found that little or no trade occurs outside official opening times. (iv) No trade occurs outside the official checkpoint (i.e., “round the gate”), which is supported by qualitative evidence. 16 The IDI sample was designed to ensure that at least half the respondents were women. The purpose was to maximize our ability to detect statistically differences between women and men despite a small sample size. Sampling weights are thus used systematically in the results presented in this chapter. 17 The LFS and IDIs were fielded four years apart. But, if anything, we would expect female participation in expanding activities, such as own-account trading, to have increased in recent years. 18 The frequency of such reports is not statistically significantly different between female and male interviewers. 157 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 3.4 Self-selection and economic potential for socioeconomic status by asset ownership. Detailed information about 20 assets was gathered in the IDIs Stark differences can be noticed within the population of and a wealth score computed using principal component interest in terms of income from SSCBT. As can be seen analysis (results not reported). We find that traders’ from Figure 5, SSCBT income is much higher in Bavet; it households are significantly wealthier than brokers’, and is lowest in Poipet.19 Mean income from SSCBT is always female SSCBTers are significantly wealthier than male higher for women, but median income is often lower than SSCBTers. This may be the upshot of positive selection of for men.20 The difference between mean SSCBT incomes women into SSCBT based on unobserved characteristics, by gender is statistically significant only in Vangtao. The e.g., knowledge of the market, entrepreneurial qualities, Lao checkpoint is also the only one where median SSCBT etc., which in turn may be evidence of specific challenges income is larger for women than for men. The absence that women have to cope with, leading to the exclusion of significance and reversal of patterns between mean of more vulnerable “hippos” from the “Sahara.” The and median incomes in the Cambodian case come from analysis also suggests that being a trader is a preferable the higher dispersion of female SSCBT incomes. Income or more sought-after activity and confirms that SSCBT is more unequal for women than for men in Vangtao as can be a valuable source of revenue for women. Male and well. female SSCBTers also differ in the role that SSCBT income plays in their households. Although noisy estimates often The higher dispersion of female SSCBT income may hint result in a lack of statistical significance, household at untapped economic opportunities. To the extent income is always more reliant on SSCBT earnings in that moderate income inequality signals a potential for female than in male SSCBTers’ households—see Figure upward mobility, SSCBT may be a worthwhile avenue 6. This highlights the importance of SSCBT for female- for income generation, in particular for women. The headed households, which are typically more financially coefficient of variation of SSCBT income is highest for vulnerable, and the fact that female traders are often the women in Bavet, where it is twice as large as for men. primary breadwinners in their households, so that trade Data on household-level socioeconomic status provide may be instrumental in empowering women. further suggestive evidence of a higher earnings Participation in SSCBT is further associated with a higher potential for women in SSCBT. It is also consistent with household socioeconomic status for women but not positive selection of women into this activity. We proxy Figure 5: SSCBT income varies widely by checkpoint and gender SSCBT income in past 12 months (USD) 200 8,000 SSCBT income in past 7days (USD) 7,000 150 6,000 5,000 100 4,000 3,000 50 2,000 1,000 0 0 Male Female Male Female Male Female Male Female Male Female Male Female SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers SSCBTers Bavet Poipet Vangtao Bavet Poipet Vangtao Mean Median Mean Median Source: Authors’ calculations. 19 This holds true whether we look at SSCBT income in the past 7 days, which is arguably a more accurate but perhaps dispersed measure, and or in the past 12 months, which is potentially more subject to measurement error. 20 Conversely, we find that gross profit as a share of the total purchase value of the shipment is at the same level for male and female traders. Differences in total income are thus not due to a higher profit rate for male traders. 158 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 6: Female SSCBTers’ households rely more heavily on into asset ownership, for women as small-scale, cross- SSCBT income than male SSCBTers’ border traders. Data limitations make it impossible 95 to distinguish between these two explanations, and endogeneity precludes a causal interpretation. However, Share of SSCBT in household income in past 7 days (%) 90 under both interpretations, the comparison suggests 85 that SSCBT is a valuable avenue for income generation for women. 80 75 4. Gender-related constraints to 70 women’s small-scale, cross-border Bavet Poipet Vangtao trade Men Women Source: Authors’ calculations. Both the literature and our data suggest that small- scale, cross-border trade offers a potential for income for men. Non-wage income data are missing from some generation and the empowerment of women. Selection of the nationally representative data that we would is however likely, which combined with the lower share otherwise use for this comparison. We can however rely of women in SSCBT than in comparable activities in on a subset of the IDI assets that are also present in the Cambodia and Lao PDR hints at binding constraints 2002–03 and 2007–08 Lao Expenditure and Consumption affecting women more than men. This section investigates Surveys (LECS). Figure 7 compares asset ownership in such gender-related constraints to women’s small-scale, Champasak-province households for respondents in the cross-border trade. same age group as the IDI respondents in Vangtao. The comparison suggests that the level of asset ownership 4.1 Capital constraints that we would expect for men in Champasak in 2014 is slightly lower than observed for male traders in the IDI One of the constraints on women’s entrepreneurship and data. However, asset ownership is much higher among female entrepreneurs’ revenues most often singled out female traders in Vangtao. This may be evidence either in the literature is women’s limited access to capital. We of positive selection on household wealth of women into find no significant difference in startup capital between SSCBT or of a higher earnings potential, which translates male and female traders in our data.21 Figure 8 however Figure 7: Wealth comparison using nationally representative data 3.5 3.0 Asset ownership index 2.5 2.0 1.5 1.0 0.5 0.0 Same age group, Same age group, Same age group, SSCBTers Champasak province Champasak province Champasak province Vangtao IDI (2014) LECS 3 (2002-2003) LECS 4 (2007-2008) Extrapolation for 2014 based on LECS 3 and 4 Mean for men Mean fro women Source: Authors’ calculations. 21 Controlling for age and the year the trader started their activity does not alter the picture. 159 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES shows that startup capital comes from a wider variety 4.2 Time constraints of sources for female than male traders, and Figure 9 presents a similar picture for how traders finance their Both SSCBTers and other stakeholders, and both daily activities. Since men and women have similar levels female and male interviewees, primarily attributed the of startup capital, this diversification of finance sources prevalence of men among brokers to physical strength. by women may hint at capital constraints: It may be Time endowment may however be a crucial determinant possible but more difficult for women to take a loan from of women’s selection into own-account trading rather relatives, hence a need to look for alternative lenders. than brokering. Since women are usually expected Figure 8: Women have to knock on more doors than men to mobilize a similar level of startup capital 80 70 60 Share of crossers (%) 50 40 30 20 10 0 Own savings Loan from Bank Relatives’ savings Money Loan from NGO Goods supplier relatives lender neighbors/friends credit Female SSCBTers Male SSCBTers Source: Authors’ calculations. Figure 9: Female traders’ day-to-day financing capital comes from a wider variety of sources 100 80 Share of crossers (%) 60 40 20 0 Own savings Loan from Bank Relatives’ savings Goods supplier relatives credit Female SSCBTers Male SSCBTers Source: Authors’ calculations. 160 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES to take care of the household and accomplish more purchased or received and that where the trader sells chores than men, women are often found to be more them (usually, her place of residence) or where the broker time-constrained, which may in turn affect their activity stops taking care of them. We see that being a trader is choices. positively correlated with the total distance traveled. The intuition behind this is that brokers usually take care In Figure 10, time endowment is proxied by the total of the goods just for dealing with border authorities.22 distance traveled by the SSCBTer in her activity, i.e., Interestingly, distance enters the regression negatively the distance between the place where the goods are when interacted with the female indicator variable. Although the coefficients on total distance and the Figure 10: Time constraints as proxied by distance partly explain interaction just miss the 10% significance cutoff, this is SSCBT activity choice consistent with the idea that time endowment—and thus 40 distance and transportation—are more of a concern for women. Marginal effect on probability 30 to be a trader (%) 20 Further evidence of more severe time constraints for 10 women can be gathered from the data. First, stakeholder 0 interviews provide some qualitative evidence that women are overrepresented among the few small- -10 scale “absentee” traders. A common story put forward -20 Female Total distance Female* total by interviewees is that mothers cannot afford to leave traveled distance traveled their homes for extended periods of time, especially Source: Authors’ calculations. as controls and negotiations with border officials make Note: The figure displays coefficients from an OLS regression. Country and checkpoints are controlled for. Whiskers represent 95% confidence intervals. “Total distance traveled” the length of a trip across the border difficult to predict. refers to the distance between where the goods were purchased (received from the trader) and where they were sold (delivered) by the trader (broker). Second, female SSCBTers are less likely to negotiate taxes and fees at the border—see Figure 11—which is consistent with women’s incentive to minimize the time Figure 11: Female traders are less likely to report negotiable they spend at the border. Third, there are important taxes and fees differences between the way female and male traders 80 are taxed at border and in the transport costs that they 70 incur, which we now discuss. Share of crossings with 60 tax negotiable (%) 50 Goods may be taxed per unit (e.g., by the number of 40 packs or boxes), by “visual assessment” of the quantity 30 and value of the goods, through a “fixed” fee per vehicle 20 whatever the quantity transported, or through a lump 10 sum paid on a daily, weekly or monthly basis. As can be 0 Male traders Female traders seen from Figure 12, most crossings are reported to give rise to a per-unit or “visual assessment” tax. The data Source: Authors’ calculations. Note: Whiskers represent 95% confidence intervals. however show that women are significantly more likely to be charged a lump sum per period or a fixed amount per vehicle. One rationale for this gender gap could be that they are more time-constrained.23 The number of different goods that one carries across the border indeed 22 A second rationale is that own-account traders make the most of a comparative advantage in connecting sellers and buyers in remote villages, where they often sell goods directly to final consumers. 23 Women are also found to be more risk-averse than men in our data. 161 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 12: Most shipments are taxed based on unit of value 100 80 Share of crossings (%) 60 40 20 0 Men Women Pooled Men Women Pooled Men Women Pooled Bavet Poipet Vangtao Tax paid per unit or visual assessment of value Tax paid by per cart or time period Source: Authors’ calculations. prolongs customs clearance, as officers are supposed to is forbidden on the Thai side. This may reduce the browse through and count the goods to calculate duties. importance of time constraints for women in Bavet. Another interpretation may be that women have weaker SSCBTers complain about the high level of taxes and bargaining power and cannot make customs officers fees24 more than about anything else, and next about the go through all goods—as they should—to calculate the uncertainty in taxes and fees, interactions with border correct taxes. The difference is significant for Poipet officials and transportation or the length of the crossing and Vangtao; interestingly, the difference goes in the process. Unsurprisingly, reducing taxes is SSCBTers’ main opposite direction in Bavet, although it is not statistically recommendation to improve border crossings, followed significant. Distances are shorter in Bavet and SSCBTers by suggestions to improve roads and transportation can cross the Vietnamese border by motorbike, which Figure 13: SSCBTers’ recommendations to improve small-scale cross-border trade conditions 80 Share of SSCBTers making the recommendation (%) 70 60 50 40 30 20 10 0 Reduce taxes Reduce the Signs showing A parking lot Weighing Receipts An ATM at Better roads One-stop Use of number of tax rates scales for taxes the border window scanners authorities and fees Bavet Poipet Vangtao Source: Authors’ calculations. 24 This pattern is common in surveys about business constraints. 162 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES infrastructures—see Figure 13.25 Only in Vangtao does “reduce taxes” come third, after “better roads” and “a parking lot.”26 The recommendations are supported by A number of qualitative evidence. There is a visible rift in Vangtao recommendations relate between traders who have a pick-up truck and can load goods onto it immediately after the physical border—see to the enhancement of Map 4—and the others who share minivans and tuk-tuks, SSCBTers’ knowledge of the parked downhill at the entrance of the checkpoint zone. laws and regulations applicable Road quality is also a major concern in Poipet and Vangtao, but field observations suggest that the recommendation and of their bargaining power, pertains primarily to breadth (to avoid traffic jams). which in turn would help reduce the time wasted in Calling for a “reduction in the number of authorities” and “one-stop windows” reflects both transportation negotiations. issues and informal taxation that is not justified as duties. Checkpoint zones often cover a large crowded area (see Maps 2 through 4), that SSCBTers must cross in A number of recommendations relate to the several directions to pay taxes and obtain the necessary enhancement of SSCBTers’ knowledge of the laws and documents, e.g., a day ticket to cross the border. This is regulations applicable and of their bargaining power, particularly strenuous, as SSCBTers often lack a motor which in turn would help reduce the time wasted in vehicle because of the cost (gas and/or additional fees) negotiations. Such recommendations include signs or because of regulations. Motorbikes are forbidden to showing tax rates (which were not displayed in any of cross the Thai border with goods, and on the Vietnamese the checkpoints visited), weighing scales and receipts for border one must dismount and walk across the wide no- taxes and fees. Female traders and brokers are expected man’s land—see Map 2. Map 2: Spatial organization of Bavet international checkpoint Source: Map data © 2015 Google. Note: Text and lines in red added by the authors. 25 It is important to note that many recommendations are put forward by a relatively small percentage of SSCBTers. The upshot is that there might be no easy fixes to improve small-scale, cross- border activities, and efforts in several directions should be combined. 26 This finding jars with the much higher taxes found in Vangtao (Figure 15). We see this as a further illustration of the discrepancy between actual and perceived challenges. People often lack a point of comparison, which makes barriers—discrimination, poor institutions, corruption, etc.—“invisible.” 163 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES to benefit the most from a faster, simpler and more that SSCBTers put forward. Figure 14 shows that on predictable border clearance. average female traders spend more than twice as much as men per crossing on transportation costs, which eats Female SSCBTers’ aversion to long border crossings is also into their business margins.27 Another option for time- obvious from transaction-level data on transportation constrained women is to transport smaller quantities. costs. Greater needs for transportation services, e.g., Despite these costly fixes, delays are apparently more hiring help to pull carts or a motorbike to transport a frequent for women and customers are reported to avoid shipment faster, are the main challenge specific to women entrusting female brokers with their goods. Map 3: Spatial organization of Poipet international checkpoint Source: Map data © 2015 Google. Note: Text and lines in red added by the authors. Map 4: Spatial organization of Vangtao International checkpoint Source: Map data © 2015 Google. Note: Text and lines in red added by the authors. 27 It is important to note that the crossing-specific cost data collected in the IDIs were extremely detailed. Transportation costs are thus distinct from fees imposed on vehicles, duties determined based on the number of carts, or bribes related to transportation. 164 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 14: Female SSCBTers spend more per crossing on more likely to be controlled although they do not bend transportation costs the rules more often than men. 4.5 First, while contrary to complaints voiced by traders 4.0 3.5 and brokers the tax burden on SSCBT is relatively light Transportation cost per crossing (USD) 3.0 except in Vangtao (Figure 15), tax rates are higher for 2.5 female than male traders.28 This holds true whether we 2.0 look at tax payments as a share of total purchase price 1.5 1.0 or total profit—see Figure 16. It also holds true whether 0.5 we consider averages, medians or—as in Figure 17—the 0 whole distribution. Higher tax rates on female traders Male traders Female traders cannot be explained by goods quality or scale economies captured by male traders, as male and female traders Source: Authors’ calculations. Note: Traders only. Whiskers represent 95% confidence intervals. enjoy similar gross profit rates. Second, female traders are significantly more likely to be 4.3 Discriminatory treatment controlled by quarantine officers, which hints at deliberate The constraints identified in the first two parts of this targeting. The difference, driven by the Vangtao sample, section may affect women disproportionately but cannot remains significant when one controls for perishable food be directly blamed on interactions between border products, which make up almost half of the crossings crossers and officials. The actors in the border economy in SSCBT and are more often traded by women. The seldom acknowledge taxation practices as harming regression results displayed in Figure 18 strongly hint at particularly female SSCBTers. Econometric analysis discrimination or at least deliberate targeting of women. however reveals that women pay higher taxes and are Quite strikingly, men are not statistically significantly Figure 15: Tax payments as a share of gross profits are high in Vangtao 60 50 40 Tax payments 30 20 10 0 Bavet Poipet Vangtao Bavet Poipet Vangtao Tax payments as share of total purchase value Tax payments as share of total gross profit Mean Median Source: Authors’ calculations. 28 The IDI data contain unique information on shipment values, taxes and fees that enable us to shed light on the tax burden faced by female SSCBTers relative to men. Tax rates remain moderate when compared with the value of the goods, but most of Lao traders’ profits vanish in taxes and fees levied by border officials when looking at tax payments as a proportion of total gross profits (total sale minus purchase price)—see Figure 15. The marked difference in Lao PDR between the two tax rate definitions shows that profit margins are very thin in Vangtao. Since brokers are not always able to put a figure on the value of the shipments they are taking care of, results about the tax burden are based on traders’ answers. 165 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 16: Female traders face higher tax rates than their male counterparts 25 20 15 (%) 10 5 0 Men Women Men Women Tax payments as share of total purchase value Tax payments as share of total gross profit Mean Median Source: Authors’ calculations. more likely to go through quarantine when they deal in hurts female traders, as they are more concentrated perishable foods (as shown by the small and insignificant in terms of types of goods traded. Traders dealing in coefficient on the “perishable food” indicator variable). perishable goods are also more vulnerable to delays and It must be noted that given the small sample size and confiscation—a common practice at Cambodian and Lao endogeneity in the regression, we cannot be positive checkpoints—as market days are often fixed and the that Figure 18 provides evidence of discrimination. It is goods must be sold fresh. possible that quarantine officers “target” female traders Female SSCBTers’ higher tax burden and greater for tax payments and controls because women are 29% interaction with quarantine cannot be directly linked with more likely (keeping activity and checkpoint constant) female traders and brokers indulging in illegal practices to deal in perishable foods. This practice nevertheless Figure 17: Tax rates faced by female traders are higher at almost every level 20 Male traders Female traders Density 10 0 0.0 0.1 0.2 0.3 0.4 Tax payments as share of total shipment value Source: Authors’ calculations. Note: The figure displays univariate kernel density estimations of tax rates for male and female traders. The Epanechnikov kernel is used. 166 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Figure 18: Women dealing in perishable foods are more likely to downwards. We decided to include in the IDI survey interact with quarantine additional questions to take into account SSCBTers’ reticence and adjust estimates of the incidence of tax 20 evasion accordingly. The methodology implemented to interact with quarantine (%) Marginal effect on probability 15 relies on Kraay & Murrell (2013). SSCBTers’ direct answers 10 to the binary question about tax evasion in the past 12 months are contrasted with the reticence-adjusted 5 estimates. We leave the presentation of the relevant 0 survey items to Appendix A and of the methodology and results to Appendix B. The main finding is that although -5 Female Perishable food Female* perishable rates of tax evasion are higher when taking reticence into food account, they are neither significantly nor qualitatively Source: Authors’ calculations. Note: The figure displays coefficients from an OLS regression using data from all three different between men and women. checkpoints. Country and checkpoints are controlled for. Whiskers represent 95% confidence intervals. Third, women are charged higher taxes on larger shipments, whereas male traders face no disincentive more than men, for example due to a low-trust equilibrium to expanding their activity. As can be seen from the between border authorities and female crossers. The IDI regression coefficients displayed in Figure 19, female data contain information on whether in the 12 months SSCBTers face progressive taxation, while their male preceding the survey the respondent had “omitted to counterparts do not. Tax exemption30 is indeed less likely, declare goods or underreported their quantity or value the higher the value of a female SSCBTer’s shipment (as on purpose in order to avoid taxes.”29 We find that about shown by the significantly negative interaction of the one tenth of SSCBTers admit to deliberate tax evasion “female” indicator variable with “total purchase value”). in the past year, but stakeholder interviews revealed Conversely, no such effect is found on male SSCBTers that SSCBTers are reluctant to confess tax evasion. Their (“total purchase value” on its own is insignificant). answers on this topic are therefore likely to be biased Figure 19: Female brokers are more likely to benefit from tax exemption, not female traders 150 100 Marginal effect on probability of tax exemption (%) 50 0 -50 -100 -150 Female Trader Total purchase Female* total purchase Female* trader value (standardized) value (standardized) Source: Authors’ calculations. Note: The figure displays coefficients from an OLS regression. Country and checkpoints are controlled for. Whiskers represent 95% confidence intervals. “Total purchase value” is standardized by checkpoint. The reference category is male brokers. 29 Some goods, e.g., alcohol, are subject to declaration in some checkpoints and simply forbidden to cross the border in others. A common practice for traders who want to minimize taxes is to hide high-duty goods under low-duty ones. This is particularly effective under the “visual assessment” method, which remains usually superficial. 30 Tax exemption is a widespread tool to introduce progressive taxation in the three checkpoints. Since brokers are not always able to put a figure on the value of the shipments they are taking care of, results about the tax burden are based on traders’ answers. 167 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Fourth, female SSCBTers are discouraged from upgrading obvious implication is to streamline taxation at border to own-account trading, while male brokers face no such checkpoints and ensure that only legal taxes are levied. disincentive. Figure 19 indeed further shows that the Signs displaying tax levy rules and the tariffs applicable probability of tax exemption is significantly lower for should be posted at each checkpoint and kept up to female traders than for female brokers. This can be seen date.33 The display of tariffs could be complemented from the large negative coefficient on the interaction of by the equipment of checkpoints with weighing scales, the “female” and “trader” binary variables. Male traders a recommendation some SSCBTers put forward. They do not differ from male brokers in their probability to would indeed avoid time-consuming negotiations and benefit from tax exemption (see the insignificant “trader” help reinforce SSCBTers’ bargaining power. Finally, one- variable, male brokers being the reference category in stop windows should help reduce border clearance time. this regression).31 This generates a tax wedge that may Women’s time constraints also resonate with another discourage women from upgrading to own-account major hurdle in trading goods across the border: trading, which we saw is a more profitable cross-border transportation. SSCBTers’ top concerns include narrow activity. roads that cause traffic jams and delays, the lack of 5. Discussion and policy implications public transportation and parking lots, and restrictions on the types of vehicles allowed across the border with Based on a mix of qualitative and quantitative data, we goods. The first way to tackle the issue of transportation have highlighted major constraints that compress female consists of investing in transportation infrastructure. The SSCBTers’ profits and that are likely to deter other women most crowded of the checkpoints visited is undoubtedly from engaging in small-scale, cross-border trade. Some Poipet, where large trucks, private cars, SSCBTers’ carts are visible to the actors in the border economy, while and tourists on foot all go through the same gate. others, invisible, work as “glass barriers” to female cross- Infrastructure improvements could also include parking border trade and entrepreneurship. The main barriers lots that would enable better-off SSCBTers to invest in are: (i) capital constraints, (ii) time constraints, and (iii) a higher tax burden. We consider that (ii) and (iii) are—at Figure 20: Bribe payments are widespread at Cambodian and least partly—due to border checkpoint infrastructures Lao checkpoints and take a variety of forms and interactions with border officials, and thus ought to be tackled by border authorities. 70 60 SSCBTers’ most frequent complaints pertain to the high 50 number and volatility of taxes and fees. Traders and Share of crossers brokers are usually unaware of the legality, tariff rate and 40 purpose of levied taxes and a high proportion reports 30 paying bribes—see Figure 20. 32 We also realized during 20 field observations that virtually all border agencies tax 10 traders and brokers, sometimes in blatant violation of 0 their mandates, and SSCBTers must travel back and forth Bavet Poipet Vangtao within checkpoint zones to make payments. Women are Cash bribe In-kind bribe Negotiation to reduce taxes likely to be particularly harmed by informal taxation, Source: Authors’ calculations. as negotiations and multiple payments take time. An 31 The gender gap in tax exemption is not due to differences in shipment values, since Figure 19 controls for that. 32 Note that the reported incidence of bribe payments does not significantly differ between women and men. 33 Whereas SSCBTers in FGDs complained about the level of taxes, they were seldom inclined to reject negotiability, probably out of fear that the application of tariffs set in stone would be detrimental to them. But if the average tax rate were to remain unchanged, predictable tariffs would improve SSCBTers’ welfare, especially for brokers, who are found in the data to be more risk-averse. Moreover, we saw that women are less attached to negotiations since they do not have time for them. 168 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES larger vehicles, or a bus service to the border proper. Women’s time constraints and high transportation costs Our findings that women may also be effectively tackled by relieving them of some are subject to higher tax of their household duties, for instance through a better rates, are more often provision of child care—the cost of which could be shared through a female traders’ association.34 controlled by quarantine Border-crossing rules need to be altered to facilitate and incur much larger SSCBT. In Vangtao, SSCBTers refrain from using pick-up transportation costs trucks because of the fees attached to crossing the border with goods in four-wheeled vehicles. Reducing those fees are suggestive of “glass might be an alternative to enlarging the existing parking barriers” to female cross- lot. On the border with Thailand, motorbikes are not border trade—either hidden allowed to cross with goods, which prevents SSCBTers from using a cost-effective transportation option and yet real discrimination means that only human-powered carts are available. or challenges that affect On the border with Vietnam riding a motorbike with goods is forbidden in the checkpoint zone, which forces women disproportionately. traders and brokers to push the heavy shipment for long distances. Our findings that women are subject to higher tax rates, are more often controlled by quarantine and incur much small-scale, cross-border traders and brokers—of their larger transportation costs are suggestive of “glass rights and obligations could be instrumental in improving barriers” to female cross-border trade—either hidden border-crossing conditions. The Charter, the exact yet real discrimination or challenges that affect women contents of which should be discussed with stakeholders, disproportionately. This is part of the explanation for would list the authorities allowed to operate at border the lower share of women in SSCBT compared with checkpoints, and which authorities can collect taxes and similar jobs in Cambodia and Lao PDR. Such practices fees and which can not. The Charter would also state are not justified by a higher propensity to evade taxes, rules of conduct, including no discrimination in tax rates since women appear no different from men in that by gender and no verbal or physical violence. Its main respect, even after adjusting reports for differential benefits would be to enhance the bargaining power reticence. Training of officials and traders on gender- of the most vulnerable categories of border users and based challenges at the border, backed up by monitoring raise the low level of knowledge of cross-border trade of the performance of officials could be established to rules and regulations that characterizes SSCBTers. The address these issues together with the establishment of design of the Charter could benefit from experience a complaints and dispute settlement mechanism to allow from the World Bank’s initiative piloted at the Mwami/ women who feel they have been unfairly targeted to Mchinji crossing between Malawi and Zambia (World seek redress. Bank 2014b). As in southern Africa, the Charter would be displayed at strategic locations, translated into local A Charter for Cross-Border Traders and Brokers languages and disseminated to stakeholder groups. reminding all parties in small-scale, cross-border trade— authorities on both sides of the border, transporters, and 34 No small-scale traders’ associations exist in Cambodian and Lao checkpoints, but in open discussions about ways to foster small-scale cross-border trade during FGDs, SSCBTers agreed associations might have a positive impact on their activities. They would however not list them among their recommendations, as they were reluctant to suggest solutions they had never tried out. Associations have been proven effective in defending female traders’ rights, improving information about prices, providing training and ensuring smooth relationships with border officials in East Africa. 169 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Cambodia Development Resource Institute. 2005. “Cambodia’s Cross Border Economy: An Exploratory Study.” Working Paper 32 by K.A.S. Murshid and Tuot Sokphally, Phnom Penh. Cambodia Ministry of Agriculture, Fisheries and Forestry. n.d. Agricultural Market Information: Gender Roles in Agricultural Marketing. Accessed June 6, 2014. http://www.agriculturalmarketinformation.org.kh/en/ information-resources/gender. EMC. 2012. Time Release Study 2012 Lao PDR. Vientiane: Emerging Markets Consulting, Ltd. Financial Ministry of the Lao PDR. 2005. “Financial ministerial instruction on the implementation of the law and enforcing decree to implement the Custom Law No. 05/NA.” Vientiane, May 25. Hausmann, Ricardo, Bailey Klinger, and Rodrigo Wagner. 2008. “Doing Growth Diagnostics in Practice: A ‘Mindbook’.” CID Working Paper No. 177, September. Kraay, Aart, and Peter Murrell. 2013. “Misunderestimating Corruption.” World Bank Policy Research Working Paper 6488. Kusakabe, Kyoko, Prak Sereyvath, Ubolratana Suntornratana, and Napaporn Sriputinibondh. 2008. “Gendering Border Spaces: Impact of Open Border Policy Between Cambodia-Thailand on Small-scale Women Fish Traders.” African and Asian Studies 7: 1-17. Lesser, Caroline, and Evdokia Moisé-Leeman. 2009. “Informal Cross-Border Trade and Trade Facilitation Reforms in Sub- Saharan Africa.” OECD Trade Policy Working Paper No. 86. Ministry of Economy and Finance of the Kingdom of Cambodia. 2005. “Circular on Management of Imported Non- Commercial Goods No. 009 SHV.” Phnom Penh, December 1. ——. 2010. “Prakas on Procedures of Payment of Duties, Taxes and Other Levies on Imported and Exported Goods No. 571 MEF.BK.” Phnom Penh, August 19. National Institute of Statistics, Ministry of Planning, Kingdom of Cambodia. 2014. “Cambodia Socio-Economic Survey 2013.” Phnom Penh, July. Prud’homme, R. 1992. “Informal Local Taxation in Developing Countries.” Environment and Planning C: Government and Policy 10 (1): 1-17. UN Women. 2012. Walking in the Dark: Informal Cross-Border Trade in the Great Lakes Region. New York City: UN Women. UNIFEM, WB, ADB, UNDP and DFID/UK. 2004. “A Fair Share for Women: Cambodia Gender Assessment.” Phnom Penh. USAID. 2012. “Women in Cross-Border Agricultural Trade.” Enabling Agricultural Trade Project Policy Brief. no. 4. October. Walker, Andrew. 1999. The Legend of the Golden Boat: Regulation, Trade and Traders in the Borderlands of Laos, Thailand, China and Burma. 170 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Warner, S. 1965. “Randomized-response: A survey technique for eliminating evasive answer bias.” Journal of the American Statistical Association 60: 63-69. World Bank. 2014a. Lao PDR Trade and Transport Facilitation Assessment. Vientiane: World Bank. World Bank. 2014b. The Republic of Zambia Diagnostic Trade Integration Study (DTIS). Washington, D.C.: World Bank. World Bank. 2012a. Cambodia Lao PDR Trade and Transport Facilitation Assessment. Phnom Penh: World Bank. World Bank. 2012b. Lao PDR - Mapping the gender dimensions of trade: A preliminary exposition. Vientiane: World Bank. World Bank. 2012c. Toward Gender Equality in East Asia and the Pacific: A Companion to the World Development Report. Washington, D.C.: World Bank. World Bank. 2011. Facilitating Cross-Border Trade between the DRC and Neighbors in the Great Lakes Region of Africa: Improving Conditions for Poor Traders. Washington, D.C.: World Bank. ——. 2005 and 2009. World Development Indicators. Accessed June 6, 2014. http://databank.worldbank.org/. 171 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix The RRQs included in the IDI questionnaire are presented in the Appendix A. Some of the questions come from Kraay & Murrell’s battery, which was used in Cambodia among other developing countries to estimate the share of the population who had been in a situation where a bribe was expected in the past year. The others were developed based on preliminary observations and qualitative interviews of SSCBTers. They were then fielded along with the other sections of the survey during pilots in all three checkpoints. Appendix A. Random-response Procedure included in the IDI Survey 5.8 In the last 12 months, have you ever omitted to declare goods or underreported their quantity or value on purpose in order to avoid taxes? □ 1 Yes □ 0 No □ 997 Doesn’t know □ 998 Refuses to answer Please read out the following script,a making sure the interviewee understands the procedure: Enumerator: Please make sure the card deck is well shuffled (do it a minimum of 5 times before the interview). I am going to read out a set of questions that describes acts or behaviors that people have expressed. Unlike other questions where you would just respond with a “yes” or “no,” this set has a slight variation to it. Before you answer each question, you will pick a card from this deck. There are 50% of black (spades/clubs) and 50% of red (hearts/diamonds) cards, randomly mixed. Based on which color you pick, I will give you an instruction to provide the appropriate response. Are you ready? I will now read the first question. Please pick a card from the deck, and if it’s a red card, just say YES regardless of whether you have done this or not. If it’s black, please just answer the question. Please do not let me see the card and do not put it back into the deck. This is very important. 1 Yes 0 No 5.9 Have you ever lied to protect yourself? □ □ 5.10 Have you ever deliberately spoken ill of a member of your family or a friend? □ □ 5.11 Have you ever deliberately tried to cheat another person? □ □ 5.12 Have you ever broken a promise? □ □ 5.13 Have you ever taken something that is not yours without permission and kept it? □ □ 5.14 Have you ever bought, sold, bartered or been given something that you knew was stolen? □ □ 5.15 Have you ever mistreated someone because they did not share your opinions or values? □ □ 5.16 Have you ever been nice to a person only because you thought it would bring you some benefit?b □ □ 5.17 If you received some extra money that your family did not know about, would you ever hide it from them and spend it on your own enjoyment? □ □ 5.18 Have you ever insulted your parents, relatives or other elders? □ □ 5.19 Have you ever bribed a policeman because you did something wrong on the road? □ □ 5.20 Have you ever damaged somebody’s property to hurt them? □ □ 5.21 Have you ever stolen money from a member of your household? □ □ Please mix in front of the respondent the cards s/he has picked with the rest of the deck. Do not look at the cards. a  Adapted from Kraay and Murrell (2013). b  This question was not exploited in the reticence adjustment procedure because pilot interviews revealed a very high proportion of “Yes,” suggesting a different rate of guilt. 172 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Appendix B. Adjusting Answers to Sensitive Questions Answers to survey questions about sensitive topics, e.g., corruption or illegal activities, are notoriously unreliable because respondents put little faith in guarantees of survey data anonymity or want to avoid negative judgments from enumerators. Kraay & Murrell (2013) have designed a new methodology that uses randomization but does not assume that reticence decreases—contrary to Warner (1965), for instance. Besides the “conventional question” (CQ), i.e., a sensitive question (set in binary terms) that the respondent is asked to answer directly, their approach includes a set of “random-response questions” (RRQs), for each of which the respondent privately tosses a coin (we used playing cards instead). She is instructed to answer “Yes” whatever the true answer to the sensitive question if the coin comes up heads and to answer the question otherwise. The RRQs are also sensitive—a crucial assumption is that the proportion of respondents who have done the sensitive action, which Kraay and Murrel call the “rate of guilt,” is the same across the CQ and RRQs,— and far from assuming that randomization elicits more truthful answers, the known probability of a “Yes” is used to estimate the incidence of reticence in the sample thanks to the generalized method of moments (GMM). Reticence- adjusted rates of tax evasion are presented in Table B.1. Table B.1: Reticence estimation and adjustment of answers to sensitive questions following Kraay & Murrell’s (2013) methodology Pooled Bavet Poipet Vangtao Men Women Brokers Traders Guilt 0.189a 0.121c 0.183c 0.270b 0.165b 0.221a 0.202C 0.109b Reticence .796a 0.712a 0.779a 1.02a 0.749a 0.871a 0.708a 0.756a Probability reticent person answers question reticently 0.476a 0.596a 0.439a 0.361a 0.508a 0.432a 0.540a 0.508a Effective reticent 0.379 a 0.424 a 0.342 a 0.369 a 0.381 a 0.376 a 0.382 a 0.384a Number of observations 157 55 54 48 65 91 45 63 Naïve guilt rate estimated directly from survey responses 0.117a 0.070b 0.120c 0.170a 0.102b 0.138a 0.125c 0.113a Source: Authors’ calculations. The Stata code for estimating these parameters was graciously shared with us by Peter Murrell. a p<.01 b p<.05 c p<.1 173 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Are the “Poor” Getting Globalised? Adelina Mendoza, World Trade Organization Gaurav Nayyar, World Bank Roberta Piermartini, World Trade Organization 1. Introduction In analysing the effect of trade opening on greater economic inclusion, this literature has focused on G lobalization is under fire. Public perceptions the tariff structure in the domestic economy. But the and recent policy debates increasingly appear conditions of access to foreign markets, as determined by to indicate that trade liberalization has been tariff policies of trading partners, are also key to capturing accompanied by rising income inequality in export opportunities and generating employment and/or developed and developing economies. The fact that trade wage gains for poor households. There are, in fact, a few liberalization creates both winners and losers has never studies which show that improving conditions of market been in question. While international trade enhances access reduce poverty rates.2 economic growth in the aggregate, the distribution of its What then are the market access conditions for the poor? benefits may vary by income group, location, gender, and Are there ample opportunities for them to reap benefits the formal-informal divide. from exporting? This is a particularly relevant question The literature on the subject discusses why poor given evidence suggesting that individual countries often households may have only marginally benefitted from protect their own “poor” (or declining sectors) by raising trade opening, both as producers and consumers. On tariffs and/or non-tariff barriers on the goods these the production side, skilled-biased technological change households/individuals (or declining sectors) produce. associated with trade and FDI is likely to have dampened But, when all countries protect the sectors where the the increase in the demand for unskilled labor in poor work and if the poor are employed in similar kind of developing economies expected after trade liberalization. sectors in different countries, a “coordination problem” Further, high reallocation costs across sectors, firms and arises: the goods (and services) produced by the poor will geographical locations that are particularly burdensome face higher barriers to trade than those provided by the for poor households affects their ability to move from non-poor, and the resulting decline in the global demand contracting to expanding areas of economic opportunity. will lower the price of goods and services that the poor On the consumption side, the pass through of lower produce. prices (resulting from trade liberalization) from the Tariff and non-tariff barriers faced by the poor, if relatively border to consumers has been affected by high domestic higher, may therefore impair income distribution by transport costs and a range of market frictions.1 keeping poor workers disconnected from global markets. 1 For the impact of high domestic transport costs see Nicita (2009), Emran and Hou (2013), and Atkin and Donaldson (2012). For the impact of market frictions see Campa and Goldberg (2002), Atkin and Donaldson (2012), Ural Marchnad (2012) and Han et al. (2016). 2 See, for example, Porto (2010) and McCaig (2011). 174 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES While international trade enhances economic growth in the aggregate, the distribution of its benefits may vary by income group, location, gender and the formal-informal divide. 175 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES The potential problem is then not too much globalization, the food basket more expensive, which works towards but too little “inclusive” globalization. This view is in stark increases in poverty, but would boost labor demand contrast to the rhetoric that is usually seen and heard. and wages, which works towards poverty alleviation. Therefore, if the household is a net seller—whether of In light of the above, the objective of the paper is to labor, goods or services—price increases will raise its real assess the obstacles that “poor” households in a given income and vice-versa. country face when trying to export goods to the rest of the world. In doing so, it investigates the overall tariff The literature has focused on explaining why the profile on exports from India. Tariff data is matched with distributional effects of trade opening, as described consumption/income data by the industrial classification by the theory, depend on a range of other factors in of each household member’s sector of employment. practice. On the production side, the transfer of skill- “Poor” households are not identified by a pre-defined biased technologies associated with trade could reduce cut-off in their level of income or consumption, but are the wages of unskilled labor even in a labor-abundant instead analysed along a continuum of decile groups that country, thereby widening the gap between the rich and form the entire distribution. The scope of “poverty” in the poor. Similarly, despite shifting low-skilled activities the paper also extends beyond the income dimension from rich to poor countries, foreign direct investment to focus on groups that are often excluded from the may increase the demand for skilled workers because growth process and perhaps are more disconnected from jobs which were low skill-intensive in the former may global trade. These include women, those working in the be relatively skill-intensive in the latter (Wood, 1997). informal sector, or those working in rural areas. Further, as trade liberalization reallocates economic activity across sectors, industries and firms, the short- The structure of the paper is the following: Section 2 run adjustment costs can be high with the burden falling provides a review of the literature. Section 3 outlines the disproportionately on poor households (Banerjee and dataset and methodology. Section 4 describes the results Newman, 2004). and Section 5 presents conclusions. Given the high reallocation costs across geographical 2. Review of the literature regions within countries, the poor may also only 2.1 Trade opening and distributional effects marginally benefit from greater openness due to sectoral variation in patterns of trade liberalization combined Trade liberalization, through its impact on prices in with spatial variation in the industrial composition of both product and factor markets, affects members of a the labor force. Take, for instance, evidence from India household as both producers and consumers (Winters which suggests that rural areas with a high concentration et al., 2004). In most developing countries, a majority of of industries that were disproportionately affected by poor households rely on labor markets for the bulk of tariff reductions experienced slower progress in poverty their income. Standard trade theory predicts that trade reduction (Topalova, 2010). With perfect factor mobility opening increases the demand for the relatively abundant across regions, labor would migrate in response to wage factor, which suggests that unskilled labor in developing and price shocks, equalizing the incidence of poverty countries would benefit most from globalization through across regions, but the low incidence of internal migration a resulting increase in wages or employment or both. in India is striking (Kone et al., 2016). Similarly, local labor As producers, farm households for example can gain markets in Brazil where workers were concentrated in by selling their output in hitherto unavailable overseas industries facing the largest tariff cuts were generally markets, which may also yield a better return. As affected more negatively (Kovak, 2013). consumers, trade liberalization can be beneficial to the extent that it reduces the price for imported goods. At The extent to which households benefit from trade times, these effects can go in opposite directions. For liberalization on the consumption side depends on a example, higher prices of agricultural exports would make range of factors that influence the pass-through of 176 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES price changes from the border to consumers. Owing to transport and other costs of distribution, the geographic While geography is hard to characteristics of localities, such as the distance to the change, access to foreign border matter. Nicita (2009), for example, finds that tariff pass-through was significantly higher in the Mexican markets is also likely to states closest to the United States border, and thus, be influenced by a set of households living in these states benefited relatively more from the reductions in tariffs. Similarly, Atkin and international trade policies— Donaldson (2015) find that the costs of intra-national such as tariffs, non-tariff trade are approximately 4 to 5 times larger in Ethiopia and Nigeria compared to the United States. This reduces measures and services trade the amount of potential surplus consumers in remote locations—far from a country’s major port, for example— restrictions—employed by a can derive from falling international trade barriers. Other country’s trading partners. spatial characteristics, such as the relative isolation of households from functioning product markets, may also matter for price transmission. Pass-through estimates for India suggest that reductions in tariffs increased above has focused on domestic factors, such as domestic domestic consumer welfare more in urban areas than in tariffs and market frictions. But the conditions of access rural areas (Ural Marchand, 2012). to foreign markets, as determined by tariff structures and non-tariff measures of trading partners, are also Market frictions are another relevant factor. If domestic key to capturing export opportunities and generating industries are imperfectly competitive, changes in tariffs employment and/or wage gains for poor households. may be absorbed by profit margins or mark-ups (Campa and Goldberg, 2002). Atkin and Donaldson (2012) have The literature is somewhat scant in this regard, but it further shown how the market power of intermediaries in does point at the importance of market access conditions domestic industries affects the mark-ups, which results in for the distribution of the gains from trade. Market different rates of tariff pass-through within sub-Saharan access is naturally influenced by geography and distance. Africa. Similarly, a heavily regulated domestic industry For example, evidence from China suggests that reduced that is dominated by state-owned enterprises would distance to (domestic and) international markets confers have limited flexibility to adjust to the changing cost substantial benefits on per capita consumption of rural conditions (Szamosszegi and Kyle, 2011). Evidence from households (Emran and Hou, 2013). While geography is China suggests that a 10 percentage point increase in the hard to change, access to foreign markets is also likely size of the private sector across cities is associated with to be influenced by a set of international trade policies 2 percentage points higher tariff pass-through, with the —such as tariffs, non-tariff measures and services trade share of the private sector among intermediaries being restrictions—employed by a country’s trading partners. particularly important (Han et al., 2012).3 For example, Porto (2010) predicts that the elimination of trade barriers on exports of agro-manufactures to 2.2 Market access, “pro-poor” trade policy industrialized countries would cause poverty to decline and coordination failures in Argentina. Similarly, McCaig (2011) analyzes the United In analysing potential explanations for why poor States-Vietnam free trade agreement to show that households have benefited less from trade opening than provinces in Vietnam that were more exposed to U.S. trade theory would predict, the literature referred to tariff cuts experienced greater declines in poverty rates. 3 The average pass-through rate is found to be 22% in a city where all enterprises are state-owned, while a city with an average size of the private sector has an approximate tariff pass-through rate of 31%. 177 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES One aspect that the existing literature has neglected export restrictions and food price volatility (Abbott, is how domestic trade policies of individual countries, 2012; Ivanic and Martin, 2014; Gouel, 2016). Countries when considered together, affect access of “poor” and frequently use export restrictions to protect poor “rich” households to foreign markets differentially. This consumers from high or volatile prices on the world is particularly important in light of evidence of countries market. However, when countries simultaneously implementing trade (and trade-related) policies that respond to higher international prices for food, say favor poor households. For example, relying on data rise, by unilaterally imposing export restrictions, the for six Sub-Saharan African (SSA) countries, Nicita et al. international price of the commodity in question will only (2014) finds that SSA’s trade policies have a systematic further increase. Consequently, the “poor” consumers “pro-poor” bias, that is, trade policies redistribute income 4 of this food commodity in countries that imposed the from rich to poor households. This is mostly explained by export restrictions may be unable to benefit from lower protection granted to agricultural products that are sold domestic prices. by poor households—the positive labor income effect dominates the impact of higher consumption prices and 3. Data and methodology other forces that benefit skilled over unskilled workers. 3.1 Matching household survey data with The question is what if all countries apply “pro-poor” tariffs and NTMs trade policies and if the poor are employed in similar kinds of sectors in different countries? The paper uses a novel approach to assess barriers that individual producers across the income distribution face in Strategies that use trade protection to assist poor international markets. In doing so, it combines household households are unlikely to be effective if one also survey data on income and consumption from India (see considers general equilibrium effects.5 There is the Box 1) with information on lowest applied tariffs in the possibility of a “coordination problem” whereby poor top 15 destination markets (EU counted as one) for households are less able to benefit from globalization Indian products in 2012.6 This is done by matching India’s because they are excluded from the process. Consider National Industrial Classification (NIC)—which is based the following. In order to protect producers from low on the International Standard of Industry Classification prices of competing products from the rest of the world, (ISIC)—of an individual’s sector of employment, with countries increase import tariffs and other non-tariff tariffs faced in India’s major export markets at the barriers on sectors that employ a large proportion of Harmonized System subheading (6-digit) level.7 The those classified as “poor.” If this done by a sufficiently same concordance was used to determine the variability large number of countries, which together constitute a of non-tariff measures (NTMs) imposed on imports from “large” country that is a “price-maker” in the world market, India by the same top importers for the same reference the resulting decline in global demand will lower the price year across different sectors.8 The number of (NTMs) of goods and services that the poor produce. As a result, applied to India in this analysis includes those applied by “poor” households will either not be price competitive to these partners on an MFN basis, e.g., those NTMs that are export to markets that have implemented measures of not targeted specifically at India but at all exporters to protection or will receive a lower price in export markets that country. that have not implemented these measures. The ISIC - HS correlation is not a full match concordance table. The coordination problem with regard to import There are ISIC codes for which there is no corresponding HS restrictions described here mirrors the literature on 4 The authors analyse how reductions in tariffs and non-tariff barriers affect consumer and producer prices, which in turn affect welfare of the average household in the top and bottom 40% of the income distribution defined by household production, household consumption, labor earnings and government transfers. 5 The literature also discusses foregone opportunities driven by dynamic effects. Using protectionist policies to assist the poor may actually harm the poor by limiting productivity increases and structural transformation. 6 These importers account for more than 75% of world imports from India. 7 The tariffs at the HS subheading level matched with the ISIC code are averages weighted by the value of imports from each partner. We also look at how much the tariffs faced have declined from their 1996 level for specific identified sectors. To calculate the reduction in tariffs, the same methodology was used to match tariffs in 1996 for the same partners. 8 The data on NTMs were extracted from the WTO’s NTM database—I-TIP. The database shows the number of NTMs by type—SPS, TBT, ADP, CV, SSG, SG, QR, TRQ and XS—in each HS chapter. 178 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES code. Hence the analysis only includes activities in which for income owing to the difficulty of getting reliable the associated good/s can be identified with the HS code. quantitative data on household income through direct For example, some ISIC codes which refer to construction enquiries in household surveys (see Box 1). Additional (e.g., ISIC “4100 - Construction of buildings”) do not have variables in the data enabled us to also look at groups a matching HS code, and such observations were dropped traditionally excluded from the process of economic from the calculation of the statistics. For the analysis of 9 growth—women, those working in rural areas and those NTMs, the statistics were done only on total NTMs without employed in the informal sector. breakdown by type of NTM. 4. Results 3.2 Defining the “poor” or “excluded” 4.1 Tariffs faced in the export market are In this study, economic “poverty” is measured as a inversely related to income relative phenomenon based on the analysis of income/ Using household level data, the average income (using consumption for the entire distribution by decile group. consumption as proxy) and tariff faced in export Consumption expenditure is widely accepted as a proxy Box 1: Household Survey Data on Employment, Income and Consumption from India The Government of India’s National Sample Survey Office (NSSO) regularly conducts national household surveys on the subjects of employment and consumption. For the present exercise, the data are taken from the 68th round conducted during the period from July 2011 to June 2012, specifically survey data collected from the questionnaire referred to as “Schedule 10: Employment and Unemployment”. The number of households surveyed was 101,724 (59,700 in rural areas and 42,024 in urban areas). These sample households correspond to 456,999 individuals (280,763 in rural areas and 176,236 in urban areas). In the employment survey, an individual’s sector of employment is based on activities pursued during certain specified reference periods: one year, one week and each day of the reference week. The activity status determined on the basis of each reference period is referred to as “usual status,” “current weekly status” and “current daily status,” respectively. The “usual status” approach is generally preferred as it is based on a relatively longer time horizon. Nonetheless, the current weekly status approach is important in any analysis of wages because these are expressed in weekly terms. Each approach is guided by the “majority time spent” criterion. For instance, under the “usual status” approach, an individual is considered “employed” in a particular industry if he or she spent the majority of his or her time in the last 365 days on that economic activity, rather than being unemployed or engaged in non-economic activities. The paper analyzed two correlated datasets from the survey in order to identify the relevant sector of employment. 1. The first relates household-level data based on usual status of the sector of employment of the household head with the corresponding household weekly consumption. Each observation refers to a household with its attendant characteristics, including the rural versus urban distinction. Household weekly consumption was used as proxy for income. 2. The second relates individual-level data based on members within the household reporting an economic activity and corresponding income from such activity during the reference week. The economic activity is the usual principal activity of the household member. Additional information on gender and type of the enterprise (formal or informal based on the number of employees) is also available from this dataset, which provides an added dimension in the analysis of the data. The advantage of using household consumption levels as a proxy for income is that wage-based data do not include income received by self-employed individuals. Analysis of household-level data was purely based on consumption, even if there are individuals within the household that reported income. For the dataset based on individual respondents, individuals (15 years and above, excluding those studying full time) who had an activity during the week and reported its corresponding earnings are included. 9 Only observations of selected respondents who work in ISIC sectors where a correlation with HS (hence tariffs) exists (32% of respondents) are included in the study. The other 68% work in ISIC sectors that do not have ISIC-HS correlation code. This is mainly an issue related to services sectors. Our study, therefore, has to be seen as an analysis of merchandise sectors only. 179 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES markets for each income decile were calculated (Table except for two neighboring deciles—the difference 1). Households in the poorest decile have an average between tariffs for deciles 2 and 3, and for deciles 6 and weekly consumption of only 511 rupees ($9.60), while 7, are not statistically significant. households in the richest decile consume ten times that The tariffs for the agriculture sector, in which 62% of amount. the respondents are classified, mirror the same inverse The results show that goods associated with higher relationship between tariff faced and income decile, income households face lower tariffs while goods but at a much higher tariff magnitude (see Figure 1). associated with low income households face higher The sectors where a large percentage of poor are tariffs. In the table, the tariff faced by the next higher concentrated include the production of cereals and income decile is consistently and significantly lower manufacturing of wearing apparel. Table 1: Average tariff faced by income decile based on household weekly consumption Income Household Weekly Income Average Tariff Faced (%) Decile Rupees US $ Simplea 1 511 9.6 24.4 a 2 783 14.8 22.9 b 3 962 18.2 22.4 b 4 1,130 21.3 21.5 c 5 1,305 24.6 20.3 d 6 1,504 28.4 19.2 e 7 1,761 33.2 18.4 e 8 2,118 40.0 17.5 f 9 2,713 51.2 16.0 g 10 5,112 96.5 14.5 h a  Using Duncan’s test on the average tariff faced. Average tariffs are significantly different across deciles except when they are tagged with the same letter of the alphabet. Figure 1: Tariff faced by income decile in agriculture and averaged across all sectors 35 All sectors Agriculture 30 25 20 Tariff faced (%) 15 10 5 0 1 2 3 4 5 6 7 8 9 10 Income decile 180 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 4.2 Rural households, women and informal Distance to an urban area makes a difference workers face higher tariffs in the Households within 0 km from the state capital or the export market nearest recognized city or urban area , face the lowest Rural vs urban tariffs in international markets at 9.6%.10 More than 10% of the respondents live within such an urban zone. Households in rural areas face an average tariff which Next, those living within one to fifty kilometers from the is 10.9 percentage points higher than their urban closest urban center faced an average tariff of 14.3% counterparts (22.6% versus 11.7%). Rural households (Table 3). The average tariff for the center dwellers is consistently face higher tariffs, with the widest 14.8 percentage points lower than the 24.4% tariff faced differences in the lower income deciles (Table 2). by those who are farther than 600 km from the city Table 2: Average tariff faced by income decile in rural and urban areas Income Tariff Faced (%) Decile Rural Urban Differencea 1 26.4 15.0 11.4 2 25.0 13.4 11.6 3 24.5 13.6 10.9 4 23.6 13.3 10.3 5 22.5 12.4 10.1 6 21.3 13.1 8.2 7 20.9 11.7 9.2 8 20.5 11.0 9.5 9 19.5 10.0 9.5 10 19.2 8.8 10.4 Overall 22.6 11.7 10.9 a  The difference in all income deciles is statistically significant. Table 3: Average income and tariff faced by distance to the nearest urban area Distance (km) to Average Income Number of Duncan’s test nearest urban area (Rupees) respondents Average Tariff (%) outcomea GT 600 1,002 428 24.4 a 401–600 1,116 1,883 17.3 bc 301–400 1,199 3,509 16.3 dc 201–300 1,235 4,992 15.7 d 151–200 1,406 2,806 17.5 b 101–150 1,429 3,215 17.5 b 51–100 1,415 2,444 17.5 b 1–50 1,881 1,685 14.3 e 0 (Center) 2,185 2,610 9.6 f Duncan’s test on the average tariff faced is used to assess whether averages across deciles are statistically different. Average tariffs are significantly different across deciles when they a  are tagged with different letters. They are not significantly different when they are tagged with the same letter of the alphabet. 10 For example, Bagalkot in the state of Karnataka is 518 km to Bangalore, which is the state capital. However, Bagalkot is only 406 km to Pune, a prominent urban area in the state of Maharashtra; this is therefore the closest distance to an urban area, which is used in the analysis. This measure is taken from Das et al. (2015).. 181 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES (less than 2% of respondents). Except for those living that lower-income groups face higher tariffs. However, the farthest at more than 600 kilometers, any other the number of women in the higher-income deciles is distance from the nearest urban center does not make much fewer than the number of men. In the 10th income any significant difference in tariffs faced (averages with decile, for example, women only account for 6% of the overlapping letters b to d). This confirms the finding respondents. shown above that people living in rural areas face higher Formal vs informal workers tariffs than people living in urban areas, and emphasizes that it is distance to an urban center, not just the inherent The household surveys reporting the principal economic classification of each district, that matters. The results activity during the reference week also identified the type also affirm that the poor face higher tariffs since, as of enterprise individuals were employed in. From this shown in Table 3, incomes are lower for those who live information, each enterprise was classified as belonging far from urban conglomerates. either to the formal or to the informal sector. The survey’s definition of the informal sector is an enterprise Men vs women employing less than 10 workers. Unfortunately, only 54% Data on individuals within the sample household of respondents reported the number of employees in the reporting a principal economic activity and income during enterprise in which they worked. Nonetheless, workers in the reference week complemented the household-level informal sector enterprises face an average 9.8% tariff, data. Using actual wages, the same methodology was which is significantly higher than the 7.2% for enterprises used to assign each individual to the appropriate income classified as in the formal sector. decile. 4.3 Globalization did not narrow the bias Overall, the average tariff facing men is 6 percentage against the poor points lower than that facing women (Table 4). Women While global tariffs fell from 1996 to 2012, the tariffs consistently face higher tariffs across all deciles, except facing the poor, workers in rural areas and women for the highest decile where men face tariffs that are on remained higher than those facing the rich, workers in average one percentage point higher than those faced urban areas and men, respectively (Table 5). In fact, the by women. Interestingly, in the two highest deciles the reduction in tariffs was slower for workers in rural areas average income of women is higher than that of men. and for women, while the tariff gap between the highest This further bolsters the finding in the previous section Table 4: Weekly wage and tariff faced by gender and income decile Wage Weekly Wage (Rupees) Tariff Faced (%) Decile Men Women Men Women Differencea 1 208 206 20.4 21.7 -1.3 2 386 382 22.2 22.6 -0.4 3 529 522 19.9 21.5 -1.6 + 4 666 663 19.9 20.2 -0.4 5 767 744 18.1 21.7 -3.6 + 6 934 920 15.8 18.2 -2.4 + 7 1,113 1,091 14.7 19.7 -4.9 + 8 1,419 1,401 12.0 15.1 -3.1 + 9 2,190 2,254 7.8 8.6 -0.8 10 8,268 8,508 4.6 3.6 1.0 + Overall 1,675 720 14.4 20.4 -6.0 + The sign + denotes that the difference between the average tariff faced by men and by women is significant (based on a t-test). a  182 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES and lowest income decile remained unchanged over the 4.5 Preliminary evidence that a 16-year period. coordination problem exists Table 5: Average tariff reduction between 1996 and 2012 Higher tariffs on goods produced by poor workers likely (change in percentage points) reflects efforts by India’s trade partners to protect sectors where their own poor work. The same pattern can be Sample Reduction (%) seen in the tariffs on India’s imports, by matching India’s Decile 1 (lowest 10%) 2.6 own applied MFN tariffs in 2012 by sector to the income Decile 10 (highest 10%) 2.5 data from the survey. Goods produced by the richest Rural 2.4 workers (in the 10th decile) face a tariff of less than one- Urban 3.3 Women 2.5 third the level on goods produced by the poorest workers Men 2.7 (the lowest 2 deciles of the population—Table 7). 4.4 Non-tariff measures are biased against Table 7: India’s import MFN applied tariffs in 2012 by income decile the poor Income In addition to higher tariffs, the products produced Decile Applied MFNa by poor workers face a greater number of non-tariff 1 35.6 c measures. On average, the poorest (income deciles 1 and 2 39.0 a 2) face some 200 different types of NTMs while workers 3 37.5 b belonging to the top income decile face only 127 (72 less 4 35.3 c NTMs compared to the first decile). Unlike for tariffs, 5 33.8 d which consistently decline across the income deciles, 6 29.9 e the total number of NTMs for the poorer 5 deciles are 7 28.0 f not significantly different; the total count varying from 8 23.3 g 192 to 201. On the other hand, the number of NTMs falls 9 16.9 h significantly between the 9th and 10th deciles, with the 10 11.6 i Using Duncan’s test on the average count of MFNs faced. Averages tagged with a  latter facing 39 NTMs less than the former (Table 6). the same letter of the alphabet are not significantly different. Table 6: Total count of all types of NTMs faced, by income decile Similarly, the applied MFN tariffs in the United States and China (taken from the WTO Integrated Database Income Number of based on the Member’s own data notification) tend to Decile NTMs Faced Duncan's testa be higher for goods produced by poorer workers (based 1 199 ab on average wage data from the UNIDO database). Goods 2 201 a produced by workers in the fifth, or richest, quantile face 3 201 a tariffs that are markedly lower than on goods produced 4 192 b by workers from the first, or poorest, quantile across all 5 194 ab three countries (Table 8). While the average tariff falls 6 180 c 7 182 c consistently across income groups in China and India, 8 169 d the average U.S. tariff for sectors corresponding to the 9 166 d richest quantile (3.7%) is close to that of the second- 10 127 e poorest quantile (3.8%). However, this appears to be a  Using Duncan’s test on the average count of NTMs faced. Averages tagged with the driven by the very high tariff on tobacco.11 Since the same letter of the alphabet are not significantly different. 11 In the US some tobacco products have tariffs as high as 350%, corresponding to ISIC 1200, while the average wage in the tobacco sector is very high, with individuals employed in the sector being included in the top quantile of the income distribution. 183 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES U.S. tariffs on in-quota tobacco imports from selected In order to assess the obstacles that the “poor” face in partners are lower than the tariffs quoted here, the accessing international markets, this paper analyzed actual applied average duty is lower. Excluding tobacco, tariffs faced on goods produced by Indian workers, by the average tariff for the richest quantile is only 0.8%, sector of occupational activity. It then calculated the less than a fifth of the average tariff of the poorest average tariff faced by individuals classified according quantile. These results reinforce the view that there is to their position in the overall income distribution. a coordination problem—the “poor” are disadvantaged Individuals also were classified according to other in terms of market access because trading partners characteristics, such as gender, whether they work in the appear to impose trade restrictions disproportionately formal or informal sector or whether they work in urban on sectors that employ poor households. or rural areas. The results show that tariffs faced in destination markets Table 8: Import tariffs by wage or income quantiles for selected countries are higher for goods produced by individuals in lower income groups. Households in rural areas face higher Wage/ Average MFN for indicated year (%) average tariffs than their urban counterparts, with Income tariffs consistently higher for households located further Quantile USA 2008 China 2010 India 2012 away from an urban center. Women consistently face Q1 4.5 12.7 37.3 higher tariffs than do men. Small informal enterprises Q2 3.8 10.9 36.4 Q3 3.3 9.8 31.8 also face higher tariff barriers than do large formal Q4 2.3 9.3 25.6 enterprises. Moreover, the sectors the poor work in are Q5 3.7 a 9.2 14.3 also disproportionately burdened by non-tariff measures. a  This goes down to 0.8% if the high tariff on ISIC 1200 is excluded from the analysis. These findings underline the fact that the poor could be paying the highest penalty if efforts to reduce barriers to trade stall, or worse, countries retreat from the 5. Conclusions liberalization already achieved. Facilitating access to This paper shows that tariffs tend to be higher and non- external markets for the goods that the “poor” produce tariff measures more prevalent for the poor, thus limiting is key to maximize the potential benefits of trade for their opportunities to access international markets. poverty. Countries individually may seek to protect their own The recent debate on globalization and income inequality “poor” from foreign competition by raising tariffs on the has often indicated that the reduction of trade costs has goods these households/individuals produce.12 There is contributed to rising inequality. This paper reveals a new indeed evidence suggesting that trade policy is biased aspect in this debate. The problem may not be too much towards imposing barriers on sectors that employ poor globalization, but too little “inclusive” globalization. Many individuals. But, when all countries protect the sectors sectors that employ a large proportion of those classified where the poor work and if the poor are employed in as “poor” still face higher barriers to trade. More research similar kind of sectors in all countries, a coordination is needed to assess the general equilibrium effects of problem arises. That is, the goods (and services) removing this unbalanced access to international markets produced by the poor will face higher barriers to access and whether this would help reduce income inequality. international markets than those produced by the non- However, this paper explains why the reduction of the poor. This will depress global demand for the goods (and trade costs for the goods that the poor (rural workers services) that the poor produce, thus worsening their and women) produce would require international income prospects. cooperation. 12 This can be the outcome of legitimate concerns, such as the need to protect jobs of workers where labor market frictions or other reasons prevent them from easily moving to more competitive sectors. 184 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES References Abbott, P. C. (2012). “Export restrictions as stabilization responses to food crisis.” American Journal of Agricultural Economics 94(2): 428–434. Atkin, David and Dave Donaldson (2015). “Who’s Getting Globalized? The Size and Implications of Intra-national Trade Costs.” NBER Working Paper No. 21439. Banerjee, Abhijit, and Andrew Newman (2004). “Inequality, growth and trade policy”. Massachusetts Institute of Technology, Department of Economics, and University College London, Department of Economics. Unpublished Manuscript. Campa, J. and L. Goldberg (2002), “Exchange Rate Pass-Through into Import Prices: A Macro or Micro Phenomenon?”NBER working paper No. 8934. Das, Samarjit, Chetan Ghate, and Peter E. Robertson (2015). “Remoteness, Urbanization, and India’s Unbalanced Growth.” World Development 66, Pages 572-587. Donaldson, Dave (2016). “Railroads of the Raj: Estimating the impact of transportation infrastructure.” American Economic Review, forthcoming. Emran, M. Shahe, and Zhaoyang Hou (2013). “Access to Markets and Rural Poverty: Evidence from Household Consumption in China.” The Review of Economics and Statistics 95, No. 2, Pages 682-697. Freund, Caroline and Caglar Ozden (2016).”Loss Aversion and Trade Policy”, World Bank Policy Research Working Paper No. 3385. Gouel, Christophe (2016). “Trade Policy Coordination and Food Price Volatility.” American Journal of Agricultural Economics 98 (4): 1018-1037 Grossman, Gene M. and Elhanan Helpman (1994). “Protection for Sale,” American Economic Review 84:4, pp.833–50. Han, Jun, Runjuan Liu, and Junsen Zhang (2012). “Globalization and wage inequality: Evidence from urban China” Journal of International Economics 87: 288–297. Ivanic, M. and W. Martin (2014). “Implications of Domestic Price Insulation for Global Food Price Volatility,” Journal of International Money and Finance 42:272-88. Kone, Z.L., M.Y. Liu, A. Mattoo, C. Ozden and S. Sharma (2016). “Internal Borders and Migration in India,” mimeo, The World Bank, Washington D.C. Kovak, Brian K. (2013). “Regional Effects of Trade Reform: What is the Correct Measure of Liberalization?” American Economic Review 103(5): 1960–1976 McCaig, Brian (2011). “Exporting Out of Poverty: Provincial Poverty in Vietnam and US Market Access.” Journal of International Economics 85 (1): 102–13. 185 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES Nicita, A., M. Olarreaga and P. Guido (2014). “Pro-poor trade policy in Sub-Saharan Africa.” Journal of International Economics 92 (2014) 252–265. Nicita, A. (2009). “Who Benefited from Trade Liberalization in Mexico?” Journal of Development Economics, forthcoming. Porto, Guido (2010). “International Market Access and Poverty in Argentina.”Review of International Economics 18(2), 396–407. Szamosszegi , A., and C. Kyle (2011). An Analysis of State-owned Enterprises and State Capitalism in China, US-China Economic and Security Review Commission. Topalova, Petia (2010). “Factor immobility and regional impacts of trade liberalization: evidence on poverty from India.” American Economic Journal: Applied Economics 2 (4), 1–41. Ural Marchand, Beyza (2012). “Tariff pass-through and the distributional effects of trade liberalization” Journal of Development Economics 99, issue 2: 265-281. Winters, L. Alan, Neil McCulloch and Andrew McKay (2004). “Trade Liberalization and Poverty: The Evidence So Far.” Journal of Economic Literature 42(1): 72-115. Wood, A. (1997). “Openness and Wage Inequality in Developing Countries: The Latin American Challenge to East Asian Conventional Wisdom.” World Bank Economic Review 11, no. 1: 33-57. 186 TRADE AND POVERTY REDUCTION: NEW EVIDENCE OF IMPACTS IN DEVELOPING COUNTRIES 187 Trade and Poverty Reduction: New Evidence of Impacts in Developing Countries Photo credits: Chhor Sokunthea (World Bank) : page 151 World Bank Flickr Images: pages 27, 103, 114, 121 and 175. Bigstock Images: pages 10, 17, 19, 20, 22, 37, 38, 39, 49, 59, 71, 76, 85, 86, 99, 119, 123, 127, 149 and 187. Global trade has contributed strongly to reducing poverty but important challenges remain in making trade work for the poorest. This publication presents eight case studies to reveal how trade can help to reduce poverty in developing countries. It focuses on four constraints faced by the extremely poor – namely that they tend to live in rural areas, work in the informal sector, live in fragile and conflict-affected regions and face gender inequality. The case studies identify ways to overcome these constraints, including through the adoption of policies that maximize the contribution of trade to poverty reduction. The studies also highlight the ongoing gaps in data and research that constrain policy-making. The publication is a follow-up to The Role of Trade in Ending Poverty, co-published by the WTO and the World Bank in 2015, which examined the challenges the poor face in benefiting from trade opportunities. The country-specific approach of this new publication complements the global perspective of the previous report. ISBN 978-92-870-4522-5