WPS6266 Policy Research Working Paper 6266 Monitoring Export Vulnerability to Changes in Growth Rates of Major Global Market Claire H. Hollweg Daniel Lederman José-Daniel Reyes The World Bank Poverty Reduction and Economic Management Network International Trade Department November 2012 Policy Research Working Paper 6266 Abstract Interest in assessing the impacts on developing countries for commodities and differentiated products. This of changes in major markets’ economic performance has methodology is applied to six developing countries, risen in tandem with global economic uncertainty over one from each World Bank region, selected to be short- and medium-term growth prospects. This paper otherwise similar yet differ in terms of the level of proposes a methodology to measure the vulnerability exposure to major global markets as well as the product of a country’s exports to fluctuations in the economic composition of their export basket. Although the results activity of foreign markets. Export vulnerability depends suggest differences in elasticity estimates across regions first on the overall level of export exposure, measured as well as product categories, the principal source of as the share of exports in gross domestic product, and international heterogeneity in export vulnerability results second on the sensitivity of exports to fluctuations in from differences in export exposure to global markets. foreign gross domestic product. The authors capture This result calls for developing countries to diversify this sensitivity by estimating origin-destination specific their export markets rather than shielding themselves elasticities of exports with respect to changes in foreign from international markets, which would actually raise gross domestic product using a gravity model of trade. economic risk and vulnerability. Furthermore, export vulnerability is computed separately This paper is a product of the International Trade Department, Poverty Reduction and Economic Management Network. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank. org. The author may be contacted at jreyes2@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Monitoring Export Vulnerability to Changes in Growth Rates of Major Global Markets Claire H. Hollweg Daniel Lederman José-Daniel Reyes * Keywords: Economic Integration, Empirical Studies of Trade, Macroeconomic Impacts of Globalization JEL Classification Codes: F14, F15, F62 Sector Board: Economic Policy (EPOL) * The authors are, respectively, Consultant, Lead Economist, and Economist of the International Trade Department of the World Bank (PRMTR). The authors thank Bernard Hoekman, Mona Haddad, and Emmanuel Pinto Moreira for helpful comments. All remaining errors are our responsibility. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not represent the view of the World Bank, its Executive Directors, or the countries they represent. 1. Introduction Although trade openness promotes economic growth and poverty reduction, fears that increasingly open markets enhance the vulnerability of developing countries to global economic fluctuations have risen lately, as trade has become a primary channel of transmission of external shocks. 1 In fact, during the 2008-2009 global crisis, world GDP fell by 5.5 percentage points while world exports dropped by 20 percentage points. It is therefore not surprising that governments and multilateral agencies are interested in assessing the potential impacts on developing countries of changes in economic performance in major global markets. These concerns have risen in tandem with global economic uncertainty over the short- and medium- term growth prospects of major world markets such as the Eurozone economies, the slowdown in China, and recent declines in global trade. It is also worth noting, however, that closed economies would face economic risks emanating from domestic markets or domestic policies, and thus openness actually reduces overall economic risk by diversifying the sources of risk. This paper proposes a standard methodology to assess the vulnerability of a country’s exports to fluctuations in the economic activity of foreign markets. The trade transmission of external shocks depends not only on a country’s level of export exposure through bilateral trade flows with specific markets, but also on the sensitivity of exports to foreign demand. As such, the analysis combines two key indicators to construct an overall measure of export vulnerability. First, a country’s level of export exposure to world markets is captured as its share of exports in gross domestic product (GDP). Second, the sensitivity of exports to demand of foreign markets is the elasticity of each country’s exports with respect to fluctuations in foreign-market GDP. This elasticity is estimated following the micro-founded gravity model proposed by Helpman et al. (2008). Although our base specification controls for zero trade flows and self-selection of firms into export markets, we acknowledge that it does not properly account for time-varying factors affecting trade costs – the so called multilateral resistance terms in Anderson and van Wincoop (2003). To address this concern we check the robustness of our results using rolling-window regressions and find that over time, differences in the estimated elasticities are statistically insignificant in most if not all of the cases. We take this to suggest that unobservable factors affecting trade costs are constant over 2003-2010. As such, we argue that failing to control for the multilateral resistance terms does not result in biased estimates, lending support to our empirical specification. The export vulnerability measure is computed for six developing countries, one from each World Bank region, selected to be otherwise similar yet differ in terms of the level of exposure to major 1 The literature on trade and growth is deep and growing. Among the studies that find a statistically significant positive relationship between trade openness or trade policy liberalization and economic growth include Dollar (1992), Edwards (1992, 1998), Harrison and Hanson (1999), Frankel and Romer (1999), Foster (2008), Wacziarg (2001), Feyrer (2009), Kim (2011), and Brückner and Lederman (2012). 2 global markets as well as the product composition of their export basket. 2 We consider two groups of major global markets: The first group composed of the so called ‘old growth poles’ includes the United States, Europe, and Japan. 3 The second group, the ‘new growth poles’, consists of Brazil, Russia, Indonesia, India, China, and Korea. The changing dynamics of global trade indicate that in the future emerging markets will be the likely drivers of export growth and diversification of developing countries (Haddad and Shepherd 2011). We also disaggregate a country’s total merchandise exports into differentiated products and commodities, as we expect their elasticities to be different (Feenstra 2004). As such, the analysis allows for a comparison of export vulnerability across exporting countries, growth poles, and product types. Our main findings can be summarized as follows. Although our results suggest differences in elasticity estimates across countries as well as product categories, we find that the main source of international heterogeneity in export vulnerability is export exposure. It cannot be overstated, however, that openness itself provides risk diversification away from domestic shocks. Hence these findings could be interpreted as a call to diversify markets rather than to raise barriers to trade, which would actually increase economic risk. We also find differences in the sensitivity of differentiated product exports versus commodity exports, although the differences in magnitude depend on the country of interest. These differences could be explained by whether the products are final goods or intermediate inputs. This paper relates to the literature on the response of trade to economic downturns. Much of the recent literature is in the context of the collapse of international trade during the 2008-9 global crisis. Levchenko et al. (2010) use import demand functions derived from a standard international business cycle model and note significant shortfalls in actual US imports relative to what the model would predict, given observed overall economic activity and prices. The authors additionally find that sectors used as intermediate inputs as well as those with greater reductions in domestic output experienced significantly higher percentage reductions in both imports and exports. Similarly, to reconcile the fact that world trade fell dramatically more than world GDP, Baldwin (2009) points to international supply chains that amplified the initial income falls through ‘compositional’ and ‘synchronicity’ effects. However, a few papers consider the income elasticity of trade in their analysis, including Irwin (2002), Freund (2009), and Hooper et al. (2000). Irwin (2002) uses an auto-regressive distributed lag specification for world export volumes of high income countries and finds that since the mid 1980s trade has been more responsive to income, with an estimated short-run elasticity of 1.55 2 Although only six countries were selected, the methodology allows the analysis to be conducted for any country in the world and has already been conducted for each developing country in the Latin America and Caribbean (LAC) and Europe and Central Asia (ECA) regions. 3 Europe consists of 27 countries: Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, and United Kingdom. 3 for the years 1985-2000. One potential explanation the authors provide is the changing composition of world trade from commodities to manufactured goods. Freund (2009) analyzes the response of trade to global downturns and estimates the relevant elasticity using a regression of world trade growth on world real GDP growth. Freund (2009) also finds that the elasticity has increased in recent decades, with an estimate of approximately 3.7 for the 2000s. Furthermore, the author notes significant variation in declining trade across industries and regions, with food and beverages the least affected and crude materials and fuels the most affected, while countries in the Europe and Central Asia region were relatively less affected and Middle East and North Africa countries were the most affected. Hooper et al. (2000) estimate short-run and long-run trade elasticities using co-integration techniques for all of the G-7 countries. Long-run trade elasticities are found to be close to 1, but with smaller short-run elasticities. To our knowledge the only paper to use the gravity model of trade to estimate an income elasticity of imports is Kwack et al. (2007). The authors find income elasticity estimates in the range of 1.05 to 3.10 with the income elasticity of US imports higher than other major trading partners. However these results might be weak since the pooled time series are estimated with OLS and zero trade flows are not controlled for. Furthermore, the authors consider the average elasticity estimate across all trading partners and products. Thus our paper advances the existing literature by not only using a methodology that considers both exposure and sensitivity but also by enhancing the validity of the elasticity estimates through a robust gravity model of trade. The remainder of this paper is organized as follows. Section 2 details the methodology as well as the empirical specification of the gravity model of trade. Section 3 describes the data, and Section 4 presents the results of the exposure, sensitivity, and vulnerability analyses. Robustness checks using rolling-window elasticity estimates as well as an alternative categorization of products are presented in Section 5. Section 6 concludes. 2. Methodology and Empirical Model Export vulnerability refers to the magnitude of the response of merchandise exports to changes in demand in major global markets. More specifically, export vulnerability is the product of two indicators: The first one is a measure of the degree of exposure of exports to major global markets; the second is an estimate of the sensitivity of exports to changes in income in these major global markets. The former indicator is defined as total exports to a particular destination as a share of the exporter’s GDP, whereas the latter is an estimate of the elasticity of exports with respect to fluctuations in major global markets’ GDP. The multiplication of these two indicators provides an estimate of the impact of a 1 percent change in a given major global market’s GDP on exports to that market as a share of the country of interest’s GDP. Export vulnerability is computed separately for differentiated product and commodity exports. There are reasons to expect the elasticity to differ across product types. Demand may likely fall 4 more for manufactured items such as cars and clothing than commodities such as food and energy for a given fall in income, because the latter are essential for households’ daily life. We also allow elasticities to differ across export destinations, as consumer preferences may differ across cultures or high income countries may not respond the same as low-income countries depending on the composition of the import basket. For example, high income countries may import relatively more luxury food items than food staples. Conversely, more advanced capital markets in high income countries may allow consumption to be smoothed over time through borrowing and savings. While the exposure indicator is straightforward to compute, the estimation of the export elasticity requires a methodological discussion. Motivated by Feenstra et al. (2001), we make use of the gravity model of trade to estimate destination-specific elasticities for homogeneous and differentiated products separately. 4 The gravity model has been extensively used in international trade due to its intuitive empirical and theoretical appeal. Anderson and van Wincoop (2003), Feenstra (2004), and Baldwin and Taglioni (2006), among others, present exhaustive literature reviews on the gravity equation as applied to international trade. Our empirical specification of the gravity model follows the micro-founded gravity model of Helpman et al. (2008). This methodology controls not only for zero trade flows but also for self-selection of firms into export markets. It involves a two-stage estimation procedure that uses an equation for selection into trading partners in the first stage and a trade flow equation in the second stage. We estimate regional gravity models and use interaction terms to compute exporter-importer specific income elasticities. 5 For each World Bank region, the second stage equation of our empirical model is given by: Equation 1 ln�𝑋𝑖𝑗𝑡 � = 𝛽0 + 𝛽1 ln�𝐷𝑖𝑗 � + 𝛽2 ln(𝐺𝐷𝑃𝑖𝑡 ) + 𝛼1 ln�𝐺𝐷𝑃 𝑗𝑡 � + 𝛽3 �𝑜𝑛𝑡𝑖𝑗 + 𝛽4 𝑙𝑎𝑛𝑔𝑖𝑗 ∗ � +β z + 𝛽5 �𝑜𝑙𝑖𝑗 + 𝛽6 �𝑜𝑚�𝑜𝑙𝑖𝑗 + � 𝛼𝑘 [γi ∗ ln(𝐺𝐷𝑃𝑘𝑡 )] + β7 λ � ∗ + β9 z ∗2 �ijt ijt 8 ijt 𝑘∈�𝑜𝑙𝑒𝑠 ∗3 + β10 z �ijt + γi + γj + γt + 𝜇𝑖𝑗𝑡 where 𝑋𝑖𝑗𝑡 is the total export value of country 𝑖 to country 𝑗 in year 𝑡. The set of exporter countries are those classified as low- and middle-income countries within the region by the World Bank (see Appendix 1 for the list of included countries). Importer countries are the 4 Despite the fact that the standard theoretical underpinning of the gravity model entails that countries trade in differentiated products (see Feenstra 2004), the gravity equation also empirically works for trade in homogeneous products. Feenstra et al. (2001) reconcile this conflict by deriving a gravity relationship from a Brander and Krugman (1983) reciprocal dumping model with purely homogeneous products. 5 We have chosen to estimate the gravity model separately for each World Bank region as opposed to a single dataset containing all countries in the world. 5 complete set of world economies (182 countries). 𝐷𝑖𝑗 is the “great circle� distance between the capital of the exporter and the capital of the respective importer. 𝐺𝐷𝑃𝑖𝑡 and 𝐺𝐷𝑃 𝑗𝑡 are the exporter’s and the importer’s GDP in year 𝑡, respectively. 𝐶𝑜𝑛𝑡𝑖𝑗 , 𝑙𝑎𝑛𝑔𝑖𝑗 , �𝑜𝑙𝑖𝑗 , and �𝑜𝑚�𝑜𝑙𝑖𝑗 are dummy variables that are equal to 1 if the countries shares a border, have a common language, have ever had colonial ties, and had a common colonizer after 1945, respectively. In order to determine if the elasticity of our country of interest with respect to the change in income of a given growth pole (𝜀ik ) is different from the regional average elasticity across all destination markets (𝛼1 ), we interact the country dummy (γi ) with each growth pole’s GDP in year 𝑡. 6 The destination-specific elasticity is therefore computed as 𝜀ik = 𝛼1 + 𝛼 ∀ 𝑘 ∈[USA, EUN, JPN, BRA, RUS, IND, IDN, CHN, KOR]. λ � ∗ is the standard inverse mills 𝑘 ij ratio that takes into account the possible selection bias given that we only observed bilateral flows with positive exports. 7 The last cubic polynomial controls for the underlying unobserved firm-level heterogeneity. 8 Finally, γi , γj , and γt are sets of exporter, importer, and year fixed effects. The time period of our empirical exercise is 2006-2010. We run six regional gravity equations, one for each World Bank region: East Asia and Pacific (EAP), Europe and Central Asia (ECA), Latin America and the Caribbean (LAC), Middle East and North Africa (MENA), and Sub- Sahara Africa (AFR). Within each region, we run two regressions: one for commodities and another for differentiated products. This brings the total of gravity models to 12 and the total of estimated elasticities to 216 (six regions by two types of products by three countries of the old growth poles and six countries of the new growth poles). While our empirical specification is based on a micro-founded gravity model where both the selection bias due to many zeros in the trade data and the underlying unobserved firm level heterogeneity are properly accounted for, a caveat should be noted. The empirical procedure developed by Helpman et al. (2008) is design for cross-sectional data, not for panel data. A potential source of bias arises because our chosen set of fixed effects do not control for time- variant factors affecting trade costs – the so-called multilateral resistance terms in Anderson and van Wincoop (2003). A proper estimation of the model in a dynamic context would control for 6 Let 𝑖 represent the country of interest. 𝛾𝑖 is the country dummy variable associated with 𝑖 . 7 The inverse mills ratio (𝜆̂∗ ̅𝑖𝑗𝑡 ) is obtained from the selection equation, or the first stage estimation. This is a probit model where we regress the probability of observing bilateral exports between country pairs on the same set of ∗ covariates used in the second stage (𝑧𝑖𝑗𝑡 ). The interaction terms are not part of the first stage estimation. Our exclusion restriction is a dummy variable that equals 1 if countries were the same country at some point of time, since this information should explain the existence of historical bilateral trade ties but, arguably, not the level of ∗ 𝜙�𝑧𝑖𝑗𝑡 � ̂∗ exports. 𝜆 ̅𝑖𝑗 is defined as ∗ . 𝛷�𝑧𝑖𝑗𝑡 � 8 It controls for the potential important effects of trade barriers and country characteristics on the share of exporting ̂𝑖𝑗𝑡 firms (see Helpman et al. 2008). 𝑧̅ ∗ � is defined as 𝑧 ∗ ̂∗ ̅ � ∗ 𝚤𝚥𝑡 + 𝜆𝑖𝑗𝑡 , where 𝑧𝚤𝚥𝑡 is the error from the first stage. 6 time-varying country dummies and time-invariant pair country dummies (Baldwin and Taglioni 2006). Unfortunately, we cannot pursue this approach because the estimated coefficient for the importer GDP would be dropped from the regression due to collinearity, making it impossible to compute elasticities. The literature proposes alternative ways to handle this problem: First, Bergstrand (1985, 1989) and Baier and Bergstrand (2001) add country prices to the gravity equation by including GDP deflators as proxies. Although this technique applied in our dynamic context would allows us to compute elasticities, an objection to using published indexes is that the myriad of costs involved in cross border transactions are unlikely reflected in aggregate price indexes. Second, Martin et al. (2008) exploit a convenient feature of the CES demand structure of the underlying theoretical model that makes relative imports from a given exporter independent of the characteristics of third countries. By choosing the imports from a given country (in their case, the United States) as a benchmark for all imports of each importing country, this methodology allows the authors to eliminate price indexes in the bilateral trade equation. The disadvantage of this approach for our purposes is that the importer GDP cancels out in the relative version of our gravity model. Rather, we check the robustness of our results using rolling-window regressions in section 5. This allows us to determine if the elasticity estimates are statistically different over time, as constant estimates could suggest that unobservable factors affecting trade costs are not that important to meaningfully bias our estimates. The use of rolling regressions is ubiquitous in the macroeconomics time-series literature covering a vast number of topics, but, as far as we know, there are no applications to the gravity model of trade. 3. Data Bilateral trade in current USD comes from the United Nations’ (2012) COMTRADE database, accessed through WITS. 9 Exporter and importer GDPs are from the World Bank’s (2012) World Development Indicators and, consistent with the underlying theory, are measured in nominal USD. 10 Bilateral country characteristics come from the CEPII distance dataset (Mayer and Zignago 2011). The analysis focuses on a country’s total merchandise exports disaggregated into differentiated product exports and commodity exports. A commodity is defined according to the International Monetary Fund’s (2012a) Primary Commodity Prices and is coded at the six-digit HS 1988/92 9 Instead of combining exports and imports for each country pair to construct symmetric trade flows, we use the unidirectional value of trade from country 𝑖 to country 𝑗 and introduce both importing and exporting country fixed effects. The exports from country 𝑖 to country 𝑗 are inferred using the imports reported by country 𝑗 from country 𝑖 . This technique is commonly used to minimize the risk of underreporting due to the fact that customs agencies usually monitor imports more closely than exports. 10 On the use of trade and GDP data in current USD, see Baldwin and Taglioni (2006). 7 level. 11 The same analysis using Rauch’s (1999) ‘conservative’ classification of commodities is presented for robustness, defined as products traded on international exchanges as well as products not traded on international exchanges but with relevant “reference� prices. 12 Differentiated product exports are then defined as the subset of total merchandise exports that are not commodities. Countries were selected to be similar in terms of three criteria, but differ in terms of export composition. The similarity criteria were: trade openness defined as exports plus imports to GDP over 50 percent; population between 5 and 15 million; and lower or upper middle-income according to the World Bank classification. Table 1 presents the countries selected from each region. In the sample, Bulgaria and Tunisia are the most open in trade, each with a trade openness over 100 percent, while Sri Lanka is the least open with a trade openness just over 50 percent. Given that the South Asian region is composed of very large countries, we chose Sri Lanka although its population is around 20 million. Lao PDR is the smallest with 6.1 million followed closely by Bulgaria. Bulgaria and Tunisia are classified as upper middle-income. Table 1: Country Selection Criteria Trade Income High Commodity Links to Region Country Population Openness Level Exporter? Growth Poles East Asia & Pacific Lao PDR 6,111,680 76 1 yes China, EU (copper, wood, electrical energy) Europe & Central Asia Bulgaria 7,534,000 122 2 no EU (copper, wheat, apparel) Latin America & Caribbean Guatemala 14,037,799 61 1 yes US (coffee, bananas, sugar, apparel) Middle East & North Africa Tunisia 10,439,200 103 2 no EU (petroleum, electronic equipment, apparel) South Asia Sri Lanka 20,667,630 55 1 no US, EU, India (tea, textiles) Sub-Saharan Africa Zambia 12,676,589 72 1 no China (copper, tobacco, cobalt) Notes: This table lists countries selected from each World Bank region. Countries were chosen according to three criteria: i) population between 5 to 15 million, ii) openness to trade measured as exports plus imports to GDP greater than 50 percent, and iii) countries classified as middle-income countries according to the World Bank classification. Given that the South Asian region is composed of very large countries, we chose Sri Lanka although its population is around 20 million. Population and trade openness values are the average for the period 2008-2010. Income level and the composition of the export basket are for 2010. An income level of 1 (2) indicates a low (upper) middle-income country according to the 2010 World Bank classification of countries, defined as a GNI per capita between USD 1,026-4,035 or USD 4,036-12,475, respectively. Source: Authors’ calculations based on data from the World Bank’s (2012) World Development Indicators. 11 The commodities consist of the following HS 1988/92 classification codes: 0201, 0202, 0203, 0204, 0206, 0207, 0209, 0210 (meats); 03 (fish, excluding 030110); 0803 (bananas and plantains); 080510, 080512 (oranges); 0901 (coffee); 0902 (tea); 1001, 1003, 1005, 1006 (grains); 1201, 1020, 1026 (oil seeds / groundnuts); 1507, 1509, 1510, 1511, 151211, 151219, 151321, 1514 (vegetable oils); 1701 (sugar); 1801 (cocoa); 230120 (fishmeal); 2304 (soybean meal); 2601, 2603, 2604, 2606, 2607, 2608, 2609, 261210 (ores); 2701 (coal); 2709 (petroleum oils); 271111 (natural gas); 284410, 284420, 284430 (uranium); 4001-4006 (rubber); 41 (hides); 4401-4412 (timber, excluding 4402, 4404, 4406); 5101-2110 (wool); 5201 (cotton). 12 See Rauch (1999) for the ‘conservative’ classification of commodities coded at the four-digit SITC level. Concordances available in WITS are used to map these codes to the six-digit HS 1988/92 level. 8 4. Results 4.1 Export Exposure to Major Global Markets A key component of a country’s vulnerability to foreign economic fluctuations is its level of export exposure. The purpose is to provide a selection of countries that are otherwise similar, including overall openness, but differ in export exposure, in terms of product composition as well as destination. Table 2 highlights how the selected countries differ in export exposure, measured as the share of exports in GDP, across all markets as well as each growth pole destination for differentiated product and commodity exports as of 2010. Table 2: Exposure Measure: Exports over GDP (%), 2010 Old Growth Poles New Growth Poles US Japan EU27 Brazil Russia Indonesia India China Korea Bulgaria Differentiated 0.55 0.12 21.68 0.08 1.12 0.08 0.11 0.63 0.18 Commodities 0.00 0.00 2.50 0.00 0.01 0.00 0.00 0.04 0.07 Guatemala Differentiated 5.03 0.10 0.74 0.03 0.01 0.00 0.01 0.03 0.04 Commodities 3.24 0.34 0.58 0.01 0.09 0.00 0.09 0.05 0.28 Lao PDR Differentiated 0.82 0.40 1.91 0.00 0.01 0.01 0.00 1.23 0.02 Commodities 0.02 0.11 0.46 0.00 0.00 0.00 0.27 7.01 0.26 Sri Lanka Differentiated 3.58 0.29 6.04 0.05 0.14 0.16 0.95 0.16 0.11 Commodities 0.09 0.15 0.60 0.00 0.48 0.00 0.10 0.05 0.01 Tunisia Differentiated 0.60 0.21 24.88 0.28 0.19 0.03 0.68 0.27 0.07 Commodities 0.32 0.06 4.80 0.00 0.01 0.00 0.01 0.01 0.01 Zambia Differentiated 0.18 0.33 0.97 0.01 0.05 0.00 0.17 15.59 2.33 Commodities 0.00 0.00 0.32 0.00 0.00 0.02 0.13 0.33 0.00 Notes: This table shows the share of exports in GDP in 2010. This measure is computed separately for differentiated product and commodity exports across all markets, old growth poles, and new growth poles. Source: Authors’ calculations based on data from the United Nations’ (2012) COMTRADE and World Bank’s (2012) World Development Indicators. These countries differ first in their links to new and old growth poles. Lao PDR and Zambia have strong trade links with China, with more than half of Zambia’s exports absorbed by China. Europe is an important export destination for many of these countries, particularly Bulgaria and Tunisia capturing over two thirds of total exports, as well as Lao PDR and Sri Lanka. Guatemala’s main export destination is the United States. India is also an important export destination for Sri Lanka, but not as important as Europe or the United States. Lao PDR has significant trade flows with smaller non-growth pole economies, including regional destinations such as Vietnam. 9 These countries also differ in terms of export products. Lao PDR and Guatemala are high commodity exporters with more than one third of all exports classified as commodities. For Lao PDR, copper, wood, and electrical energy are the most important export products. For Guatemala, coffee, bananas, and sugar are the top export products followed by apparel. Seventy percent of Zambia’s exports in 2010 were of copper (refined and unrefined), primarily to China, with these codes classified as commodities under the Rauch classification but not the International Monetary Fund’s classification. Thus Zambia is considered a high commodity exporter depending on the classification. Bulgaria, Sri Lanka, and Tunisia export primarily differentiated products, although certain commodities are still significant. Bulgaria’s most important export products in 2010 included copper and wheat as well as apparel. Tunisia exports petroleum, electronic equipment, and apparel. Sri Lanka’s most important export product is tea, followed by textiles. 4.2 Sensitivity of Exports to Foreign GDP Fluctuations Export vulnerability depends not only on overall level of export exposure but also on the sensitivity of exports to fluctuations in foreign GDP. Even if an economy is extremely open, exports may not be affected if foreign demand is income inelastic. Conversely, an economy that is relatively closed with the exception of a few specialized exports could be extremely affected if its main export destination has elastic demand for those products. We capture this sensitivity by estimating Equation 1, separately for differentiated product and commodity exports. The coefficient on the importer GDPs reflects the regional average elasticity of exports with respect to all trading partners’ GDP. Dummy variables for the country of interest interacted with growth pole GDPs are then included to capture how this elasticity varies for the country of interest across each growth pole. Tables 3 and 4 report the estimation results of Equation 1 for differentiated product and commodity exports, respectively. 13 13 Results for the first stage estimation are available from the authors upon request. 10 Table 3: Differentiated Product Exports (2006-2010) Notes: This table presents the second stage estimation results of each regional gravity model given by Equation 1. The regression includes year, importer, and exporter fixed effects. Robust standard errors are in parentheses. *** Significant at the 1% level; ** significant at the 5% level; * significant at the 10% level. Coefficients for regression constant and fixed effects are suppressed. Source: Authors’ calculations. 11 Table 4: Commodity Exports (2006-2010) Notes: This table presents the second stage estimation results of each regional gravity model given by Equation 1. The regression includes year, importer, and exporter fixed effects. Robust standard errors are in parentheses. *** Significant at the 1% level; ** significant at the 5% level; * significant at the 10% level. Coefficients for regression constant and fixed effects are suppressed. Source: Authors’ calculations. 12 All standard gravity model variables have the expected signs and are significant in most regressions. Distance is negative and statistically significant except in the regressions for commodity exports in SAR and AFR regional estimations. Contiguity and common language are positive and significant in many of the regressions. Colony is rarely significant while common colonizer is statistically different from zero and positive in all cases except in SAR. There is significant heterogeneity in the regional elasticities to foreign demand across geographic regions and across product type within regions. For trade in differentiated products, regional elasticities are close to one. EAP, a region with high involvement in the Global Value Chains, exhibits the largest elasticity to changes in foreign demand. Commodities appear to be more responsive than differentiated products to changes in foreign demand only for the ECA and EAP regions. For the remaining regions, the commodity elasticity is lower and sometimes not statistically different from zero. The main results concerning the calculations of the relevant elasticities are reported in Table 5, calculated by summing the average regional elasticity and the interaction terms of the country of interest dummy and the growth pole’s GDP. Table 5: Sensitivity Measure: GDP Elasticity of Exports to Major Global Markets (2006- 2010) Old Growth Poles New Growth Poles US Japan EU27 Brazil Russia Indonesia India China Korea Bulgaria Differentiated 0.85 0.87 0.87 0.88 0.85 0.89 0.87 0.86 0.88 Commodities 1.22 1.20 1.28 1.20 1.22 1.27 1.28 1.28 1.26 Guatemala Differentiated 0.96 0.91 0.90 0.93 0.92 0.91 0.91 0.88 0.91 Commodities 0.63 0.60 0.56 0.59 0.59 0.63 0.54 0.55 0.66 Lao PDR Differentiated 1.14 1.13 1.16 1.13 1.13 1.01 1.04 1.18 1.14 Commodities 1.55 1.51 1.60 1.52 1.54 1.53 1.50 1.62 1.57 Sri Lanka Differentiated 1.00 0.96 0.98 0.96 1.00 0.99 0.95 0.98 0.99 Commodities 0.36 0.36 0.36 0.29 0.51 0.25 0.39 0.33 0.24 Tunisia Differentiated 0.85 0.87 0.89 0.92 0.91 0.89 0.89 0.85 0.83 Commodities 0.44 0.41 0.37 0.30 0.44 0.25 0.25 0.31 0.34 Zambia Differentiated 0.79 0.89 0.82 0.90 0.91 0.78 0.80 0.95 0.99 Commodities 0.12 0.20 0.17 0.08 0.07 0.24 0.29 0.25 0.14 Notes: This table presents the estimated elasticities of each country of interest’s exports with respect to foreign GDP of selected markets. The elasticities are the sum of the free standing importer GDP coefficient and the interaction term of each country of interest with each selected market’s GDP. Source: Authors’ calculations. 13 Interestingly, few differences in elasticity estimates are observed between major export destinations within each region and product category. This result contrasts with Kwack et al. (2007) who find elasticity estimates to vary across major destinations. For example, the import elasticity for the United States was found to be 1.89 in contrast with Korea of 1.38, India of 1.47, and China of 1.57, although closer to Japan of 1.99 and most European countries of around 2 for the 1990s. This result could point to the importance of disaggregating elasticity estimates across both exporters and importers, as variation in the relative importance of different export partners may be driving the differences found in Kwack et al. (2007). However, some variation does exist in elasticity estimates across regions and product categories. For Bulgaria and Lao PDR, commodity exports are more sensitive to fluctuations in major export destinations’ GDP than differentiated product exports. For Guatemala, another commodity exporter, it is the opposite. For Zambia, the difference between the two is striking, with commodities far less sensitive. Although Irwin (2002) predicts that commodity exports could have lower elasticities than differentiated product exports, this trend is not clear in the results. A more likely explanation is consistent with Levchenko et al. (2010) who found that sectors used as intermediate inputs experienced significantly higher reductions in trade in both imports and exports. Guatemala’s commodity exports are mainly final products, such as coffee, bananas, and sugar, whereas its differentiated product exports are textiles, such as fabrics, which are likely used as intermediate inputs into other garments. On the other hand, Bulgaria’s and Lao PDR’s commodity exports are mainly intermediate inputs, such as copper, wood, and wheat, while its differentiated product exports are final textiles such as apparel. Furthermore, these results are also consistent with Freund’s (2009) observations that trade in food and beverages were the least affected while trade in crude materials and fuel were the most affected. Lao PDR’s exports show the greatest sensitivity to demand of foreign markets with estimated elasticities all above 1. For example, if the GDP of China were to fall by 1 percent (relative to a hypothetical benchmark), Lao’s total exports to China would fall by 1.17 percent. Tunisia and Guatemala also have estimated elasticities close to one. Zambia’s, however, are noticeably lower than other regions, with elasticities between 0.5 and 0.7. This is due to the result that the average regional elasticity for Sub-Saharan Africa is lower than other regions, but also because most of Zambia’s interaction terms are negative. Overall, our elasticity estimates are all close to 1, which is predicted by the gravity model literature. Furthermore, our elasticity estimates are much lower than those found by Freund (2009) but close to what others have found in the literature (including Irwin 2002, Kwack et al. 2007, and Hooper et al. 2000). 4.3 Vulnerability to Foreign Fluctuations through Trade Finally, exposure indicators can be combined with estimates of elasticities to construct an overall measure of export vulnerability. Table 6 reports the estimated change in exports as a share of GDP in response to a one percent change in the GDP of each growth pole. 14 Table 6: Vulnerability of Exports to Major Global Markets (% of GDP), 2010 Old Growth Poles New Growth Poles US Japan EU27 Brazil Russia Indonesia India China Korea Bulgaria Differentiated 0.005 0.001 0.188 0.001 0.009 0.001 0.001 0.005 0.002 Commodities 0.000 0.000 0.032 0.000 0.000 0.000 0.000 0.001 0.001 Guatemala Differentiated 0.048 0.001 0.007 0.000 0.000 0.000 0.000 0.000 0.000 Commodities 0.020 0.002 0.003 0.000 0.001 0.000 0.000 0.000 0.002 Lao PDR Differentiated 0.009 0.005 0.022 0.000 0.000 0.000 0.000 0.015 0.000 Commodities 0.000 0.002 0.007 0.000 0.000 0.000 0.004 0.113 0.004 Sri Lanka Differentiated 0.036 0.003 0.059 0.000 0.001 0.002 0.009 0.002 0.001 Commodities 0.000 0.001 0.002 0.000 0.002 0.000 0.000 0.000 0.000 Tunisia Differentiated 0.005 0.002 0.222 0.003 0.002 0.000 0.006 0.002 0.001 Commodities 0.001 0.000 0.018 0.000 0.000 0.000 0.000 0.000 0.000 Zambia Differentiated 0.001 0.003 0.008 0.000 0.000 0.000 0.001 0.148 0.023 Commodities 0.000 0.000 0.001 0.000 0.000 0.000 0.000 0.001 0.000 Notes: This table presents the overall measure of export vulnerability, calculated by multiplying the exposure and sensitivity indicators. The numbers represent the impact of a 1 percent change in a given major global market’s GDP on exports to that market from each country of interest as a share of GDP. Source: Authors’ calculations. A 1 percent reduction in the GDP of Europe (relative to a benchmark) translates into a contraction of differentiated product exports of Bulgaria to Europe equivalent to 0.2 percent of the country’s GDP. This translates to roughly 90 million USD using the country’s 2010 GDP level. In contrast, for Guatemala, the contraction of differentiated product exports to Europe is equivalent to less than 0.01 percent of the country’s GDP, or 3 million USD using the country’s 2010 GDP level. This source of international heterogeneity in the vulnerability of exports, both across regions as well as growth poles, comes from exposure indicators, despite noticeable differences in elasticity estimates. As further illustration of these effects, the International Monetary Fund’s (2012b) World Economic Outlook predicts zero growth for Europe for 2012, resulting in a 2 percent contraction from its 2010 growth level. In such a scenario, the numbers in the table would need to be multiplied by a factor of two, translating into 0.4 percent of Bulgaria’s GDP or 190 million USD. 5. Robustness Checks This section discusses the robustness of our results, first to time-varying elasticity estimates using rolling-window regressions, and second to an alternative definition of commodities. 15 5.1 Rolling-Window Elasticity Estimates An alternative method to confront the assumption of time-invariant fixed effects in the baseline specification is to use rolling-window elasticity estimates. Although this does not solve for the fact that we cannot control for time-varying factors affecting trade costs in our base specification, if rolling-window elasticity estimates are constant over time, this could suggest that the fixed effects are constant over time. Such a result would help justify our baseline specification. On the other hand, if rolling-window elasticity estimates are not constant over time, we acknowledge that time-varying factors, not captured in our baseline specification, may be biasing our results. However, other things could also be changing. For example, the elasticity estimates may be different in crisis and non-crisis periods, and since our baseline estimates contain a crisis period, this could be a potential explanation for any observed changes over the last decade. We present recursive estimates of Equation 1 for four-year windows beginning in 2000-2003 through 2007-2010 and plot the elasticity estimates for the United States over time along with the 95 percent confidence bands in Figure 1. 14 Figure 1: Rolling-Window Elasticity Estimates Guatemala Differentiated product exports Commodity exports 3 3 2 2 1 1 Elasticity, USA Elasticity, USA 0 0 -1 -1 -2 -2 -3 -3 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 Year Year 14 Results for other exports markets are available from the authors upon request. 16 Sri Lanka 3 Differentiated product exports Commodity exports 3 2 2 1 1 Elasticity, USA Elasticity, USA 0 0 -1 -1 -2 -2 -3 -3 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 Year Year Lao PDR Differentiated product exports Commodity exports 3 3 2 2 1 1 Elasticity, USA Elasticity, USA 0 0 -1 -1 -2 -2 -3 -3 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 Year Year Tunisia Differentiated product exports Commodity exports 3 3 2 2 1 1 Elasticity, USA Elasticity, USA 0 0 -1 -1 -2 -2 -3 -3 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 Year Year 17 Zambia 3 Differentiated product exports Commodity exports 3 2 2 1 1 Elasticity, USA Elasticity, USA 0 0 -1 -1 -2 -2 -3 -3 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 Year Year Bulgaria Differentiated product exports Commodity exports 3 3 2 2 1 1 Elasticity, USA Elasticity, USA 0 0 -1 -1 -2 -2 -3 -3 2003 2004 2005 2006 2007 2008 2009 2010 2003 2004 2005 2006 2007 2008 2009 2010 Year Year Notes: This figure plots rolling-window elasticity estimates. We recursively run Equation 1 for four-year windows beginning in 2000-2003 through 2007-2010 and plot the elasticity estimates over time along with the 95 percent confidence bands. Rolling window elasticity estimates have been calculated for each growth pole. However, due to the similarity between the estimates among growth poles, only the elasticity estimates to the United States are shown. Source: Authors’ computations. The graphs indicate that over time, differences in the estimated elasticities are statistically insignificant in most if not all of the cases even. Lao’s commodity elasticity estimates, which appear to vary over time, seem to have overlapping 95 percent confidence intervals. We take this to suggest that unobservable factors affecting trade costs are constant over time. As such, we argue that failing to control for the multilateral resistance terms does not result in biased estimates, lending support to our empirical specification. Indeed, since each estimation window of four years includes country-specific (and regional) intercepts, the rolling regression approach implicitly controls for four-year time-varying multilateral resistance terms. 15 15 Rolling-window estimates of the intercept by exporting country are available from the authors upon request. 18 5.2 Alternative Product Classification We now present the export vulnerability results using an alternative classification of differentiated product and commodity exports following Rauch’s (1999) classification. For the sake of brevity, Table 7 presents the resulting vulnerability indicators, but not the elasticities. 16 Table 7: Vulnerability of Exports to Major Global Markets (% of GDP), 2010 Rauch Classification Old Growth Poles New Growth Poles US Japan EU27 Brazil Russia Indonesia India China Korea Bulgaria Differentiated 0.003 0.001 0.130 0.000 0.008 0.000 0.000 0.001 0.000 Commodities 0.002 0.000 0.087 0.000 0.002 0.001 0.001 0.005 0.002 Guatemala Differentiated 0.041 0.003 0.007 0.000 0.000 0.000 0.000 0.000 0.001 Commodities 0.033 0.001 0.004 0.000 0.001 0.000 0.001 0.001 0.003 Lao PDR Differentiated 0.007 0.005 0.028 0.000 0.000 0.000 0.000 0.009 0.000 Commodities 0.003 0.001 0.001 0.000 0.000 0.000 0.004 0.111 0.003 Sri Lanka Differentiated 0.035 0.002 0.054 0.000 0.001 0.001 0.007 0.001 0.001 Commodities 0.001 0.001 0.004 0.000 0.002 0.000 0.001 0.000 0.000 Tunisia Differentiated 0.007 0.002 0.240 0.001 0.002 0.000 0.000 0.002 0.000 Commodities 0.001 0.000 0.043 0.002 0.000 0.000 0.004 0.001 0.000 Zambia Differentiated 0.000 0.000 0.003 0.000 0.000 0.000 0.001 0.000 0.000 Commodities 0.001 0.003 0.007 0.000 0.000 0.000 0.002 0.127 0.021 Notes: This table presents the overall measure of export vulnerability, calculated by multiplying the exposure and sensitivity indicators. The numbers represent the impact of a 1 percent change in a given major global market’s GDP on exports to that market from each country of interest as a share of GDP. Source: Authors’ calculations. The results change little when using Rauch’s alternative classification of commodities. The first noticeable difference (not reported in Table 7) is that Zambia is now considered a high commodity exporter due to the fact that copper is now classified as a commodity. The elasticity estimates remain similar across export destinations. For Bulgaria and Lao PDR, commodities are still more sensitive than differentiated products, however this is also now the case for Guatemala. For Zambia, the estimated elasticities are very close in magnitude for the two product types. And the estimates for Lao PDR (EAP) are still higher than any other country (region). Overall, however, we continue to find that the principal source of international heterogeneity in export vulnerability results from differences in export exposure. 16 The complete set of estimations is available from the authors upon request. 19 6. Concluding Remarks More countries than ever before are relying on export-led growth strategies for long-run economic development. Yet fears that increasingly open and interlinked global markets enhance the vulnerability of developing countries to foreign shocks have risen lately, although exposure to foreign markets is itself a source of risk diversification as domestic shocks might be uncorrelated with foreign shocks. This paper proposes a simple measure of export vulnerability of a country’s exports to fluctuations in the economy activity of foreign markets. We applied this methodology to six developing countries, one from each World Bank region, selected to be otherwise similar yet differ in terms of the level of exposure to major global markets as well as the product composition of their export baskets. We acknowledge that real sector transmission of external shocks depends not only on a country’s level of export exposure through bilateral trade flows with specific markets but also on the sensitivity of exports to foreign demand. Consequently, we combined two indicators to construct an overall measure of export vulnerability: exposure, measured as the export share in GDP, and export sensitivity, estimated as the elasticity of each country’s exports with respect to fluctuations in foreign- market economic activity. This elasticity was estimated using robust techniques that follow the micro-founded gravity model proposed by Helpman et al. (2008). Although our results suggest differences in elasticity estimates across countries as well as product categories, we found that the principal source of international heterogeneity in export vulnerability results from differences in export exposure. We also found differences in the sensitivity of differentiated product exports versus commodity exports, although the differences in magnitude depend on the country of interest. These differences could be explained by whether the products are final goods or intermediate inputs. These findings call for developing countries to pursue a risk diversification by focusing on increasing the number of foreign markets with which they trade rather than through shielding themselves from international markets, which would actually raise economic risk. A final word of caution is warranted, however. The analyses presented herein focused on bilateral trade relations, but countries can be affected by fluctuations in major global markets even in the absence of bilateral trade flows. For instance, if income in major markets affects global prices of tradable goods, then all countries that produce such goods would be affected. Hence our estimates would be interpreted as floor estimates of the global transmission of economic fluctuations through exports of developing countries. 20 References Anderson, J. E. and E. van Wincoop (2003). “Gravity with Gravitas: A Solution to the Border Puzzle.� American Economic Review, 93(1): 170–92. Baier, S. L. and J. H. Bergstrand (2001). “The Growth of World Trade: Tariffs, Transport Costs, and Income Similarity.� Journal of International Economics, 53(1): 1-27. Baldwin, R. (Ed.) (2009). “The Great Trade Collapse: Causes, Consequences, and Prospects�. Centre for Trade and Economic Integration, The Graduate Institute, Geneva. Baldwin, R. and D. Taglioni (2006). “Gravity for Dummies and Dummies for Gravity Equations.� NBER Working Paper No. 12516, National Bureau of Economic Research, Cambridge, MA. Bergstrand, J. H. (1985). “The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence.� The Review of Economics and Statistics, 67(3): 474-81. – (1989). “The Generalized Gravity Equation, Monopolistic Competition, and the Factor- Proportions Theory in International Trade.� The Review of Economics and Statistics, 71(1): 143- 53. Brander, J. and P. Krugman (1983). “A ‘Reciprocal Dumping’ Model of International Trade.� Journal of International Economics, 15(3-4): 313-21. Brückner, M., and D. Lederman (2012). “Trade Causes Growth in Sub-Saharan Africa.� World Bank Policy Research Working Paper 6007, Washington, DC. Dollar, D. (1992). “Outward-oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-1985.� Economic Development and Cultural Change, 40(3): 523- 544. Edwards, S. (1992). “Trade Orientation, Distortions and Growth in Developing Countries.� Journal of Development Economics, 39(1): 31-57. Edwards, S. (1998). “Openness, Productivity and Growth: What Do We Really Know?� Economic Journal, 108(447): 383-398. Feenstra, R. C. (2004). Advanced International Trade: Theory and Evidence. Princeton, NJ: Princeton University Press. 21 Feenstra R., J. R. Markusen, and A. Rose. 2001. “Using the Gravity Equation to Differentiate Among Alternative Theories of Trade.� Canadian Journal of Economics, 34(2), 430-47. Feyrer, J. (2009). “Trade and Income -- Exploiting Time Series in Geography.� NBER working paper 14910, National Bureau of Economic Research, Cambridge, Massachusetts. Foster, N. (2008). “The Impact of Trade Liberalisation on Economic Growth: Evidence from a Quantile Regression Analysis.� Kyklos, 61(4): 543-567. Frankel, J. A. and D. Romer (1999). “Does Trade Cause Growth?� American Economic Review, 89(3): 379-399. Freund, C. (2009). “The Trade Response to Global Downturns: Historical Evidence.� Policy Research Working Paper No. 5015, The World Bank, Washington, DC. Haddad, M. and B. Shepherd (Eds.) (2011). Managing Openness: Trade and Outward-Oriented Growth After the Crisis. Washington, DC: The World Bank. Harrison, A. and G. Hanson (1999). “Who Gains from Trade Reform? Some Remaining Puzzles.� Journal of Development Economics, 59(1): 125-154. Hooper, P., K. Johnson, and J. Marquex (2000). “Trade Elasticities for the G-7 Countries.� Princeton Studies in International Economics Working Paper No. 87, Department of Economics, Princeton University, Princeton, NJ. Helpman, E., M. Melitz and Y. Rubinstein (2008). “Estimating Trade Flows: Trading Partners and Trading Volumes.� The Quarterly Journal of Economics, 113(2): 441-87. International Monetary Fund (2012a). Primary Commodity Prices. Washington, DC. International Monetary Fund (2012b). World Economic Outlook. Washington, DC. Irwin, D. A. (2002). “Long-run Trends in World Trade and Income.� World Trade Review, 1(1): 89-100. Kim, D.-H. (2011). “Trade, Growth and Income.� Journal of International Trade & Economic Development, 20(5): 677-709. Kwack, S. Y, C. Y. Ahn, Y. S. Lee, and D. Y. Yang (2007). “Consistent Estimates of World Trade Elasticities and an Application to the Effects of Chinese Yuan (RMB) Appreciation.� Journal of Asian Economics, 18: 314-30. 22 Levchenko, A. A., L. T. Lewis, and L. L. Tesar (2010). “The Collapse of International Trade during the 2008-09 Crisis: In Search of the Smoking Gun.� IMF Economic Review, 58(2): 214- 253. Martin, P., T. Mayer and M. Thoenig (2008). “Make Trade Not War?� Review of Economic Studies, 75(3): 865-900. Mayer and Zignago (2011). “Notes on CEPII’s Distances Measures (GeoDist).� CEPII Working Paper 2011-25, French Research Center in International Economics, Paris, France. Rauch, J. E. (1999). “Networks versus Markets in International Trade.� Journal of International Economics, 48: 7-35. United Nations (2012). COMTRADE. New York, NY. Wacziarg, R. (2001). “Measuring the Dynamic Gains from Trade.� World Bank Economic Review, 15(3): 393-429. World Bank (2012). World Development Indicators. Washington, DC. 23 Appendix 1 List of Countries by World Bank Region East Asia Europe and Latin America and Middle East and South Sub-Saharan and Pacific Central Asia the Caribbean North Africa Asia Africa Cambodia Albania Antigua and Barbuda Algeria Afghanistan Angola China Armenia Argentina Djibouti Bangladesh Benin Fiji Azerbaijan Belize Egypt, Arab Rep. Bhutan Botswana Indonesia Belarus Bolivia Iran, Islamic Rep. India Burkina Faso Kiribati Bosnia and Herzegovina Brazil Iraq Maldives Burundi Lao PDR Bulgaria Chile Jordan Nepal Cameroon Malaysia Croatia Colombia Lebanon Pakistan Cape Verde Marshall Islands Czech Republic Costa Rica Morocco Sri Lanka Central African Republic Micronesia, Fed. Sts. Estonia Dominica Syrian Arab Republic Chad Mongolia Georgia Dominican Republic Tunisia Comoros Palau Hungary Ecuador Yemen, Rep. Congo, Dem. Rep. Papua New Guinea Kazakhstan El Salvador Congo, Rep. Philippines Kyrgyz Republic Grenada Côte d'Ivoire Samoa Latvia Guatemala Eritrea Solomon Islands Lithuania Guyana Ethiopia Thailand Macedonia, FYR Haiti Gabon Tonga Moldova Honduras Gambia, The Vanuatu Montenegro Jamaica Ghana Vietnam Poland Mexico Guinea Romania Nicaragua Guinea-Bissau Russian Federation Panama Kenya Serbia Paraguay Lesotho Slovak Republic Peru Liberia Slovenia St. Kitts and Nevis Madagascar Tajikistan St. Lucia Malawi Turkey t. Vincent and the Grenadines Mali Turkmenistan Suriname Mauritania Ukraine Uruguay Mauritius Uzbekistan Venezuela Mozambique Namibia Niger Nigeria Rwanda São Tomé and Principe Senegal Seychelles Sierra Leone South Africa Sudan Swaziland Tanzania Togo Uganda Zambia Zimbabwe Notes: This table lists the exporter countries included in the estimations of the regional gravity models. Each country is classified as low-income or middle-income according to the World Bank. Countries in boxes were chosen for the analysis. 24