99399 ETHIOPIA’S GREATRUN The Growth Acceleration and How to Pace It 2014 2004 ETHIOPIA’S GREAT RUN THE GROWTH ACCELERATION AND HOW TO PACE IT February, 2016 iii TABLE OF CONTENTS ACKNOWLEDGEMENTS................................................................................................................................ vii LIST OF ABBREVIATIONS............................................................................................................................... ix EXECUTIVE SUMMARY................................................................................................................................... xi PART A: EXPLAINING GROWTH.................................................................................................................... 1 THE GROWTH ACCELERATION.................................................................................................................... 3 1.1 Recent Economic Growth in Perspective.............................................................................................................3 1.2 Rapid Growth in the Context of Development Progress......................................................................................3 1.3  Proximate Growth Determinants.........................................................................................................................6 ECONOMIC STRATEGY — ‘THE ETHIOPIAN WAY’................................................................................ 13 2.1 Main Elements of Economic Strategy................................................................................................................13 2.2 Did Ethiopia Fo he Insights of ‘The Growth Commission’?...............................................................................16 2.3 Fast Growing Non-resources Rich African Peers................................................................................................16 2.4  An East Asian Strategy?.....................................................................................................................................20 EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS.................. 23 3.1 Introduction......................................................................................................................................................23 Explaining Ethiopia’s Recent Growth Performance............................................................................................24 3.2  Annex 3.1  Methodology..........................................................................................................................................30 Annex 3.2  Model Robustness..................................................................................................................................35 GROWTH AND STRUCTURAL CHANGE................................................................................................... 39 4.1 Introduction......................................................................................................................................................39 4.2  Structural Change in Ethiopia...........................................................................................................................41 4.3  The Rising Services Sector.................................................................................................................................46 4.4 Ethiopia’s Experience in the International Context............................................................................................47 4.5 A Regional Perspective and Potential Implications for Ethiopia.........................................................................49 Annex 4.1  Selected Structural Change Indicators.....................................................................................................54 DRIVERS OF AGRICULTURAL GROWTH................................................................................................... 57 5.1 Introduction......................................................................................................................................................57 The Growth of Agriculture 2004–2014.............................................................................................................58 5.2  iv ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT 5.3 Land Intensification and Adoption of Improved Agricultural Technologies .......................................................60 5.4  Drivers for Change............................................................................................................................................63 5.5  Evidence on Changes in Drivers........................................................................................................................65 5.6  Conclusion and Further Challenges...................................................................................................................73 PART B: SUSTAINING GROWTH.................................................................................................................. 77 MANAGING GROWTH EXPECTATIONS.................................................................................................... 79 6.1 Introduction......................................................................................................................................................79 6.2 Growth Accelerations: The International Experience.........................................................................................79 6.3 Growth Tailwinds and Headwinds.....................................................................................................................81 6.4  Cyclical Factors..................................................................................................................................................83 6.5 Benchmarking...................................................................................................................................................83 6.6  Scenario Analysis...............................................................................................................................................86 6.7 Summary...........................................................................................................................................................89 ETHIOPIA’S FINANCING CHOICE: PUBLIC INFRASTRUCTURE OR PRIVATE INVESTMENT?... 91 7.1 Introduction......................................................................................................................................................91 7.2 Is Ethiopia’s Current Level of Public Infrastructure Investment Optimal?..........................................................91 7.3 Firm-level Constraints: Infrastructure or Credit?...............................................................................................95 7.4  Domestic Finance Reform.................................................................................................................................97 7.5 Complementary Infrastructure Financing Options..........................................................................................100 Annex 7.1  Public and Private Investment: Model, Empirical Results and Calibration............................................107 GROWTH AND STRUCTURAL REFORMS............................................................................................... 113 8.1 Introduction....................................................................................................................................................113 8.2 Trends and Status in Structural Reforms..........................................................................................................114 8.3 The Potential Growth Impact of Reforms........................................................................................................117 8.4 Productivity Impacts of Services Trade Liberalization.......................................................................................124 8.5 Reform Sequencing: Best Practice and Ethiopia’s Experience...........................................................................126 8.6  Reform Risks...................................................................................................................................................127 8.7  Quo Vadis Ethiopia?........................................................................................................................................129 8.8 Growth Model Sustainability Monitoring........................................................................................................130 Annex 8.1  Methodology for Estimating Growth Impact of Reforms......................................................................134 Annex 8.2  Robustness Checks...............................................................................................................................135 Annex 8.3  Estimating the Impact of Services Inputs Quality on Firms’ Productivity ............................................138 REFERENCES.................................................................................................................................................. 141 LIST OF FIGURES Figure 1.1: Recent Economic Growth in Perspective.................................................................................................4 Figure 1.2: Ethiopia’s Development Performance.......................................................................................................5 Figure 1.3: Growth Characteristics............................................................................................................................7 Table of Contents v Figure 1.4: Growth Accounting and Decompositions................................................................................................9 Figure 15: Ethiopia: Real GDP Growth..................................................................................................................11 Figure 2.1: Financial Sector Indicators.....................................................................................................................18 Figure 3.1: Regression Results: Key Growth Drivers in Ethiopia (Real GDP per Capita).........................................24 Figure 3.2: Trends in Key Growth Drivers...............................................................................................................26 Figure 3.3: Infrastructure Growth Rates: Ethiopia in the Global Context (1970–2010)...........................................28 Figure 3.4: Prediction Performance of the Model....................................................................................................29 Figure A3.1: Definition of Time Periods Used in the Study........................................................................................33 Figure A3.2: Growth Impact of Infrastructure............................................................................................................36 Figure 4.1: Ethiopia: Output and Employment by Sector, 1999–2013....................................................................43 Figure 4.2: Ethiopia: Labor Productivity, 1999–2013..............................................................................................44 Figure 4.3: The Ethiopia Services Sector..................................................................................................................48 Figure 4.4: Labor Productivity, Growth and Employment by Sector........................................................................50 Figure 4.5: Output and Employment Shares across Countries, 2012.......................................................................51 Figure 5.1: Agricultural Growth in Ethiopia, 2004–14............................................................................................59 Figure 5.2: Input Use: Fertilizer, Improved Seeds, Irrigation and Pesticides..............................................................62 Figure 5.3: Input use by Educational Level and Government Agricultural Expenditure...........................................66 Figure 5.4: Extension Service, Transport to Markets, and Farmer’s Education..........................................................68 Figure 5.5: Modern Input Adaption, Prices and Rainfall.........................................................................................70 Figure 5.6: Microfinance Activity, Fertilizer Constraints, Crop Damage and Real Wages.........................................72 Figure 6.1: Potential GDP Projection, Results of Benchmarking and Policy Simulations.........................................84 Figure 7.1: Public and Private Investment in Low Income Countries (1980–2012..................................................93 Figure 7.2: Ethiopia: Public and Private Investment and their Returns....................................................................95 Figure 7.3: Numerical Simulations with Alternative Values of Government Infrastructure.....................................102 Figure 7.4: Numerical Simulations with Alternative Values of the Deposit Rate....................................................103 Figure 8.1: Ethiopia: Structural Reform Indices, 1973–2005.................................................................................114 Figure 8.2: Structural Reform Indices by Country and Income Groups.................................................................116 Figure 8.3: Services Trade Restrictiveness Index by Sector and Mode.....................................................................117 Figure 8.4: Structural Reform Indices for East Asian Countries.............................................................................129 LIST OF TABLES Table 1.1: Capital Growth Rates and Estimated Returns to Capital (percent)..........................................................9 Table 2.1: Key Characteristics of the Ethiopian Economic Model..........................................................................15 Table 2.2: Growth Commission Recommendations and Ethiopia’s Experience.......................................................19 Table 3.1: Parameter Values, Changes and Predicted Growth Effects (real GDP per capita)...................................25 Table A3.1: Regression Baseline Results....................................................................................................................34 Table A.1: Gross Value Added by Sector.................................................................................................................54 Table A.2: Employment by Sector..........................................................................................................................54 Table A.3: Labour Productivity by Sector...............................................................................................................55 Table A.4: Sectoral Decomposition of GVA Per Capita Growth (1999–2013)........................................................55 Table A.5: Sectoral Decomposition of GVA Per Capita Growth (2005–2013)........................................................56 Table A.6: Demographics and Employment Rate...................................................................................................56 vi ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Table 5.1: Probit Model: Adoption of Improved Seeds or Chemical Fertilizer in Cereal Production.......................64 Table 6.1: Benchmarking Structural, Stabilization and External Factors (2006–10 data)........................................85 Table 6.2: Assumed Annual Growth Rates of Policy Variables by Scenario (percent)..............................................87 Table 6.3: Illustrative Scenarios and Growth Projections until Early 2020s............................................................88 Table 6.4: Summary of Growth Expectations for Ethiopia.....................................................................................90 Table 7.1: Estimated Excess Returns of Public Investment.....................................................................................94 Table 7.2: Most Binding Constraints to Doing Business in Ethiopia, Various Rankings.........................................96 Table 7.3: Predicted Effects of Financial Sector Reforms........................................................................................99 Table 7.4: Alternative Infrastructure Financing Options.......................................................................................104 Table A7.2.1: The Response of Private Investment to Public Investment....................................................................108 Table A7.2.2: The Response of Output to Public Investment.....................................................................................109 Table A7.3.1: Calibration Results..............................................................................................................................110 Table 8.1: Average Values of Structural Reform Indices........................................................................................119 Table 8.2: Baseline Growth Regressions (Dependent variable: Real GDP Per Capita, Growth Rate).....................120 Table 8.3: Coefficient Estimates and Potential Growth Impact of Reforms...........................................................121 Table 8.4: Coefficient Estimates and Potential Long-run Growth Impact of Reforms...........................................121 Table 8.5: Reforms, Growth, and Distance to the Technology Frontier................................................................123 Table 8.6: Labor Productivity Determinants Based on Perception of Services’ Performance..................................125 Table 8.7: International Best Practice Guidance on Reform Sequencing...............................................................127 Table 8.8: Ethiopia Growth Model Sustainability Indicators................................................................................131 Annex Table A8.1: GMM Regression: Coefficient Estimates and Growth Impact of Reforms.....................................136 Annex Table A8.2: Control Variable Check: Coefficient Estimates and Growth Impact..............................................136 Annex Table A8.3: Coefficient Estimates and Potential Growth Impact (1973–1989)................................................137 Annex Table A8.4: Coefficient Estimates and Potential Growth Impact (1990–2006)................................................137 LIST OF BOXES Box 1.1: When Did Ethiopia’s Economy Take Off?....................................................................................................11 Box 2.1: The Banking Sector in Ethiopia....................................................................................................................17 Box 4.1: The Demographic Dividend.........................................................................................................................45 Box 6.1: Stylized Facts about Growth Accelerations and their Aftermath....................................................................80 Box 7.1: Literature Review: Does Financial Repression Inhibit or Facilitate Economic Growth?................................98 Box 7.2: A Theoretical Model of Financial Repression and Interest Rate liberalization.............................................101 Box 8.1: Structural Reform Indices: Data and Definitions........................................................................................115 vii ACKNOWLEDGEMENTS T he World Bank greatly appreciates the close 8. Fiseha Haile (2015): Chapter 8 collaboration with the Government of Ethiopia 9. Claire Hollweg, Esteban Rojas and Gonzalo Varela (the Ministry of Finance and Economic (2015): Chapter 8 Cooperation, in particular) in the preparation of this report. The report was prepared by a team led by Excellent research assistance was provided by Lars Christian Moller (Lead Economist and Program Mesfin Girma (Economist, GMFDR), Eyasu Tsehaye Leader, AFCE3). It draws upon a series of background (Economist, GPVDR), Ashagrie Moges (Research paper commissioned for the study, including: Analyst, GMFDR), and Fiseha Haile (Consultant, GMFDR). Helpful comments were provided by: 1. Shahid Yusuf (2014): Chapters 2 and 6. Kevin Carey (Lead Economist, GMFDR) and Michael 2. Lars Moller and Konstantin Wacker (2015): Geiger (Senior Country Economist, GMFDR). The Chapters 3 and 6. report was prepared under the overall guidance of 3. Pedro Martins (2014, 2015): Chapter 4 Albert Zeufack (Practice Manager, GMFDR), Guang 4. Ejaz Ghani and Stephen O’Connell (2014): Zhe Chen (Country Director, AFCS1), Carolyn Turk Chapter 4 (Country Director, AFCE3), Pablo Fajnzylber (Practice 5. Fantu Bachewe, Guush Berhane, Bart Minten, Manager, GPVDR) and Sajjad Shah (Country Program and Alemayehu Taffesse (2015): Chapter 5 Coordinator, AFCE3). The peer reviewers were: 6. Maya Eden (2015ab): Chapter 7 Barry Bosworth (Brookings), Cesar Calderon (Lead 7. Aart Kraay and Maya Eden (2014ab): Chapter 7 Economist, AFRCE) and Andrea Richter (IMF). ix LIST OF ABBREVIATIONS ADLI Agricultural Development Led FY Fiscal Year Industrialization GCI Global Competitiveness Index AGP Agricultural Growth Program GDP Gross Domestic Product AGSS Agricultural Sample Survey GMM Generalized Method of Moments AISE Agricultural Input Supply GNI Gross National Income Enterprise Govt C Government Consumption ATA Agricultural Transformation Agency GTP Growth and Transformation Plan ATVET Agricultural Technical and GVA Gross Value Added Vocational Education and Training HDI Human Development Index CA Capital Account HIPC Heavily Indebted Poor Country CBE Commercial Bank of Ethiopia Initiative CIMMYT International Maize and Wheat IFDC International Fertilizer Improvement Center Development Center CPIA Country Policy and Institutional IFPRI International Food Policy Research Assessment Institute CSA Central Statistical Agency IMF International Monetary Fund DA Development Agent LFS Labor Force Survey DB Doing Business LIC Low Income Country DBE Development Bank of Ethiopia LMIC Lower Middle Income Country DSA Debt Sustainability Analysis M2 Broad Money Aggregate EES Ethiopian Electric Service MDG Millennium Development Goals EEU Ethiopian Electric Utility MDRI Multilateral Debt Relief Initiative EEU Ethiopia Economic Update MOA Ministry of Agriculture EIAR Ethiopian Institute of Agricultural MOFED Ministry of Finance and Economic Research Development EPRDF Ethiopian Peoples’ Revolutionary NASA National Aeronautics and Space Democratic Front Administration ERA Ethiopia Railways Corporation NBE National Bank of Ethiopia (The ERHS Ethiopia Rural Household Survey Central Bank) ESE Ethiopian Seed Enterprise OLS Ordinary Least Squared ESLSE Ethiopian Shipping and Logistics PASDEP Plan for Accelerated and Sustained Services Enterprise Development to End Poverty FAOSTAT Food and Agriculture Organization PIM Public Investment Management Statistics Polity2 Governance variable FDI Foreign Direct Investment PPP Purchasing Power Parity FtF Feed the Future PPP Public Private Partnership x ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT PPTs Percentage Points SSA5 Burkina Faso, Mozambique, PSNP Productive Safety Net Program Rwanda, Tanzania, Uganda q/ha Quintal per hectare TFP Total Factor Productivity REER Real Effective Exchange Rate TOT Terms of Trade RER Real Exchange Rate UN United Nations RuSACos Rural Saving and Credit UNDP United Nations Development Cooperatives Program SOE State Owned Enterprise WDI World Development Indicators SSA Sub-Saharan Africa WTO World Trade Organization xi EXECUTIVE SUMMARY This report addresses two questions: What explains Ethiopia’s growth acceleration? How can it be sustained? In brief, we find that Ethiopia’s rapid economic growth, concentrated in agriculture and services, was driven by substantial public infrastructure investment and supported by a conducive external environment. To sus- tain high growth, three policy adjustments are proposed, including: identifying sustainable ways of financing infrastructure, supporting private investment through credit markets, and, tapping into the growth potential of structural reforms. PART A: EXPLAINING GROWTH The Growth Acceleration Ethiopia’s economic growth has been remarkably rapid and stable over the past decade. Real GDP growth averaged 10.9 percent in 2004–2014, according to official data. By taking into consideration popula- tion growth of 2.4 percent per year, real GDP growth per capita averaged 8.0 percent per year.1 The country moved from being the 2nd poorest in the world by 2000 to the 11th poorest in 2014, according to GNI per capita, and came closer to its goal of reaching middle income status by 2025. This pace of growth is the fastest that the country has ever experienced and it also exceeds what was achieved by low-income and Sub- Saharan African countries in that period. Recent growth was also noticeably stable, as the country avoided the volatility by spells of drought and conflict which had plagued growth in the past. Accelerated economic progress started in 1992 with a shift to an even higher gear in 2004. Econometric analysis supports a story of two growth accelerations as average growth increased from 0.5 percent in 1981–92 to 4.5 percent in 1993–2004 and to 10.9 percent in 2004–14. The first ‘gear shift’ took place shortly after the political and economic transition of 1991 with the downfall of the communist Derg regime and the introduction of a more market-oriented economy. The subsequent Ethiopian People’s Revolutionary Democratic Front (EPRDF) government, in turn, implemented a series of structural economic reforms dur- ing the 1990s which paved the way for the second growth acceleration starting in 2004 (the subject of this report). Interestingly, structural economic reforms have been largely absent from Ethiopia’s recent story of success, though they offer a promising growth potential if implemented. The recent growth acceleration was part of a broader and very successful development experience. Poverty declined substantially from 55.3 percent in 2000 to 33.5 percent in 2011, according to the interna- tional poverty line of US$1.90. Despite rapid growth, Ethiopia remained one of the most equal countries in the world with a Gini coefficient of consumption of 0.30 in 2011. But progress went beyond monetary dimensions. Life expectancy increased by about one year annually since 2000 and is now higher in Ethiopia 1 Using UN population estimates and applying them to the official national accounts data in constant factor prices. xii ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT than the low income and Sub-Saharan Africa averages. In fact, Ethiopia also surpassed these peer groups in several other key development indicators, including child and infant mortality. As a result, the country has attained most of the Millennium Development Goals. That said, Ethiopia faces a challenge in promoting shared prosperity as the poorest 15 percent of the population experienced a decline in well-being in 2005–11 mainly as a result of high food prices. Growth was concentrated in services and agriculture on the supply side, and, private consumption and investment on the demand side. While agriculture was the main economic sector at the beginning of the take-off, the services sector gradually took over and was complemented, in recent years, by a construc- tion boom. Out of an average annual growth rate of 10.9 percent in 2004–14, services contributed by 5.4 percentage points followed by agriculture (3.6 percentage points) and industry with 1.7 percentage points. Private consumption contributed to most growth on the demand side with public investment becoming increasingly important. Growth decompositions reveal relatively high contributions from total factor productivity and structural change. While capital and labor accumulation was important for growth, Ethiopia stands out from other non-resource rich fast-growing Sub-Saharan African countries (SSA5) by its very high total fac- tor productivity growth of 3.4 percent per year.2 Similarly, while most labor productivity growth came from within sectors (as in other countries), inter-sectoral labor shifts (structural change) explain a quarter of decadal of Ethiopia’s recent per capita GDP growth (which is higher than in most other countries). Still, Ethiopia remains at an early stage of development as reflected by continued high returns to capital. Economic Strategy – ‘The Ethiopian Way’ Ethiopia stands out in many ways, including in the economic strategy that paved the way to success. In brief, economic strategy focused on promoting agriculture and industrialization while delivering substantial public infrastructure investment supported by heterodox macro-financial policies. Ethiopia’s strong commit- ment to agricultural development is noteworthy as reflected by high government spending and the world’s biggest contingent of agricultural extension workers. While a strong push for infrastructure development at the early stage of development is far from unique, the way in which Ethiopia achieved this sets it apart. Heterodox financing arrangements supported one of the highest public investment rates in the world. Even if Ethiopia generally did not follow the recommendations of the Growth and Development Commission (2008), it did deliver the recommended impressive rates of public investment with the purpose of crowding-in the private sector. Despite low domestic savings and taxes, Ethiopia was able to finance high public investment in a variety of orthodox and heterodox ways. The former include keeping government consumption low to finance budgetary public infrastructure investment as well as tapping external conces- sional and non-concessional financing. Three less conventional mechanisms stand out: First, a model of financial repression that kept interest rates low and directed the bulk of credit towards public infrastructure. Second, an overvalued exchange rate 2 The IMF (2013) identifies six fast-growing Non-Resource Rich Sub-Saharan African countries, including Burkina Faso, Ethiopia, Mozambique, Rwanda, Tanzania and Uganda. Excluding Ethiopia, we refer to this group as SSA5. Although natural resources is becoming increasingly important in some of these countries they were not natural resources dependent at the relevant period of analysis (1995–2010). List of Abbreviations xiii that cheapened public capital imports. Third, monetary expansion, including direct Central Bank budget financing, which earned the government seignorage revenues. Ethiopia’s economic strategy was unique. Although Ethiopia gradually moved in the direction of a market-based system, it continued to intervene in most sectors of its economy thereby not adopting some of the key recommendation of the Growth Commission of ‘letting markets allocate resources efficiently’. Indeed, apart from market oriented reforms implemented during the 1990s, structural economic reforms have been absent from Ethiopia’s growth strategy in part because of initial economic success. Although it was inspired by the East Asian development state model and shares some common features, it is also differ- ent from these countries both in conception and outcomes. Agriculture, for instance, features much more prominently in the Ethiopian strategy than in East Asia. Critically, also, Ethiopia’s economic success thus far has not been derived from the success of numerous firms drawn from the private sector as in East Asia. Explaining Growth: The Role of Structural, External and Stabilization Factors Using a cross-country regression model, we are able to distinguish between key drivers of growth. Our approach avoids tweaked Ethiopia-specific results because we use an existing regression model originally con- structed to investigate growth elsewhere. The model is estimated on 126 countries for the 1970–2010 period, including low income countries. Ethiopia’s per capita real GDP growth rate is predicted using Ethiopian values of the underlying growth determinants for three different periods: Early 2000s, Late 2000s, and Early 2010s. We distinguish between structural, external and stabilization factors. The model predicts Ethiopia’s growth rate quite accurately thereby underscoring its relevance as a useful analytical tool for our purposes. Economic growth was driven primarily by structural improvements. When measured at Purchasing Power Parity (PPP), the model predicts a real GDP per capita growth rate for Ethiopia of 4.3 percent in 2000–13 compared with an observed rate of 4.8 percent. The contribution of structural factors is estimated at 3.9 percentage points. The growth acceleration was also supported by a conducive external environment. Exports quadrupled in nominal terms, while volumes doubled, reflecting a substantial positive commodity price effect. Macroeconomic imbalances in the form of exchange rate overvaluation and high inflation held back some growth. Public infrastructure investment, facilitated partly by restrained government consumption, was the key structural driver of growth. In contrast to many countries in the region, the government deliberately emphasized capital spending over consumption within the budget and this was key for supporting growth, according to the model. This shift was facilitated by declining military spending following the 1998–2000 war with Eritrea giving rise to a ‘peace dividend’. Increased openness to international trade also supported growth as did the expansion of secondary education, though these effects were less pronounced. The strong contribution of infrastructure investment arises from a substantial physical infrastructure expansion combined with their high returns. Ethiopia stands out during the 2000s for having registered very rapid infrastructure development. Using the data for 124 countries over four decades, the country was among the 20 percent fastest in terms of infrastructure growth over the past decade. Although this is partly the result of starting from a very low level, these infrastructure growth rates also exceed those of fast grow- ing regional peers with comparable income levels. As we do not know the true economic return to infra- structure investment in Ethiopia, their average returns are estimated from the country sample. Given that xiv ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT public investment was concentrated in providing basic infrastructure, such as energy, roads, and telecom, this growth effect seems plausible. Macro-financial policies held back some growth, though the effect was small. Based on the expe- rience of other countries, the model predicts growth to fall when credit to the private sector declines, the exchange rate appreciates and inflation is high. Ethiopia experienced all three trends over this period and this gives rise to an estimated negative macro-financial growth effect. What stands out, however, is that the quantitative effect is quite small (0.44 percentage points). This result helps explain how Ethiopia was able to achieve high economic growth in the presence of seemingly sub-optimal macro-financial policies. In fact, it raises the question of whether growth was able to accelerate precisely because of this heterodox policy mix, which supported growth-inducing infrastructure investment. Although it is hard to conclude firmly either way, Ethiopia’s experience supports the impression that ‘getting infrastructure right’ at the early stage of development can go a long way in supporting growth. The Role of Structural Change A modest shift in labor from agriculture to services and construction can explain up to a quarter of Ethiopia’s per capita growth in 2005–13. This result illustrates the strong potential of structural change as a driver of economic growth as discussed in the literature (e.g. McMillan et al. 2014). Although Ethiopia has experienced high economic growth and some structural change in production away from agriculture towards services, the similar shift in employment has been much more modest. Nevertheless, agricultural employment did decline from 80 to 77 percent between 2005 and 2013 and because agricultural labor productivity is so low, this shift gave rise to static efficiency gains as relative labor shares increased in construction and services where the average value added of a worker is up to five times higher. The nature of structural change taking place in Ethiopia differs notably from the vision of government policy. Specifically, economic strategy in Ethiopia aims to promote the kind of structural change first described by Sir Arthur Lewis (1954) in which workers move out of agriculture and into manufacturing. Ethiopia has followed this ‘trodden path of development’ only partially as economic activity (output and jobs) have shifted from agriculture and into construction and services, largely by-passing the critical phase of industrialization. In response, the government has strengthened its institutional, legal and regulatory framework focusing on promoting light manufacturing FDI, especially in the form of industrial parks (see World Bank 2015 for details). The growth acceleration period marked the rise of the services sector in Ethiopia. Services overtook agriculture to become the largest economic sector, the biggest contributor to economic growth, and is the second biggest employer. Within services, commerce, ‘other services’ and the public sector were the most important contributors to output and jobs. On the other hand, the Ethiopian services story is predominantly one of a rise in traditional activities, which require face-to-face interaction, rather than modern activities such as ICT or finance. Ethiopia’s growth acceleration was also supported by positive demographic effects. The economic take-off coincided with a marked increase in the share of the working-age population giving a positive boost to labor supply. Up to thirteen percent of per capita growth in 2005–13 can be attributed to this ‘demo- graphic dividend’ effect. A continued rise in the working age population will support potential economic growth in the coming decades, but for the country to fully reap these benefits it must accelerate the ongoing List of Abbreviations xv fertility decline and equip workers with marketable skills to be attractive to prospective employers. Both the manufacturing and services sectors would play an important role in absorbing this additional labor. Drivers of Agricultural Growth Ethiopia’s agricultural sector has recorded remarkable rapid growth in the last decade and was the major driver of poverty reduction. The sector is, by far, the biggest employer in Ethiopia, accounts for most merchandise exports and is the second largest in terms of output. The sector also contributed to most of the employment growth over the period of analysis. Although some labor shifted out of agriculture, sub- stantial shifts are likely to take a long time. Critically, agricultural growth was an important driver of poverty reduction in Ethiopia: Each percent of agricultural growth reduced poverty by 0.9 percent compared to 0.55 percent for each percent of overall GDP growth (World Bank, 2015). For these reasons, the report takes a deeper look at the drivers of agricultural growth. Agricultural output increases were driven by strong yield growth and increases in area cultivated. Yield growth averaged about 7 percent per year while area cultivated increased by 2.7 percent annually. A decomposition of yield growth reveals the importance of increased input use as well as productivity growth. As in the Green Revolution, increased adoption of improved seeds and fertilizer played a major role in sus- taining higher yields. While starting from a low base, these inputs more than doubled over the last decade. Total factor productivity growth averaged 2.3 percent per year. The factors associated with agricultural production growth include extension services, remoteness and farmer’s education. A regression model was used to identify the likelihood of adopting modern tech- nology. Farmers that received extension visits, less remote households and more educated farmers were more likely to adopt improved agricultural technologies. Recent agricultural growth is largely explained by high government spending on extension services, roads, education as well as favourable price incentives. First, Ethiopia has built up a large agricultural extension system, with one of the highest extension agent to farmer ratios in the world. Second, there has been a significant improvement in access to markets. Third, improved access to education led to a significant decrease in illiteracy in rural areas. Fourth, high international prices of export products as well as improv- ing modern input—output ratios for local crops have led to better incentives. Other factors played a role as well, including good weather, better access to micro-finance institutions in rural areas, and improved tenure security. Recent poor rains in Ethiopia during 2014 and 2015 pose a major challenge to the country and the impact of climate change stresses the importance of continued investment in irrigation to reduce reli- ance on rain-fed production. PART B: SUSTAINING GROWTH Managing Growth Expectations What should we expect in terms of Ethiopia’s growth rate over the next decade? Following a decade-long spell of double digit growth on the back of a strategy and performance that seemingly emulates the East Asian developmental states, including China, one might assume that such high growth rates can be sustained in xvi ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT the future on the back of the same strategy that worked so well in the past. The second part of this report takes a deeper look at this issue on the basis of available international and country-specific evidence. We begin by highlighting the exceptional nature of the past decade performance by drawing upon the objective statistical experience of growth accelerations elsewhere. According to Pritchett and Summers (2014), cross-country experiences of per capita GDP growth since the 1950s has been an average of 2 per- cent per year with a standard deviation of 2 percent. Episodes of per capita growth of above 6 percent tend to be extremely short-lived with a median duration of nine years. China’s experience from 1977 to 2010 is the only instance of a sustained episode of per capita growth exceeding 6 percent and only two other coun- tries come close (Taiwan and Korea). In other words, these country experiences are statistically exceptional. A country specific analysis of growth head- and tailwinds suggest a balance of factors at play in Ethiopia. These factors were derived on the basis of the stylized facts and conceived wisdom emanating from the most recent growth and economic development literature. The likelihood of continued high growth in Ethiopia is buoyed by five factors: productivity-enhancing structural change, within-sector productivity gains (including agriculture), technological catch-up, urbanization, and FDI. The demographic transition and a large domestic market offer important potential. These factors would need to be balanced against a number of ‘growth headwinds’ factors. Exogenous factors include geographical disadvantages and a slowdown of world trade. Endogenous factors include: lagging agricultural productivity, low export size and diversifica- tion, a small financial sector, low levels of human capital and poor trade logistics. Most of these ‘inhibitors’ do not pose insurmountable hurdles but collectively they could dampen Ethiopia’s chances of maintaining its growth rate over the course of the next decade. Cyclical analysis suggests that a slowdown is pending. By the very nature of having experienced a growth acceleration, Ethiopia’s real GDP per growth rate has exceeded the potential rate of GDP growth for the past decade. Potential GDP growth, in turn, is a function of capital, labor and TFP growth. Investment has been exceptionally high the past years and is thus likely to slow down. A rising working age population provides some growth impetus, but total factor productivity growth will be hard to sustain at its current high levels. Additionally, economic activity has been strongly supported by a construction boom in the past 3 years (2011/12–2013/14). Even if government policy drives part of this boom, the private component is cyclical in nature and will not last indefinitely. Regression model simulations indicate a growth slowdown under alternative policy scenarios. We use the abovementioned regression model to identify growth drivers and simulate three scenarios. The first scenario assumes continued infrastructure investment that comes at the cost of private sector crowding-out in the credit market, the buildup of inflationary pressures due to supply constraints, and, a policy of con- tinued real exchange rate appreciation to keep capital imports cheap. The second scenario aims to promote accelerated private sector investment and reduce macroeconomic imbalances. Specifically, the pace of public infrastructure investment slows down but is partially substituted by private sector involvement. The third assumes an acceleration public infrastructure investment at the cost of growing macroeconomic imbalances. All three policy scenarios yield comparable annual real GDP per capita growth rates of about 4 percent in PPP terms, which is well below the rate of 6.5 percent observed in the Late 2000s. Put together, alternative approaches suggest a likely range of GDP growth between 4.5 and 10.5 percent over the next decade. In per capita terms, this is equivalent to a range of 2.0 to 8.0 percent assum- ing a population growth rate of 2.5 percent per year. The lower bound is given by international experience of growth accelerations and Ethiopia’s 1993–2004 growth rate. The upper bound is given by the maximum List of Abbreviations xvii achieved in Ethiopia and elsewhere. We note that a decadal growth projection based on Ethiopia’s level of Hausmann-Hidalgo concept of ‘economic complexity’ is at the lower range at 4.4 percent per year. The challenge confronting policy makers is to make sure that growth remains at the higher end of this range. To sustain high growth, three policy adjustments are proposed. This includes (1) supporting private investment through credit markets; (2) identifying sustainable ways to finance infrastructure, and; (3) tap- ping into the growth potential of structural reforms. We discuss each of these in turn. Supporting Private Sector Led Growth with Credit While public infrastructure investment helps firms to become more productive, Ethiopian firms appear more concerned with getting access to credit. According to six different survey instruments, credit is men- tioned as the more binding constraint for firms. This matters, because it suggests that the government may have made progress in addressing the infrastructure constraint and now needs to pay more attention to alle- viating other constraints important to firms. It is indicative that the marginal return to private investment may be higher than the marginal return to public infrastructure investment. Indeed, empirical estimates of these relative returns presented in this report support this assessment. This results arises from the fact that Ethiopia has the third highest public investment rate in the world and the sixth lowest private investment rate combined with the economic logic of diminishing marginal returns. Arguably, the Ethiopian economy would benefit from a shift of domestic credit towards private firms. If the aim of government policy is to enhance the productivity of private firms, then it is important to understand what the firm-level constraints are. If firms really need credit more than access to new roads or better telecommunication to grow and prosper, then government policy would need to support the allevia- tion of the credit constraint at the firm level. Since public infrastructure investment is partially financed via the same domestic savings pool, it is clear that infrastructure financing competes directly with the financing of private investment projects. Two policy reforms could potentially address the challenge of private sector credit. The first would be to continue the existing system of financial repression, but to direct more credit towards private firms. In that way Ethiopia’s financial system would become more similar to Korea, where the bulk of credit was directed towards private priority sectors. A second reform involves a gradual move towards a more liberal- ized interest rates that better reflect the demand and supply for savings/credit and encourage more savings. Policy reforms should be informed by two criteria: the relative returns of public and private invest- ment, and, the savings rate. This insights were derived from a simple theoretical model developed for the purposes of this study. Financial repression with more private credit is attractive in situations where the mar- ginal return to private investment is much higher than the marginal return to public investment. Interest rate liberalization is attractive if the saving rate of the country is low as welfare would rise by increasing the deposit rate towards more market-determined levels. The report presents evidence that both constraints are binding in Ethiopia. Ethiopia needs to provide more access to credit to the private sector and this can be done either within or outside the existing policy paradigm. The theoretical analysis and the empirical evidence suggest that there are welfare enhancing effects of either option. Given Ethiopia’s preference for financial repression, the less substantive reform would be to maintain this system, but to follow South Korea’s footsteps and direct the bulk of domestic credit to priority private sectors. xviii ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Identifying Alternative Infrastructure Financing Sources Continued infrastructure development remains one of Ethiopia’s best strategies to sustain growth, but the current financing model is not sustainable. Infrastructure was the most important driver of economic growth during the growth acceleration. This is because the economic returns to infrastructure were high and the physical infrastructure expansion in Ethiopia was substantial. Since Ethiopia continues to have the 3rd largest infrastructure deficit in Africa, it is not surprising that the cross-country regression model also predicts this policy to be the best going forward. However, the past infrastructure expansion was financed via a range of mechanisms that will begin to show their limits in the future in terms of external debt sustainability, private sector crowding out in the credit markets and a strong exchange rate that undermines external competitive- ness. Going forward, Ethiopia needs more infrastructure, but it would need new mechanisms to finance it. There are a range of alternative financing mechanisms to continue and the report briefly discusses their merits. Some options are consistent with current government strategy and thinking. This includes raising tax revenues, increased private sector involvement (including PPPs) and improved public investment management. Other options deviate from the existing paradigm, including: increasing domestic savings and developing capital markets via a higher real interest rate; greater selectivity and prioritization of investments; securitization of infrastructure assets, and; improved pricing, including higher electricity tariffs. The Growth Potential of Structural Reforms Ethiopia lags behind Sub-Saharan African peers in most reform dimensions. This is especially the case for domestic finance, the current account, the capital account, and services trade restrictiveness. On the posi- tive side, Ethiopia has done well in reducing trade tariffs and is at par with peers here. What would be the impact on growth if Ethiopia closed the reform gap with its peers? To address this question, we perform a benchmarking exercise using an existing regression model that links reform with growth (Prati et al., 2013). Even modest structural reforms that close gaps with peers would potentially have considerable impact on GDP per capita growth. The results presented are only indicative and do not constitute a comprehensive appraisal of reforms that have actually been introduced. If Ethiopia were to catch up with the average Sub- Saharan Africa country in terms of financial liberalization, its per capita GDP growth rate would be boosted by 1.9 percentage points per year. These substantial effects arise because this type of reform is highly potent for growth and owing to a substantial reform gap. Similar reforms of the current account and opening the capital account are estimated to increase real GDP growth rates by 0.8 and 0.7 percentage points, respectively. There are considerable firm level gains to be reaped from services sector liberalization in Ethiopia, especially in credit access, energy and transport services. For example, if the access to credit conditions of Ethiopia were to match those of Rwanda, then firm labor productivity would increase by 4.3 percent, keeping all else equal. Similarly, if electricity conditions were to also match those of Rwanda, the labor productivity gains would be close to a 2.2 percent. Finally, matching China’s transportation services would imply productivity gains of 4.2 percent. These results were derived using a similar benchmarking method based on an existing regression model. In terms of reform sequencing, Ethiopia has already followed international best practice through its ‘trade-first’ approach, although it has proceeded very slowly. Economic theory, country experience and best international practice would generally suggest the following sequence of reforms: (1) trade liberalization; List of Abbreviations xix (2) financial sector liberalization, and (3) capital account opening. That said, every country experience has been unique and reforms have to be customized to their specific country setting. It is noteworthy that Ethiopia has so far liberalized its merchandise trade, but not yet its services trade. The next possible step on the reform path may be to engage in services trade and financial sector liberalization. Although there are economic benefits to reforms as well as an emerging consensus about their sequencing, policy makers are often concerned about risks. While the average longer term net benefits seem to be positive, there is no guarantee that all countries will automatically benefit from reforms. Ethiopia has the added advantage of being in a position to learn the lessons of successful as well as painful experiences of other countries. Still, there are important pitfalls on the reform path (e.g. regulatory frameworks need to be well developed before liberalizing domestic finance) and these would need to be studied more carefully if Ethiopia were to re-initiate the structural reform agenda. Monitoring Growth Model Sustainability Finally, the report propose a series of indicators that would be worth monitoring going forward to capture the many trade-offs that are embedded in the current growth strategy. At some point, we argue, the costs of pursuing the current policy would outweigh its benefits. For example, the loss of external com- petitiveness associated with an overvalued exchange rate may outweigh the benefits in the form of cheaper public capital imports. A deterioration in these indicators may precede a slowdown in growth and provide early warning to policy makers that the current growth model has run its course. Policy makers are encour- aged to be proactive and initiate reform efforts now as opposed to waiting until growth slows down. With the recent launch of the Second Growth and Transformation Plan and the recent appointment of a new economic team, the timing is right to consider the proposals of this report. Encouragingly, the GTP2 envisions a strong increase in the tax revenue to GDP ratio in a bid to raise domestic savings and identifying alternative and more sustainable ways to finance infrastructure. In a similar vein, the private sector is expected to play an important role in supporting infrastructure provisions in ways that reduced the need for public borrowing. The new strategy also stresses the role of the private sector as the ultimate engine of growth and emphasizes the need to maintain a competitive real exchange rate. Moreover, domestic savings are to be mobilized by ensuring that the real interest rate remains positive. The analysis and proposals put forward in this report are aimed to support the Government of Ethiopia in achieving these goals in its quest towards becoming a lower middle income country by 2025. PART A: EXPLAINING GROWTH Part A is structured as follows: Chapter 1 highlights the key characteristics of the growth accel- eration. Chapter 2 describes the economic strategy that supported high growth. Chapter 3 identi- fies key growth determinants distinguishing between structural, external and stabilization factors. Chapter 4 explains the take-off of the agriculture sector. Chapter 5 utilizes a structural change framework to gain further insights about the determinants of growth. 3 THE GROWTH ACCELERATION 1 consistently exceeded low-income and Sub-Saharan Ethiopia has experienced a growth acceleration since 2004, enabling a catch-up with the rest of the world, as a part of Africa averages as well as SSA5 (Figure 1.1.2). a very successful broader development performance. While As a result, Ethiopia’s real GDP has tripled since agriculture was the main growth contributor at the beginning 2004 although it remains well below regional and of the take-off, the services sector gradually took over and has been complemented, in recent years, by a construction low-income levels. Figure 1.1.3 illustrates the dramatic boom. Private consumption contributed to growth on the rise in real GDP observed over the past decade while demand side with public investment becoming increasingly Figure 1.1.4 puts this performance into perspective by important. A Solow growth decomposition shows that growth was driven by factor accumulation along with very high total comparing with relevant peers showing that although factor productivity growth. A Shapley decomposition reveals Ethiopia is catching up with peers, its income level that most of the increase in value added per person came remains low. Comparisons with China are also insight- from higher within-sector labor productivity supported by ful (Figure 1.1.5). While China and Ethiopia had similar structural and demographic change. Still, Ethiopia remains at an early stage of development as reflected by continued levels of income in the 1980s, China is now 14 times high returns to capital. richer than Ethiopia. Ethiopia managed to grow ‘at Chinese rates’ for about decade, but China itself experi- enced a growth acceleration that lasted for three decades. 1.1  Recent Economic Growth in Encouragingly, Ethiopia has moved from being the 2nd Perspective poorest to the 11th poorest country in the world since 2000, according to GNI per capita (Atlas Method). It also Economic growth has been remarkably rapid and moved closer to its goal of becoming a middle income stable over the past decade. Real GDP growth aver- country by 2025 gradually narrowing the gap to the rel- aged 10.9 percent in 2004–2014, according to offi- evant income threshold (Figure 1.1.6). In sum, Ethiopia cial data. By taking into consideration population made a lot of progress, but it remains a poor country. growth of 2.4 percent per year, real GDP growth per capita averaged 8.0 percent per year in this period. Rapid Growth in the Context of 1.2  This substantially exceeds per capita growth rates Development Progress achieved in the first decade after the country’s transi- tion to a market-based economy (1992–2003: 1.3 Ethiopia’s growth performance over the past decade percent; 1993–2004: 4.5 percent), under the com- was part of a broader and very successful develop- munist Derg regime (1974–91: –1.0 percent), and ment experience. From 2000 to 2011 the wellbeing during monarchy (1951–73: 1.5 percent). Droughts of Ethiopian households improved on a number of and conflict produced volatile growth patterns prior dimensions. In 2000, Ethiopia had one of the high- to 2004, but growth has been rapid and stable since est poverty rates in the world, with 55.3 percent of then—an impressive performance from a historical the population living below the international poverty perspective (Figure 1.1.1). Ethiopia’s growth rate line of US$1.90 2011 PPP per day (Figure 1.2.1) and also exceeded regional and low-income averages over 44.2 percent of its population below the national the past decade. Since taking off in 2004, growth has poverty line. By 2011, 33.5 percent lived on less 4 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 1.1: Recent Economic Growth in Perspective 1. Ethiopia, Real GDP growth 2. Real GDP growth in Ethiopia, SSA and LICs 16 1st Take-off 2nd Take-off 14 14 12 12 10.6 10 10 8 8 5.6 6 4 6 2 0.5 4 0 –2 2 –4 0 –6 Drought War –8 –2 –10 –4 –12 Famine Drought 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 –14 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 Ethiopia SSA5 LIC SSA 3. Real GDP in Birr (left axis) and US$ (right axis) 4. Real GDP per Capita (2005 US$) 700 30 1200 600 25 1000 500 20 800 400 15 600 300 10 400 200 100 5 200 0 0 0 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 Birr US$ Ethiopia Sub-Saharan Africa Low income (developing only) 5. GNI per Capita (Atlas Method): Ethiopia & China 6. GNI per Capita (Atlas Method) 7000 1400 6000 1200 5000 1000 Gap: 4000 54% 800 3000 600 Gap: 2000 400 85% 1000 200 0 0 1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2000 2005 2010 2013 2014 Ethiopia China Ethiopia Lower middle income threshold Source: World Bank (WDI). Note: SSA5 include Burkina Faso, Mozambique, Rwanda, Tanzania and Uganda. The Growth Acceleration 5 than the international poverty line and 29.6 percent (Figure 1.2.3). As explained in more detail in World of the population was counted as poor by national Bank (2014a), this can largely be explained by the measures. Ethiopia is one of the most equal countries effect of rising food prices in 2011 which hurt the in the world and low levels of inequality have, by and real incomes of marginal farmers and urban dwellers. large, been maintained throughout this period of rapid The average household in Ethiopia has better economic development (Figure 1.2.2). health, education and living standards today than Nevertheless, Ethiopia faces a challenge in in 2000. Life expectancy increased by about one year terms of promoting shared prosperity. Promoting per year that passed since 2000 and is now higher in shared prosperity requires fostering the consumption Ethiopia than the low income and regional averages growth of the bottom 40 percent. Prior to 2005, (Figure 1.2.4). Substantial progress was made towards Ethiopia made good progress on sharing prosperity: the attainment of the Millennium Development Goals consumption growth of the bottom 40 percent was (MDG), particularly on extreme poverty, undernour- higher than the top 60 percent in Ethiopia. However, ishment, gender parity in primary education, infant this trend was reversed in 2005 to 2011 with lower and child mortality (Figure 1.2.5), maternal mortal- growth rates observed among the bottom 40 percent ity, HIV/AIDS, malaria and water access, though FIGURE 1.2: Ethiopia’s Development Performance 1. Share of population below international poverty line Incidence of Monetary Poverty in Ethiopia and other African Countries (Percentage of the population at $1.25 PPP poverty line) 100 90 80 70 60 50 40 30 20 10 0 00 04 11 9806 97 05 94 03 05 10 04 10 03 11 06 11 05 11 00 07 06 09 04 Ethiopia Ghana Kenya Lesotho Madagascar Malawi Nigeria Rwanda Senegal Tanzania Uganda Zimbabwe 2. Gini coefficient of consumption 70 60 50 40 30 20 10 0 00 05 11 9806 97 05 94 03 05 10 04 10 04 10 06 11 05 11 00 05 06 09 95 Ethiopia Ghana Kenya Lesotho Madagascar Malawi Nigeria Rwanda Senegal Tanzania Uganda Zimbabwe (continued on next page) 6 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 1.2: Ethiopia’s Development Performance (continued) 3. Growth of household consumption by groups 4. Life expectancy (total, years) 4 65 3 60 2 55 1 50 45 0 40 –1 35 –2 30 –3 1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 1996–2000 2000–2005 2005–2011 Bottom 10% Bottom 40% Top 60% Ethiopia Sub-Saharan Africa Low income (developing only) 5. Child and Infant Mortality in Ethiopia and SSA 6. Share of Population with no Education, by Gender 300 80 250 70 200 60 150 100 50 50 40 0 30 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2000 2005 2011 Ethiopia SSA LIC Male Female Total Source: 2.1–2.3 and 2.6: World Bank (2015). 2.4–2.5: World Bank (WDI). progress is lagging in primary enrolment and sanita- 1.3  Proximate Growth Determinants tion. Women are now having fewer births—the total fertility rate fell from 7.0 children per women in 1995 The growth acceleration was driven by services to 4.1 in 2014. At the same time, the prevalence of and agriculture on the supply side, and, private stunted children was reduced from 58 percent in 2000 consumption and investment on the demand side. to 40 percent in 2014. The share of population with- More recently, there is evidence of a boom in invest- out education was also reduced considerably from 70 ment and construction activity. Figure 1.3 decomposes percent to less than 50 percent (Figure 1.2.6). Finally, growth and output into major supply and demand side the number of households with improved living stan- components since 1980. The following trends emerge: dards measured by electricity, piped water and water in residence doubled from 2000 to 2011. Despite this  Growth was driven primarily by services and agri- impressive progress, the country faces deep challenges culture (Figure 1.3.1). in every dimension of development. One key challenge  The services sector has overtaken agriculture as is to sustain rapid economic growth. the largest in terms of output. This shift has been The Growth Acceleration 7 FIGURE 1.3: Growth Characteristics 1. Real GDP Growth (supply side), 1980/81–2013/14 2. Real GDP Growth (demand side), 1981–2013/14 12 16 10.8 14 1.1 10 12 10 4.7 8 5.4 8 1.3 6 2.4 6 7.7 4.4 4 3.2 1.7 2 4 1.4 2.8 0 0.3 –2 –2.7 –3.0 2 0.7 3.6 0.0 –4 –0.9 0.9 0 –6 1980/81–1990/91 1991/92–2002/03 2003/04–2013/14 1980/81–1990/91 1991/92–2002/03 2003/04–2013/14 Agriculture Industry Services GDP Gov. cons Priv. cons Total inv. Imports Stat discre Exports GDP growth 3. Real GDP Shares (supply side), 1980/81–2013/14 4. Real GDP Shares (demand side), 1981–2013/14 80 45 80 70 40 70 64.0 60 35 60 52.3 30 50 45.5 50 25 40 40 40.2 20 30 26.6 36.8 30 15 20 14.3 10 20 9.5 10.9 10 5 10 0 0 0 1980/81 1982/83 1984/85 1986/87 1988/89 1990/91 1992/93 1994/95 1996/97 1998/99 2000/01 2002/03 2004/05 2006/07 2008/09 2010/11 2012/13 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 Total investment Public consumption Agriculture Industry Services Private consumption (RHA) 5. Real GDP Growth (supply side), 2003/04–2013/14 6. Real GDP Growth (demand side), 2004–2013/14 14 20 12 15 2.0 4.5 5.1 3.3 6.1 10 3.6 3.0 8.2 1.3 4.7 5.8 10 10.1 2.5 8.5 6.3 5.7 14.2 4.1 8 1.0 4.0 5.3 1.2 11.6 10.2 5.9 7.0 5 9.2 1.1 4.4 7.6 6 2.3 6.0 6.9 5.4 6.3 4.0 1.0 1.0 1.5 2.8 0 –0.1 4 8.5 7.1 5.8 5.0 1.0 1.1 2.1 2.8 –0.8 –2.4 –0.3 –0.5 –5 –3.8 –3.0 2 3.9 3.2 4.1 4.8 –6.0 –4.7 2.5 2.2 3.1 2.3 –4.5 0 –10 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 Agriculture Industry Services GDP Gov. cons Priv. cons Total inv. Net Export GDP growth Source: Staff Estimates based on data from the Ministry of Finance (National Accounts Department). 8 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT ongoing for a decade, and it accelerated since decomposes gross value added per person into four com- 2004 (Figure 1.3.3). ponents: labor productivity gains within sectors, labor  Since 2004, the sectoral drivers of growth have productivity gains between sectors (structural change), shifted further towards services and, lately, indus- demographic gains, and increases in the employment try. The recent rise of industry is due to a con- rate. For the 1999–2013 period as a whole, more than struction boom and not because of a rise in the seventy percent of growth is attributed to within-sector manufacturing sector which remains very small labor productivity gains, especially in agriculture and at about 4 percent of GDP. commerce (Figure 1.4.4). The other three components  Private consumption and investment were contribute to varying degrees depending on the time the major demand side contributors (Figure period. The structural change and demographic effects 1.3.2). are particularly pronounced in the later period (2005–  The investment rate has increased substantially 13). The employment effect, in comparison, is negative since mid-1990s with a commensurate decline owing to a rise in the student population. in public consumption (Figure 1.3.4). Despite substantial capital accumulation, the  Agriculture is no longer the major driver of returns to capital have remained high. Table 1.1 growth. In 2004, about a quarter of growth was shows estimates of the return to capital (change in due to agriculture. By 2014, less than a quarter output as a result of a change in capital) using two of growth came from this sector. alternative methods. The estimated return on capi-  The growth contribution of investment activity tal ranges from 18 to 24 percent depending on the has increased in recent years (Figure 1.3.6). period of analysis and method. Interestingly, returns to capital increased during the high growth period High total factor productivity growth and (2004–12) according to both methods, in spite of the factor accumulation account for most economic fact that the capital growth rate also increased. growth. Growth can come from two sources: using This finding is consistent with the fact that more factors of production or inputs (labor and capi- Ethiopia is still in the early phases of development tal) to increase the amount of goods and services that as the economy continues to be capital scarce. Two an economy is able to produce or combining inputs theories of growth and economic development sup- more efficiently to produce more output for a given port this observation. First, in the neoclassical growth amount of input. Decomposing into these two sources model (Solow, 1956), starting from a low level of yields insights into the proximate causes of growth. output per worker, saving and investment take place As illustrated in Figure 1.4.1, the growth accelera- and the capital-labor, and thus the output-labor ratio, tion period (in this case: 2000–10) was character- rises. The economy experiences diminishing returns ized by substantially higher total factor productivity to capital: the marginal product of capital, and thus (TFP) growth, and, accumulations of capital and labor the market-determined rate of profit of capital, falls. compared to previous decades.3 The contribution of Second, within a competitive market economy of the human capital, in comparison, was modest and did Lewis (1954) model, it is only when the economy not increase in the 2000–10 period. TFP growth was particularly high in Ethiopia compared with fast- 3 The TFP estimates presented here are comparable to other recent growing regional peers not dependent on natural estimates despite differences in decomposition method, assumptions and time periods. The IMF (2012) estimates TFP growth at 5.2 percent resources (SSA5), as shown in Figure 1.4.2. for the 2006/07–2010/11 period with contributions of 2.6 percent Rising labor productivity was a major con- and 3.2 percent for labor and capital, respectively. Merotto and Dogo (2014) decompose a real GDP growth rate of 11.1 percent for the tributor to growth with positive contributions from 2003/04–2011/12 period into (percentage points): TFP (4.3), capital structural and demographic change. Figure 1.4.3 (4.3), labor (2.0), and education (0.4). The Growth Acceleration 9 FIGURE 1.4: Growth Accounting and Decompositions 1. Ethiopia: Solow Decomposition of Real GDP 2. SSA5: Solow Decomposition of Real GDP 10 10 8 8 3.4 6 6 2.1 0.3 0.3 4 4 2.1 2.1 0.2 1.1 0.7 0.9 0.2 2 0.8 0.4 2 1.2 1.1 1.0 2.4 2.5 1.7 1.3 1.6 0.7 0 0 –1.1 –0.6 –2 –2 1980–90 1990–2000 2000–10 1980–90 1990–2000 2000–10 TFP Education Adjusted labor Capital TFP Education Adjusted labor Capital 3. Shapley Decomposition (GVA per capita growth) 4. GVA per capita growth (%, 2005–2013) 140 Agriculture 120 6.2 13.0 Commerce 100 5.1 42.5 24.5 Structural change 80 17.2 Construction 60 8.8 Manufacturing 40 71.6 76.2 56.8 Other services 20 Public services 0 –8.0 Mining –13.7 –20 Utilities –40 Transport 1999–2013 1999–05 2005–13 Finance Demographic effect Employment rate Structural change Within sector productivity –5 5 15 25 35 Source: 4.1–4.2: IMF (2013). 4.3–4.4: Martins (2015). TABLE 1.1: Capital Growth Rates and Estimated Returns to Capital (percent) Full period Sub Periods 1983–2012 1983–2001 1991–2001 2002–2003 2004–2012 Real GDP Growth Rate 5.2 3.0 5.2 –2.2 10.8 Method 1: Initial-year gross fixed capital formation  Capital Growth Rate 7.1 6.0 5.1 6.4 9.2 Average Rate of Return to Capital 23.5 24.7 21.8 20.0 21.7 Method 2: ‘Rule of Thumb’ capital output ratio Capital Growth Rate 5.0 3.2 5.3 5.3 8.4 Average Rate of Return to Capital 18.7 17.9 17.9 17.9 20.5 Source: Merotto and Dogo (2014). 10 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT emerges from the first, labor surplus and capital scarce deviation of Ethiopia’s growth rate dropped from 6.0 classical stage of development and enters the second, in 1992–2003 to 1.4 in 2004–14. First, there was an labor scarce and more capital abundant neoclassical absence of major droughts and the weather was rela- stage that real incomes begin to rise generally. tively favorable. Second, there was relative political When did the Ethiopian economy take-off? stability and an absence of wars and conflict. Third, Interestingly, the econometric evidence is suggestive of the Ethiopian economy is relatively closed and external two recent take-offs, as discussed in Box 1.1. The first, events tend to have less impact than in other countries in 1992, when there was a change in economic and in the region. In particular, the trade-to-GDP ratio is political systems. The second, in 2004, when economic quite low, the capital account is closed and there are growth rates became consistently high and stable. It can no foreign banks operating in Ethiopia. During the therefore be argued that the economy changed to higher 2008/09 global financial crisis, for instance, Ethiopia gears both in 1992 and 2004. For the purposes of this was mainly affected by a decline in exports rather than study, we shall focus mainly on the period since 2004, the GDP growth. which is henceforth termed ‘the growth acceleration’. In summary, Ethiopia has experienced a growth Why has growth volatility declined since 2004? acceleration since 2004 in the context of a very suc- There are three main reasons for why the standard cessful development performance. The Growth Acceleration 11 BOX 1.1: When Did Ethiopia’s Economy Take Off? The concept of an economic take-off was conceptualized by Rostow (1960). He proposed a historical model of growth whereby economies undergo five stages of growth as follows: (1) traditional society; (2) preconditions for take-off; (3) take-off; (4) drive to maturity; and (5) age of high mass consumption. Although Rostow’s theory has faced much criticism, the concept he introduced remains useful as it identifies a point in time in the early stages of development where an economy starts growing at high and self-sustained rates. Subsequently developed econometric methods help identify and analyze sustained growth take- offs (e.g. Hausmann et al. 2005)). Two alternative years emerge as candidates for a take-off in Ethiopia: 1992 and 2004. To motivate the discussion consider the following average annual growth rates: (1981–1992: 0.5%), (1993–2004: 4.5%) and (2004–2014: 10.9%). This box briefly discusses the arguments in favor of either of these two interpretations and attempts to reconcile them. We follow the definitions of Hausmann et al. (2005): (a) a growth acceleration should be sustained for at least 8 years and the change in growth rate has to be at least 2 percentage points; (b) a country can have more than one instance of growth acceleration as long as the dates are more than 5 years apart; (c) trend breaks were selected at the 1% level of significance (α = 0.01) in the Autometrics options in the software package OxMetrics 7 (Doornik et al., 2013). The test results illustrated below are sufficiently robust to changes in these specifications. Econometric tests reveal a break in GDP growth in 1992 when the focus is on the 1980–2014 period and extreme observations are taken into account. This result is derived using the algorithm developed in Doornik et al. (2013) which identifies the existence, timing and significance of breaks in mean growth rates. In the years 1998 and 2003 GDP growth witnessed sharp contractions, which coincide with the war with Eritrea and a period of severe drought, respectively. Accounting for these two extraordinary observations, statistical tests singled out the year 1992 as a period marking a turning point in growth performance (see Figure 1.5.1). This essentially reflects the fact that Ethiopia’s GDP growth rate surged from 0.5% in 1981–92 to 7.7% in 1993–2014. Identifying 1992 as a break point is consistent with the hypothesis that the economy took off around the time of political and economic regime change. The shift in growth performance around 1992 is associated with the introduction of market-oriented economic reforms that ensued the demise of the socialist Derg regime (1974–1991). In addition, the period since 1992 was preceded by political regime change and the end of a major civil war. Note, however, that economic growth was relatively unstable in 1992–2004 compared to 2004–2014. This prompts the need for studying the period since 1992 separately. Econometric tests identify 2004 as the turning point in growth performance when the focus is on the 1992–2014 period. The result, illustrated in Figure 1.5.2, unveils that that the period 1992–2014 comprises two ‘distinct’ growth regimes: 1993–2003 (when GDP growth was significantly higher than that in 1980–1992), and 2004–14 (when growth accelerated further and exhibited more stability). In sum, it can be argued that the economy changed to higher gears in both 1992 and 2004. The year 1992 marked the shift from a command economy to a more market-oriented economy. Growth was higher, but somewhat unstable. 2004, in turn, marked the first year of an unprecedented sustained high-growth period. FIGURE 15: Ethiopia: Real GDP Growth 1. 1980–2014 2. 1992–2014 0.125 0.125 0.100 0.100 0.075 0.050 0.075 0.025 0.050 0.000 –0.025 0.025 –0.050 0.000 –0.075 –0.100 –0.025 1985 1990 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 GDP growth rate SIS + IIS α = 0.01 Note: Red line: actual GDP growth rate. Green line: fitted line derived from the statistical test. Source: Haile (2015). 13 ECONOMIC STRATEGY — ‘THE ETHIOPIAN WAY’ 2 Which economic strategy did the Ethiopian government Main Elements of Economic 2.1  pursue? In brief, economic strategy focused on promoting Strategy agriculture and industrialization while delivering substantial public infrastructure investment supported by heterodox Since 1991, Ethiopia has pursued a policy of macro-financial policies. Overall, there was substantial government intervention in many aspects of the economy. Agricultural Development Led Industrialization Ethiopia’s economic strategy was unique. It differed markedly (ADLI). ADLI builds on the development theories from other strategies, such as the recommendations of the from the 1960s in which (smallholder) agriculture Growth Commission (2008) as well as the experience of other fast growing African countries. Although it was needs to be developed first to facilitate demand for inspired by the East Asian development state model and industrial commodities and inputs for industrializa- shares a few common features, economic strategy also tion. The policy aims to increase agricultural produc- differed from this model both in conception and outcomes. tivity to increase overall production, as well as invest in those industries with most production linkages to rural areas. The strategy assumes that inter-sectoral Ethiopia is a unique country and its economic linkages will reinforce the growth impetus derived growth strategy is no exception. The country from increasing productivity in both sectors with the prides itself of many special characteristics, includ- agricultural sector obtaining machinery, chemicals ing not being colonized, the use of the Julian and consumption goods from industry in exchange calendar (with 13 months), of being the cradle of for food and raw material. Since the 1990s, ADLI mankind and origin of coffee, and having its own implementation was rather interventionist as agri- worldwide known cuisine. On the economic front, cultural productivity increased and linkage devel- there are also many unique characteristics, includ- opment requires substantial public investment and ing on the economic policy side. As argued in this direct support policies, but initially it was done rather chapter, Ethiopia’s policy mix is an interesting cautiously. hybrid of alternative economic models, but most Starting in the mid-2000s, ADLI was gradu- of all it is unique. ally complemented by efforts to promote light The chapter is structured as follows: Section 2.1 manufacturing to support structural transforma- describes the main elements of Ethiopia’s economic tion and exports. The 2005 PASDEP 5-year plan strategy. Section 2.2 compares it briefly with the focused on boosting agricultural production via recommendations of the Growth and Development intensification and yield growth and an industrial Commission (2008). Section 2.3 briefly juxtaposes and export earnings strategy based around indus- Ethiopia’s strategy with that of non-resource rich, tries with linkages to agriculture. Horticulture was fast-growing Sub-Saharan African countries (SSA5). encouraged with great success, but attempts to Finally, Section 2.4 makes comparisons with East boost leather processing and other industries were Asian Developmental States. initially less successful. Under the Growth and 14 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Transformation Plan (2010–15), the country’s indus- infrastructure deficit remains the third largest in trialization process has been promoted by empha- Africa, according to the AfDB. sizing light manufacturing in key sectors where the Heterodox financing arrangements to support country has a perceived comparative advantage (e.g. public investment is at the heart of the strategy. apparel, leather, agribusiness, wood, and metal). Public investment projects are implemented over The process is supported through industrial policy the national budget and through State Owned (e.g. directing scarce credit and foreign exchange Enterprises (SOEs) using domestic and external towards selected sectors).4 More recently, the govern- sources of financing. Domestic financing draws upon ment has emphasized the importance of developing a range heterodox arrangements, including: direct industrial zones aimed at attracting foreign investors. lending and bond purchases by state-owned banks to Indeed, the second Growth and Transformation finance SOE projects, and, compulsory purchases by Plan (2015–20), is expected to put an even stron- private banks of bonds whose proceeds partly finance ger emphasis on structural transformation (along long term development lending to the private sector the lines of the ‘Lewis Model’), industrialization, by DBE (the ’27 percent rule’). The dominance of urbanization, and export promotion. state-owned banks (accounting for about two thirds Massive public infrastructure investment has of banking system deposits) and credit rationing in been at the center of economic strategy since the the presence of negative real interest rates imply cheap mid-1990s. The public investment rate rose from sources of financing for public projects, but also lead about 5 percent in the early 1990s to 18.6 percent of to the exclusion of many private investment projects. GDP in 2011, making it the third highest in in the This partly helps explain why the private investment world (World Bank, 2013). Starting from very low rate in Ethiopia is the sixth lowest worldwide. In levels of infrastructure, Ethiopia has invested heavily sum, this model can be described as ‘financial repres- in the energy, transport, communications, agriculture sion’ (see Box 2.1). In addition, monetary policy has and social sectors. Power generation capacity increased occasionally been relatively loose, contributing to from 473 MW in 2002 to 2,268 MW in 2014, and higher than necessary inflationary pressures, includ- is projected to reach 4,138 MW in 2015. More than ing through regular direct central bank financing of 10,000 MW generation capacity will be available by the budget. 2020 once major ongoing hydro, geothermal and The role of government in the Ethiopian ‘mixed wind projects have been completed. Chief amongst economy’ is substantial compared to market econo- these is the construction of Africa’s largest dam, mies. In welfare economic textbook versions of a ‘mar- the Grand Ethiopian Renaissance Dam at a cost of ket economy’, government intervention is justified in US$4.2 billion (8.5 percent of GDP). The federal the context of market failure, including information and regional road network increased from 26,500 asymmetry, externalities, monopoly, and, to meet km in 1997 to 60,000 km in 2014. Railway lines social policy objectives. Government interventions connecting Addis Ababa with the Port of Djibouti in the Ethiopian ‘mixed economy’ model includes as well as a Light Railway line in the capital are near welfare economic justifications, but goes beyond this, completion. The customer base of Ethio Telecom (a as summarized in Table 2.1. state monopoly) rose from 7 to 26 million (mobile, fixed, internet) in 2011–14 and projects to upgrade the existing networks are ongoing. Ethiopia has also invested heavily in the agriculture, education, health, 4 Rodrik (2008) describes Ethiopia’s industrial policy framework as simply consisting of: (a) a list of priority sectors (export-oriented agribusiness, and water & sanitation infrastructure over the past textiles and garments, processed leather, and so on); and (b) a list of decades. In spite of this impressive growth, Ethiopia’s incentives (cheap land, tax incentives, technical support). Economic Strategy — ‘The Ethiopian Way’ 15 TABLE 2.1: Key Characteristics of the Ethiopian Economic Model Role of Government Intervention Production The government produces some goods and services and its rationale include: (a) to encourage competition (e.g. wholesale markets), (b) there is ‘insufficient capacity’ in the private sector (e.g. sugar production), and (c) to meet social objectives (e.g. keeping some retail prices low). Credit and The State channels the majority of credit and foreign exchange through state-owned banks, mainly the Com- foreign mercial Bank of Ethiopia (CBE) and Development Bank of Ethiopia (DBE). The former largely supports public exchange investment and the latter supports long term private investment. Protected Key services sectors (finance, telecom, trade logistics, retail) are protected from foreign competition on the basis sectors of ‘infant industry’ arguments that the domestic private sector is too underdeveloped to withstand foreign com- petition (e.g. retail) or the government regulator insufficiently prepared (e.g. financial sector). State monopoly Despite privatization of some SOEs, major state monopoly companies remain, including electricity production enterprises (EET) and distribution (EES), telecoms (Ethio Telecom), railways (ERC), sugar (Sugar Corporation), trade logistics (ESLSE), and air transport (Ethiopian Airlines). Sometimes profits are transferred between SOEs (e.g. from tele- coms to railways). Capital account Closed. This implies that domestic residents and banks do not have access to foreign capital markets. More- over, repatriation of profits from Foreign Direct Investment (FDI) is difficult. Land The Government owns all land. Land users can buy and sell lease rights. Promoting A key objective of government intervention is to promote ‘value creation’ and minimize ‘rent seeking’. To il- ‘value creation’ lustrate, if a private investor acquires land and builds a plant that converts a raw material (say leather) into an and avoiding intermediate or final product (say shoes) and employs labor in this process then the activity is ‘value creating’. ‚rent-seeking‘. If, on the other hand, the land is not put into productive use and the investor sells the land use rights at a profit five years later then the activity is termed ‘rent seeking’. An alternative term for the latter may be ‘speculation’. Economic Development Policy Public investment Substantial public investment is facilitated through a heterodox policy mix of: low or negative real interest rates, credit and foreign exchange allocations, real currency appreciation, recurrent expenditure restraint, and low international reserves. External In addition to the concessional multilateral credits, Ethiopia is relying substantially on bilateral non-concessional borrowing credits, especially from China. Sovereign bond financing is also used. ADLI Agriculture Development Led Industrialization (ADLI) which emphasizes smallholder agricultural growth to stimulate growth in other sectors of the economy, most notably industry. Sectorial Emphasis is put on the development of agriculture and manufacturing. The services sector receives less atten- policies tion (except exports). Key sectors (leather, textile, metal, cut flower and agro industry) are actively favored owing to their potential comparative advantage (labor intensive production drawing upon domestic resources base). Structural The Government vision of structural transformation follows the Lewis model, whereby the process of industri- transformation alization gradually absorbs surplus labor from the agriculture sector. This is associated with labor productivity growth, urbanization, and reduced population growth. Financial and Monetary Sectors Absence of key Negative real interest rates imply excess demand for credit, so the credit market clears via rationing as opposed financial markets to the price mechanism. The dollar is not depreciating fast enough compared to the domestic-foreign price dif- ferential. As a result there is excess demand and a black market premium. Given a fixed, low nominal interest rate, the Treasury Bill market also does not clear via the price mechanism. There is no stock market. A very lim- ited market exist for corporate and subnational bonds. The recent sovereign bond market marks an exception. Printing money Ethiopia has experienced very high levels of inflation over the past decade. The resulting seignorage provided and the the Government a substantial source of finance. The federal budget continues to be partly financed through seignorage direct advances from the central bank. Savings The Government has raised money for a major infrastructure project (the Grand Renaissance Dam) using bonds. A housing savings scheme, promoted by CBE, is also in place. Source: Own elaboration based on The Growth Report (2008). 16 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Did Ethiopia Fo he Insights of ‘The 2.2  narrowly coincides, including on labor markets and Growth Commission’?5 equality. On the other hand, in most other dimen- sions there is limited coincidence. Differences in key Ethiopia has emphasized public investment, but dimensions such as capital flows, financial sector de-emphasized the promotion of a vibrant pri- development, regional development and the quality vate sector. International experience underscores of the public debate, stand out. Similarly, Ethiopia the relevance of emphasizing public investment for implemented a number of what the Growth Report growth because it expands the range of opportuni- cites as suboptimal policies, such as energy subsidies, ties and returns on private investment. According to open-ended protection of some sectors, price controls the Growth and Development Commission (2008): and export bans. ‘No country has sustained rapid growth without also keeping up impressive rates of public investment—in Fast Growing Non-resources Rich 2.3.  infrastructure, education, and health. Far from crowd- African Peers ing out private investment, this spending crowds it in. It paves the way for new industries to emerge and raises Ethiopia’s growth experience stands out from a the return to any private venture that benefits from group of similar, high-performing regional peers, healthy, educated workers, passable roads, and reliable analyzed by the IMF. Five other non-resource electricity.’ However, the same report also notes that dependent African economies experienced high ‘Government is not the proximate cause of growth. growth rates, though not as high as Ethiopia: Burkina That role falls to the private sector, to investment and Faso, Mozambique, Rwanda, Tanzania, and Uganda entrepreneurship responding to price signals and mar- (SSA5). According to the IMF (2013a), these coun- ket forces … Government provides the environment tries (including Ethiopia) share several key character- for growth, but it is the private sector that invests and istics which help explain their growth performance, creates wealth for the people.’ Ethiopia followed the such as improved macroeconomic management, first part of this advice, but faces a challenge in fol- stronger institutions, increased aid, and higher invest- lowing the second. ment in human and physical capital. This leads the Remarkably, Ethiopia achieved high growth report authors to conclude that ‘their experience despite generally not following consensus view on demonstrates that improvements in macroeconomic how to achieve it. Aside from public infrastructure policy combined with structural reforms and reliable investment and a few other polices, Ethiopia generally external financing, can foster productive investment did not follow the recommendations of the Growth and stimulate growth’ (ibid). Commission. Table 2.2 offers a systematic compari- A closer inspection of the Ethiopian experience, son of the common characteristics of the 13 high however, reveals that it fits the IMF narrative only growth economies studied by the Commission and partially. What sets Ethiopia markedly apart is its the experience of Ethiopia. Of the five major char- emphasis on a state-led model of growth (which most acteristics, Ethiopia’s experience coincided in terms peers moved away from), the strong focus on agricul- of ‘committed, credible, and capable government’, tural development and uneven progress in macroeco- high investment, and relative macroeconomic stabil- nomic management and institutions. The absence of ity. On the other hand, Ethiopia did not fully exploit major recent structural reforms in Ethiopia also stands the world economy, let markets allocate resources, or mustered high rates of savings. In terms of the addi- 5 The Commission, headed by Nobel Laureate Michael Spence, based its tional ingredients of high growth experiences else- growth recommendations on the experience of 13 high growth economies where listed in the table, Ethiopia’s experience only (9 from East Asia). Economic Strategy — ‘The Ethiopian Way’ 17 BOX 2.1: The Banking Sector in Ethiopia Ethiopia’s banking sector consists of two state-owned commercial banks, one state-owned development bank and 16 private banks. Three public banks constitute 77 percent of total assets of the banking sector. Within this group are the Commercial Bank of Ethiopia (CBE) and the Development Bank of Ethiopia (DBE). CBE holds 80 percent of the total outstanding loans and DBE is a large holder of treasury bills. There are no foreign-owned banks. On average, banks appear to be well capitalized and profitable. The system-wide capital adequacy ratio was 17.2 in June 2014 compared to the 8 percent minimum requirement. The profitability of the banking sector remains high with return on assets and return on equity at 3.1 and 44.6 percent, respectively and well above regional averages (2 and 17 percent, respectively). Asset quality is also good with nonperforming loans at less than 3 percent of banks’ total loan portfolio (March 2014). The system-wide liquidity ratio, however, is only slightly above the 15 percent minimum requirement. Ethiopia’s banking sector is characterized by ‘financial repression’. The term financial repression was initially coined by McKinnon (1973), who defined it as government financial policies strictly regulating interest rates, setting high reserve requirements on bank deposits, and mandatorily allocating resources. Economists are divided as to whether financial repression is good or bad for economic growth (see Chapter 7, Box 7.1, for a discussion). In Ethiopia, the key characteristics include: (1) below market-clearing deposit rates; (2) relatively high reserve and liquidity requirements of banks; (3) allocation of the bulk of credit by state-owned banks, especially CBE and DBE. Nominal interest rates are low and rigid. The minimum deposit rate, regulated by NBE, has remained constant for the past five years at 5.0 percent. The lending rate is fully liberalized, but has been relatively unchanged over the same period with minimum and maximum observed lending rates unchanged at 7.5 and 16.25 percent, respectively. Moreover, the spread between the minimum deposit rate and the observed maxiumum lending rate has been constant at around 11 percent. As a result, changes in the real interest rates have exclusively been a product of changes in inflation (Figure 2.1.3). Domestic credit has declined over time, as a share of the economy, as a result of low real interest rates. Figure 2.1.1 reveals a substantial reduction in domestic credit as a share of GDP over the past decade from 35.7 percent of GDP in 2004 to 28.6 percent of GDP in 2014. This demonetization trend is mirrored in a similar decline of broad money (M2) as a share of GDP , which declined from 38.5 percent in 2004 to 28.4 percent in 2009. The most plausible explanatory factor for this trend has been the low real interest rate observed over this period. Credit growth has been increasingly concentrated in public projects rather than private ones. This is reflected in the composition of domestic credit stock (Figure 2.1.4). The share of private credit in total outstanding credit has declined from 37 percent in 2007/08 to 28 percent in 2014/15. Conversely, the share of loans to State Owned Enterprises increase from 21 to 62 percent over this period. Private sector credit, as a share of GDP , has declined because of declining domestic credit and a policy preference for financing SOEs. Put differently, the private sector has been taking a smaller share of a shrinking cake. Private sector credit declined from about 16 to 11 percent of GDP in 2004–14, according to the NBE definition, which counts all DBE loans as private. In sum, Ethiopia’s credit market is characterized by financial repression with an emphasis of financing public infrastructure projects. Figure 2.1.5 illustrates this in a simple investment-savings diagram. The savings curve slopes upward as a higher real interest rate encourages households to save. The investment curve is downward sloping because more investment projects are profitable at low real interest rates. In a market-based system, the real interest rate is determined where the two curves intersect. In the case of Ethiopia, the minimum deposit rate is effectively set by the Government (NBE) and this will have a strong bearing on the level of the real interest rate. Since it is set below the market rate, the market clears via quantity rationing. Some agents get access to credit at the below market rate. Other agents are completely excluded from credit access giving rise to ‘unsatisfied demand’. out, as discussed in Chapter 8, whereas African coun- and accountability’, ‘political stability and absence of tries, including SSA5, reformed substantially. Ethiopia violence/terrorism’, and ‘regulatory quality’. On the also did not improve much on standard quantitative other hand, foreign investors often cite economic and measures of institutional development compared with political stability as a reason to locate in Ethiopia. high-performing regional peers. Ethiopia lags consid- Though there have been improvements in ‘rule of law’, erably behind on three dimensions of the Worldwide Ethiopia still scored lower than peers. Encouragingly, Governance Indicators (Kaufman et al., 2010): ‘voice it was at par with others in government effectiveness 18 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 2.1: Financial Sector Indicators 1. Domestic Credit (percent of GDP) 2. Private Sector Credit (percent of GDP) 40 90 25 80 35 70 20 60 30 15 50 40 25 10 30 20 20 5 10 15 0 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Ethiopia LIC SSA (RHA) Ethiopia LIC SSA (RHA) Ethiopia (IMF definition) (NBE definition) 3. Real Interest Rates 4. Composition of Domestic Credit Stock (percent) 20 100% 10 36 37 36 39 36 36 33 31 30 80% 0 –10 9 8 60% 9 14 –20 23 42 37 33 –30 40% 49 58 61 62 –40 50 20% 41 –50 27 29 21 14 –60 0% Jun-02 Dec-02 Jun-03 Dec-03 Jun-04 Dec-04 Jun-05 Dec-05 Jun-06 Dec-06 Jun-07 Dec-07 Jun-08 Dec-08 Jun-09 Dec-09 Jun-10 Dec-10 Jun-11 Dec-11 Jun-12 Dec-12 Jun-13 Dec-13 Jun-14 Dec-14 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 2014/15 Real maximum lending rate Real minimum deposit rate SOEs CG credit Private credit 5. Quantity Rationing in Credit Markets 6. Banking Sector Market Share (deposits) Real interest Savings (S) CBB OIB ZB Others rate (r) CBO NIB WB Market BOA determined UB rate Below AIB market rate DB Investment (I) CBE S, I Satisfied Unsatisfied demand demand Source: World Bank (WDI), NBE, IMF and CSA. NBE definition includes DBE. IMF definition includes DBE. Economic Strategy — ‘The Ethiopian Way’ 19 TABLE 2.2: Growth Commission Recommendations and Ethiopia’s Experience Common characteristics of fast growth Ethiopia’s Strategy and Experience They fully exploited the world economy Has a very low export-to-GDP ratio, is not a WTO member, has high services restrictiveness levels and overvalued real exchange rate. They maintained macroeconomic stability Has maintained some degree of macroeconomic stability, but has often chosen to forgo some stability in the interest of pursuing high growth through public investment. They mustered high rates of saving and investment Investment: Yes. Savings: Not until very recently. They let markets allocate resources Ethiopia has a mixed economy where the markets allocate some resources, while the Government allocates others. Committed, credible, and capable governments Yes. Ingredients Technology transfer Limited. Also not taking full advantage of FDI in services sector. Competition and structural change Limited. Efficient labor markets Yes Export promotion and industrial policy Investment in airline services and horticulture were successful. Other areas have been less successful. Capital flows and Financial Market Openness Closed capital account. Financial sector development Financial repression. Urbanization and rural investment Urban infrastructure investments are implemented. Rural invest- ment is high. Equity and equality of opportunity Very low income inequality. Strong progress on equality of oppor- tunity, but low levels remain. Regional development There is a need to develop ‘second cities’ to complement the capi- tal, Addis Ababa. The environment and energy use Ethiopia is a regional leader on promoting the green economy. Energy subsidy encourages ‘over consumption’. The quality of the debate Limited. Examples of Sub-optimal growth policies Ethiopia’s Strategy and Experience Subsidizing energy except for very limited subsidies targeted Has some of the lowest tariffs in the world with across the board at highly vulnerable sections of the population. subsidies benefitting primarily the better-off who actually have ac- cess to energy. Providing open-ended protection of specific sectors, indus- Many services sectors benefit from the absence of foreign competi- tries, firms, and jobs from competition. tion. Imposing price controls to stem inflation, which is much better Temporarily adopted in January-May 2011. handled through macroeconomic policies. Banning exports for long periods of time to keep domestic Grains exports are currently banned. prices low for consumers at the expense of producers. Resisting urbanization and as a consequence underinvesting Ethiopia has very low urbanization rates and the Government’s in urban infrastructure. interest in urbanization is relatively recent. Source: Own elaboration based on The Growth Report (2008). 20 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT and control of corruption. Finally, Ethiopia has government leadership in the development process, consistently scored lower than SSA5 on the World high rates of investment, the promotion of light manu- Bank Country Policy and Institutional Assessment facturing and structural change. However, there are a (CPIA) index (3.4 vs. 3.8 in 2005–13). number of important differences of economic strategy: However, there is something that sets Ethiopia apart institutionally that is not well captured in  First, agriculture features much more promi- most commonly used governance indices. Ethiopia nently in Ethiopia than in the Asian growth has, to some extent, followed the ‘authoritarian devel- strategy. A sizable agricultural economy was opmental state model’ with a mixed state-market a point of departure for Japan, Korea, Taiwan approach with country specific aspects. One very (China) and China and in the latter case, the suc- important aspect of such a model is that rents are cessful reform of the rural economy lent impetus managed with a view to increasing productivity and to and helped win political backing for the reform competiveness over time, and that individual rent- process. Resource transfers from the rural sector seeking does not crowd out this orientation. Ethiopia’s through terms of trade effects, may have contrib- main strength is having institutionalized how power uted modestly at the start of the industrialization is organized based on a development oriented vision, but their role diminished rapidly. It was the trans- including an understanding of how development fer of workers from the agricultural sector into needs to be shared among competing groups to industrial jobs in rapidly urbanizing economies keep the country unified. These systems are not just that raised productivity and enlarged the GDP. heterodox from an economics point of view, they This is the model of structural change Ethiopia is are also heterodox institutionally—i.e. they do not now trying to follow, though it has yet to succeed conform to the Western consensus of what is a good as workers are moving to services instead. institutional model—and consequently don’t show  Second, East Asia’s performance in the last quarter up as particularly good on World Governance indica- of the 20th century is inseparable from the suc- tors. The experience in Africa is different—efforts to cess of numerous firms mainly drawn from the pursue state-led development have failed where they private sector, specialized in manufacturing and were tried and mostly ended in the 1980s (debt cri- the penetration of Asia’s manufactured exports sis/structural adjustment). The main reasons for such into the markets of western countries. failure include poor investment choices, a failure to  Third, the acceleration of growth in East Asia invest in a ‘Green Revolution’ to improve agriculture, was followed quickly by an increase in domestic and rent-seeking and corruption. saving (which bankrolled development) and in investment with the private sector in the forefront 2.4. An East Asian Strategy?6 because expanding opportunities and government incentives created profitable opportunities, which When Ethiopia’s economic strategy is viewed companies were eager to capitalize on and not through the East Asian or Chinese mirror, there because of a change in the business environment. are similarities but also important differences. Despite recent increases in savings in Ethiopia, Ethiopia, in its policy pronouncements, highlights it still has a much larger savings-investment gap manufacturing, underscores the desirability of using than the East Asian countries. exports as a lever to raise the growth rate and relies on state owned or controlled firms to promote industrial change. It often compares itself with China, Vietnam 6 This sub-section draws heavily upon a background paper by Shahid and Korea. There are similarities, including substantial Yusuf (2014) commissioned for this report. Economic Strategy — ‘The Ethiopian Way’ 21  Fourth, infrastructure did not drive growth and China, the structure of the Ethiopian economy, in East Asia although it undoubtedly played a its low level of urbanization, its export composition supporting role. All East Asian countries invested (notably the large share of unprocessed commodi- in infrastructure to accommodate the needs of ties and of transport services), its heavy dependence industry, trade and urbanization, but such spend- on foreign financing, and the continuing salience ing followed rather than led the development pro- of the state in economic decision-making, does not cess. In Ethiopia, recent growth has been driven resemble that of the East Asian “tigers” nor for that by infrastructure as discussed later in Chapter 3. matter does the quantum of resources (including More recently, China has sustained its growth at savings) mobilized domestically, the investment by great cost, by investing lavishly in infrastructure the private sector in tradable activities, the number and urban real estate for which there is little of Ethiopian firms active in the global marketplace immediate demand, but this was not the East and the influx of FDI. Asian norm during its growth heyday. In summary, more than anything, Ethiopia’s  Fifth, East Asian countries actively used real growth strategy stands out for its uniqueness rather exchange rate undervaluation to gain com- than its resemblance to other strategies. It is char- petitiveness and promote growth and exports. acterized by a focus on agricultural and industrial By contrast, Ethiopia’s real exchange rate has development with a strong public infrastructure drive remained overvalued over the past several decades. supported by heterodox macro-financial policies. It has a few characteristics in common with regional fast Ethiopia’s strategy, while it incorporates ele- growers, draws selectively from East Asian policies, ments of the East Asian model, diverges signifi- but bears little resemblance to the conceived wisdom cantly in terms of conception and outcomes from derived from the Growth Commission. Finally, it that of East Asian. After a decade of growth that stands out for a relative absence of structural economic matches the highest rates achieved by Japan, Korea policy reforms—a topic we re-visit in Chapter 8. 23 EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS7 3 The econometric method is a System Generalized Growth was driven by public infrastructure investment and restrained government consumption supported by a Method of Moments (GMM). Internal instruments conducive external environment. Using a cross-country are used to avoid endogenity biases of lagged depen- regression model, we are able to distinguish between dent and explanatory variables. The model is estimated structural, external, and stabilization factors. The growth acceleration is mainly explained by structural factors, on 126 countries for the 1970–2010 period, includ- including infrastructure and low government consumption. ing low income countries and using 5 year averages Increased trade openness and the expansion of secondary of non-overlapping averages. We divide our period of education also helped, but the effects were modest. Macroeconomic factors held back some growth, owing to analysis as follows: Early 2000s, Late 2000s, and Early declining private credit, real currency overvaluation, and 2010s (see Annex A3.1 for a definition). relatively high inflation. The model has three types of variables as well as a persistence effect. Structural factors include variables such as human capital, private sector credit, trade open- 3.1 Introduction ness, infrastructure and government consumption. Stabilization factors include inflation, the real exchange The purpose of this chapter is to classify growth rate and the presence of a banking crisis. External determinants into structural, external, and stabili- factors are commodity prices and terms of trade. The zation factors. In previous chapters we described the model also includes a persistence effect, capturing the characteristics of the Ethiopian growth process and fact that changes in the underlying variables can also the economic strategy that supported it. We found affect growth in periods after they were implemented. that growth was concentrated in the services and The model accurately predicts Ethiopia’s agriculture sectors, and characterized by high total growth over the period of analysis and is reasonably factor productivity and substantial capital and labor robust. Our growth predictions are quite similar to accumulation. Economic strategy was characterized observed rates and the model has a better predictive by high public investment and heterodox financing power than similar studies on Ethiopia. Moreover, it arrangements, among others. This chapter brings these passes a series of robustness tests, including the choice pieces together by linking growth performance directly of infrastructure variable which emerges as one of the to economic policies. key explanatory factors. We apply a cross-country regression model to The chapter focuses on presenting the main results derive insights about the determinants of growth. of the analysis while deferring the methodology to a In particular, we use an existing regression model technical annex. Specifically, Section 3.2 presents the developed by Brueckner (2013) and originally con- main results of the analysis. Annex A3.1 describes the structed to investigate growth in Latin America. This methodology used to derive these results. Annex A3.2 approach avoids tweaked Ethiopia-specific results and examines model robustness. Additional details on meth- helps address the following question: Can we explain odology are available in Moller and Wacker (2015). Ethiopia’s recent growth performance by factors also observed to influence growth elsewhere? 7 This chapter is a summary of Moller and Wacker (2015). 24 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 3.1: Regression Results: Key Growth Drivers in Ethiopia (Real GDP per Capita) 2000–2013 Early 2000s Late 2000s Early 2010s 7 7 7 8 Annualized GDP p.c. growth (in %) 6 6 6 5 5 5 6 4 4 4 4 3 3 3 2 2 2 2 1 1 1 0 0 0 0 –1 –1 –1 –2 –2 –2 –2 Persistence Structural Stabilization External Residual (model error) Note: A positive residual implies that actual growth was higher than predicted growth. Source: Table 3.1. Explaining Ethiopia’s Recent Growth 3.2  (Figure 3.2.2). As previously mentioned, public invest- Performance ment was also financed by extension of credit through the domestic banking sector and foreign borrowing, Economic growth in 2000–13 was driven primarily and partly executed through off-budget state owned by structural improvements, according to model enterprises. It also helped for growth (0.3 ppts) when results. The results are illustrated in Figure 3.1, Ethiopia opened more to international trade, espe- which summarizes more detailed results contained cially in the Early 2000s, as trade of goods and ser- in Table 3.1. The left panel of Figure 3.1 shows that vices increased its share of GDP from 37.5 percent in almost all of the (predicted) growth performance of 2002 to 48.7 percent in 2012 (World Bank, 2014a).8 the 2000–2013 period can be attributed to structural Although Ethiopia made substantial efforts in improv- factors: the model estimates these factors to have ing educational attainment of its population over the contributed 3.9 percentage points (ppts) of average past decade, the growth enhancing effects hereof have annual per capita growth rate of 4.3 percent (measured been somewhat modest (0.3 ppts), as reflected in the in Purchasing Power Parity, PPP). The positive effect regression results and the Solow decomposition results of the external environment was outweighed by the presented in Chapter 1 (Figure 1.1.5), though this may negative effects associated with macroeconomic imbal- be a measurement problem as discussed by Pritchett ances (each accounting for 0.3 ppts). The remaining (2001). Secondary education (the explanatory vari- growth impetus was explained by persistence effects. able) remains limited, even if gross enrolment improved Finally, the residual captures the difference between from 13.1 to 32.8 percent over the period of analysis.9 model predicted growth and actual growth. Financial disintermediation held back some Public infrastructure investment and restrained growth. An expansion of credit to the private sector government consumption were the key structural enables firms to invest in productive capacity, thereby drivers of growth. These two factors were linked as an expansion in budgetary infrastructure investment 8 See also Chapter 2 Figure 2.2.1 on the trade policy reform side. was facilitated by reduced government consumption, 9 See World Bank (2012) for an analysis of the challenges related to secondary education in Ethiopia. There is also evidence of a substantial arising partly as a result of a ‘peace dividend’ effect mismatch between skills supply and demand, for instance in manufactur- following the end of the 1998–2000 war with Eritrea ing (World Bank, 2014b). EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 25 TABLE 3.1: Parameter Values, Changes and Predicted Growth Effects (real GDP per capita) 2000–13 Early 2000s Late 2000s Early 2010s Predicted Predicted Predicted Predicted Parameter Change effect Change effect Change effect Change effect Persistence 0.781 0.005 0.28% –0.010 –0.82% 0.039 3.07% 0.041 3.20% Structural: 3.84% 2.52%   3.01% 1.20% Δln(schooling) 0.018 0.064 0.20% 0.094 0.17% 0.090 0.16% 0.008 0.01% Δln(credit/GDP) 0.074 –0.013 –0.16% 0.012 0.09% –0.009 –0.07% –0.042 –0.31% Δln(trade/GDP) 0.082 0.020 0.28% 0.052 0.42% 0.019 0.16% –0.012 –0.10% Δln(govt C) –0.262 –0.040 1.79% 0.009 –0.24% –0.041 1.08% –0.086 2.26% Δln(tele lines) 0.141 0.072 1.74% 0.147 2.07% 0.119 1.68% –0.051 –0.72% Δln(institutions) –0.003 0.000 0.00% 0.000 0.00% 0.000 0.00% –0.000 0.00% Stabilization: –0.28% 0.09% –0.58% 0.01% Δln(inflation) –0.011 0.052 –0.10% 0.055 –0.06% 0.285 –0.32% –0.182 0.21% Δln(exch rate) –0.064 0.016 –0.18% –0.023 0.15% 0.040 –0.26% 0.031 –0.20% Δln(bank crisis) –0.040 0.000 0.00% 0.000 0.00% 0.000 0.00% 0.000 0.00% External: 0.34% –0.55% 0.95% 0.18% Δln(TOT change) 0.118 0.016 0.32% –0.047 –0.55% 0.081 0.95% 0.014 0.16% Δln(commodity 10.482 0.000 0.01% 0.000 0.00% 0.000 0.00% 0.000 0.02% prices) Predicted 4.18% 1.24% 6.45% 4.53% average annual GDP per capita growth rate Source: Authors’ calculation, obtained by inserting Ethiopian values for the explanatory variables and using the regression coefficients (parameters) of the baseline model presented in Annex Table A3.1 (column 1). laying the foundation for a sustainable growth path. Structural improvements were particularly However, Ethiopia is falling behind its peers in this important in the late 2000s supported by a posi- area (Figure 3.2.3). Regression results suggest a small tive external environment, but held back by macro negative growth effect of financial repression policies imbalances. How did growth drivers change over whereby rationed credit is channeled through state- time? While structural factors dominate in the Early owned banks primarily towards public investment. 2000s (2.5 ppts) and Late 2000s (3.1 ppts), they It is noteworthy that the negative quantitative effect become less pronounced in the Early 2010s (1.2 ppts). was substantially smaller than what is generally per- The diminishing effect in the last period, however, is ceived (–0.2 ppts in 2000–13). On the other hand, affected by the choice of the infrastructure variable, as the negative growth effect shows an increasing trend discussed in Annex A3.2. The growth contribution of (rising to –0.3 ppts in the Early 2010s), suggesting that financial repression policies may become more 10 See Huang and Wang (2011) for a review of the effects of financial repression policies on growth and for empirical evidence suggesting costly if maintained. Indeed, this would be consistent that such policies were conducive in the early (but not later) stages of with the experience of China.10 growth in China. 26 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 3.2: Trends in Key Growth Drivers 1. Infrastructure Trends 2. Public investment and consumption (% of GDP) 100 1.4 20 18 Telephone (Right Axis) 80 1.1 16 14 60 0.8 12 Road density 10 40 0.5 8 6 20 0.2 4 Mobile 2 0 –0.1 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 Mobilesubscriptions (per 100 people) Public investment Public consumption Road density (km per 1,000 sq.km) Telephone lines (per 100 people, RHS) 3. Private sector credit (% of GDP) 4. Exports: Value and Volume, 2000–2011 25 600 20 500 15 400 10 300 200 5 100 0 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Ethiopia (IMF definition) LIC average Value Volume SSA average Ethiopia (NBE definition) 5. CPI Inflation (average, percent) 6. Real Effective Exchange Rate (REER) 50 130 45 125 40 120 35 115 30 110 25 105 20 100 15 95 10 90 5 85 0 80 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 Ethiopia SSA LIC Source: 1: Moller & Wacker (2015). 2: World Bank (2013). 3: World Bank (WDI). 4: World Bank (2014a). 5.5–5.6: IMF (WEO). EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 27 the external environment increased over the period, other countries in Africa achieved better monetary shifting from a negative (–0.5 ppts) to a positive control by shifting toward reliance on indirect instru- contribution (0.9 ppts) between the Early 2000s to ments like open market operations to soak up liquidity. the Late 2000s and a weak positive effect in the Early Monetary policy instruments in Ethiopia, in contrast, 2010s (0.2 ppts). The deteriorating macro environ- are limited to changes in bank reserve requirement ment was a drag on growth only during the late 2000s and sales/purchases of foreign exchange reserves. In (–0.6 ppts), but not in the other two periods. The addition, the real exchange rate was allowed to appre- substantial persistence effect observed in the late 2000s ciate substantially since 200412, owing to insufficient and early 2010s (about 3 ppts) echoes the growth nominal depreciation in a system of foreign exchange drivers during the early and late 2000s, respectively. rationing (Figure 3.2.6).13 This made the import of We interpret this as the lagged effect of the structural expensive public infrastructure equipment relatively improvements implemented in earlier periods.11 cheaper but undermined export competitiveness. The A conducive external environment also sup- net effect on growth was negative (0.3 ppts in the ported the growth acceleration. As argued in World Late 2000s), though not as substantial as estimated Bank (2014a), a strong rise in exports helped support elsewhere.14 Nevertheless, what stands out from these the economic boom. Since 2003 exports quadrupled simulations is that the negative growth effect of some in nominal terms, while volumes doubled, reflecting a macro policies was not that substantial. High inflation substantial positive price effect (Figure 3.2.4). This is and an overvalued exchange rate cut little over half a consistent with findings by Allaro (2012) that exports percentage point from economic growth rates. This ‘Granger caused’ growth and with Gebregziabher helps explain why Ethiopia was able to grow fast in the (2014) who finds strong causal effects from exports presence of some sub-optimal macro policy choices. on real output. The strong growth effect is somewhat Would economic growth have been higher in the surprising given that merchandise exports account for absence of heterodox macroeconomic policies? It is only 7 percent of GDP—the lowest among populous hard to give a precise answer to this question. However, developing countries (services exports account for an it should be pointed out that financial repression, a additional 7 percent of GDP). strong real exchange rate and monetary policy induced The growth drag of macroeconomic imbalances inflation all helped support high public infrastructure sets Ethiopia apart from high performing regional peers, though the effect was modest. Growth was 11 The subsistence effect should not be conflated with the unexplained held back by high inflation and an overvalued exchange residual as per Equation (2) in Annex A3.1. rate in the Late 2000s. The country experienced much 12 The degree of real exchange rate overvaluation fluctuated over time. World Bank (2013) shows that the real exchange rate of Ethiopia has larger inflationary impacts of the two commodity remained overvalued throughout the 1951–2011 period. By 2011, the price shocks in 2008 and 2011 than other low income RER over-valuation was 31 percent. The IMF (2014), using alternative methods and measurement, finds that the real effective exchange rate and African countries (Figure 3.2.5). This is partly was overvalued by 10–13 percent in 2014. explained by expansionary policies in the form of high 13 The de facto exchange rate arrangement is classified as a crawl-like ar- rangement by the IMF (2013b). The authorities describe it as a managed growth of the monetary base owing to credit expan- float with no predetermined path for the exchange rate. The annual pace sions to state owned enterprises and direct central of nominal depreciation, however, has been stable at 5 percent in recent years. The NBE continues to supply foreign exchange to the interbank bank lending to the government. Following the 2010 market based on plans prepared at the beginning of the fiscal year, which devaluation, the monetary authorities also allowed the take into account estimates of supply and demand. 14 The World Bank (2014a) estimates that 10 percent real exchange rate unsterilized accumulation of foreign exchange reserves overvaluation in Ethiopia holds back the growth rate by 2.2 percentage arising from the ensuing rise in exports and this con- points. This is generally consistent with our coefficient derived from the full cross-country data set, which implies that an immediate 0.6 percent- tributed to additional inflation (inducing a negative age point decrease in the short run will cumulate to a 3 percentage point growth effect of 0.3 ppts in the Late 2000s). Most effect over the long run. 28 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 3.3: Infrastructure Growth Rates: Ethiopia in the Global Context (1970–2010) Telephone lines Roads 3 1.0 2 Log change over decade Log change over decade 0.5 ETH ETH 1 BFA RWA UGA TZA 0 TZA 0 MOZ MOZ –1 –0.5 –2 –1.0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 Rank of country on (0–1) scale Rank of country on (0–1) scale Mobile phones 15 Log change over decade 10 ETH BFA UGA 5 TZA MOZ RWA 0 0 0.2 0.4 0.6 0.8 1.0 Rank of country on (0–1) scale All countries and decades Ethiopia 2001–2010 SSA5 2001–2010 Source: Author’s calculation based on WDI data. investment. By providing full access to credit at negative Why did infrastructure contribute so strongly real rates and foreign exchange at below market prices, to growth in Ethiopia? In brief, because of the sub- the cost of public investment was reduced substantially. stantial expansion in physical infrastructure that Moreover, direct central bank financing of the budget took place over the past decade combined with the deficit and seignorage allowed more public investment estimated high returns to infrastructure investment to be financed, thus supporting growth. In the presence (derived from 124 countries in 1970–2010). To put of more orthodox macro policies, public infrastructure Ethiopia’s infrastructure performance into perspec- investment would have been lower, which would have tive, we use the underlying data set to identify decadal lowered growth. The net growth effect of these two infrastructure growth rates for the 124 countries over alternatives is difficult to estimate with precision, but 4 decades (1970s–2000s). Figure 3.3 illustrates the it is unlikely that that public infrastructure investment results for three infrastructure variables. Each obser- could be maintained at similar high levels on the back vation represent a country and each country is repre- of orthodox macro policies. sented 4 times. Ethiopia’s performance in the 2000s EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 29 FIGURE 3.4: Prediction Performance of the Model 2000–2013 Early 2000s Late 2000s Early 2010s 9 9 9 9 8 8 8 8 7 7 7 7 6 6 6 6 5 5 5 5 4 4 4 4 3 3 3 3 2 2 2 2 1 1 1 1 0 0 0 0 Actual Predicted Predicted alternative infrastructure Source: Authors’ calculation. Note: ‘predicted’ is based on the baseline model, ‘predicted alternative infrastructure’ subtracts the effect of phone lines from the baseline result and instead adds the average effect of roads and cell phones. is marked in red. We note that infrastructure growth cross-country OLS regression model reported in (measured in log changes over decades) is among the IMF (2013a), where the residual effect predominates. 20 percent fastest in all 3 cases. That infrastructure Our residuals are quite small as depicted previously growth is high for countries at low levels is not sur- in Figure 3.1 (gray bars). prising. However, the results resonate further as SSA5 Our key findings are summarized as follows: countries are also plotted (in blue). Ethiopia outper- Overall, the model provides good predictions of formed these countries in infrastructure growth, which Ethiopia’s growth performance since 2000 and our helps explain recent economic growth. results offer plausible quantitative explanations for Our growth predictions are quite similar to key economic trends and policy developments over observed rates and the model has a better pre- this period. We find evidence that public infrastruc- dictive power than similar studies on Ethiopia. ture investment—financed in part by restrained Figure 3.4 compares actually observed per capita government consumption—was the key structural growth with the predictions from the model. As illus- driver of growth. While this policy mix enjoys broad trated, the model slightly under-predicts growth in support among mainstream macroeconomists, het- the entire period (2000–2013) while almost perfectly erodox policies pursued by the Ethiopian govern- capturing the differing magnitude of the slow growth ment such as financial repression, an overvalued real period in the Early 2000s and the growth acceleration exchange rate, and monetary policy induced infla- of the Late 2000s. Predictions for the Early 2010s tion (including ‘printing money’ to finance public are the least precise, but this can be corrected by a spending) are more controversial. Interestingly, the substitution of infrastructure variables, as discussed empirical results show that the growth dividend of later. While discrepancies for country-specific results the former set of policies outweighed the drag from derived from cross-country studies are not excep- the latter. This helps shed light on the question of tional in the literature (see results for LAC countries how Ethiopia could achieve high economic growth in in Araujo et al., 2014), our residual compares very the presence of seemingly growth-inhibiting macro- favorably to the Ethiopia results derived from the financial policies. 30 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Annex 3.1 Methodology countries and allow us to decompose those correlates of growth in more detail. Our analysis is based on previous studies investigat- ing the determinants of growth in developing coun- Model and Data tries. We use an empirical growth model originally set up by Loayza et al. (2005) which was improved and The underlying model expresses domestic income updated by Brueckner (2013). These cross-country as function of key growth drivers. Our goal is to growth regression models were originally constructed estimate the impact of certain variables Xct on domes- to investigate growth in Latin America and the tic income, measured as the natural log of real PPP Caribbean (Araujo et al, 2014). GDP per capita (lnyct for country c in period t). More Such cross-country growth regressions may formally, the estimated equation can be written as the have their limitations, but their insights should not dynamic (‘steady-state’) process: be neglected. We are well aware that cross-country growth regressions have their limitations. While some lnyct = θlnyct–1 + Γln(X)ct + ac + bt + ect(1) strands in the literature neglect them altogether, we at least view them as one possible approach to gain where ac and bt are country and time fixed effects, respec- insights into the dynamics of growth (see also Durlauf, tively; and ect is an error term that remains unexplained 2009), especially in a country like Ethiopia where data by the model (‘residual’, i.e. the difference between coverage is relatively scarce. Furthermore, our econo- predicted and observed growth). Note that a time metric approach tries to address the most conventional period t is the average over (non-overlapping) 5-year methodological issues that can arise in cross-country periods to smoothen short-run and cyclical effects. growth regression exercises. These drivers of growth are grouped into the This approach avoids tweaked Ethiopia- categories of structural, stabilization, and external specific results. Taking an existing cross-country effects. Following the above-mentioned studies, the regression model to analyze Ethiopia’s recent growth individual variables in the vector Xct are assigned to performance potentially runs the risk that Ethiopia- those three categories. This facilitates an interpreta- specific factors might not be well-reflected in the tion whether growth was driven by “good policies” model. However, this is essentially our goal as set- (structural, stabilization) or “good luck” (external). ting up a new cross-country model for our purpose We briefly discuss the individual variables (which are will be prone to model selection that produces the taken from the Brueckner (2013) dataset) and their ‘best’ results for the specific case of Ethiopia. In intuition below. A more detailed technical description, essence, our approach helps address the following including the original data sources and variable values question: can we explain Ethiopia’s recent growth for Ethiopia is presented in Moller and Wacker (2015). performance by factors also observed to influence Stabilization variables contain inflation, bank- growth in other countries? Or is the Ethiopian case ing crises and the exchange rate. Capturing the idea a specific one? Even if one remains skeptical of cross- that macroeconomic fluctuations can influence growth country growth regressions, a failure of the model to over an extended period, we control for the number of appropriately predict observed growth in Ethiopia banking crises in each period, the inflation rate, and would suggests that factors that correlate with growth the exchange rate. A decrease in the latter variable is in most countries cannot explain Ethiopia’s growth equivalent to a currency depreciation.15 acceleration. On the other hand, a good predictive performance would imply that Ethiopia’s growth 15 As the interpretation of the exchange rate variable is somewhat difficult acceleration is in line with experiences of other in the cross-country context from a policy perspective, it should rather be EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 31 Structural variables capture a broad set of capital than attainments. Unfortunately, data on such fundamental country characteristics. This includes a level of quality is not available for a broad range of secondary school enrollment as a proxy for human countries. Researchers thus face a trade-off between data capital, a measure for trade openness (trade-to-GDP sophistication and country coverage. Since data quality ratio adjusted for population), an institutional variable is especially poor in lower income countries, while our (polity2), and private credit-to-GDP as a measure of goal is to include them in the sample to obtain most financial development. Our baseline variable for infra- appropriate estimates for Ethiopia, we are thus limited to structure is fixed telephone lines per capita but given the mentioned data at hand. On the upside, this leaves us the substantial impact obtained for this variable, we also with a panel of 126 countries for the 1970–2010 period perform alternative specifications with mobile phone (see Moller and Wacker (2015) for country coverage). and road coverage. Furthermore, our model includes government size (government consumption to GDP). Estimation Although several government expenditures can have a beneficial effect on income (especially in areas like We use System GMM estimation with a limited set health, education, or public infrastructure), the essen- of internal instruments. This method is appropriate tial idea of this variable is to capture the negative effects as some of the explanatory variables, Xct, may them- that an excessive government and associated taxes can selves be a function of the dependent variable and have on private activity. As our model describes long because dynamic panel estimation in the presence run growth, this should not be confused with the of country fixed effects generally yields biased esti- positive stimulative effects that increased government mates for the lagged dependent variable (e.g. Nickel, consumption can have during economic downturns. It 1981; Wooldridge, 2010). Our estimator uses inter- should also be noted that our model is conditional on nal instruments to avoid endogeneity biases. More other variables, i.e. the positive effect of government specifically we use one first-differenced lag of the spending for education and infrastructure, e.g., will be explanatory variables as instrument for those variables captured by those variables (and lagged GDP).16 Finally, in levels (see Arellano and Bover, 1995, and Blundell a high level of consumption (i.e. recurrent) expen- and Bond, 1998).18 To addresses most conventional ditures limits fiscal space to counter cyclical shocks. Generally welcome counter-cyclical measures can then seen as a control variable, i.e. controlling for the fact that an undervalued only be financed with relatively distortionary taxes (see exchange rate might boost growth temporarily. 16 See also Loayza et al. (2005: 40–41). Optimally, one would like to also Afonso and Furceri, 2010, on the effect of expen- subtract such expenditures like health, education, or infrastructure diture volatility) or an increased debt burden.17 but this is not feasible for the wide range of countries included in the sample. Following the line of reasoning above, one should also note that External factors are reflected in terms of trade the negative effect of government size is considerably smaller when the and commodity prices. Net barter terms of trade and model is estimated unconditionally (i.e. without controlling for other variables), reflecting the fact that it then implicitly captures the positive the country-specific commodity export price index of effects of education, infrastructure etc. (see the results in Araujo et al., Arezki and Brueckner (2012) are used to capture the 2014, especially Tables 3 and A.3). 17 For an alternative view on the effects of government consumption on most important effects of the global environment on output in neoclassical growth models, see e.g. Aiyagari et al. (1992) and growth. Furthermore, global conditions will also be the literature therein. 18 We limit the instrument set to one lag in the baseline model in order reflected in the time dummies. to ensure that the number of instruments does not grow too large in the Although our variables are not perfect, they allow System GMM estimation and furthermore avoid over-fitting the model by using the ‘collapse’ sub-option in the STATA xtabond2 command. Com- for a large coverage of developing countries. In the modity prices, terms of trade and time dummies are treated as exogenous. best case, one would have more sophisticated variables to We also use the one-step estimator as the two-step estimator is infeasible given the dimension of our data set. This also avoids severely downward reflect the underlying economic rationale. E.g., educa- biased standard errors associated with the two-step estimator (Blundell and tional achievements might be a better proxy for human Bond, 1998). See Brueckner (2013) for further details and discussions. 32 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT methodological pitfalls, our model includes country time-invariant. To illustrate: growth in the Late 2000s fixed effects which avoids unobserved heterogeneity can be explained by a dynamic persistence effect from across countries, while the use of internal instruments the Early 2000s, the change in explanatory variables avoids endogeneity biases. By limiting the instrument X in the Late 2000s, and a global time-specific shock set to one lag we avoid the well-known problems asso- (relative to the previous period). In this setting, the ciated with too many instruments (Roodman, 2009).19 persistence effect should be interpreted as an ‘echo’ Despite the limitations and concerns about (or fading out) of previous improvements. Finally, the System GMM, of which we address several in residual part Δect remains unexplained by the model. robustness checks, we find it to be the most For calculating Ethiopia’s drivers of growth, appropriate estimation method for our purpose. we extend its explanatory variables by one 5-year System GMM can incorporate a wide range of (lin- period. As the original dataset contains average 5-year ear) relationships among current and lagged values values across countries until 2010 only, we update the of economic variables. It also helps isolate exogenous Ethiopia values using 2013 data. We treat the latter changes in a variable from automatic reactions of that period (the ‘Early 2010s’) as if it represented a 5-year variable to other variables in the system and is careful period average for 2011–2015. This can be moti- not to confuse changes in a variable with a temporary vated by the consideration that the 2013 value is the shock. Furthermore, the identification strategy over mid-point for this period and most macro variables time variation makes it appropriate to assess our peri- are highly persistent. Since not all of the 2013 data ods of interest for one country, as opposed to identi- were available from sources fully consistent with the fication strategies using cross-country variation (like original dataset we added to the logged original series the between-effects estimator) that also potentially the log changes between the 2006–2010 averages suffer from unobserved cross-country heterogeneity. and the 2013 values of those data we had available.21 Given the dimension of our panel data set (especially Despite some caveats, this approach allows us to also the focus on a relatively short period) and some data analyze the most recent period of Ethiopia’s growth gaps, we also find it superior to cointegration meth- performance. Furthermore, it should be noted that ods. To address remaining concerns, we also perform this extension does not affect the model estimation several robustness checks in addition to the battery of checks applied by Brueckner (2013), as discussed 19 Since the preparation of the Brueckner (2013) study, recent findings in Annex 3.2. have shown that this might give rise to the opposite problem of suspi- ciously weak instruments (Bazzi and Clemens, 2013; Kraay, in progress) which we briefly address in the robustness section. Calculating Growth Contributions in 20 Our calculation differs slightly from the one in Brueckner (2013) and Araujo et al. (2014) as we do not use the actual lag of the growth rate for Ethiopia calculating the persistence effect from equation (2) but instead take the growth rate as predicted from the model. Furthermore, we take second differences of external factors as they enter the estimated levels equation Growth contributions over each time period can be already in first differences. To calculate effects for the 2000–2013 period, calculated by first-differencing equation (1): we proceed as follows to accommodate dynamic effects: we calculate the changes over the full 15-year period and multiply them by the respective coefficient Γ times (3+2θ+θ2)/3. This assumes that the change has been Δlnyct = θ(Δ lnyct–1)+ ΓΔln(X)ct + Δbt + Δect(2) uniform over time and accommodates their dynamic effects. Similarly, the persistence effect is calculated as (θ+ θ2+ θ3)/3 times the growth rate in the period prior to 2000. as log-changes approximate growth rates of a variable. 21 Where series were not identical (e.g. education), adding log (i.e. per- centage) changes still provides a good proxy. For the commodity price I.e. growth can be explained by a persistence effect index growth, we took the 90th percentile of previous index changes in (θ[Δ lnyct–1]), changes in the explanatory variables Ethiopia. This was fairly consistent with our attempts to construct a similar index for this period. For the real exchange rate, we used IMF X, and a period-specific global shock (Δbt).20 Note data for the Real Effective Exchange Rate (using the fiscal year 2013/14 that the country fixed effect cancels out because it is as reference) as a proxy. EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 33 FIGURE A3.1: Definition of Time Periods Used in the Study 1996 2000 2005 2010 2013 Avg. 1996–2000 Avg. 2001–2005 Avg. 2006–2010 2013 t0 t1 t2 t3 Early 2000s Late 2000s Early 2010s 2000–2013 results as it was performed after estimation (and for rate. A one-time increase in human capital in the Early Ethiopia only). 2000s, for example, will thus be captured as Δln(x) > To facilitate an analysis of the growth accel- 0 in equation (2) and impact the growth rate in the eration period since 2004 with the available same period with parameter Γ. This growth-enhancing data, it is useful to define time periods precisely. effect will be echoed in the Late 2000s via the lagged As illustrated in Figure A3.1, our analyzed period dependent variable effect as θ Γ < Γ (which is captured 1996–2013 consists of 4 data points (t0, t1, t2 and t3), as persistence effect in our model) and eventually fade each capturing the average values of the following out over time. time periods: 1996–2000, 2001–2005, 2006–2010, Our model is well specified and consistent and 2011–2015 (proxied by the 2013 value). In the with economic theory. Table A3.1 summarizes the remainder of the paper, we refer to ‘Early 2000s’ as regression model. Overall, there is no indication that the change between the 1996–2000 and 2001–2005 the model is mis-specified and parameters show the averages, the ‘Late 2000s’ is defined as the perfor- expected signs, except from the institutional variable mance between the 2001–05 and 2006–10 averages, Policy2 which is statistically insignificant.22 Parameter and the ‘Early 2010s’ comparing 2006–2010 averages estimates are either statistically significant or at the and 2013 values. borderline of significance, except for schooling which It is also instructive to recap the implica- is a well-known issue in growth regressions (e.g. tions and dynamics of this empirical neoclassical Pritchett, 2001) and should not lead to neglect of growth model. The level equation (1) implies that education policies. the (log) level of GDP changes with the (log) level of the explanatory variables X. Any change or innova- tion in the (log) level of X will thus have a permanent effect on (log) GDP and the effect is intermediated 22 Institutional quality usually does not vary as much over time, so it is through a temporary (transitory) effect on the growth difficult to identify the according parameter in this context. 34 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE A3.1: Regression Baseline Results (1) (2) (3) (4) VARIABLES log of GDP per capita (in PPP) Persistence 0.781*** 0.784*** 0.726*** 0.746*** (0.0569) (0.0563) (0.0491) (0.0392) ln(exch rate) –0.0640 –0.0622 –0.0553* –0.0172 (0.0404) (0.0392) (0.0332) (0.0355) ln(schooling) 0.0178 0.0445 0.0104 –0.0266 (0.0503) (0.0502) (0.0463) (0.0452) ln(credit/GDP) 0.0743** 0.0542* 0.0432* 0.0238 (0.0311) (0.0304) (0.0221) (0.0245) ln(trade/GDP) 0.0824 0.0609 0.0916*** 0.0968 (0.0502) (0.0490) (0.0350) (0.0584) ln(govt C) –0.262*** –0.259*** –0.215*** –0.127 (0.0442) (0.0423) (0.0359) (0.0810) ln(tele lines) 0.141*** 0.129*** 0.0769*** 0.0816*** (0.0309) (0.0297) (0.0216) (0.0261) ln(inflation) –0.0113 –0.0145 –0.00523 –0.0128 (0.0118) (0.0110) (0.00886) (0.0112) Δln(TOT change) 0.118*** 0.123*** 0.116*** 0.110*** (0.0286) (0.0277) (0.0264) (0.0339) ln(bank crisis) –0.0399 –0.0430 –0.0414 –0.0461* (0.0317) (0.0314) (0.0259) (0.0236) Δln(commodity prices) 10.48*** 11.11*** 7.507*** 6.963 (2.686) (2.546) (2.391) (4.943) ln(institutions) –0.00265 0.00190 –0.00549 (0.0330) (0.0247) (0.0255) Constant 2.502*** 2.829*** 3.203*** 2.469*** (0.708) (0.465) (0.600) (0.453) Observations 464 502 464 464 Number of countries 126 141 126 126 Estimation SysGMM SysGMM SysGMM FE Note: Baseline w/o Polity2 lags 1–3 as baseline instruments as FE No of instruments 153 166 171 AB(1) 0.023 0.024 0.033 AB(2) 0.102 0.045 0.062 Sargan test 0.131 0.017 0.001 Note: Based on Brueckner (2013). Standard errors in parentheses. ***, **, and * indicate statistical significance on the 1, 5, and 10 percent level, respectively. AB(1) and AB(2) is the p-value of the Arellano and Bond test for first and second order autocorrelation, respectively. Sargan test reports p-values. EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 35 Annex 3.2 Model Robustness growth drivers. As a robustness check, we estimate an Ethiopia-specific coefficient—parameter by parameter. Our model passed a series of robustness checks. This stepwise procedure is chosen to keep the instru- The original model of Brueckner (2013) also under- ment set in the GMM framework at a reasonable size. went a series of standard robustness checks. E.g. it was Our results do not show any statistically significant shown that main results are robust to taking 10-year deviation of Ethiopia’s determinants of growth from non-overlapping panel data, balanced panel data, the overall sample of countries included (See Moller time-varying coefficients, or alternative specifications. and Wacker, 2015, for results). This is consistent with Furthermore, unconditional models were estimated the findings of Brueckner (2013) that the underlying variable by variable as this limits the weak-instrument model is largely robust to parameter heterogeneity. We problem in the case where various instruments appear only find a strong and positive country-specific effect strong in isolation but are highly correlated so that of Polity2 but as this variable is trending upwards in they are weak when used together (see Dollar and Ethiopia between 1970 and 2000, it only captures Kraay, 2003). Further to these results, we present three the growth of income over time: after controlling for other specifications in Table A1.1: column (2) reports a country-specific time-trend, this Ethiopia-specific the results without the Polity2 variable that had a variable is no longer statistically significant. Moreover, counter-intuitive sign in the baseline specification. we exclude Ethiopia from the baseline model to avoid The results (which also allow to include a wider set of the possibility that Ethiopia itself is driving the results, countries for which Polity2 is not available) are almost in which case the good performance of the model to identical to the baseline model in column (1). In col- explain the country’s growth would be tautological. umn (3) we expand the instrument set for the explana- Although the broad country coverage of our sample tory variables to include lagged differences for lags 1, makes this unlikely, the results of Warner (2014) war- 2, and 3 (as opposed to including only the first lag). rant some caution. The results reported in column (2) of While not a fully sophisticated check for instrument Annex Table A3.1 however, confirm that the exclusion robustness, this should still convince the reader of the of Ethiopia from the sample has almost no effect. robustness of the results to the used instrument set and Some effects differ in lower income countries that we do not use too few instruments. Results are but these differences are not significant. As one again similar to the baseline, though the infrastructure may argue that the overall model includes several parameter is (statistically significantly) smaller but still high-income countries and is thus not appropriate significant and large. The results of the Sargan test also for Ethiopia, we re-estimate the model using only indicate that the instrument set is not as appropriate countries that were below the median or mean of as in the baseline model. In column (4) we also report GDP p.c. in 1995. Detailed results are presented in fixed-effect results for comparison, again with similar Moller and Wacker (2015) and they provide some results to the baseline but a somewhat smaller (but weak evidence that our benchmark model somewhat positive and significant) infrastructure parameter. The overestimates the positive impact of infrastructure fact that the lagged dependent variable parameter of (as proxied by telephone lines). We also estimated this model is somewhat smaller than in the baseline the model for landlocked countries only but due to also confirms that our baseline model is well-specified the small sample size (25 countries with a total of 85 because of the downward bias of this parameter in observations), results were mostly insignificant and fixed effect estimation (Nickell, 1981). thus not very informative. We also examine whether some variables had a Overall, these results support the baseline significantly different effect on growth in Ethiopia, model but suggest that the infrastructure results but find no substantial Ethiopia-specific effects for may be on the higher side. As demonstrated, our 36 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT model is robust to a wide range of alternative speci- Growth Impact of Infrastructure FIGURE A3.2:  fications. Where results differ in magnitude, this dif- 7% ference is rarely statistically significant and thus often 6% reflects random sample effects. The only case where 5% infrastructure is statistically insignificant arises in a 4% considerably reduced sample and the quantitative 3% result is still in line with parameter estimates from 2% other robustness checks. In total, these robustness 1% checks suggest, however, that our estimate for infra- 0% structure in the baseline model may be on the higher –1% side. However, we find no evidence that any of these 2000–2013 Early 2000s Late 2000s Early 2010s alternative models would be more appropriate to cap- Telephone Mobile Roads ture growth in Ethiopia over the recent past. Moreover, Source: Author’s calculation. even if we assume one of the lower parameter values (which themselves might be on the lower side), the associated contribution of infrastructure to Ethiopia’s in 2000–13 is nearly identical, despite minor differ- growth acceleration would still be substantial and in ences in magnitudes and with respect to the timing the range of one percentage point per year. across sub-periods.24 Differences only arise in the last The results are also robust to the choice of period (the Early 2010s), which is then also reflected alternative infrastructure variables, apart from in the 2000–2013 predictions. Over this period, diverging trends in Early 2010. Given the critical fixed telephone lines were declining but this does not explanatory importance of infrastructure and the adequately capture infrastructure developments as difficulty of appropriate measurement and interpre- some substitution towards cell phones took place and tation, we also tested whether results are sensitive to other infrastructure—such as roads—was increasing alternative variable specification. The model results considerably as well (see Figure 3.2.1). reported in Table A3.1 use fixed telephone line cover- When correcting for alternative measurements, age. We tested two alternative infrastructure variables, we are able to predict growth in Early 2010s with namely mobile phone subscribers and road coverage greater accuracy. To assess the impact of these alter- and found nearly identical results up to 2010.23 We native infrastructure measures on the overall pre- note that all three types of infrastructure are provided dicted growth rate, we subtract the effect of phone exclusively by the public sector in the case of Ethiopia lines from our baseline predictions and instead add giving us greater confidence in drawing inferences the average effect of roads and cell phones, with the between public infrastructure investment and growth. results depicted in the light-red bar of Figure A3.2 Both alternative infrastructure variables have a strong (‘predicted alternative infrastructure’).25 As one can positive impact on growth in our model, although the effect of roads is only at the borderline of statistical significance (z statistic 1.55). The results are depicted 23 As our original data set only covered roads data until 2007 in Ethiopia, in Figure A3.3 as annualized contributions to the we replaced this data using national sources. overall predicted growth rate (in percentage points). 24 For a firm-level analysis showing the positive effect of road infrastruc- ture on firm location choice and startup size in Ethiopia, see Shiferaw The bars for telephone lines simply reproduces the et al. (2013). results depicted in Table A3.1 to facilitate comparison. 25 Note that this calculation is possible because variables are measured in logs but it is only correct if one assumes that the substitution of the We can further see that the effect of fixed telephone infrastructure variables in the regression model would have no effect on coverage and roads for explaining growth in Ethiopia the estimated parameters of other variables. EXPLAINING GROWTH: STRUCTURAL, EXTERNAL AND STABILIZATION FACTORS 37 see, this corrects for this Early-2010s specific effect. low levels) in the Early 2000s, stagnating in 2007. Especially if one assumes an average of our baseline It is unlikely, that all these improvements are cap- prediction and the one with alternative infrastruc- tured in the GDP growth rate of the Early 2000s but ture, our predictions come very close to the actually they would also be reflected in the Late 2000s (via observed growth rate. persistence effects). During the Late 2000s, mobile Infrastructure improvements in the early phones started to substitute for landlines, which 2010s paid off during later years as well. As one also explains why the growth effect for landlines is can see from Figure 3.2.1, our baseline variable of considerably smaller than for mobile phones in the telephone lines saw a considerable pickup (from very Early 2010s. 39 GROWTH AND STRUCTURAL CHANGE26 4 replicated in recent decades by East Asia. Farmers Ethiopia has experienced growth from structural change as labor shifted from agriculture into services and construction. moved into higher-productivity manufacturing or About a quarter of Ethiopia’s recent economic growth can agro-processing; economies diversified and began to be explained by a sectoral shift of just three percent of export more sophisticated goods. The share of the its worker. However, structural change in Ethiopia did not follow the desired path of expanding the share of its small labor force employed in manufacturing peaked at manufacturing sector and this remains a major challenge. 25 to 45 percent in countries like the UK, U.S. and International and regional experience suggest that all Sweden before these countries de-industrialized. Even economic sectors are of importance at different stages of development. In Ethiopia, agriculture matters because of Korea, where the manufacturing employment share poverty and size, manufacturing because it creates urban was in the single-digit range in the 1950s, peaked at jobs, and, services because it helps manufacturing become nearly 30 percent before decreasing in the 1980s. more competitive and absorb the rapidly growing labor force. Premature deindustrialization makes it harder for today’s developing countries to follow this trod- den path. Low- and middle-income countries are 4.1. Introduction beginning to deindustrialize at lower shares of industry in output and employment than their predecessors did. Structural change is vital for sustaining economic This stylized fact has become known as ‘premature dein- growth. In simple terms, structural change can be dustrialization’ (Rodrik, 2015). India has been pointed defined as the reallocation of labor from low-pro- out as the paradigmatic case of a country in which the ductivity sectors to more dynamic (higher-produc- size of its manufacturing sector declines relatively early tivity) economic activities.27 For most developing on, after employment in the sector reached 13 percent countries, this would usually require shifting labor of the total workforce. Other examples include Brazil, from subsistence agriculture to commercial agricul- where manufacturing employment peaked at 16 per- ture, manufacturing, and modern services. ‘The speed cent, and Mexico where it peaked at 20 percent. This with which this structural change takes place is the compares with early industrializers that managed to key factor that differentiates successful countries from place at least 30 percent of its labor force in manufac- unsuccessful ones’ (McMillan and Rodrik, 2011). turing before the sector started declining. The traditional path of economic development Services are now playing the role manufactur- involved processes of export-oriented industrial- ing did in the past and there is an ongoing debate ization and deindustrialization. One of the oldest over the implications of premature deindustrial- ideas in development economics is that the route ization for countries’ development. Take Africa as towards development involves structural change, following first a process of industrialization where 26 This chapter draws upon background papers prepared for this report by Ghani and O’Connell (2014), Hollweg, Rojas, and Varela (2015), workers leave the agricultural sector for the higher- and, Martins (2015). productivity manufacturing sector and second one of 27 Structural change can also refer to the changing composition of output. However, since shifts in production tend to precede shifts in employ- deindustrialization where workers move into services. ment, this transformative process is arguably only under way once labor This path was first taken by Western countries and starts to relocate. 40 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT an example. Nearly two decades of strong growth are rural areas. As a result, the bulk of excess labor in low- transforming the structure of Africa’s economies, but income countries is absorbed in non-tradable services not as expected. Sectoral composition of output has operating at low levels of productivity. Second, because shifted in favor of services, with this sector’s growth of the non-tradability of these sectors, partial productiv- outpacing that of agriculture. Commensurate labor ity gains are self-limiting as they cannot expand without shifts are taking place very slowly (World Bank, 2014). inducing a negative terms of trade shock against them- On the one hand, optimists argue that services selves. Put differently, since demand in these sectors is may have the potential to become the new growth constrained to within national borders, productivity escalator for developing countries. Ghani and improvements can only result in a price reduction. In O’Connell (2014) show that countries furthest away manufacturing, instead, exports provide an opportunity from the frontier of productivity show the fastest pro- to avoid that outcome, since exporters face, potentially, ductivity growth in services, converging independently almost infinite demand for their products. of their structural characteristics (known in develop- Moreover, the manufacturing sector has a ment economics as ‘unconditional convergence’).28 The strong potential to lead the growth process. Making authors do not argue that service is superior to manu- this case, Rodrik (2013a) develops an analytical frame- facturing, or the other way round. Rather, they make work identifying four distinct channels for growth the point that the late comers to development now have consistent with empirically observed stylized facts. many more levers to pull. Arguably, services can also be The first, the ‘fundamentals’ channel is a process of dynamic and contribute to growth and jobs. These argu- convergence that accompanies the accumulation of ments build on previous work by Ghani (2010) showing fundamental capabilities arising from broad-based how developing countries can take advantage of mod- investment in human capital and institutional arrange- ern services exports where they actually have a higher ments. The second channel is the forces of uncondi- revealed comparative advantage compared to their tional convergence operating within manufacturing, own goods exports and compared also to high-income which refers to the empirical fact that poorer coun- countries. Although India is the most famous case of tries achieve higher labor productivity growth rates in services-based growth, there are a dozen other examples manufacturing than richer countries (Rodrik, 2013b). including Bangladesh, Mozambique and Rwanda. Those two dynamic effects are potentially augmented On the other hand, premature deindustrializa- by two effects of reallocating labor from traditional tion could instead imply a movement of labor to activities to higher-productivity manufacturing (third prospectively less dynamic sectors, thus diminishing channel) and modern services (fourth channel).29 developing countries’ growth potential going for- Africa’s structural change experience has been ward. Rodrik (2014) argues that services are different one of realizing static gains and dynamic losses. from manufacturing in two important ways, which make the sector unlikely to play the role of a growth 28 Rodrik has raised questions regarding the robustness of this ‘uncon- escalator. First, with the rise of technology, many seg- ditional convergences in services’ result with respect to the period of analysis (Project Syndicate, October 13, 2014). By comparison the Rodrik ments of services are themselves tradable and becom- (2013b) result regarding unconditional convergence in manufacturing ing important in global commerce. While these are is statistically more robust. 29 While intuitive appealing and insightful, this framework falls somewhat high productivity, high wage, and high skill-intensive short in comprehensively explaining Ethiopia’s growth acceleration since sectors, they require highly trained workers, which are 2004. The fundamentals channel was only partially at play in Ethiopia as in- stitutional improvements were modest as were human capital accumulation unlikely to be those exiting agriculture in Ethiopia or effects. Static efficiency gains emerged as some agricultural labor shifted other developing countries. As manufacturing world- into construction and services. Although unconditional convergence in manufacturing is an empirical regularity, Ethiopia did not experience a wide has become more capital and skill-intensive, it has very high rate of labor productivity growth in this sector. Nor were there diminished its potential to absorb abundant labor from any much sign of productivity-enhancing labor shifts into modern services. GROWTH AND STRUCTURAL CHANGE 41 An analysis of Africa’s experience by Timmer et al. deindustrialization’ and ‘the potential of services as a (2014) show that the expansion of manufacturing growth escalator’, questions have emerged both with activities during the early post-independence period respect to the ability to Ethiopia successfully pursue led to a growth enhancing reallocation of resources. manufacturing-led growth and benefit from Chinese This process of structural change stalled in the mid- exports of low wage jobs. Shahid Yusuf (2014) in 1970s and 80s. When growth rebounded in the 1990s, a background paper to this report, argues that the workers mainly relocated to market services which had smallness of the sector make such prospects highly above-average productivity levels, but productivity challenging and that China will fight hard not to lose growth was low and increasingly falling behind the even unskilled manufacturing jobs. More recently, The world frontier. This pattern of static gains but dynamic Economist (2015) has argued that Asia’s dominance losses of reallocation since 1990 is found for many on manufacturing will endure, making development African countries. It is comparable to patterns observed harder for others, including Africa, and that ‘just wait- in Latin America, but different from those in Asia. ing for higher Chinese wages to push jobs their way is These considerations are highly relevant for a a recipe for failure’. When China loses market share country such as Ethiopia, which is experiencing for industrial exports, it does so to countries such as structural change from agriculture directly to ser- Vietnam, Indonesia and Cambodia rather than to vices. As in the rest of Africa, output shifts have been low-wage countries in Africa. pronounced and the employment shifts modest. Because With this backdrop, this chapter seeks to it is a poor country, Ethiopia’s employment structure address the following questions: What was the role (A: 78%; I; 7%; S:15%) resembles more that of Africa’s of structural change during Ethiopia’s growth accel- poor people (A: 78%; I; 5%; S:16%) than the regional eration? How does Ethiopia’s experience of structural average (A: 37%; I: 24%; S: 40%) (World Bank, 2014). change compare with other countries? Which sectors In line with the conventional wisdom, the offer Africa and Ethiopia the best hope of growth and Government of Ethiopia has been pursuing a strat- transformation? egy of industrialization in recent years. This strategy The remainder of the chapter is structure as is consistent with the recommendations of Rodrik, as follows: Section 4.2 presents an analysis of structural well as those of Dinh et al (2012) and Lin (2011). Hinh change in Ethiopia in the 1999–2013 period, includ- Dinh and co-authors argue that Ethiopia has abundant ing changes in output, employment and labor produc- low-cost labor, which gives it a comparative advantage tivity. Section 4.3 documents the rise of the services in less-skilled, labor-intensive sectors such as light sector. Section 4.4 summarizes the international pat- manufacturing. This is held back by constraints such terns of structural change and places Ethiopia’s experi- as shortage of industrial land, poor trade logistics and ence in this context. Section 4.5 outlines the challenges limited access to finance. To unleash the comparative faced by Africa in terms of the sectoral emphasis of advantage, therefore, policy should focus on addressing their growth paths and brings these considerations these sector-specific constraints. Justin Lin (2011), in into a discussion of Ethiopia’s options. turn, argues that rising wages in China will ultimately imply the export of millions of unskilled manufacturing 4.2 Structural Change in Ethiopia30 jobs to low-wage countries such as Ethiopia, as China moves up the ladder of development and specializes in Ethiopia’s output tripled in real terms over the more skills-intensive manufacturing. past 14 years—driven primarily by services and But the results of this industrialization strat- egy are yet to materialize, both for Africa and 30 The period of analysis (1999–2013) is determined by the use of three Ethiopia. In addition to arguments of ‘premature consecutive Labor Force Surveys in 1999, 2005, and 2013. 42 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT agriculture. Real gross value added increased from role of the agriculture, manufacturing and commerce 184 billion birr in 1999 to 571 billion birr in 2013, sectors. as illustrated in Figure 4.1.1. Services contributed to Labor productivity growth in commerce was half of output growth in this period, while agriculture twice as high as in manufacturing and construc- contributed by one third (Figure 4.1.3). Industry, tion. In general, labor productivity (gross value added which apart from manufacturing also includes con- per worker) increased by 4.8 percent per year in struction and utilities, contributed about 15 percent 1999–2013, while accelerating over this period from to growth. The contribution of manufacturing was 2.7 percent in 1999–2005 to 6.5 percent in 2005–13 just 4 percentage points. (annual growth rates). Figure 4.1.2 shows sectoral dif- The structure of output shifted from agricul- ferences of labor productivity growth. Most sectors ture towards services while the corresponding exhibited strong positive productivity growth, except employment shift was modest. The output share mining, finance and utilities all of which have small of agriculture declined from 57 percent in 1999 to employment shares. 42 percent in 2013 (Figure 4.1.5). Services output, Economic growth can be understood and meanwhile, increased from 33 to 45 percent, while analyzed in terms of rising labor productivity. industry increased from 10 to 13 percent. Agriculture Figure 4.2.4 presents a decomposition of output (value continued to dominate employment, as illustrated in added per person), which can increase for various Figure 4.1.6, though its employment share declined reasons, including: rising labor productivity within from 80.2 to 77.3 percent between 2005 and 2013. each sector (if each worker produces more), structural Workers moved mainly into services (1.8 percentage change (if workers move from low- to higher produc- points) and construction (0.7 percentage points). tivity activities), demography (if the relative share of Total employment increased by 15 million the working age population rises) and employment people since 1999 reaching 40 million in 2013 (if a higher share of the working age population is (Figure 4.1.2). Agriculture accounted for three employed). quarters of employment growth (11 million people). Ethiopia has experienced strong labor pro- Manufacturing employment increased from 1.1 mil- ductivity growth. Between 1999 and 2013, real lion in 1999 to 1.8 million in 2013, but remains value added per person exhibited an average annual relatively small (about 5 percent of employment). growth rate of 4.5 percent. Aggregate labor produc- The services sector accounted for 16 percent of total tivity growth accounted for nearly 90 percent of employment growth (2.4 million people), as shown this increase, with 72 percent due to within-sector in Figure 4.1.4. improvements and 17 percent due to structural Labor productivity levels are highest in sectors change. Changes in the employment rate and the such as finance, utilities, mining, and transport, demographic structure contributed with 5 percent and is lowest in agriculture and manufacturing. and 6 percent, respectively. If labor had not relocated Output per worker per year ranges from 126,700 across sectors, output per capita growth would have birr (2010/11 prices) in finance to just 6,000 birr been nearly one fifth lower. in agriculture and 9,200 birr in manufacturing Structural and demographic change has accel- (Figure 4.2.1). It is useful, however, to put these fig- erated since 2005. The contribution of structural ures into perspective, since their ultimate impact on change was increasing over time—from 9 percent in the economy greatly depends on the relative employ- 1999–2005 to 25 percent in 2005–2013. The negative ment weight of each sector. Figure 4.2.3 combines impact of the employment rate in 2005–2013 might information on labor productivity and sectoral shares be partly explained by young people staying longer of employment in 2013. This highlights the important in education—since the working-age population GROWTH AND STRUCTURAL CHANGE 43 FIGURE 4.1: Ethiopia: Output and Employment by Sector, 1999–2013 1. Value Added (billion birr, 2010/11 prices) 2. Employment (million people) 600 45 40 500 6.1 35 3.0 400 30 4.1 25 2.1 300 3.6 20 1.4 200 15 30.8 10 25.2 100 19.9 5 0 0 1999 2005 2013 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Services Industry Agriculture 2013 Services Industry Agriculture 3. Value Added Growth Contribution, 99–13 (%) 4. Employment Growth Contribution, 99–13 (%) Agriculture Services Agriculture 35.2% 16% 73% Services 50.2% Industry 11% Industry 14.5% 5. Value Added Shares (%) 6. Employment Shares (%) 100% 100% 90% 90% 14.7 13.2 15.2 5.5 6.6 7.5 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 79.8 80.2 77.3 30% 30% 20% 20% 10% 10% 0% 0% 1999 2005 2013 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Services Industry Agriculture Services Industry Agriculture Source: National Accounts Directorate, MoFED. Central Statistical Agency (CSA): LFS 1999, LFS 2005, LFS 2011. 44 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 4.2: Ethiopia: Labor Productivity, 1999–2013 1. Labor Productivity Level, 2013 (1,000 birr) 2. Labor Productivity Growth, 1999–2013 (%) Other services 35.1 Other services 3.8 Public services 33.1 Public services 2.6 Finance 126.7 Finance –2.0 Transport 64.2 Transport 2.6 Commerce 22.2 Commerce 9.1 Construction 29.6 Construction 4.3 Utilities 78.4 Utilities –0.7 Manufacturing 9.2 Manufacturing 4.4 Mining 60.3 Mining –6.6 Agriculture 6.0 Agriculture 3.0 0 20 40 60 80 100 120 140 –10 –5 0 5 10 3. Labor Productivity and Labor Shares, 2013 4. Decomposing GVA Growth per Capita (99–13) 110 140 120 90 13.0 Productivity ('000 birr) 100 6.2 5.1 42.5 24.5 70 80 17.2 60 8.8 50 40 71.6 76.2 56.8 30 20 0 –8.0 10 –13.7 –20 –10 –40 Employment Share (%) 1999–2013 1999–05 2005–13 Agriculture Mining Manufacturing Utilities Demographic Effect Employment Rate Construction Commerce Transport Finance Structural change Within sector productivity Public services Other services 5. GVA per capita growth contribution (%, 99–13) 6. Productivity and Employment Changes, 2005–13 Log of sectoral productivity / total productivity 1.0 Structural change 17.2 y = 0.2129x + 0.4159 R² = 0.5084 Finance Other services 6.7 Utilities Transport Public services 3.9 0.5 Commerce Mining Construction Finance –0.9 Other services Transport Public services 1.9 Commerce 28.8 Construction 3.6 0.0 Manufacturing Utilities –0.1 Agriculture Manufacturing 3.5 Mining –2.4 –0.5 –3.5 –3.0 –2.5 –2.0 –1.5 –1.0 –0.5 0 0.5 1.0 1.5 Agriculture 26.6 –10 0 10 20 30 40 Change in the employment share (percentage points) Source: Martins (2015) using data from National Accounts Directorate, MoFED. Central Statistical Agency (CSA): LFS 1999, LFS 2005, LFS 2011. GROWTH AND STRUCTURAL CHANGE 45 includes those aged 10 and above. This might actu- a rising dependency ratio. In recent times, however, ally constitute a positive development, to the extent demographic change has been playing a more positive that young people are acquiring relevant skills that role. As a youth bulge enters the labour force, it lowers can boost growth and structural change in the future. the (child) dependency ratio and potentially delivers Changes in the demographic structure had a negative a demographic dividend (See Box 4.1). Commerce effect on growth in 1999–2005 as a consequence of and agriculture provided the strongest contributions BOX 4.1: The Demographic Dividend Ethiopia’s growth acceleration was supported by positive demographic effects. The economic take-off coincided with a marked increase in the share of the working-age population giving a positive boost to labor supply. Up to thirteen percent of per capita growth in 2005–13 can be attributed to this demographic effect. A continued rise in the working age population will support potential economic growth in the coming decades. The demographic dividend describes the interplay between changes in a population’s age structure due to the demographic transition and rapid economic growth. Declines in child mortality, followed by declines in fertility, produce a ‘bulge’ generation and a period when a country has a large number of working age people and fewer dependents. Having a large number of gainfully employed workers per capita gives a boost to the economy (World Bank, 2014). The economic benefits of a demographic dividend, however, are not automatic. For the youthful workforce to add value they must be equipped with education and skills and the business environment must be such as to generate jobs sufficient to productively absorb the available labor. Ethiopia experienced rapid demographic change over the past three decades. Child and infant mortality started declining in the mid-1980s while the total fertility rate fell rapidly a decade later in the mid-1990s, partly in response to lower death rates. Population growth rates declined and life expectancy increased. In 1980, Ethiopia was doing worse on these indicators than the average Sub-Saharan African country. By 2010 this situation was reversed. Declining mortality followed by declining fertility produced a shift in the age-structure of the population as more people were able to work. According to UN data, the share of the working-age population started increasing in 2005—at the same time when the economy took off. Having been constant since 1985 at about 50–51 percent, the working age population share increased to 52.3 percent in 2010 to 55.1 percent in 2015 with a projected peak of 67.5 percent in 2055. A similar pattern is observed for the dependency ratio which remained relatively constant in 1990–2005 at around 97–98 percent, but then declined to 91.3 in 2010 and 81.5 in 2015. A rising share of the working age population accounts for thirteen percent of Ethiopia’s per capita growth in 2005–13. This result is illustrated in the decomposition of value added per person presented in Figure 4.2.4. Changes in the demographic structure had a negative effect on growth in 1999–2005 (by 8.0 percent) as a consequence of a rising dependency ratio. However, during the 2005–13 economic boom period, a rising working age population can account for 13.0 percent of per capita growth. By comparison, demography accounted for 20 percent of per capita growth in the Republic of Korea in 1970–90 (Martins, 2014; 2015). Ethiopia’s demographic transition is taking place faster than in the rest of Africa. Africa’s share of the working age population started rising in 1990, but will not peak until ninety years later in 2080. The same process will take only 50 years in Ethiopia (2005–55). This is the result of a more rapid decline in mortality and fertility in Ethiopia compared to the rest of the region. Improvements in female education had a particularly important impact on rapidly declining fertility in Ethiopia. The 1994 education reform removed school fees, instituted school lunches in rural areas, increased the education budget, and allowed classes to be taught in the local language rather than Amharic (World Bank, 2014). This lead to a substantial increase in female education by 0.8 years, on average. This, in turn, reduced the probability of teenage birth and teenage marriage (by 7 and 6 percentage points, respectively, per year of additional schooling). For Ethiopia to reap the benefits of the demographic dividend it must put in place policies to further accelerate the fertility decline and for the economy to absorb a rapidly increasing labor force. Speeding up the fertility decline require policies to reduce child mortality and improve child health, support female education and empowerment, addressing social norms on fertility, reducing child marriage and expanding comprehensive family planning programs. Reaping the ensuing economic benefits, in turn, require measures on the labor supply and demand side. Workers must be endowed with marketable skills to be attractive to prospective employers and employment of women outside the home must be encouraged. Demand for labor can be boosted by attracting foreign direct investment through an improved business environment (Galor and Weil, 1996; World Bank, 2014). 46 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT to within-sector productivity (Figure 4.2.5). This is low productivity agriculture and into services (the partly explained by their large employment shares. share of services employment increased by 1.5 Ethiopia experienced growth-enhancing struc- percent while the share of agriculture declined by tural change, especially since 2005. Figure 4.2.6 about 3 percent). plots changes in employment shares and the relative  This structural transformation produced effi- productivity of sectors—the latter is measured by the ciency gains because labor productivity in services log of the ratio between sectoral productivity and total is higher than in agriculture (and manufacturing). productivity in 2013. In a classic pattern of structural Commerce, for instance, is five time more produc- change, we would expect to find agriculture in the tive than agriculture. bottom-left quadrant—with relatively low labour pro-  Services sector exports are as large as merchan- ductivity and a declining labour share—and the more dise exports, each accounting for roughly 7 per- dynamic economic sectors in the top-right quadrant— cent of GDP (World Bank, 2014). with relatively high labour productivity and a rising  The contribution of services to poverty reduc- labour share. The figures provides some evidence of tion has been modest. Agriculture contributed growth-enhancing structural change in Ethiopia, with the most to poverty reduction (World Bank, the labour share declining in the sector with the low- 2014). est labour productivity (i.e. agriculture) and increas- ing in more dynamic sectors—albeit slowly. On the Drilling down further into the Ethiopian ser- other hand, commerce and manufacturing observed vices sector we divide it into five sub-sectors, as a decline in their labour shares, while manufacturing follows: has relatively low productivity levels.  Commerce: Wholesale & retail trade. Hotels & 4.3 The Rising Services Sector restaurants.  Transport: Transport; storage & communication. The services sector was one of the driving forces  Finance: Financial intermediation. behind Ethiopia’s growth acceleration. In this sec-  Public services: Public administration & social tion, we analyze the performance of the sector in more services. Education. Health & social work. detail, starting by highlighting the stylized facts:  Other services: Real estate, renting & business. Other community social and personal services.  The services sector is the largest in terms of eco- Activities of private households. Extraterritorial nomic output accounting for 45 percent of value organizations and bodies. added in 2013/14.  The sector accounts for about half of economic Commerce, ‘other services’ and the public sec- growth generated during the 2004–14 growth tor are the most important services sub-sectors acceleration period. in terms of output and employment in Ethiopia.  Services are the second biggest employer in Put together, they account for 85 percent of sector Ethiopia, accounting for 15 percent of total value added and 92 percent of jobs. Specifically, each employment (6.1 million people). account for roughly a half, a quarter and a fifth of value  It helped absorb a rapidly growing labor force added and jobs in the services sector (Figures 4.3.1 and by creating 2.4 million new jobs between 1999 and 2013 (16 percent of all new jobs).  The services sector supported structural trans- 31 Note the similarity in value added and employment shares of these formation in the form of labor shifts away from three sectors, which is not the case for the economy as a whole. GROWTH AND STRUCTURAL CHANGE 47 4.3.2).31 The remaining services sectors are transport Services have made the largest contribution to (10 percent of services output and 6 percent of ser- growth in Ethiopia, just like in most other coun- vices jobs) and finance (5 percent of services output tries. Figure 4.3.2 compares the contribution of differ- and 2 percent of services jobs). Output shares hardly ent sectors to growth in Ethiopia with other countries. changed over time, though the employment share Indeed, services have made the largest contribution to of ‘other services’ and public services have increased growth in both developing and developed economies. while commerce declined (Figures 4.3.3 and 4.3.4). Once again, China is an exception to this trend. Labor productivity levels are highest in finance Labor productivity growth in Ethiopian ser- while the commerce sector saw the strongest vices was relatively high. Figure 4.3.4 plots growth increase in labor productivity growth. Figures in service labor productivity for Ethiopia and other 4.3.5 illustrates services sector labor productivity lev- countries on the vertical axis (from late 1990s to late els. Output per worker in finance is twice as high as 2000s), and initial service labor productivity on the transport and 4–5 times larger that of the remaining horizontal axis (early 1990s). The fitted line is a down- services sectors. Commerce labor productivity growth ward sloping line implying that low income countries was 3–4 times stronger than other services sectors, like Ethiopia that started with a lower level of labor except finance which experienced a decline. productivity in services, and were further away from The Ethiopian services story is predominantly the global labor productivity frontier, have experienced one of a rise of traditional rather than modern a much faster catch up and growth in service labor activities. Figure 4.3.6 illustrates trends in value productivity. This is good news for low income coun- added between 1998 and 2011 dividing the services tries in Africa as they have more room to catch-up. sector into two components. Modern activities include As discussed in more detail in Ghani and O’Connell finance as well as communications. The remaining sec- (2014) this is suggestive evidence of a potential phe- tors (commerce, transport, public services, and other nomenon of unconditional convergence in services. services are classified as traditional services, which Note that Ethiopia is above the trend line, implying typically require more face-to-face interaction.32 As higher productivity growth than expected at its level shown in the graph, traditional services rose from 31 of income. to 42 percent of value added in 1998–2011, while Conversely, Ethiopian manufacturing labor modern services increased their share from 2 to 3 productivity growth was relatively low. Figure 4.3.3 percent mainly on account of finance. plots growth in manufacturing labor productivity on the vertical axis and initial labor productivity on Ethiopia’s Experience in the 4.4  the horizontal axis. The fitted line is also downward International Context sloping implying that late comers to development that started with a lower level of labor productivity in Services have grown much faster than other sectors manufacturing have also experienced a faster growth in in Ethiopia, just like in other low income countries. productivity. This graph is comparable to the one used Figure 4.3.1 compares economic growth rates by sector by Rodrik (2013b) to demonstrate unconditional con- in Ethiopia data from other countries (both developed vergence in manufacturing. Unfortunately, Ethiopia is and developing) during the last two decades. It shows below the trend line, implying a much slower progress that both services and industry have experienced faster in the manufacturing sector compared to the East growth rates than agriculture. Service has experienced Asian Tigers which are above the line. the fastest growth rate in Ethiopia, as well as for other low income countries. Services grew fast than other 32 There is no internally agreed definition of modern versus traditional sectors in all country groups, except China. services (Goswani and Saez, 2014). 48 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 4.3: The Ethiopia Services Sector 1. Value Added (billion birr, 2010/11 prices) 2. Employment (million people) 300 7 250 6 64 1.5 5 200 46 4 1.2 0.8 150 14 0.6 0.1 26 3 0.6 0.7 0.4 0.0 0.1 0.0 0.1 100 31 2 26 2.8 15 5 110 50 15 3 11 1 2.3 2.4 6 27 39 0 0 1999 2005 2013 1999 2005 2013 Other services Public services Finance Other services Public services Finance Transport Commerce Transport Commerce 3. Value Added Shares (%) 4. Employment Shares (%) 100% 100% 22.7 27.2 24.6 16 19.8 25 90% 90% 80% 80% 16 70% 70% 17.6 17.6 0.5 3.4 20 23.2 0.9 3.5 60% 23 60% 5.2 2.2 50% 4.1 50% 6.2 4.8 10 9 40% 10 40% 30% 30% 64.1 58.1 47 20% 41 35 42.6 20% 10% 10% 0% 0% 1999 2005 2013 1999 2005 2013 Other services Public services Finance Other services Public services Finance Transport Commerce Transport Commerce 5. Labor Productivity Level, 2013 (1,000 birr) 6. Value added: modern and traditional services 70 Other services 35.1 60 50 Public services 33.1 40 30 Finance 126.7 20 10 Transport 64.2 0 1997/98 1998/99 1999/00 2000/01 2001/02 2002/03 2003/04 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 Commerce 22.2 Agriculture Services – Traditional 0 50 100 150 Industry Services – Modern Source: National Accounts Directorate, MoFED. Central Statistical Agency (CSA): LFS 1999, LFS 2005, LFS 2011. GROWTH AND STRUCTURAL CHANGE 49 As countries grow richer, their services output fact, the predicted manufacturing output share given and job shares increase at the expense of agricul- Ethiopia’s income per capita is about 10 percent com- ture. Figure 4.4 plots output and employment shares pared to the actual one of 5 percent. The job share for for the three major sectors (agriculture, manufacturing manufacturing did exhibit the expected rising trend and services)33 against income per capita across a sam- and level. However, when combined with the output ple of about 100 countries. The fitted regression lines share, it is clear that some manufacturing labor pro- are positive for services and negative for agriculture, ductivity growth was held back. implying increasing and declining shares, respectively, as income rises. The inverted U shape of manufactur- A Regional Perspective and 4.5  ing suggest a rising contribution to output and jobs Potential Implications for Ethiopia at early stages of development, but declining shares at later stages of development where manufacturing Rodrik (2014) argues that Africa has four options gives way for services. to generate sustained, rapid growth in the future. The global pattern has changed over time as The first one is to revive manufacturing and put indus- services are now creating more jobs and manufac- trialization back on track, so as to replicate as much turing less (premature deindustrialization). Figures as possible the traditional route to convergence. The 4.3.5 compares the relationship between the share of second is to generate agriculture-led growth, based a country’s total employment in the industrial sector on diversification into non-traditional agricultural against its level of income. This relationship is shown products. The third is to generate rapid growth in for three different points in time, 1988 (blue), 2000 productivity in services, where most of the people (green), and 2010 (red). It shows that the job curves will end up in any case. The fourth is growth based in industry have shifted downwards over time. This on natural resources, in which many African countries means that industrial sectors are creating fewer and are amply endowed. Given that the natural resources fewer jobs over time. Put another way, the point at sector is still in its infancy in Ethiopia (World Bank, which de-industrialization begins is happening earlier 2014), we dwell primarily on the first three strategies. in the development process. Improvements in tech- Africa’s nascent industrialization process is nology have made manufacturing much more capital- held back by its poor business climate and may intensive. This is happening even at the low-quality not overcome the global challenge associated with end of the spectrum (Rodrik, 2012). So the capacity ‘premature de-industrialization’. Chinese greenfield of manufacturing sector to absorb labor is shrinking investments in manufacturing in countries such as over time. Figure 4.3.6 for the services sector shows Ethiopia, Nigeria, Ghana, Tanzania offers hope that the opposite trend as service job curves have shifted Africa is well poised to taking advantage of rising upwards over time. This means that services are creat- costs in Asia, but the aggregate data do not yet show ing more jobs and also at earlier stages of development. something like that is happening. The consensus Ethiopia’s structural change pattern is con- view on what holds African manufacturing back is sistent with these international trends, except for a ‘poor business climate’, including costs of power, manufacturing output which remained a small share transport, corruption, regulations, security, contract of the economy. Reverting now to Figure 4.4, note enforcement, and policy uncertainty. According to that it also compares Ethiopia’s 1992–2012 trends with Rodrik, an undervalued real exchange rate may be those across the world in 2012. While the services share of output and employment increased at the expense of agriculture, the manufacturing output share did 33 Note that this presentation omits industry sectors different from not rise as expected at this stage of development. In manufacturing, including construction and utilities. 50 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 4.4: Labor Productivity, Growth and Employment by Sector 1. GDP Growth Rate by Sector, 1990–2012 2. Growth rate contribution by Sector, 1990–2012 14% 6% 12% 5% 10% 4% 8% 3% 6% 2% 4% 2% 1% 0% 0% LIC LMIC OECD SSA SSA ETH IND CHN LIC LMIC OECD SSA SSA ETH IND CHN (non-oil) (oil) (non-oil) (oil) Agriculture Industry Services Agriculture Industry Services 3. Manufacturing Sector Productivity, 1990–2010 4. Services Sector Productivity, 1990–2010 2 1.4 Change in ln(manufacturing VA per worker), China change in ln(services VA per worker), early China 1.2 1.5 India 1 early 1990s to late 2000s 1990s to late 2000s 1 0.8 Ethiopia Thailand Malaysia India 0.6 0.5 Malaysia 0.4 Ethiopia Tanzania Brazil 0 0.2 Brazil 0 Thailand –0.5 –0.2 Tanzania –1 –0.4 5 7 9 11 6 8 10 12 ln (manufacturing VA per worker), early 1990s ln (services VA per worker), early 1990s 5. Industry Employment Share by Income Level 6. Services employment share by income level Share of employment in industry vs. GDP per capita Share of employment in services vs. GDP per capita late 80s, late 90s, late 00s late 80s, late 90s, late 00s 50 100 40 80 30 60 20 40 10 20 5 7 9 11 7 8 9 10 11 ln (GDP per capita, const USD @ PPP) ln (GDP per capita, const USD @ PPP) 1987–1990 1997–2000 2007–2010 1987–1990 1997–2000 2007–2010 Source: World Bank (WDI) reproduced from Ghani and O’Connell (2014). GROWTH AND STRUCTURAL CHANGE 51 FIGURE 4.5: Output and Employment Shares across Countries, 2012 Services VA % of GDP vs. GDP per capita, 2012 % Employment in Services vs. GDP per capita, 2012 Ethiopia & Comparator Countries Ethiopia & Comparator Countries 100 100 Russian Federation Singapore 80 80 Brazil KoreaRep Singapore Malaysia Brazil 60 Philippines 60 Russian Thailand Federation Korea Rep India Philippines Malaysia 40 40 Ethiopia–2012 Thailand India 20 Ethiopia–2012 Ethiopia–1992 Ethiopia–1992 20 0 .5 5 10 15 20 25 30 40 50 .5 5 10 15 20 25 30 40 50 GDP per capita, 000s (in constant 2005 international $ at PPP) GDP per capita, 000s (in constant 2005 international $ at PPP) Data source: ILOSTAT database & World Bank World Development Indicators, 2014. Data source: ILOSTAT database & World Bank World Development Indicators, 2014. Ethiopia data from national accounts. Ethiopia data from Authors' calculations using labor force surveys. 2012 GDP share regressed on log (2012 GDP per capita) & log (2012 GDP per capita) 2012 employment share regressed on log (2012 GDP per capita) & log (2012 GDP squared. per capita) squared. Manufacturing VA % of GDP vs. GDP per capita, 2012 % Employment in Manufacturing vs. GDP per capita, 2012 Ethiopia & Comparator Countries Ethiopia & Comparator Countries 40 30 Thailand KoreaRep 30 20 Malaysia KoreaRep Singapore Malaysia Thailand RussianFederation Brazil India Philippines 20 Singapore 10 Philippines India RussianFederation Brazil Ethiopia–2012 10 0 Ethiopia–1992 Ethiopia–2012 Ethiopia–1992 0 –10 .5 5 10 15 20 25 30 40 50 .5 5 10 15 20 25 30 40 50 GDP per capita, 000s (in constant 2005 international $ at PPP) GDP per capita, 000s (in constant 2005 international $ at PPP) Data source: ILOSTAT database & World Bank World Development Indicators, 2014. Data source: ILOSTAT database & World Bank World Development Indicators, 2014. Ethiopia data from national accounts. Ethiopia data from national accounts. 2012 GDP share regressed on log(2012 GDP per capita) & log(2012 GDP per capita) 2012 GDP share regressed on log(2012 GDP per capita) & log(2012 GDP per capita) squared. squared. Agriculture VA % of GDP vs. GDP per capita, 2012 % Employment in Agriculture vs. GDP per capita, 2012 Ethiopia & Comparator Countries Ethiopia & Comparator Countries Ethiopia–1992 100 60 Ethiopia–1992 80 Ethiopia–2012 Ethiopia–2012 40 60 India Thailand 40 Philippines 20 India Thailand Malaysia Philippines Malaysia 20 Brazil KoreaRep KoreaRep Brazil Singapore 0 RussianFederation 0 RussianFederation Singapore .5 5 10 15 20 25 30 40 50 .5 5 10 15 20 25 30 40 50 GDP per capita, 000s (in constant 2005 international $ at PPP) GDP per capita, 000s (in constant 2005 international $ at PPP) Data source: ILOSTAT database & World Bank World Development Indicators, 2014. Data source: ILOSTAT database & World Bank World Development Indicators, 2014. Ethiopia data from national accounts. Ethiopia data from national accounts. 2012 GDP share regressed on log(2012 GDP per capita) & log(2012 GDP per capita) 2012 GDP share regressed on log(2012 GDP per capita) & log(2012 GDP per capita) squared. squared. Source: World Bank (WDI) reproduced from Ghani and O’Connell (2014). 52 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT the most effective tool in overcoming these obstacles farm workers are turned into programmers or even for spurring industrialization as a real exchange rate call center operators. Contrast this with turning a depreciation of, say 20 percent, is effectively a 20 farmer into a factory worker in producing garments percent subsidy on all tradable industries. Sustaining or shoes. So raising productivity in services requires this, of course, would require an appropriate monetary steady and broad-based accumulation of capabilities in and fiscal framework. Yet it remains much harder for human capital, institutions, and governance. Services Africa to industrialize today than previously as global technologies also seem less tradable and more context- demand has shifted from manufacturing to services specific. Productivity gains in narrow segments may and because of fiercer global competition, including be easily established (e.g. by letting in Walmart or from Asia. Carrefour in retailing), but productivity gains along Since so much of Africa’s workforce is still in the entire retail sector is much more challenging. What agriculture, it may make sense to prioritize agricul- do these regional trends imply for Ethiopia? tural development as a part of a growth strategy. Given the large size of the agriculture sector in Without question, there are many unexploited oppor- Ethiopia today and in the future, it is imperative tunities in African agriculture, whether in perishable that continued efforts are made to make the sector non-traditional products such as fruits and vegetables more productive. The agriculture sector is, by far, or perishable cash crops such as coffee. Agricultural the biggest employer in Ethiopia, accounts for most diversification is equally hampered by a ‘poor busi- exports and is the second largest in terms of output. ness climate’ in addition to policy challenges associ- The sector also accounted for most of employment ated with extension, land rights, standard setting, growth over the period of analysis. Although some and input provision. Once again, the exchange rate labor shifted out of agriculture, substantial shifts can be an important compensatory tool. The main are likely to take a long time. Moreover, agricultural counter-argument is that it is very difficult to identify growth has been an important driver of poverty reduc- historical examples of countries that have pulled such tion in the last decade. Each percent of GDP growth a strategy off. In fact, one of the strongest correlates of reduces poverty by 0.55 percent, but each percent of economic development is export diversification away agricultural growth reduces poverty by 0.9 percent from agriculture. Moreover, even if this strategy were (World Bank, 2015). As a result, further labor pro- to succeed it would not reverse the process of migra- ductivity improvements in the sector are indispensable tion to urban areas, implying the need for strategies for Ethiopia’s future prospects. addressed towards urban job generation. Ethiopia faces considerable challenges in terms Despite encouraging examples, competitive- of achieving growth and structural change through ness in the services sector may depend on skills manufacturing or ‘industrialization.’ A major chal- and institutions that Africa is yet to acquire. A lenge relates to that of scale: the manufacturing share strategy emphasizing services productivity growth of output is remarkably small at 4 percent and has draws encouragement from success cases observed remained at this level since 1980. By comparison, in industries such as mobile telephony and mobile manufacturing account for about 9 percent of output banking. Though Rodrik argues that services have in SSA—an already low level. Lack of manufacturing not traditionally acted as a ‘growth escalator’ like sector growth during the take-off has not been an manufacturing, there are important counter-examples issue, but since the rest of the economy has grown at a emerging from South Asia as documented by Ghani similar rate of about 10–11 percent there has been no (2012). One challenge for Africa is the relatively high structural change in output in favor of manufacturing. requirement on worker skills. The IT sector requires Similarly, only 5 percent of the labor force is engaged long years of education and institution building before in manufacturing—a share that has hardly changed GROWTH AND STRUCTURAL CHANGE 53 since the mid-1990s. The sector has the second lowest of labor productivity are relatively high and labor labor productivity level amongst major sectors, only productivity growth has been substantial. However, twice as high as agriculture. the services expansion has been into traditional and While encouraging, the rise of the construction non-tradable sectors. sector raises questions about sustainability of recent Achieving high growth will also require achievements on growth and structural change. acknowledging the importance of many services The construction sector is, in its very nature, a sector sectors as growth and development escalators. The highly dependent on the business cycle. The sector services and manufacturing sectors in Ethiopia are is booming now, but this boom will not last forever, more intensely linked than in most countries in the and could in a worst case scenario turn into a bust, as world: 63 percent of all inputs used for Ethiopian seen elsewhere. Over the past three years, about a fifth manufacturing exports are from services. In Ethiopia, of GDP growth is attributed to the construction sec- services are more important for manufacturing value tor. This is substantial, as the sector has limited value added than the manufacturing sector itself, which added owing to high input costs. Construction activ- contrasts with many countries in the world. This is ity is driven by a combination of public and private because manufacturing is concentrated in low-value investment which led to a rise in employment from added activities with little domestic value addition tak- 1.5 to 1.9 million workers between 2005 and 2013, ing place, and with transport and distribution services many migrating from rural areas and working as day playing a very important role. Thus, the competitive- laborer. Construction has led to recent growth, jobs ness of the services sector is crucial for manufacturing and contributed to structural change. At some point, to thrive (Hollweg et al., 2015). this impetus will fade as the business cycle turns, even Going forward, Ethiopia would need to move as the country continues to pursue an infrastructure- forward across all sectors. Agriculture productivity led growth strategy. improvements are indispensable, as the majority of the On the other hand, the services sector has labor force (including the majority of the poor) will demonstrated considerable potential for Ethiopia, continue to work in the sector. Manufacturing growth including through its contribution to structural is essential for structural change and it offers positive change and in creating jobs. The services sector is prospects for employment, exports and productivity the largest in terms of economic output and is the gains. The services sector is also of high importance second largest employer in the economy. It accounts given its employment generating potential which is for most of the structural shifts away from agriculture important to absorb the rapidly rising working age in terms output and, to a lesser extent, labor. Levels population. 54 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Annex 4.1: Selected Structural Change Indicators TABLE A4.1: Gross Value Added by Sector GVA by sector GVA by sector GVA by sector (constant 2010/11 birr, million) (% total GVA) (Annual compound growth, %) Sector 1999 2005 2013 1999 2005 2013 1999–05 2005–13 1999–13 Agriculture 101,374 133,571 238,752 55.1 49.0 41.8 4.7 7.5 6.3 Mining 1,708 2,470 8,157 0.9 0.9 1.4 6.3 16.1 11.8 Manufacturing 7,988 11,048 24,798 4.3 4.1 4.3 5.6 10.6 8.4 Utilities 2,102 3,021 6,124 1.1 1.1 1.1 6.2 9.2 7.9 Construction 5,378 10,262 34,832 2.9 3.8 6.1 11.4 16.5 14.3 Commerce 26,867 39,108 110,158 14.6 14.4 19.3 6.5 13.8 10.6 Transport 5,828 11,264 25,792 3.2 4.1 4.5 11.6 10.9 11.2 Finance 2,657 5,443 13,559 1.4 2.0 2.4 12.7 12.1 12.3 Public services 15,167 25,788 45,563 8.2 9.5 8.0 9.2 7.4 8.2 Other services 14,875 30,533 63,585 8.1 11.2 11.1 12.7 9.6 10.9 Total 183,944 272,508 571,320 100.0 100.0 100.0 6.8 9.7 8.4 Source: Calculated official national accounts data. TABLE A4.2: Employment by Sector Employment by sector Employment by sector Employment by sector (thousands) (% total employment) (Annual growth, %) Sector 1999 2005 2013 1999 2005 2013 1999–05 2005–13 1999–13 Agriculture 19,869 25,208 30,821 79.8 80.2 77.3 4.0 2.5 3.2 Mining 16 82 195 0.1 0.3 0.5 31.8 11.5 19.8 Manufacturing 1,107 1,529 1,882 4.4 4.9 4.7 5.5 2.6 3.9 Utilities 28 33 90 0.1 0.1 0.2 2.7 13.4 8.7 Construction 229 446 825 0.9 1.4 2.1 11.8 8.0 9.6 Commerce 2,342 2,406 2,845 9.4 7.7 7.1 0.5 2.1 1.4 Transport 123 146 378 0.5 0.5 0.9 3.0 12.6 8.4 Finance 20 38 134 0.1 0.1 0.3 11.6 17.1 14.7 Public services 578 729 1,212 2.3 2.3 3.0 3.9 6.6 5.4 Other services 585 818 1,492 2.4 2.6 3.7 5.7 7.8 6.9 Total 24,897 31,435 39,874 100.0 100.0 100.0 4.0 3.0 3.4 Source: Calculated from labour force surveys. GROWTH AND STRUCTURAL CHANGE 55 TABLE A4.3: Labour Productivity by Sector GVA per worker by sector (constant 2010/11 birr, GVA per worker by sector Employment thousands) (Annual growth, %) Elasticity Sector 1999 2005 2013 1999–05 2005–13 1999–13 1999–05 2005–13 1999–13 Agriculture 5.1 5.3 7.7 0.6 4.9 3.0 0.85 0.28 0.41 Mining 109.0 30.1 41.7 –19.3 4.2 –6.6 9.50 0.60 3.04 Manufacturing 7.2 7.2 13.2 0.0 7.8 4.4 0.99 0.19 0.33 Utilities 75.0 91.9 68.2 3.4 –3.7 –0.7 0.40 1.68 1.15 Construction 23.5 23.0 42.2 –0.4 7.9 4.3 1.05 0.36 0.48 Commerce 11.5 16.3 38.7 6.0 11.5 9.1 0.06 0.10 0.07 Transport 47.4 76.9 68.2 8.4 –1.5 2.6 0.20 1.23 0.61 Finance 135.2 143.6 101.4 1.0 –4.3 –2.0 0.88 1.70 1.41 Public services 26.2 35.4 37.6 5.1 0.8 2.6 0.37 0.86 0.55 Other services 25.4 37.3 42.6 6.6 1.7 3.8 0.38 0.76 0.47 Total 7.4 8.7 14.3 2.7 6.5 4.8 0.55 0.24 0.29 Source: Calculated from labour force surveys and national accounts data. TABLE A4.4: Sectoral Decomposition of GVA Per Capita Growth (1999–2013) Share of contribution from (%): Within-sector Between-sector Changes in Total contribution Sector productivity shifts employment (%) Agriculture 26.6 1.4 0.5 28.5 Mining –2.4 3.5 0.6 1.8 Manufacturing 3.5 0.0 0.6 4.1 Utilities –0.1 0.9 0.2 0.9 Construction 3.6 3.2 1.7 8.5 Commerce 28.8 –4.1 –2.7 21.9 Transport 1.9 2.7 0.7 5.3 Finance –0.9 3.5 0.4 3.0 Public services 3.9 1.9 1.1 7.0 Other services 6.7 4.1 2.1 12.9 Total 71.6 17.2 5.1 93.8 Note: The total does not add up to 100 because the demographic component (6.2 percent) cannot be disaggregated by sector. Source: Calculated from labour force surveys and national accounts data. 56 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE A4.5: Sectoral Decomposition of GVA Per Capita Growth (2005–2013) Share of contribution from (%): Within-sector Between-sector Changes in Total contribution  Sector productivity shifts employment (%) Agriculture 34.3 2.6 –16.7 20.1 Mining 0.8 1.0 0.4 2.2 Manufacturing 5.1 0.0 –1.0 4.2 Utilities –0.7 1.5 0.2 1.0 Construction 6.0 2.4 1.1 9.5 Commerce 29.6 –1.5 –2.1 26.0 Transport –1.1 5.2 0.9 5.0 Finance –1.7 4.2 0.4 2.9 Public services 1.1 3.2 1.1 5.4 Other services 3.0 5.8 1.9 10.7 TOTAL 76.2 24.5 –13.7 87.0 Note: The total does not add up to 100 because the demographic component (13 percent) cannot be disaggregated by sector. Source: Calculated from labour force surveys and national accounts data. TABLE A4.6: Demographics and Employment Rate   1999 2005 2013 Total population 54,453 63,229 80,444 Working-age population (10+) 36,022 41,018 55,629 Working-age population (% total population) 66.2 64.9 69.2 Employment rate 69.1 76.6 71.7 Source: Calculated from labour force surveys. 57 DRIVERS OF AGRICULTURAL GROWTH34 5 Agriculture is the second largest economic sec- Ethiopia’s agricultural sector has recorded remarkable rapid growth in the last decade. There have been significant tor in Ethiopia and cereal production accounts for increases—more than a doubling—in the use of modern more than a quarter of GDP. Out of the 10.7 percent inputs, such as chemical fertilizers and improved seeds, average annual growth in real GDP recorded during explaining part of that growth. However, there was also significant land expansion, increased labor use, and Total the last decade, agriculture accounted for 3.6 percent- Factor Productivity (TFP) growth, estimated at 2.3 percent age points. This compares to 5.6 percent and 1.5 per- per year. The expansion in modern input use appears to cent for services and industry, respectively. However, have been driven by high government expenditures on the agricultural sector, including agricultural extension, the contribution of agriculture to overall growth has but also by an improved road network, higher rural declined over the decade (from 7.1 percent in 2004/05 education levels, and favorable international and local to 2.3 percent in 2013/14). Within agriculture the price incentives. crop production subsector was most important repre- senting 28 percent of GDP and growing at an average annual rate of 8.8 percent. Other agricultural activi- 5.1 Introduction ties (animal farming, hunting, and forestry) jointly accounted for nearly 12 percent of GDP and grew at This chapter identifies drivers of Ethiopia’s agri- 5 percent a year, on average. Owing to its predomi- cultural modernization process. As in the Green nant importance and in light of the available data, the Revolution, increasing adoption of improved seeds analysis of agricultural growth presented here focuses and chemical fertilizer have played a major role in primarily on the cereal production of smallholder agricultural output growth. While starting from a farmers within the main season (meher). low base, the adoption of improved seeds and the The chapter is structured as follows: Section use of chemical fertilizer more than doubled over 5.2 provides evidence on growth in the agricultural the last decade. This increasing adoption of modern sector over the last decade and further decomposes agricultural inputs has been facilitated by large invest- this growth into different components. Section 5.3 ments in the agricultural sector and beyond, leading discusses the modernization of the agricultural sector to improved road and communication networks, a and looks at the increasing adoption of chemical fertil- better educated rural population, and a large agri- izer, improved seeds, and other modern and improved cultural extension workforce. To further stimulate practices. Section 5.4 identifies four major drivers that agricultural growth in the country in the last decade, have contributed to agricultural growth in the country, there were no major droughts, which Ethiopia has and discusses more in particular the role of extension, suffered from before, there were improved incentives improved marketing, rural education, and incentives. for agricultural intensification because of favorable international prices for export crops and improved 34 This chapter is based on the background paper prepared by IFPRI: modern input–output price ratios for locally con- Bachewe, Berhane, Minten, and Taffesse (2015). The World Bank sin- cerely appreciates the collaborative efforts with IFPRI in its preparation. sumed crops, and, more broadly, there was an end of Interested readers are encouraged to consult the background paper for widespread civil conflict. more detailed information about data sources and methodology. 58 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Section 5.5 presents evidence on these drivers. Section contributions of area expansion and yield growth to out- 5.6 concludes and discusses further challenges for put growth for grains. While cultivated area increased by agricultural growth in the future. 4 percent in the beginning of the decade, this declined to 0.8 percent in 2013/14. Growth in yields has consis- The Growth of Agriculture 5.2  tently been higher than area expansion over the decade, 2004–2014 but the difference has become significantly larger at the end of the decade than in the beginning. The total area cultivated increased by 2.7 percent A Solow decomposition of crop output growth per year over the past decade. In the 2013/14 main reveals the importance of increased input use, agricultural season smallholder farmers in Ethiopia including labor, as well as productivity growth. cultivated 12.9 million hectares of land compared to Specifically, labor accounted for 31 percent followed 10.1 million in 2004/05 (Figure 5.1.1). This growth by the expansion in cultivated area (13 percent), was mainly driven by expansion of area covered by increased application of chemical fertilizer (8 percent), cereals. Grains accounted for about 96 percent of the improved seeds (11 percent), returns to scale (8 per- total cropped area during 2004/5–2013/14. In par- cent) and rural roads (3 percent). The unexplained ticular, nearly three-quarters of the area was covered by residual, or total factor productivity growth, reached the five major cereals (teff, barley, wheat, maize, and 22 percent. Average annual TFP growth was 2.3 per- sorghum). Next in importance were pulses and oil- cent (Figure 5.1.3). seeds. Area allocated to pulses, vegetables, root crops, A cursory look at outcomes in other countries and fruits grew relatively faster, but from a low base. can provide a perspective. Figures 5.1.5 and 5.1.6 Agricultural output and the number of small- report on maize and wheat yield levels and growth rate holder farmers rose by 9.4 percent and 3.8 percent during 2004–2013 for selected countries. The impres- per year, respectively. By any standard, the growth in sive growth rates recorded in Ethiopia are clearly from crop output in the last decade has been quite rapid. a low base and the country has a lot of catching up to According to CSA estimates, total agricultural output do relative to those with the highest yield levels. For level during the meher in 2013/14, estimated at 32 mil- instance, in 2004 Ethiopian maize yields were less than lion metric tons, was 124 percent higher than the level a quarter of those in Egypt and a fifth of those in the in 2004/5. This was mainly driven by growth in cereals USA. By 2013, the gap narrowed and Ethiopian maize output, which accounted for 72 percent of the total. yields reached 44 person of Egypt’s and a third of the The number of smallholders grew from 11 million in USA. Nonetheless, these gaps remain considerable. 2004/5 to 15.3 million in 2013/14 (Figure 5.1.1). The performance of Ethiopia’s agriculture is Yield growth averaged about 7 percent per year. consistent with the recent recovery and growth of Yields in cereals averaged 21.4 quintals per hectare agriculture in many African countries. Nin-Pratt (q/ha) in 2013/14 and ranged from about 28 q/ha in (2015) reports that agricultural output per worker maize to 13 q/ha in teff. Averaged annual yield growth grew by 2 percent during 2001–2012. This compares by crop ranged from 8.1 percent for maize to 4.8 to 0.6 percent growth during 1990s and no growth in percent for barley, while reaching 5.2 percent for teff, the 1970s and 1980s. He also estimates annual average 5.9 percent for wheat and 7.1 percent for sorghum. TFP growth rate of 2.2 percent for the best perform- Growth in cereal yields was faster relative to other crop ers (ranging between –1.0 and 4.2 percent) during groups with the exception of root crops, which had 1995–2012. The corresponding figure reported for considerably higher variation in yields. Ethiopia is 2.6 percent. The contribution of area expansion has been The use of complementary sources of data rein- declining over time. Figure 5.1.2 illustrates the forces the impression of significant yield growth in DRIVERS OF AGRICULTURAL GROWTH 59 FIGURE 5.1: Agricultural Growth in Ethiopia, 2004–14 1. Output, area cultivated and smallholder farmers 2. Growth in area cultivated and yield of grains, %/y 18 350 12 16 300 10 14 250 12 8 10 200 6 8 150 6 4 100 4 50 2 2 0 0 0 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 Smallholders Crop Output Cultivated Area Yield (million) (Million Quintals) Area (million hectares) 3. Solow Decomposition of Crop Growth, 2004/5–2013/4 4. Yield Growth (percent per year) 8 Rural TFP roads 22% 7 RTS 3% 8% 6 Services 5 0.2% Irrigation 4 1% Labor 3 Pesticides 31% 1% 2 Improved 1 seeds 12% 0 Fertilizer Teff Maize Barley Wheat 8% Land Capital 13% 1% CSA 2005–14 Ad Hoc Surveys (2008–13) ERHS (1994–2009) 5. Maize Yield: Levels (tons/ha) and Growth (%/y) 6. Wheat Yield: Levels (tons/ha) and Growth (%/y) 12 8 10 7 8 6 6 5 4 4 2 3 0 2 –2 1 –4 0 China Egypt Ethiopia Kenya USA China Egypt Ethiopia Kenya USA 2004 2013 Growth 2004 2013 Growth Source: (1) and (2) CSA annual reports. (3) Bachewe et al (2015) using CSA and NBE data. (4) Bachewe et al (2015) using AGP, FtF, ATA and IFPRI data. (5) and (6) FAOSTAT. 60 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT the cereal sector over the past decade. We comple- and 1970s, usually referred to as the Green Revolution ment the data on agricultural change from the CSA (Evenson and Gollin, 2003). There has been since a presented so far with two complementary methods, significant effort to try to replicate this revolution in namely a comparison of data from large ad hoc house- the African continent, and in Ethiopia in particular. hold surveys over the last 6 years (see Bachewe et Since early 1990s, Ethiopia has implemented several al., 2015) and from the Ethiopian Rural Household cereal intensification programs promoting the adop- Survey (ERHS). Given that surveys were fielded in dif- tion of modern agricultural technologies. At the cen- ferent areas and with different methodologies, caution ter of these strategies has been the push for adoption is required in comparing yields over time. The results of chemical fertilizer—improved seed packages by are illustrated in Figure 5.1.4. While growth rates are smallholders (Spielman et al., 2010). In this section, similar, it is to be noted that there a differences in we assess to what extent changes in the adoption of yield levels between these different data sources and these improved technologies have happened in the survey methods.35 last decade and how they might have contributed to agricultural growth. Land Intensification and Adoption 5.3  In line with Ethiopia’s intensification efforts, of Improved Agricultural chemical fertilizer imports and use have dra- Technologies matically increased over the last decade. Fertilizer imports have increased by 124 percent and fertilizer Land and labor expansion have been important use by smallholders increased by 144 percent over the contributing factors to increased agricultural pro- past decade (Figure 5.2.1). The figures imply signifi- duction and land intensification. Cultivated land cant imported fertilizer carryovers given that imports increased by 27 percent over the last decade, according exceeded use. These trends are noteworthy as Ethiopia to CSA, while the number of smallholders increased by has one of most depleted soils in Africa (IFDC, 2012), 39 percent. This indicates smaller sizes of farms over and despite the introduction of chemical fertilizers in time (the average landholding size declined annually the late 1960s, their application levels had remained by 1.4 percent over this period) and therefore more low for decades (Rashid et al, 2013). intensive labor use per unit of land, given relatively Most fertilizer use has been on cereals, in part little off-farm opportunities in rural areas (World because of the attention given to cereal production Bank, 2014a). Headey et al. (2014) confirm these to achieve food security in Ethiopia. According stylized facts and document how already small farm to CSA, 2.1 million (46 percent) holders growing sizes have been declining quite rapidly and that young cereals used fertilizer in 2004/05 and this number farmers cultivate substantially less land than previous increased to 5.5 million (76 percent) holders in generations did. They also find that family labor use 2013/14 (Figure 5.4.2). Cereal area applied with fer- per hectare increases substantially with increasing land tilizer—which nearly doubled during the same period pressure, leading to higher gross incomes per hectare. from 2.7 million hectares in 2003/04 to 5.2 million To what extent did Ethiopia experience ele- hectares in 2013/14, or an increase from 36 percent to ments of a Green Revolution? Technological change 53 percent of the total cereal area—accounted for at in agriculture—such as the replacement of tradi- least 91 percent of total fertilized area in all years except tional seed varieties with improved cultivars and the in 2009/10. The intensity of fertilizer use on areas increased adoption of chemical fertilizers, often aided covered with fertilizer has also increased substantially by better water management through improved irriga- from 92 kg/ha in 2003/04 to 122 kg/ha in 2013/14. tion—has been the driver for a dramatic increase of agricultural output in Asian countries in the 1960s 35 Please refer to the background paper by Bachewe et al (2015) for details. DRIVERS OF AGRICULTURAL GROWTH 61 Fertilizer use on other crops has also shown companies have therefore become more important significant increases over the same period consid- over time in improved seed distribution and it is ered. However, fertilizer adoption is less prevalent for estimated that there are now more than 30 private, these crops than for cereals. For example, the propor- agricultural cooperatives, and parastatal seed produc- tion of area fertilized for other crops such as pulses, ers in Ethiopia (Benson et al., 2014). While there are oilseeds, and vegetables has also nearly doubled in the still important challenges relating to timeliness of seed same period (an increase from 7.9 percent to 14.5 delivery and the quantity and quality of seed provided percent for pulses, from 5.4 percent to 11 percent for (Benson et al., 2014; Spielman and Mekonnen, 2013; oilseeds, and from 24.9 percent to 43.1 percent for Spielman et al., 2010), seed availability has improved vegetables). The proportion of area covered with root over the last decade, leading to higher adoption rates crops that was fertilized also increased from 20 percent and higher agricultural growth. in 2003/04 to 31 percent in 2013/14. The share of farmers using improved seeds in Significantly more improved seeds have been the cereal sector has increased over the last decade. developed and released in Ethiopia in the last decade While overall adoption rates are low, the share has seen than in any period before. Improved variety release significant improvements, with more than a doubling has been particularly dynamic in the case of wheat noted over the last decade, from 10 percent of the and maize. These improved varieties of wheat and cereal producers in 2004/05 to 21 percent in 2013/14 maize were most often developed and released by the (Figure 5.2.4). This was driven by the rapid increase of Ethiopian Institute of Agricultural Research (EIAR) in improved maize seed adoption which increased from 12 collaboration with the International Maize and Wheat to 28 percent over the decade. Large increases are also Improvement Center (CIMMYT). In the case of wheat, noted in the case of teff where adoption of improved it is estimated that 54 of the available 87 improved seeds increased from 1 percent to 5 percent and of wheat varieties over the last 40 years were developed and from 4 percent to 8 percent over that same period. released in the period 2001–2011 (Figure 5.2.3). This While these official data show significant improvement compares to 33 and 45 respectively in the case of maize. in adoption over time, there might however likely be sig- For other cereals, the number of varieties were lower, nificant measurement errors in improved seed adoption possibly because of lower funding and less international in the country, as discussed in Bachewe et al. (2015). involvement. For example, 32 improved varieties of teff Access to irrigation is low and did not change were released over the last 40 years of which 20 were significantly in the last decade. We further look at released in the period after 2000 (ATA-MoA, 2014). two other modern agricultural practices, i.e. irriga- Once improved seeds have been released, there tion and pesticides, which were a major contributor has been a heavy reliance on the Ethiopian Seed to the Green Revolution in Asia. While root crops Enterprise (ESE) parastatal for the production (especially potatoes and onions) are more likely to be and distribution of improved seeds. In the past grown on irrigated land, CSA numbers illustrated in the ESE used to produce most of the new improved Figure 5.2.5 indicate that the proportion of the area varieties on its own farms, as well as on state farms, irrigated has not changed in a significant way over the with large private farms playing only a minor role last ten years for these crops as well.36 in production of new varieties. However, given the consistent shortage of base seed in the system, there 36 Other data sources on irrigation that find irrigated areas are much has been an increasing decentralization, with other larger than those reported by CSA. This includes an estimate Hagos et al. seed distributors being allowed to participate in the (2009) of 5 percent of total cultivated area in 2006 and by Mulat (2011) of about 5 percent in 2011. As explained in Bachewe et al (2015) this system to cover base seed shortages (Alemu et al., may be due to differences in seasonal coverage and the fact that irriga- 2010). Regional research institutes and private seed tion is more common in larger farms as opposed to among smallholders. 62 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 5.2: Input Use: Fertilizer, Improved Seeds, Irrigation and Pesticides 1. Fertilizer use & imports (in 1000 metric tons) 2. Fertilizer: share of holders and cereal area cover 1000 80 75.6 900 70 800 60 700 47.9 50 46.3 600 53.1 500 40 400 30 36.0 300 29.5 20 200 100 10 0 0 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 Imported Used Holders (%) Area applied (%) 3. Improved varieties of major cereals released (#) 4. Share of improved seed applying farm holders 100 30 90 25 80 70 20 60 50 15 40 10 30 20 5 10 0 0 Wheat Maize Barley Sorghum Teff 2004/05 2009/10 2013/14 1970–2000 2000–2011 Barley Sorghum Teff Wheat Maize Cereals 5. Share of area under irrigation (%) 6. Share of pesticides applied area (%) 10 30 25 8 20 6 15 4 10 2 5 0 0 2004/05 2009/10 2012/13 2004/05 2009/10 2013/14 Cereals Pulses Oilseeds Cereals Pulses Oilseeds Vegetables Root crops All crops Vegetables Root crops All crops Source: (1): Imports: AISE for 2004/05–2011/12 and Comtrade thereafter. Use: CSA-AgSS. (2): CSA: AGSS. (3): ATA-MoA (2014). 4: CSA-AGSS. 5 & 6: CSA Annual Reports. See details in Bachewe et al (2015). DRIVERS OF AGRICULTURAL GROWTH 63 Pesticides use, on the other hand, increased The regression results show large and signifi- significantly over time. While 13 percent of the cant effects for extension, remoteness, and educa- crop area was exposed to pesticides in 2004/05 this tion on improved technology adoption. Farmers increased to 21 percent in 2013/14 (Figure 5.2.6). that received extension visits are associated with higher Again, these trends were driven mainly by cereals likelihoods of adoption of improved technologies. which show relatively high adoption rates and growth Less remote households and more educated house- over this period. holds are more likely to adopt improved agricultural In sum, important changes took place in the technologies. These factors are significant and consis- adoption of improved agricultural technologies tent associates of improved technology adoption and over the last decade. There has been more than a dou- have shown large changes over the last decade (as we bling in the use of chemical fertilizer, improved seeds, will argue below) and can therefore be justified to be and pesticides over this period, illustrating the increas- main drivers for improved technology adoption in ing modernization and intensification of agriculture in the last decade. Ethiopia. The uptake of these improved agricultural Other factors show significant association technologies was particularly pronounced in the sec- with improved technology adoption as well. First, ond half of the decade (2009/2010–2013/14) suggest- larger plots are associated with a higher likelihood of ing that land expansion and TFP growth were major improved technology adoption. This holds for the contributing factors to agricultural growth in the first four cereals. Second, cultivator managed plots have a half of the decade (2004/05–2009/10), but that in lower likelihood of adoption of improved technologies. the latter half agricultural growth was explained by Third, access to credit of the households leads to higher increasing use of modern inputs. likelihoods of adoption of improved agricultural tech- nologies. However, their effects are not significant in 5.4 Drivers for Change all specifications or these factors have not shown large changes over the period studied. While these results What were the drivers associated with increased from the national agricultural sample survey illustrate agricultural production in Ethiopia? To identify that these three drivers show a significant association the drivers for the increasing adoption of improved with improved technology adoption, the results how- technologies in the last decade, two conditions have ever do not unambiguously show that productivity to be met. First, they need to be linked with sig- change can be attributed to these factors, because of nificantly increased adoption of improved practices. possible endogeneity issues. A number of authors have Second, they need to have shown major positive looked at this issue with better, but not nationally rep- changes over the last decade. In this section, we resentative, datasets. We review that literature below. focus mostly on the first issue, using primary data First, a number of studies have assessed the analysis as well as a literature review. Table 5.1 relates impact of the increased coverage by extension different associates with the adoption of improved agents. Dercon et al (2009) have shown that exten- seeds and/or chemical fertilizer for four main cere- sion has yielded significant impacts on consumption als in the country, i.e. teff, maize, wheat, and barley. growth. Others show that there is a strong association We run a Probit model using as dependent variable between increased use of technologies, mainly use of the adoption of this improved technology (yes=1; improved seeds, fertilizer and pesticides, and extension no=0) for these four crops and characteristics of the services provided (Ragasa et al., 2013; Berhane et al., household, plot characteristics, and climatic variables forthcoming; Minten et al, 2013). Using large-scale as explanatory variables. Table 5.1 presents the aver- panel data in high-potential agricultural areas, Berhane age marginal effects. et al. (forthcoming) illustrate that the extension system 64 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE 5.1: Probit Model: Adoption of Improved Seeds or Chemical Fertilizer in Cereal Production Barley Maize Teff Wheat Received extension visit (1=yes) 0.3276*** 0.5605*** 0.3589*** 0.3744*** 0.037 0.024 0.025 0.022 Avg. travel time to nearest city (pop. >=50000); hours –0.0153*** –0.0035 –0.0134** –0.0201*** 0.004 0.004 0.004 0.005 Age of head (years) 0.0001 –0.0007** 0.0004 –0.0004 0.000 0.000 0.000 0.000 Household head is female (1=yes) 0.0323** –0.0003 0.0271* 0.0346** 0.011 0.009 0.012 0.013 Education (highest grade) 0.0103*** 0.0049*** 0.0074*** 0.0073*** 0.002 0.001 0.002 0.002 Household size 0.0057** 0.0047*** 0.0010 0.0043* 0.002 0.001 0.002 0.002 Plot area in hectares 0.1357*** 0.1557*** 0.1371*** 0.2339*** 0.034 0.024 0.021 0.038 All farm plots combined –0.0020 –0.0034 0.0088 –0.0003 0.004 0.004 0.006 0.005 Cultivator owns the land (1=yes) –0.0262* –0.0730*** –0.0228* –0.0361** 0.011 0.011 0.010 0.013 Cultivator used irrigation (1=yes) –0.1212*** 0.0529 –0.1896*** –0.1441* 0.023 0.048 0.049 0.057 Cultivator rotated crops (1=yes) 0.0143 –0.0289 0.0611* 0.0590* 0.023 0.019 0.027 0.027 Cultivator has access to credit (1=yes) 0.0310* 0.0465*** 0.1163*** 0.0884*** 0.013 0.011 0.014 0.014 Average population density in woreda 0.0000 0.0004*** 0.0003** 0.0000 0.000 0.000 0.000 0.000 Technology adoption rate in woreda (last year); all crops 0.5920*** 0.3177*** 0.6641*** 0.6413*** 0.028 0.032 0.040 0.028 Rainfall, elevation, slope, and climate variables Included Included Included Included Observations 18913 38390 31247 19619 Source: Bachewe et al (2015) using CSA Agricultural Sample Survey 2008/09. Table shows (average) marginal effects. For dummy variables marginal effect is the discrete change from the base level. Clustered standard errors (at EA level) below coefficients; * p<0.05 ** p<0.01 *** p<0.001. does not increase productivity directly but that it Second, Dorosh et al. (2012) illustrate the works mostly indirectly via its effects on input use and importance of remoteness on adoption of improved input use complementarities. Krishnan and Patnam technologies for Africa as a whole. Using data from (2014) further illustrate that the effect of extension a remote area in Ethiopia, Minten et al. (2013) show, diminishes over time and that neighborhood effects using data from a quasi-experimental setting, that a become more important in stimulating improved 20 km increase in the distance from the farm to the technology adoption. modern input distribution center and output market DRIVERS OF AGRICULTURAL GROWTH 65 led to a 47 kg/ha and 6 kg/ha reduction in chemical of chemical fertilizer as well as improved seeds. The fertilizer and improved seed use, respectively. Stifel World Bank (2014) further shows that poverty reduc- and Minten (2015) further show a large association of tion in the country was linked with improved agricul- remoteness with agricultural production. These results tural performance, especially when prices were high, suggest substantial impacts of remoteness on modern access to markets was good, and fertilizers were used. input use and they therefore suggest that an improved In sum, the available evidence therefore shows transportation network contributes significantly to the that four factors (extension, roads, education and increase in modern input use. incentives) are associated with increasing adoption Third, Huffman (2001) shows that the domi- of improved technologies in Ethiopia. Evidence on nant effect of education on agriculture is techni- the changes in the last decade of these four main driv- cal change. International literature for developing ers—agricultural extension, road infrastructure, edu- countries shows that more educated farmers are more cation, and incentives—for the adoption of improved efficient and adopt modern technologies more eas- agricultural technologies as well as other factors is ily (Ogundari, 2014; Appleton and Balihuta, 1996; discussed next. Jamison and Lau, 1982). In the case of Ethiopia, several authors have illustrated the important effect 5.5 Evidence on Changes in Drivers of education on fertilizer use and innovation more broadly, especially in traditional areas in Ethiopia Government Strategy (Endale, 2011; Asfaw and Admassie, 2004; Knight et al., 2003; Weir and Knight, 2004). Illiterate farm- Cognizant of the fact that the vast majority of ers accounted for 63 percent of the total number in the Ethiopian population resides in rural areas, an average year during 2004/5–2013/14, according mostly deriving livelihoods from agriculture, the to CSA. Moreover, the proportion of farmers with Government of Ethiopia (GoE) has long put agri- informal education was 8.5 percent while those culture at the center of its national policy priorities. with formal education of grades 1–3 and 4 or higher An Agriculture Development Led Industrialization accounted for 10.5 and 18 percent of the total. Figures (ADLI) strategy was formulated in the mid-1990s 5.3.1 and 5.3.2 indicate that relative to their share, that served as roadmap to transform smallholder agri- farmers with education in Ethiopia perform superior culture in the country. The emphasis and focus given in the adoption of modern inputs, as shown by CSA to ADLI paved the way for rethinking overall growth surveys. However, out of the total number of farmers pathways and served as a blueprint of the national that adopted fertilizer and improved seeds, illiterate development agenda in the decades to come. ADLI farmers accounted for 58 percent and 54 percent, envisioned national development through the need for respectively. concerted efforts to transform Ethiopia’s traditional Fourth, the ratio of output over input prices has agriculture sector first, which according to the plan, been shown to be a major incentive for the adop- would eventually provide impetus to other sectors tion of fertilizer and other improved technologies in including manufacturing. Africa (Morris, 2007). Spielman et al. (2012) docu- To ensure a more efficient utilization of land ment to what extent incentives matter in the adop- and labor resources in rural areas, the GoE invested tion of fertilizer in Ethiopia in particular. Minten et heavily in the provision of rural public services in al. (2013) illustrate for the North of Ethiopia how the late 1990s and early 2000s. Among the top pri- distances to input distribution centers and changes in orities in the agenda were rural education and health, value-costs-ratios, driven by transportation and trans- rural infrastructure, extension services, and strength- action costs, lead to significantly lower adoption rates ening of public agricultural research. Government 66 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 5.3: Input use by Educational Level and Government Agricultural Expenditure 1. Proportion using fertilizer 2. Proportion using improved seeds Grades 4 Illiterate and higher 58% Grades 4 21% and higher Illiterate 24% 54% Grades 1–3 11% Literate (informal) 10% Grades 1–3 12% Literate (informal) 10% 3. Agricultural expenditures as a share of total government expenditures in Africa, 2003–2008 25 20 18 18 15 14 12 10 10 10 9 9 6 7 7 7 6 6 6 6 6 6 5 5 5 5 4 5 5 4 4 3 4 4 4 3 4 4 4 2 3 3 2 3 2 2 2 2 2 2 2 1 1 1 1 0 GNQ GNB MAR ZAF SSD LSO SWZ CAF CPV MUS DJI BWA COD NAM SYC GHA NGA KEN CIV SDN UGA MOZ DZA AGO STP MRT BDI TUN TZA ERI CMR GMB TCD BEN SLE GIN RWA TGO LBR SEN NER ZMB MLI ETH COG BFA MDG ZWE MWI EGY 4. Agricultural expenditures as a share of total government expenditures in Africa, 2008–13 25 20 18 16 15 15 10 10 10 10 8 8 8 9 7 6 7 7 7 5 5 6 6 6 6 6 6 5 5 4 4 4 4 4 2 3 3 3 3 3 3 3 3 3 3 3 3 3 2 2 2 2 1 1 1 0 GNQ GNB MAR ZAF SSD LSO SWZ CAF CPV MUS DJI BWA COD NAM SYC GHA NGA KEN CIV SDN UGA MOZ DZA AGO STP MRT BDI TUN TZA ERI CMR GMB TCD BEN SLE GIN RWA TGO LBR SEN NER ZMB MLI ETH COG BFA MDG ZWE MWI EGY Source: (1) and (2): CSA Annual Reports. (3) and (4): Benin (2014). DRIVERS OF AGRICULTURAL GROWTH 67 expenditures in Ethiopia have over the years been up, a new wave of training of Development Agents guided by several plans. Since 2005, a five-year plan- (DAs) was launched through Agricultural Technical and ning period was used. The period 2005–2010 was Vocational Education and Training (ATVET) centers guided by the Plan for Accelerated and Sustained newly established throughout the country. The PASDEP Development to End Poverty (PASDEP). The period plan outlined to assign at least three DAs (specializing 2010–2015 was the first phase of the Growth and in crop production, livestock and natural resources) in Transformation Plan (GTP). In these plans, Ethiopia each kebele. The new DAs were trained and mandated has consistently advanced the agricultural sector as one to train farmers and carry out agricultural extension of the important sectors to invest in. services in each kebele. Each kebele was planned to For example, Ethiopia signed the CAADP37 build one Farmer Training Center where farmers agreement in 2003 and was one of the few coun- would have access to participatory demonstrations tries to meet the targeted 10 percent expenditures for improved technologies and new farming systems. in this area. Ethiopia was one of only four countries Ethiopia achieved one of highest extension that met the 10 percent target over the two periods agent-to-farmer ratios found in the world. By 2008 (the three other countries are Malawi, Zimbabwe, and 2009, ATVET centers established throughout the and Burkina Faso), as illustrated in Figures 5.3.3 country had trained some 60,000 DAs and around and 5.3.4. Ethiopia’s agricultural expenditures made 8,500 Farmer Training Centers had been built at the up 17 percent of the total budget in the first period kebele level. As a result, there is one DA for every 476 (2003–2008). This declined to 10 percent for the farmers or 21 DAs per 10,000 farmers in Ethiopia period 2008–2013 (Benin, 2014). The sections below (Figure 5.4.1). This is significantly higher than in other illustrate the extent to which this expenditure has countries such as China, Indonesia and Tanzania, contributed to facilitate improved agricultural per- where this ratio stood at 16, 6, and 4 respectively formance and the adoption of improved agricultural (Davis et al., 2010). technologies, likely contributing to reduce some of the The number of holders reporting using the inefficiencies that have hampered improved technol- extension advisory service tripled with a doubling ogy adoption in the past (Jack, 2011). in the use of extension packages. As illustrated in Figure 5.4.2, it increased from 3.6 million in 2004/05 Changes in Informational Efficiency and the to 10.9 million in 2013/14. The number of holders Role of Agricultural Extension that participated in various crop type extension pack- ages more than doubled, from 2.6 million in 2004/05 At the core of the GoE’s investment in rural areas and to 6.6 million in 2013/14. In the same period, the agriculture has been the provision of a public agri- planted area covered by the extension package program cultural extension system which has seen unprec- increased from 1.5 million hectares in 2004/05 to 3.9 edented public expenditures since 1992 (Davis et million hectares in 2013/14. al., 2010). These expenditures largely focused on the provision of advisory and training services through a Changes in Input and Output Market public extension structure that spans from the federal Efficiency ministry to the regions and down to the woreda and kebeles through frontline extension agents. Well-functioning agricultural marketing systems In an effort to redress the challenges faced and to are important anywhere in the world, but espe- scale up best practices learned in the earlier period, the cially so in Ethiopia. This is because of the disastrous government launched a more comprehensive large- scale extension system in 2002. As part of the scaling 37 Comprehensive Africa Agriculture Development Program (CAADP). 68 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 5.4: Extension Service, Transport to Markets, and Farmer’s Education 1. Extension Agent per 10,000 Farmer Ratio 2. Extension Service Coverage 25 12,000,000 21 10,000,000 20 8,000,000 16 15 6,000,000 10 4,000,000 6 2,000,000 5 4 3 0 2 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 0 Ethiopia China Indonesia Tanzania Nigeria India Advisory Holders (under Hectares (number of holders) ext-pkg) (under ext-pkg) 3. Share of Population Connected to a Main City 4. Farmer’s Education Level 100% 14 75 Proportion of farmers educated (%) Proportion of illiterate farmers (%) 12 70 80% 10 65 60% 8 60 6 55 40% 4 50 20% 2 45 0 40 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 0% 1997/98 2006/07 2010/11 Access > 10 hours Access 5–10 hours Access 3–5 hours Illiterate Grades 1–3 Grades 7–8 Access 1–3 hours Access < 1 hour Literate (informal) Grades 4–6 Grades 9 and above 5. Travel time to a city of at least 50,000 people (1996/97 and 2010/11) 1994: Travel time to a city 2007: Travel time to a city of at least 50,000 people of at least 50,000 people < 1 Hour < 1 Hour 1–3 1–3 3–5 3–5 5–7 5–7 7–10 7–10 Over 10 Over 10 Source: (1): Davis et al (2010). (2): CSA-AASS. (3) & (5): Kedir et al (2015). (4): CSA Annual Reports. DRIVERS OF AGRICULTURAL GROWTH 69 implications in the past of badly functioning food more people are living in urban settings in Ethiopia. markets for food security, with food stocks available As urban people are much less likely to grow their in some parts of the country and widespread famine own food, this implies that commercial surplus in other parts (von Braun et al., 1998). Major reasons has increased significantly over the last ten years. for historically badly functioning food markets in Moreover, access to urban centers leads to increasing Ethiopia have been linked to lack of market informa- agricultural intensification and urbanization which tion, bad road infrastructure, and high transaction can then act as an engine of agricultural transforma- costs (e.g. von Braun and Olofinbiyi, 2007). However, tion (Schultz, 1951). For example, Figure 5.5.1–2 there have been important changes in this area in show that adoption of herbicides and chemical fertil- the last decade in Ethiopia which have improved the izer (DAP as well as UREA) for teff production was functioning of these markets. significantly higher in villages close to Addis compared Most importantly, the Ethiopian government to the more remote ones. While a significant improve- has embarked on a large road investment program ment is seen over time for most farmers, intensification in the last two decades. The total length of all-weather has however been more pronounced in villages close surfaced roads tripled in less than 15 years, from an to cities (see also Kedir et al., 2015). estimated 32,900 km in 2000 to 99,500 km in 2013 (NBE, 2014). This type of road development has Education important effects on the connectivity of agricultural markets in the country. In 1996/97, transportation Ethiopia made significant strides to achieve uni- infrastructure connected Addis Ababa to a limited versal primary education coverage, particularly in number of urban markets such as Mekelle, Bahr Dar, rural areas. This increases the number of educated Jimma, and Dire Dawa. By 2010/11, secondary cities farmers as some of these students make their liveli- linked to each other, and major corridors linking key hood afterwards in the agricultural sector. Moreover, market centers were fully constructed (Figure 5.4.5). efforts have also been made to make adult education Whereas in 1997/98, only 15 percent of the popula- accessible. Growth in the proportion of farmers with tion was within 3 hours of a city of at least 50,000 higher levels of education over the last decade is strik- people, 47 percent was within 3 hours of such cen- ing. Figure 5.4.4 shows that the share of illiterate farm- ters in 2010/11 (Figure 5.4.3). The improved road ers declined at 1.8 percent per year over this period. network has further led, among others, to a reduc- Furthermore the proportion of informally educated tion of travel times between wholesale markets in the farmers increased at an average annual rate of 1 per- country by an estimated 20 percent. However, travel cent while those in grades 1–6 and 7–8 increased at costs might have even fallen further possibly driven by about 3 percent and 5 percent. The increase in the more competition and a shift to bigger and cheaper proportion of those with at least grade 9 education trucks (Minten et al., 2014). In this regard, the grow- was remarkable standing at 15 percent. ing accessibility and expanding transport services have been shown to have positive impacts on agricultural Prices and Incentives productivity (Li, 2011). An important contribution to changes in mar- Over the last decade, prices in in input and out- ket performance has also been urbanization and put markets have changed significantly which increasing commercial surplus flowing from rural have led to improved incentives for agricultural to urban areas. Urbanization has increased rapidly, intensification. This has been the case for the trad- but starting from a low base, and it is estimated that, able and non-tradable agricultural sector. Figure 5.5.2 compared to the beginning of the decade, 3.7 million illustrates the ratios of output prices of the five main 70 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 5.5: Modern Input Adaption, Prices and Rainfall 1. Changes in adoption of chemical fertilizer in teff production by distance to Addis Ababa 100 2.0 80 1.5 60 1.0 40 0.5 20 0 0 0 50 100 150 0 50 100 150 Transport costs to Addis (Birr/quintal) Transport costs to Addis (Birr/quintal) Herbicides at survey Herbicides 10 years before DAP+urea at survey DAP+urea 10 years before 2. Output – Fertilizer Price Ratio 3. Export Price Indices (2003/04=100) 0.8 4.0 0.7 3.5 0.6 3.0 2.5 0.5 2.0 0.4 1.5 0.3 1.0 0.2 0.5 0.1 0.0 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 0.0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Teff Wheat Barley Sorghum Maize Coffee Pulses Oilseeds Chat 4. Patterns in Total Rainfall (millimeter) During meher and belg Seasons of 2004–2013 1800 1600 1400 1200 1000 800 600 400 200 0 Belg 2004 Meher 2004 Slack 2004 Belg 2005 Meher 2005 Slack 2005 Belg 2006 Meher 2006 Slack 2006 Belg 2007 Meher 2007 Slack 2007 Belg 2008 Meher 2008 Slack 2008 Belg 2009 Meher 2009 Slack 2009 Belg 2010 Meher 2010 Slack 2010 Belg 2011 Meher 2011 Slack 2011 Belg 2012 Meher 2012 Slack 2012 Belg 2013 Meher 2013 Slack 2013 Tigray Amhara Oromiya SNNP Source: (1): Minten et al (2013). (2) & (3) Berhane et al (forthcoming). (4) NASA. DRIVERS OF AGRICULTURAL GROWTH 71 cereals over chemical fertilizer prices were twice as there were droughts, government and donors have high in 2012 than in 2004, leading to improved been addressing the production shortfalls adequately incentives for chemical fertilizer use for the produc- as illustrated by interventions after the drought in the tion of these crops. This changing ratio over time Horn of Africa in 2012 (Maxwell et al., 2014). seems to have been linked to a number of factors, Other factors have contributed to improved including the fixing of margins for chemical fertiliz- agricultural productivity but their importance ers for the distributing cooperatives in order to keep has been more limited than the drivers mentioned their prices low—sometimes leading to profitability above. This includes access to credit, lower riskiness issues for these distributing cooperatives (Rashid et for technology adoption, land certification, mobile al., 2013)—, increasing exchange rate overvaluation phone technology, and safety nets. Each of these are making imports cheaper (World Bank, 2014b), a discussed in turn. decline in international fertilizer prices since 2008 Lack of access to credit is seemingly one of (FAO, 2012), and high output prices, especially from the major constraints to promoting agricultural 2008 until 2010. productivity and rural transformation overall. International prices were significantly higher Following the international microfinance revolution for most export crops at the end of the decade than in the 1980’s and 1990’s, Ethiopia has seen remark- in the beginning. Figure 5.5.3 shows that the price able progress in this sector starting in the early 2000s. of coffee was 2.5 times higher in 2012/13 than in Over the last two decades, the number of microfinance 2003/04 while the prices of oilseeds and pulses were institutions (MFIs) in the country has grown to 32 twice as high. In contrast, the price of chat stayed rather (from one in 1994). Figures 5.6.1 and 5.6.2 indicate stable over this period. This general price increase for that the MFI industry has expanded enormously in the export commodities has led to significantly higher last decade both in terms of number of active borrow- export revenues from these export crops as well as for ers—from around 500 thousand in 2003 to around increasing incentives for investments in these commod- 3.5 million borrowers in 2014—and an outstanding ities, as for example shown in the rapid expansion of loan portfolio from a little less than 2 billion in 2003 sesame cultivation in the country over the last decade. to around 11 billion birr (in 2011 prices) in 2014. In addition, significant amounts of government program Other Factors loans were also disbursed through Rural Saving and Credit Cooperatives (RuSACos) in the last decade. Over the last decade, there have been no major Data from the Ministry of Agriculture indicate that incidences of large-scale droughts that have plagued in 2014 about 600 million ETB was disbursed to Ethiopia before. Given the rain-fed character of smallholders through RuSACos (which was about 56 Ethiopia’s agriculture, its production is heavily depen- million ETB in 2008/09). However, while there has dent on timely and sufficient rainfall. Figure 5.5.4 been significant growth in this area, it is unclear how shows levels of rainfall for the four main agricultural much of the micro-finance has been used towards regions, of the meher, belg, and slack seasons for the the agricultural sector. For example, the CSA data years 2004 until 2013. It illustrates, on average, the shows that the share of farmers that used credit for relative stability of rainfall patterns seen over the last agricultural purposes has not changed significantly decade. Moreover, Ethiopia is equipped with a good over time (it varied between 22 percent and 28 per- early warning system for the crop and the livestock cent over the period 2005–2013). It therefore seems sector as well as with a large safety net program—the that credit might have been more readily available in Productive Safety Net Programme (PSNP)—as to deal rural areas, but it might not have directly impacted with the consequences of droughts. In the case that agricultural activities. 72 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 5.6: Microfinance Activity, Fertilizer Constraints, Crop Damage and Real Wages 1. MFI: Active borrowers (million) 2. MFI: Gross Loan Portfolio (bn. birr, 2011 prices) 4.0 12.0 3.5 10.0 3.0 8.0 2.5 2.0 6.0 1.5 4.0 1.0 2.0 0.5 0.0 0.0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 3. Fertilizer adoption constraints (% of hhs) 4. Share of crop area damaged (2004/05–2013/14) No problem/not relevant Weeds Pests Others Hailstone Birds Lack of credit Wild animals High price Too much rain Shortage of rain Arrived late Locust Frost or floods Shortage of supply Crop disease 0 10 20 30 40 0 1 2 3 4 5 5. Rural Real Wages (US$/day) 1.6 1.5 1.4 1.3 1.2 1.1 1 2004/05 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 2013/14 Source: (1) & (2): ASEMI. (3) AGP Survey (2011). (4) & (5): CSA. DRIVERS OF AGRICULTURAL GROWTH 73 The risk situation might also have changed in traders and brokers used mobile phones, it was ubiqui- favour of higher agricultural productivity over the tous by the end of the decade, leading to more efficient last decade. Dercon and Christiaensen (2011) show markets (Minten et al., 2014). On the other hand, that uninsured risk is a significant constraint on tech- the penetration of the mobile phone with farmers in nology adoption in Ethiopia, and although insurance Ethiopia is still limited, especially when compared to markets may be no different in 2014 than in 2004, it neighboring Kenya. The available evidence also shows is plausible that the riskiness of technology adoption that the spread of mobile phones did not lead to major has changed: (i) the doubling of output-input price changes in agricultural pricing for these farmers that ratios makes negative returns to fertilizer use in the had access to a phone (Tadesse and Bahiigwa, 2015). event of bad weather much less likely; (ii) there is now Ethiopia has started up since 2006 a large a functioning public safety net for many; (iii) the run safety net program—the Productive Safety Net of good weather may make the salience of bad weather Program—that covers a large area of the country shocks more muted (perhaps changing farmer’s expec- and benefits more than 7 million vulnerable people tations of the risk); (iv) farmers are seemingly richer, in the country. In this program, some of the partici- have more assets, and are better able to self-insure; pants are paid to participate in communal works to and (v) the micro-finance institutions may be easier help improve communal infrastructure such as roads, to default on than the government (formerly the main irrigation, and terraces. Taffesse et al. (2015) show to credit provider). what extent investments in these communal assets A large-scale land certification program was set have contributed to higher agricultural productivity. up in the country in the 1990s as to ensure more However, the area covered makes only up a small part secure land property rights. This land certification of the cultivated land of the country as a whole and program has been one of the largest, cheapest, and might therefore only partly explain agricultural growth fastest in Africa (Deininger et al., 2008) and it is in the last decade. estimated that about half of the farmers in the main four regions benefited from this (Ghebru et al., 2015). 5.6 Conclusion and Further Challenges While land stays property of the state, the certificates have, however, allowed to ensure more secure property There have been significant changes in Ethiopia’s rights as they have been found to have led to higher agricultural and food economy in the last decade. investments, more land rental market activity, higher Agricultural output more than doubled over the last productivity, and improved food security (Holden et decade driven by area expansion but more impor- al., 2007; Deininger et al., 2011; Ghebru and Holden, tantly by significant yield increases. The real value of 2013; Melesse and Bulte, 2015). As this land certifica- agricultural GDP increased by 7.6 percent per year tion program in most areas happened before the period and export earnings from agricultural commodi- under study, it might not have directly contributed to ties doubled over this period. Moreover, average per increased productivity in the last decade but it seems capita food consumption increased by more than 20 however clear that the more secure property rights percent over this period and we note a relative decline have been an enabler for the agricultural productivity of expenditures on cereals in the consumption basket, increase seen in the last decade. indicative of important changes in food habits. This In the last decade there has been a spread of the agricultural growth is further shown to have been mobile phone in urban and rural areas in Ethiopia. associated with significant poverty reduction in the This might likely have impacted agricultural trade country (World Bank, 2014). through improved access to information. While in The increased productivity is partly explained the beginning of the decade, none of the agricultural by a rapid uptake of a number of improved 74 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT agricultural technologies. Over the period studied, that this growth process will be continued in total fertilizer consumption increased by 143 percent the future. Sustainable intensification will need to and the share of cereal farmers that applied chemi- receive even more attention in the future as land con- cal fertilizers increased from 46 percent in 2004/05 straints have increasingly become binding (Headey to 76 percent in 2013/14. Other chemical use, such et al., 2014). It seems therefore that there will be as pesticides and herbicides, increased as well. More a need for more widespread adoption of adapted improved varieties for the major cereals were released modern inputs and improved technologies, especially in the period 2000–2011 than in the thirty years as there is still significant opportunity for further before and while there are still problems in distribu- growth. Chemical fertilizer use stays central to the tion of improved seeds, its use—while still relatively government effort to increase agricultural productiv- low—doubled during this period. However, part of ity. It has therefore initiated a unique soil mapping the growth cannot be explained by increasing adoption exercise in the country as to adjust fertilizer pack- of these modern inputs, and other production factors, ages to specific soil conditions. This is a promising and we therefore also note significant growth in Total development that might address soil deficiencies in Factor Productivity (TFP), on average 2.3 percent per the country more appropriately. However, emphasis year. We further note that the increasing adoption of on an efficiently functioning fertilizer distribution these modern agricultural technologies and its contri- system is also needed as shown in the stated reasons bution to agricultural growth has especially happened on constraints to adoption of fertilizer by farmers in the second sub-period of the last decade. In the first in the baseline survey of the AGP program. While sub-period, agricultural growth was relatively more the share of farmers using fertilizer in these high driven by area expansion and TFP growth. potential areas was high, there were still significant Major drivers for the increasing adoption of issues with the availability of fertilizer as more than modern inputs seem to be multiple, linked with 20 percent and 12 percent of the farmers citing lack significantly higher expenditures in the agricultural of supply and late arrival of fertilizer as a constraint sector. First, Ethiopia has built up in the last decade to adoption respectively (Figure 5.6.3). a large agricultural extension system, with one of the Higher adoption rates of improved and high- highest extension agent to farmer ratios in the world. performing seeds are also needed. To stimulate Second, there has been a significant improvement in adoption, better supply and marketing conditions are access to markets. Third, improved access to education required. On the supply side it seems that while the led to a significant decrease in illiteracy in rural areas. public sector has an important role to play and more Fourth, high international prices of export products resources should be channeled to the development as well as improving modern input—output ratios of improved seeds in local research settings given the for local crops over the last decade have led to better high return to such investments (Alston et al., 2000), incentives for the agricultural sector. These factors all a more active role of the private sector is required as show a strong association with increasing adoption well. Moreover, marketing, distribution and informa- of improved technologies, and consequently agricul- tion provision on improved seeds should be improved tural productivity. However, other factors played a to generate a more vibrant seed sector in Ethiopia. To role as well, including good weather, better access to further agricultural intensification, there will also be micro-finance institutions in rural areas, and improved a need for better water management and increasing tenure security. irrigation, more intensive use of land through double While the agricultural growth process in cropping, and more attention to reduce important Ethiopia has been remarkable, there are a number soil erosion problems in the country. This will require of challenges that should be addressed to assure an important role of the government and the private DRIVERS OF AGRICULTURAL GROWTH 75 sector as to assure that appropriate technologies are indicators. While there are still a number of unknowns developed and are available at an affordable price. As on how this nutritional transformation can be most part of this effort, the innovation system’s link with efficiently achieved, it seems that behavioral change the public extension system as well as the private sector communication, sanitation, improved market access, should be institutionalized and strengthened. and production diversity, especially in less connected Climate change is expected to have a significant areas, should have a major role to play (Hirvonen and impact on Ethiopian agriculture in the decades Hoddinott, 2014; Hoddinott et al., 2013; Stiefel and ahead. It is estimated that the shifting rainfall pat- Minten 2015). terns and increasing temperatures will lead to crop Most of the agricultural growth has happened yield decreases as well as lower incomes from livestock in the cereal sector in the last decade. However, as (Robinson et al., 2013). Moreover, the incidences of the Ethiopian population is becoming richer and more unexpected weather shocks is expected to increase. urbanized, this will likely lead to changing demands Figure 5.6.4 illustrates that weather related events were for foods, different consumption baskets, and a trans- already the major source of crop damage in the last formed agricultural sector. For example, there will be decade. Incorporating climate change in agricultural an increasing demand for livestock products—with development programs will therefore become increas- more cereals being used as feed demand—, fruits ingly important. This will have to be done through and vegetables, processed and ready-to-eat products. adaptation to possible effects of climate change as well Some of the products require the development of as mitigation where practices are pursued that will lead new and different value chains, given lack of relevant to least greenhouse gas emissions. knowledge, seeds, and other inputs. Given the high Recent poor rains during the Belg and Meher perishability of some of these emerging agricultural seasons illustrate the vulnerability of Ethiopia’s products, investments will also be needed towards rain-fed agricultural systems. At the time of writ- new off-farm technologies, such as cold storages and ing (October 2015), Ethiopia is experiencing severe processing, that might then help to fulfil that demand. drought following a failure of the Belg rains and very Gender issues are also important in agriculture unstable rainfall pattern of the Meher rains. It is too and addressing those seems needed to improve agri- early to determine the potential impact of this on cultural performance, but more importantly nutri- agricultural growth and overall economic growth. tional indicators, in Ethiopia. Empowering women Preliminary analysis of rainfall data and potential crop in agriculture has likely pay-offs for nutritional and loss suggest that the current situation is similar to the agricultural outcomes. This will require policies and one experienced in 2003, but not as bad as in 2002. interventions in different areas. For example, Kumar Further efforts to expand the irrigated area would and Quisumbing (2015) show that reforms in law thus be imperative to make agriculture less reliant on and land registration has been an important avenue the weather. to improved gender equality in Ethiopia. While agriculture has shown high growth rates Mechanization in Ethiopia’s agricultural pro- and its growth has contributed to significant wel- duction and post-harvest activities is currently low. fare improvements, there is still significant scope However, the increasing transformation of Ethiopia’s for improvement. One concern is the slow change economy is leading to higher real wages in rural areas in nutritional indicators and the high level of stunt- (Figure 5.6.5). These higher rural wages will give ing in the country, especially in rural areas (Headey incentives for an induced innovation towards labor- et al., 2014). More attention should therefore be reducing technologies (Ruttan and Hayami, 1984). paid to how agricultural growth can have enhanced This trend can already be seen by the increasing adop- beneficial effects on food diversity and nutritional tion of herbicides, a substitute of weeding labor, in 76 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT commercial agricultural areas but it will also drive the that the right machines, and spare parts, at affordable demand for more mechanization, especially for these prices will be there to alleviate that constraint is there- activities where there is a peak demand of labor, such as fore also an important further challenge for sustained during planting and harvesting periods. Making sure higher agricultural productivity. PART B: SUSTAINING GROWTH 77 PART B: SUSTAINING GROWTH Part B is structured as follows: Chapter 6 discusses how to manage our expectations to Ethiopia’s future growth performance. Chapter 7 analyzes Ethiopia’s financing choice between public infra- structure investment and private investment, and considers complementary financing mechanisms for infrastructure. Chapter 8 analyzes the role of structural economic reforms in enhancing future growth. 79 MANAGING GROWTH EXPECTATIONS 6 prospects for growth? How does this evidence help What should we expect in terms of Ethiopia’s growth rate over the next decade? Double digit growth? Given its form our growth expectations for Ethiopia? low level of income, Ethiopia has substantial potential The chapter is structured as follows: Section 6.2 for realizing the benefits of catch-up growth, especially reviews the international experience with growth accel- if it can avoid economic crises. At the same time, there are also reasons for being cautious with very high growth eration and derives stylized facts. Section 6.3 offers a expectations. The first reason is statistical: only few countries qualitative discussion of country-specific head-winds have managed to sustain a growth acceleration beyond a and tail winds expected to affect growth going forward. decade. Second, country-specific growth headwinds appear to be somewhat stronger than the tailwinds. A third reason Section 6.4 considers cyclical factors and presents is cyclical: the ongoing private construction boom will not quantitative estimates. Section 6.5 uses our cross- last forever. Finally, regression model simulations suggest country regression model to benchmark the growth a growth slowdown in three alternative policy scenarios. effect of individual policy variables. Section 6.6 uses Going forward, growth may be expected to range between 4.5 and 10.5 percent. The objective of policy is to maximize the same model to simulate growth outcomes under the changes that high growth rates can continue. alternative policy scenarios. Section 6.7 concludes. Growth Accelerations: The 6.2  6.1 Introduction International Experience What should we expect in terms of Ethiopia’s What can we learn from the empirical evidence growth rate over the next decade? This chapter will on growth accelerations? A number of statistical refrain from making explicit long term growth projec- facts can be derived about economic growth experi- tions for Ethiopia in recognition of the dismal record ences across countries. The stylized facts that appear of economists in this area. What it does instead is particularly relevant for Ethiopia are summarized in to present a range of realistic outcomes to form our Box 6.1. Overall, the evidence compels us to con- evidence-based expectations going forward. Notably, template a wider range of outcomes than are typically it argues that a continuation of the current trend, considered in Ethiopia. Many of the great economic while being the policy objective under GTP2, is by forecasting errors of the past half century came from no means a certain outcome and does in fact represent excessive extrapolation of performance in the recent an optimistic scenario. past and treating a country’s growth rate as a perma- We address the following questions: What nent characteristic rather than a transient condition. can we learn from the empirical evidence on growth How can Ethiopia’s recent growth acceleration accelerations across countries? What are the country- be interpreted in light of the empirical evidence? specific factors that either support or inhibit long-term Between 2003/04 and 2013/14, Ethiopia achieved growth? To what extent is the current growth accel- an average annual GDP growth rate of 10.9 percent eration a cyclical or permanent phenomenon? Which and 8.0 percent in per capita terms. We note that our individual growth policies will give the ‘biggest bang period choice is deliberately determined as the one for the birr’? Which policy packages offer the best which maximizes recent growth as the inclusion of 80 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT BOX 6.1: Stylized Facts about Growth Accelerations and their Aftermath What to expect about growth accelerations: • Cross-country experience of per capita GDP growth has been an average of 2 percent per year with a standard deviation of 2 percent (using Penn World Tables data since the 1950s). • Episodes of per capita growth of above 6 percent are extremely short-lived with a median duration of 9 years. • China’s experience from 1977 to 2010 is the only instance of a sustained episode of per capita growth exceeding 6 percent for more than 32 years (1977–91: 7.61 percent and 1991–2014: 8.6 percent). Only two countries come close: Taiwan (1962–1994: 6.8 percent) and Korea (1962–1982: 6.3 percent). • The end of an episode of super-rapid growth is nearly always a growth deceleration. Of the 28 episodes of above 6 percent per capita growth, only 2 ended with a shift to higher growth: Korea (1982) and China (1991). • Since 1950, only 24 economies have notched up a 4.5 percent per capita growth over 30 years (Rodrik, 2013). • Six non-resource rich economies have experienced decade long rates of per capita growth exceeding 7 percent: Japan, Korea, Hong Kong, Singapore, and China (Virmani, 2012). • Six African countries experienced GDP per capita growth of more than 3 percent in 1995–2000 without that growth being driven by the exploitation of natural resource wealth: Burkina Faso, Ethiopia, Mozambique, Rwanda, Tanzania, and Uganda (IMF, 2013). • The Commission on Growth and Development (2008) analyzed 13 economies that had grown at an average rate of 7 percent a year (total—not per capita!) or more for 25 years or longer since 1950. What to expect after a growth acceleration: • The single most robust and striking statistical fact about cross-national growth rates is regression to the global mean of 2 percent per capita growth. • Past growth is not a good predictor of future growth. The R-squared of decade-ahead predictions of decade growth varies from 0.056 (for the recent decade) to 0.13. The best coefficient prediction is around 0.3 for decade-ahead predictions. • The median of the growth episode that follows an episode of super-rapid growth is 2.1 percent per year. • The reason for the low growth on average of developing countries versus developed countries is not the lack of rapid growth—it is the lack of the growth persistence and the very low growth rates during their periods of negative growth (North, Wallis, and Weingast, 2009). Source: Authors compilation from Pritchett and Summers (2014) unless otherwise noted. earlier years would reduce average growth. As an illus- growth is a statistically reasonable expectation. Even tration of this statistically exceptional performance, we though the past is a poor predictor of the future, this start by noting that it is 3 standard deviations above the would imply a growth rate of 2.4 percent per capita global cross-country mean growth rate of 2 percent. for Ethiopia (using their regression coefficient of 0.3). What can be (statistically) expected of Ethiopia’s The statistical evidence also suggests that after 11 future growth performance? If we knew nothing years of growth acceleration (above the median of 9 about the Ethiopia economy and its prospects except years), a growth deceleration could occur. Using the for the recent growth acceleration, we could make the median experience across countries which experienced following assertions. In light of the stylized facts listed a growth acceleration would yield a per capita growth in Box 1, there is a high probability that Ethiopia’s rate of 2.1 percent. Finally, and at the most pessimis- growth rate in 2015–25 would not be as high as t tic end of the range of statistical possibilities, it is not observed in 2004–14. Ethiopia might join the world improbable that Ethiopia could even experience nega- league of countries with exceptional growth perfor- tive growth rates in the future. To illustrate, among mance, but the associated probability is very low. The 44 low income countries a negative growth rate has work by Pritchett and Summers (2014) suggests that been registered 44 percent of the time and the growth a reversion to the global mean of 2 percent per capita rate was negative by 5.4 percent (North, Wallis and Managing Growth Expectations 81 Weingast, 2009). In fact, it is often the ability to avoid expectations, it is equally important to consider ‘growth disasters’ in the form of economic crises that country specific factors. Indeed, in their analysis set apart countries that achieve high long-term growth of the future growth prospects of China and India, rates compared to those that do not. Summers and Pritchett (2014) take such an approach. Based on international cross-country experi- We identify a set of growth tailwind and head- ence it is reasonable to set a floor for our expecta- wind factors deemed relevant to a high-growth, tions to annual per capita growth in Ethiopia in the non-resources rich, low-income African country, coming decade of 2 percent. This is not a prediction such as Ethiopia. These factors were derived on the of what will happen to Ethiopia’s growth over the next basis of the stylized facts and conceived wisdom ema- decade. Rather it is a suggestion to what might happen nating from the most recent growth and economic if Ethiopia’s future growth patterns follow well estab- development literature. lished empirical regularities observed across countries. The likelihood that Ethiopia’s growth accelera- History teaches that abnormally rapid growth, such tion could persist for a decade or more are buoyed as what Ethiopia has experienced since 2004, is rarely by five key factors: persistent. While it might be the case that Ethiopia will continue for another decade of double digit growth  First, there is scope for achieving substantial this would be a statistical tail event. gains from structural change i.e. the large scale A glance at Ethiopia’s historical record suggests shifting of the population into higher value add- that this ‘expectations floor’ is not unreasonable. ing urban-industrial and services activities when In fact, it coincides with the 2.1 percent per capita combined with fundamental changes in institu- growth rate achieved between 1992 and 2001 under tional deepening and the strengthening of social EPRDF. By contrast, the 1.5 percent per capita growth capabilities.39 achieved under Monarchy (1951–73) or the –1.0 per-  Second, additional productivity can be squeezed cent per capita under the Derg (1974–91) would not out through intra-sectoral transfer of resources be a relevant basis for comparison owing to the widely from less to more productive activities. In agri- different economic and political regimes. culture, for instance, this would involve the nar- At the other extreme, a repeat of the growth rowing of the gap between yield per hectare on performance of the past decade, would represent a the most productive and the average farms and an ceiling for growth expectations. As mentioned above, increase in the share of mechanized, commercial it would be unrealistic to expect a further acceleration farming. of growth beyond the 8 percent per capita growth rate  Third, continuing productivity gains can be currently observed. After all, only 2 out of 28 growth derived from technological catch-up through acceleration episodes worldwide have followed this pat- the deepening of human capital and investment tern. Similarly, there are examples of countries, even if in Research and Development across the spectrum the group is small, which managed to continue growing of economic activities with agriculture leading the at very high rates for two decades. In sum, even if this is way during the medium term.40 a low probability outcome in light of the cross-country empirical evidence, it cannot be ruled out. 38 This section draws from the background paper prepared by Yusuf (2014) for this report. The paper is a comprehensive review of the lit- erature and more references can be found therein. Growth Tailwinds and Headwinds38 6.3  39 These are the so-called advantages of backwardness first explored by Gerschenkron (1962). See also Abramovitz (1986) and Rodrik (2013a,b). 40 See Comin and Hobijn (2010). Parente and Prescott (1999, 2003) While statistical facts from cross-country expe- maintain that technology gaps caused by barriers to the adoption of rience are insightful and critical in forming technologies are the cause of income divergences and slower growth. 82 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT  Fourth, Ethiopia can tap the growth potential South-North and South-South relations (premature inherent in large urban agglomerations, an de-industrialization). integrated domestic market and demand from a nascent middle class. The endogenous factors are several:  Fifth, there are benefits to be reaped from FDI and participation in global value chains mediat-  Lagging agricultural productivity, traceable ing the trade of both light manufactures such as to illiteracy and low levels of education, biotic garments and footwear and of services.41 factors, access to fertilizer, environmental degra- dation, crop diseases, shortcomings of extension Two additional factors may also contribute to services, and infrastructure constraints.44 the strength of the tailwind at some point in the  The size and characteristics of Ethiopia’s export future: sector is inhibiting its ability to contribute to structural transformation. Ethiopia has the low-  For the ‘demographic dividend’ to serve as a est ratio of merchandise exports to GDP among tailwind for Ethiopia, the fertility transition populous countries in the world (7 percent); it needs to be accelerated, the workforce endowed has half as many of exporting firms as Kenya with marketable skills and businesses must invest (which has half the population of Ethiopia), and in activities that create many more jobs. average exporter size is small. Unprocessed and  A large domestic market offers attractive some- undifferentiated agricultural products dominate what protected opportunities for local businesses. exports with manufactures accounting for only With a population of 90 million, Ethiopia cer- about 10 percent.45 tainly has the numbers, although its low per  A relatively small financial sector offering limited capita income its small urban middle class and its access to financial services to the vast majority fractionated domestic market currently undercut of the population especially those living in the the advantages that could be bestowed by size. rural areas.  Access to financial services in Ethiopia is improving but remains low by world standards. Offsetting the factors underpinning Ethiopia’s Financial access remains particularly poor in rural growth prospects are a number of ‘growth head- areas where only six percent of the adult population winds’ that would need to be overcome to sustain is served by rural-focused institutions. the growth momentum. These can be divided into  Low levels of human capital and IT use. two categories: exogenous factors; and those that are Ethiopia is ranked 172 out of 189 countries in endogenous and reflect historical legacies and past policies. 41 See Wacziarg and Welch (2008), Baldwin (2011), and, Baldwin and The exogenous factors include: others (1998). 42 Growth of landlocked countries trailed that of coastal ones until recently however, in the past five years, landlocked countries have The geographical disadvantages of being a land- outperformed coastal ones by 1.5 percent per annum (Hostland and Giugale, 2013). See also Collier (2006), Hausman (2001), and Sachs, locked, resource poor country in a relatively slow Mellinger and Gallup (2001). growing, conflict prone neighborhood.42 43 There are at least four contributing factors: the lingering after effects of the financial crisis; ageing of the workforce and its concomitant de- Slower growth of trade partners among devel- cline in many countries; higher energy prices; and the possibility that oped economies43 and a revival of manufacturing technological change might be slowing. 44 See Reimers and Klasen (2013) for recent cross-country empirical in these countries that dampens the performance evidence and Weir (1999) for Ethiopia. of middle income countries and could restrain both 45 World Bank (2014). Managing Growth Expectations 83 the UNDP Education HDI Index (mean years 2001 to 9.7 percent in 2010. It subsequently declined of schooling were 2.4 years) and exhibit pervasive to 9.0 percent in 2014 and is expected to decline fur- technological backwardness (including computer ther in the 2015–17 projection period. Capital growth use and internet access) that impinges upon the has been exceptionally high the past years and is thus productivity and international competitiveness likely to slow down. A rising working age population of all sectors. provides some growth impetus, but total factor pro-  Weak transport infrastructure, and slow prog- ductivity growth will be hard to sustain at its current ress at trade facilitation practices and easing high levels. Since TFP growth (estimated using an HP foreign exchange restrictions on businesses con- filter for 1970–2012) was exceptionally high during straints that discourage both domestic private and the growth acceleration, this would suggest gradually foreign investment. Trade logistics is a significant declining TFP growth going forward (owing to the hindrance because the time to export is 44 days assumed mean reverting process of TFP growth). and costs are high, especially those related to docu- A specific cyclical factor relates to the ongo- ment preparation. ing construction boom. A decomposition of recent GDP growth reveals that the contribution of con- Most of these ‘inhibitors’ do not pose insur- struction has averaged about 20 percent the past 3 mountable hurdles but collectively they may years (2011/12–2013/14) compared to an average of dampen Ethiopia’s chances of maintaining its just 5 percent in the eight years prior. International growth rate over the course of the next decade. experience suggest that construction booms are cycli- Ethiopia can minimize the disadvantages of being cal phenomena. As demand for housing and office landlocked through further investment in multimodal space rises, construction activity with medium term transport solutions, by increasing the efficiency of all gestation periods starts gradually rising to augment transport modes and raising the quality of the logistics the supply. However, this process often overshoots so as to enhance tracking and timeliness. Each of the leading to excess supply followed by declining prices. other constraints can be eased through the application Lower prices, in turn, puts a halt to construction of policies many already in effect. activity. A key unanswered question is what drove the unprecedented construction boom over the past three 6.4  Cyclical Factors years in Ethiopia? Of particular interest is whether it was public or privately driven. Estimates from 2005 A key questions is the extent to which the current suggest that about half of construction value added is growth acceleration is a cyclical or a permanent private and half is public (PSD Hub, 2010). If these phenomenon. In this section we present three argu- shares are approximately correct also today, then the ments suggesting that the cyclical component is size- growth rate may fall by one percentage points once able suggesting lower growth rates going forward. the private construction boom comes to an end. The First, projections of potential GDP growth public part of construction will continue as long as indicate a slowdown in the medium term on there is a sustainable policy in place to support public account of declining TFP growth. Figure 6.1.1 pres- infrastructure investment. ents the historical trend and projections for potential GDP growth since 2001. Potential GDP growth is 6.5 Benchmarking computed as a function of changes in the capital stock, labor force and total factor productivity along the lines We return now to the cross-country regression of the Solow decomposition presented in Chapter 1. model introduced in Chapter 3. The model can be Potential GDP growth increased from 6.8 percent in used as a tool to gain further insights about economic 84 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 6.1: Potential GDP Projection, Results of Benchmarking and Policy Simulations 1.Potential GDP and Components (growth rates) 20 18 16 14 12 10 8 6 4 2 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Capital stock Potential GDP TFP Working age population 2. Simulated Income Levels (SSA benchmark) 3. Simulated Income Levels (LMIC benchmark) 2,000 200 2,000 150 (Counterfactual) income level (Counterfactual) income level 1,500 150 1,500 % income change % income change 100 1,000 100 1,000 50 500 50 500 0 0 0 0 Actual Infrastructure Govt consumption Credit/GDP Openness Inflation Schooling Banking crises Actual Infrastructure Govt consumption Credit/GDP Openness Inflation Schooling Banking crises Counterfactual income % income change Counterfactual income % income change 4. Policy Simulations – Major Growth Drivers 5. Policy Simulations – Structural Growth Drivers 8% 3.5% 7% 3.0% 2.5% 6% 2.0% 5% 1.5% 4% 1.0% 3% 0.5% 2% 0.0% –0.5% 1% –1.0% 0% –1.5% –1% –2.0% A B C Late 2000s A B C Late 2000s External Stabilization Structural Infrastructure Government consumption Trade openness Persistence Growth Private credit Education Structural growth contribution Source: Author’s calculations. Managing Growth Expectations 85 TABLE 6.1: Benchmarking Structural, Stabilization and External Factors (2006–10 data) Lower Middle Income Countries SSA Variable Ethiopia 25% Median Average 75% 90% Real GDP per capita (PPP), US$ 599.3 1,443.3 2,257.4 2,247.3 3,575.6 7,389.2 Real Exchange Rate 39.7 40.0 47.7 49.2 60.2 40.7 Secondary enrolment (gross, %) 32.8 42.2 55.0 54.9 82.1 86.0 Private sector credit (% of GDP) 20.1 14.1 25.9 23.6 36.7 44.5 Trade openness 0.2 –0.2 0.1 0.1 0.3 0.4 Gov. consumption (% of GDP) 9.1 7.9 11.0 12.3 19.5 6.4 CPI Inflation (annual, %) 18.4 6.0 8.8 8.7 12.0 4.3 Telephone Lines (per person) 1.1 1.3 4.0 3.4 8.8 7.0 Mobile phones (subscriptions) 3.7 30.0 44.5 40.0 57.1 78.0 Roads 0.5 1.6 2.9 2.6 4.1 6.2 Banking crisis (0=no; 1=yes) 0 0 0 0 0 0 Polity2 (governance variable) 12.0 8.0 14.5 11.1 19.0 19.0 Note that the best performing country depends on the variable in question. The 75th percentile is generally the ‘best performer’ for LMICs, how- ever, for some variables that have a negative impact on growth it is the 25th percentile (exchange rate, government consumption, banking crisis and inflation). To facilitate interpretation we have underlined the relevant benchmark. Source: Authors’ calculation based on main data set. growth going forward. In this section we analyze the analysis shows that infrastructure improvements have impact of changing each policy variable at a time been important for economic growth, this does not holding all other variables constant. In the subsequent mean that Ethiopia’s infrastructure level does not lag section, we vary several policies at a time. behind and would not require further improvements. Which policies would bring Ethiopia the In fact, as shown in Table 6.1, the gap in infrastruc- Biggest Bang for the Birr? We address this question ture between Ethiopia and benchmark countries from by simulating alternative values for the explanatory Sub-Saharan Africa and with lower-middle income variables in the model—one by one. Specifically, we status is still substantial. The same is true for second- benchmark Ethiopia against Sub-Saharan African ary education, inflation, and financial intermedia- (SSA) and Lower-Middle Income (LMIC) peers. We tion. Compared to the median of low-middle income use the 90th percentile of best-performing SSA coun- countries and the best performers of SSA, Ethiopia tries (in each of the respective variables), and the 75th also underperforms on institutions. percentile of best-performing lower-middle income The most important gaps to close are not nec- countries. These benchmark values are depicted in essarily the largest ones, but the ones that matter Table 6.1. LMICs have a real GDP per capita between the most for raising the growth rate. As our econo- US$981 and 4,526 (PPP) in the Late 2000s period in metric model implies, some reforms will have a higher our data set, compared to US$599 in Ethiopia. We impact on growth than others. Our goal is thus to note that these benchmarks are ambitious, but they are identify those gaps which will bring Ethiopia the ‘big- chosen to better highlight the effects of benchmarking. gest bang for the birr’ by closing them. Building on Despite recent progress on some fronts, the approach by Araujo et al. (2014), the underlying Ethiopia still lags behind on various aspects that idea is that progress in areas where gaps are large is are important for growth. For example, while our easier to achieve and should receive relatively more 86 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT policy attention, while this should be balanced against especially the relevance of governance, which is usu- the potential payoffs in terms of income gains. The ally not very time-varying and thus hard to estimate results are depicted in Figure 6.1, showing the (coun- in fixed-effects models (which offer other advantages). terfactual) GDP per capita levels if Ethiopia reached Furthermore, human capital effects are usually difficult variable levels of the SSA (Figure 6.1.2) or LMIC to estimate in growth regressions. Finally, changes in benchmark (Figure 6.1.3) for a given explanatory a single policy variable would have to be traded-off variable. For infrastructure, we take the average effect with changes in other variables—a point we seek to of roads, telephone lines, and mobile phone subscrip- address in the subsequent section. tions. We calculate the counterfactual income by the period 2021–2025 assuming that Ethiopia closed its 6.6  Scenario Analysis gap today.46 Continued infrastructure improvements offer In this section, we use our model to gain further the single best growth prospect for Ethiopia, accord- insights about the future of Ethiopia’s growth. ing to benchmarking. This result is partly the product Specifically, we articulate three illustrative policy sce- of a substantial existing infrastructure gap and partly narios and assess their growth implications. Compared because of the high economic returns to infrastructure. with benchmarking, this approach has the advantage The simulated improvements imply that Ethiopia of incorporating trade-offs involved in changing mul- would have to achieve the same road coverage as Gabon tiple policy variables simultaneously. The exercise has (SSA) or Cote d’Ivoire (LMIC) or the mobile phone important methodological limitations implying that coverage similar to Mauritius (SSA) or Mauritania great care must be taken when interpreting results. (LMIC). Conversely, the estimated growth effects of Still, the simulations offer sufficiently meaningful education and trade openness are relatively small. insights about the future path of Ethiopia’s growth Increasing credit to the private sector and and the sources of growth under alternative policies. addressing macroeconomic imbalances also offer The three policy scenarios share a number of important growth dividend. If Ethiopia reached key background characteristics. We use Early 2010s the level of financial development (private credit to data as the base values and project one decade ahead GDP) as Zimbabwe (SSA) or Bolivia (LMIC), its for the Early 2020s. This is done to facilitate an evalu- simulated income level would be 15.1 or 11.7 percent ation of Ethiopia’s goal of reaching middle income higher, respectively. Considering that Ethiopia also status by 2025. For the external sector, we assume lags considerably behind in terms of price stability unchanged commodity export prices and terms of (inflation), this suggests that improvements in the trade. Similarly, we do not expect institutional changes country’s macro framework would potentially provide or the occurrence of a banking crises. Note that these further income gains. baseline assumptions, together with the persistence Certain caveats should be kept in mind con- effect of previous-period reforms, lead to a per capita cerning these results. The outcomes of this exercise growth rate of 3.1 percent. The individual scenarios should not be interpreted mechanically, i.e. simply are described below and summarized in Table 6.2 catching up with LMIC or SSA benchmark countries while Table 6.3 offer detailed results. (The results are in terms of infrastructure will not bring per capita also illustrated in Figure 6.1.5–6). income to a threefold level. For starters, this exer- Scenario A: ‘If it ain’t broken why fix it?’ This cise does not take into account other effects (such as question is legitimate since Ethiopia’s growth strategy external conditions). Furthermore, the nature of our model makes it more difficult to correctly identify 46 Note that an improvement today will not only have an impact in the some parameters compared to others. This concerns same period but also in future periods via the lagged dependent variable. Managing Growth Expectations 87 TABLE 6.2: Assumed Annual Growth Rates of Policy Variables by Scenario (percent) Late 2000s Early 2020s Base value A. Business as Usual B. Private Sector Reform C. Accelerated Public Investment 2000s Δ% Level Δ% Level Δ% Level Δ% Level Structural: Δln(schooling) 6.4 32.8 3.3 45.4 4.5 51.1 2.0 40.0 Δln(credit/GDP) –1.3 20.1 –1.3 17.7 1.3 22.8 –2.5 15.6 Δln(trade open) 2.0 0.2 0.0 0.2 0.7 0.21 0.0 0.2 Δln(govt C) –4.0 9.1 1.0 10.1 0.0 9.1 3.0 12.2 Δln(tele lines) 7.2 1.1 7.2 2.2 3.6 1.6 14.4 4.2 Stabilization: Δln(inflation) 5.2 18.4 0.0 18.4 –10.0 6.4 4.0 27.2 Δln(exch rate) 1.6 39.7 1.6 46.5 0.0 39.7 3.2 54.4 Note: The assumed percent changes do not imply that underlying percentage values change by the depicted percentage points, but by the percent- age. To illustrate, a 4 percent increase of an inflation rate of 10 percent leads to an inflation rate of 10.4 percent, not 14 percent. has delivered an unprecedented growth acceleration reflected in a modest increase in infrastructure and over the past decade. This is the spirit behind scenario unchanged government consumption (implying no A: ‘Business as usual’. To sketch this scenario, we additional growth drag from public debt). Increased assume continued public infrastructure investments fiscal space is used to finance an expansion in second- and improvements in education that are in line with ary education, which would then approach current developments observed over the last decade. Public average LMIC secondary enrollment rates in the investment projects come at the cost of private sector Early 2020s.47 Increased private sector development crowding-out in the credit market, the buildup of is facilitated by an increase in private credit to GDP, inflationary pressures due to supply constraints, and, which would then be slightly below the current LMIC a policy of continued real exchange rate appreciation average in the Early 2020s. Macroeconomic stabiliza- (to keep capital imports cheap). Public consumption tion leads to a decline of inflation, a realignment of increases in this scenario for the purpose of (imper- the exchange rate,48 and more trade openness which fectly) capturing the growth drag of accumulating reflects increased competitiveness from the exchange public debt to finance infrastructure investments (see rate and private sector development. Loayza et al. 2005). Scenario C: ‘If it worked so well the past, let’s Scenario B: ‘The strategy of the past may not do more of it in the future’. Scenario C represents bring growth in the future’. It is equally legitimate to ask whether the current growth strategy needs to 47 World Bank (2013) identifies additional resources as a key determinant be adjusted today to enable it to deliver tomorrow. of expansion of secondary education. The assumed values for education Scenario B simulates a reform which aims to pro- expansion in the 3 scenarios benefitted from input from Bank educa- tion specialists. mote accelerated private sector investment and reduce 48 We assume an unchanged exchange rate in this scenario, as the macroeconomic imbalances. Specifically, the pace of exchange rate is currently overvalued but due to a Balassa-Samuelson effect (assumed in all scenarios), an appreciation of the exchange rate is public infrastructure investment slows down but is expected in developing economies. The scenario thus assumes that those partially substituted by private sector involvement, two effects cancel each other out. 88 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE 6.3: Illustrative Scenarios and Growth Projections until Early 2020s Scenario A: Scenario B: Scenario C:   Business as Usual Private Sector Reform Accelerated Public Investment Predicted Predicted Predicted Parameter Change effect Change effect Change effect Persistence 0.786 0.045 3.14% 0.045 3.14% 0.045 3.14% Structural:     1.00%   1.05%   1.49% Δln(schooling) 0.023 0.033 0.11% 0.045 0.14% 0.020 0.06% Δln(credit/GDP) 0.072 –0.005 –0.13% 0.013 0.13% –0.020 –0.25% Δln(trade) 0.082 0.000 0.00% 0.007 0.08% 0.000 0.00% Δln(govt C) –0.266 0.020 –0.37% 0.000 0.00% 0.030 –1.11% Δln(tele lines) 0.140 0.072 1.40% 0.030 0.70% 0.120 2.79% Δln(institutions) –0.002 0.000 0.00% 0.000 0.00% 0.000 0.00% Stabilization:     –0.14%   0.17%   –0.35% Δln(inflation) –0.012 0.040 0.00% –0.007 0.17% 0.080 –0.07% Δln(exch rate) –0.063 0.025 –0.14% 0.000 0.00% 0.030 –0.28% Δln(bank crisis) –0.040 0.000 0.00% 0.000 0.00% 0.000 0.00% External: 0.00% 0.00% 0.00% Δln(TOT change) 0.117 0.000 0.00% 0.000 0.00% 0.000 0.00% Δln(commodity exp 10.507 0.000 0.00% 0.000 0.00% 0.000 0.00% prices) Predicted average annual 4.00% 4.35% 4.28% GDP per capita growth rate Source: Authors’ calculation. an accelerated public infrastructure investment sce- between the public infrastructure variable and the nario. While the direct growth benefits hereof are private credit-to-GDP ratio.49 As a result, we do not positive, as illustrated in the benchmarking exercise, claim internal consistency within each scenario nor this would need to be weighed against a number of do we pay great attention to the specific numerical growth-inhibiting factors, including private sector results. With those caveats in mind, we find the fol- crowding-out, more detrimental effects from elevated lowing results. public consumption and higher inflation, for the All three policy scenarios yield comparable real above mentioned reasons. GDP per capita growth rates of about 4 percent per The three scenarios are highly illustrative and year. While there are some nuances in the projected we do not claim a great degree of precision in their growth rate of each scenario, the differences are not identification. The exact numerical specification of each of the scenarios is fraught with a number of diffi- culties. Chief amongst these is the lack of quantitative 49 We tried estimating these using the available cross-country data, but this did not produce meaningful results. Instead, the choice of specific estimates of the policy trade-offs. To illustrate, we do numerical values were based largely on the historical values and relation- not know with precision the functional relationship ships observed in Ethiopia complemented with LMIC benchmark values. Managing Growth Expectations 89 substantial enough to merit attention. A number of A per capita growth rate of 4 percent is clearly not as insights emerge from this exercise. First, none of the high as the 6.5 percent rate Ethiopia achieved in the three policy scenarios stand out as superior to the Late 2000s. On the other hand, it is still one stan- others in terms of growth outcomes, suggesting that dard deviation higher than the global historical aver- either the model has limitations in terms of policy age since 1950 of 2 percent calculated by Summers guidance or that there are indeed alternative ways to and Pritchett (2014). It is also higher than any of the reach the same outcome. Second, the simulations sug- growth rates observed in Ethiopia in the post-WW2 gest that it will be challenging for Ethiopia to grow period. Thus, in an international context, a slowdown at a rate necessary to reach middle income status by in Ethiopia’s per capita growth rate to 4 percent still 2025. Third, projected per capita growth of about 4 represents a better-than-average long term growth percent is nonetheless a strong performance in light performance. of international experience. Finally, sensitivity analyses Sensitivity analysis suggest that the range of indicate a range of per capita growth outcomes in the projected future per capita growth rates lie in the interval of 3 to 6 percent per year in the three illustra- interval between 3 and 6 percent.51 The lower bound tive scenarios. We detail these points below. of this estimate is given by the persistence effect which All scenarios suggest that it will be challenging adds about 3 percentage points to growth. While it is for Ethiopia to grow at a rate necessary for reaching feasible to design scenarios with negative growth rates middle income status by 2025. Calculations under- arising from set-backs in terms of banking crisis, insti- taken by the World Bank (2013) show that Ethiopia tutions, the external environment or economic policy would need to grow by 10.7 percent per year until mix, we do not consider such scenarios realistic. The 2025 to reach middle income status, as measured by upper bound of about 6 percent, in turn, is derived GNI per capita (Atlas Method). This would be equiva- by assuming away any policy trade-offs under the lent to 8.0 percent real GDP per capita growth. Since accelerated public investment scenario C. the growth variable used here is log transformed, we note that the MIC target is equivalent to 6.5 percent 6.7 Summary per capita. None of the three policy scenarios come close to achieving this growth rate even under assump- We conclude by summarizing the insights of vari- tions of a more supportive external environment.50 The ous approaches to aligning our expectations of intuition behind this result is the presence of policy Ethiopia’s future growth. Table 6.4 presents an trade-offs in all scenarios: positive growth effects of overview. First, international experience suggest a high public infrastructure investment are outweighed likely range of 2 to 8 percent per capita. Second, the by negative effects from public consumption increases, historical experience since 1991 suggest a similar private credit and competitiveness. Moreover, where range. Third, simulation outcomes using our growth growth benefitted from low government consump- regression model suggests growth rates around 3½–4 tion in the past, all three scenarios assume that this variable would have to rise (or stay constant) in the future reflecting a higher public debt burden. In all 50 For example, an annual terms of trade improvement of 1.2 percent (similar to the historical average) would add only another 0.2 percentage cases, public consumption is unlikely to fall further points to the annual growth rate. going forward, implying that Ethiopia cannot rely on 51 The regression model allows for fairly flexible growth outcomes for Ethiopia within reasonable intervals. For instance, we derive a range of this variable as a growth driver like in the past. 0.43 to 10.6 percent per capita growth when every single variable grows at All scenarios, if realized, would nevertheless its least or most favorable pace observed over the last three 5-year intervals, respectively. Similarly, if we take the average of the median growth rate represent a remarkable growth performance for and the minimum or maximum over the same period, respectively, we Ethiopia by historical and international standards. get an interval of 3.7 to 8.8 percent for per capita growth. 90 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE 6.4: Summary of Growth Expectations for Ethiopia Real GDP Per Capita Real GDP Approach Lower bound Upper bound Lower bound Upper bound International experience 2.0 8.0 4.5 10.5 Historical (Ethiopia) 2.0 8.0 4.5 10.5 Simulations (outcome) 4.0 4.4 6.5 6.9 Simulations (sensitivity) 3.0 6.0 5.5 8.5 Overall 2.0 8.0 4.5 10.5 Note: Assuming population growth of 2.5 percent per year. percent with more extreme assumptions increasing of Harvard University for Ethiopia is 4.4 percent per this range to 3 to 6 percent. When combined, these year. The highest projection is made for India (7.9 alternative approaches suggest a plausible overall range percent) followed by a group of four East African of between 2 and 8 percent per capita growth. countries (Uganda, Kenya, Malawi and Tanzania) in How does this range compare to other avail- the 6.5–7.0 percent range. Ethiopia’s growth rate was able long term growth projections for Ethiopia? 27th of 124 countries. The authors use a measure eco- Growth projections using a measure of ‘economic nomic complexity that captures the productive capa- complexity’ fall within the lower bound of our pro- bilities embedded in a country’s exports (Hausmann, posed range. The decadal (2013–23) real GDP forecast Hidalgo et al, 2014). 91 ETHIOPIA’S FINANCING CHOICE: PUBLIC INFRASTRUCTURE OR PRIVATE INVESTMENT? 7 think through the opportunity costs of financing To sustain high economic growth, Ethiopia faces a critical choice: Should it continue to direct the bulk of domestic infrastructure. financing towards infrastructure or should it allocate more This chapter addresses the following questions: to support private investment? We argue that since firms Is Ethiopia’s current level of public infrastructure appear more constrained in credit than in infrastructure, it may be time to focus more on alleviating credit constraints. investment optimal vis-à-vis private investment? This view is supported by empirical estimates indicating What are the marginal returns to public infrastruc- relatively higher marginal returns to private investment in ture and private investment in Ethiopia? What is the Ethiopia. A range of policy options are considered. Within Ethiopia’s existing financial repression model, policy biggest constraint for Ethiopian firms: credit or infra- makers could simply avail more credit to the private sector structure? What domestic finance reform options are thereby resembling more closely the Korean model. Another relevant for Ethiopia if the objective is the enhance approach would be to initiate a process of interest rate finance to the private sector? If infrastructure remains liberalization, which would enhance the overall savings pool to the benefit of both infrastructure and private investment. so important for future growth, then how else can it Infrastructure investment is critical for growth, but it needs be financed? to be financed in a way that reduces trade-offs with private The chapter is structured as follows: Section 7.2 investment, which is equally critical for growth. estimates the marginal return to public infrastructure and private investment. Section 7.3 examines briefly examines the constraints faced by firms. Section 7.4 7.1 Introduction explores options for making more credit available to the private sector. Section 7.5 explores alternative Few policy choices are more critical and more con- options for financing infrastructure. tested in Ethiopia than the allocation of domestic credit to public infrastructure and private use. Is Ethiopia’s Current Level of 7.2  Ethiopia’s system of financial repression and quantity Public Infrastructure Investment rationing (as described in Chapter 2, Box 2.1) lends Optimal?52 itself to a relatively straightforward question: Suppose you had an additional unit of saving or credit and We begin by stressing the importance of infrastruc- you wished to get the highest growth return. Should ture development for Ethiopia’s growth. A key result you finance an infrastructure project, such as a road, from our regression model in Chapters 3 and 6 is that or, should you channel it to finance the expansion infrastructure has driven economic growth in the past of capacity in a private firm? The road project may and must continue to be an important driver of growth crowd-in private sector activity if an infrastructure in the future. This is because of the high economic bottleneck is alleviated. At the same time, it denies access to credit of another firm that is ready to expand and create jobs. Striking the right balance is a delicate 52 This section presents the key findings of a research paper prepared Eden and Kraay (2014a) and discusses its application to Ethiopia. Interested task and economic analysis can only take you so far. readers are encouraged to consult this work for further technical details However, such analysis can give help policy makers of the underlying theoretical and empirical models. 92 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT returns to infrastructure and the fact that Ethiopia’s Consider an example in which public and infrastructure deficit is one of the largest in the world. private capital are perfect substitutes. In this case, So when we ask the question of whether the current public capital will always crowd out private capital. level of public investment in infrastructure is optimal, However, this does not necessarily mean that public this is not a question of whether we need more infra- investment is undesirable, since it may be more pro- structure or not. Clearly, in the long run, we need ductive than private capital because of some com- much more. The problem, however, arises when there parative advantage that the government has relative is a financing constraint. Especially, when financing to the private sector (for example, the government infrastructure imposes a trade-off via domestic credit may have a comparative advantage in providing secu- markets to private investment financing. So to phrase rity through the police rather than private security the question more carefully: is Ethiopia’s current level services). Similarly, consider a case in which public of public infrastructure investment optimal compared investment fully complements or crowds-in private to the current level of private investment? To address investment. This does not necessarily mean that more this question we draw upon a comprehensive theoreti- public investment is desirable, since there is an optimal cal and empirical analysis contained in the background level of investment (both public and private), which papers prepared for this report by Eden and Kraay depends on the cost of financing. (2014ab) and Eden (2015b). From a policy perspective, the right question The recent surge in public investment in is whether the returns to public investment exceed Ethiopia has prompted renewed research interest the costs of financing. Optimality requires that the regarding the optimal scope of public investment. return to public investment is equated with the interest It gives rise to the question of whether Ethiopia is rate, and this is true regardless of the effect of public investing too much in public capital, or whether other investment on private investment. In Ethiopia, the developing countries investing too little. The work by supply of credit may be effectively fixed: the economy Eden and Kraay (2014a) suggests that both questions is relatively closed and the external borrowing con- can be answered affirmatively: the vast majority of straint of the government is somewhat binding. Under low income countries are investing too little in public a binding credit constraint, the government can be capital. However, it is likely that Ethiopia may be over- thought of as jointly choosing the bundle of public investing in public capital relative to the optimal level, and private investment, subject to an external credit which also takes into account the marginal returns to constraint. This is because the choice to increase pub- private investment. lic investment corresponds to a reduction in private An important aspect of this debate relates to investment through the increased availability of credit the potential spillovers from public investment for the private sector (i.e. the oft-cited crowding-out to private investment. Part of the benefits of public effect in Ethiopia). investment may come from raising the productivity of When a theoretical model is developed, it private investment. However, public investment may emerges that the marginal product of public capital also crowd out private investment, either by reducing should equal the marginal product of private capi- the supply of savings or by directly substituting for tal, in the optimum. To shed more light on whether private enterprises. While the relationship between public investment levels are optimal in Ethiopia, we public and private investment is important for deter- use a theoretical model for that is adequately calibrated mining the optimal scope of public investment, the with data relevant for Ethiopia. The parameters of relationship is more nuanced than simply crowding the model presented in Annex A7.1 were calibrated in vs. crowding out, which has tended to dominate using empirical estimates of (a) the extent to which policy discussions in Ethiopia. public investment increases private investment, and Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 93 (b) the extent to which public investment increases FIGURE 7.1: Public and Private Investment in output. Estimates were derived from a sample of 39 Low Income Countries (1980–2012) IDA-eligible low-income countries using a small open 0.15 GUY economy framework.53 Intuitively, the response of Public Investment (% of GDP) MOZ private investment is informative regarding the degree of substitutability between public and private invest- 0.10 ETH MWI DJI ment (σ): if they are more substitutable, public invest- MDG RWA SEN MLI BDI BGD NPL MRT HND NIC ment will have a larger crowding out effect on private COM BOL BFA BEN TGO KHM YEM GHAUGA COG MDA 0.05 CAF UZB SLE CIV NER GIN PAK investment (or a lower crowding in effect). Similarly, TZA KEN LKA ARM SDN the higher the estimate for the relative productivity of CMR public investment (γ), the larger is the first-order effect 0 on output. It turns out that the parameters σ and are 0.05 0.10 0.15 0.20 0.25 0.30 jointly determined by the equilibrium responses of Private investment (% of GDP) private investment and output. Source: Eden and Kraay (2014a). To estimate the effect of public investment on private investment and output, the exogenous vari- ation from predetermined disbursement of loans countries with countries with higher public invest- from official creditors is exploited. Empirically, ment rates also having lower private investment rates. focusing on changes in public investment induced The empirical analysis finds general evidence by predetermined disbursement of official loans is of ‘crowding-in’ of public investment among low- useful for establishing a causal relationship, because income countries. The main regression results derived it abstracts from changes in public investment that in Eden and Kraay (2014) are summarized in Annex are either (a) responses to changes in private invest- 7.2. On average, an extra dollar of public investment ment, or (b) responses to changes in the economic raises private investment by roughly two dollars and environment that affect both public and private invest- output by 1.5 dollars. These results are derived from ment. Using the predetermined disbursement as an a two-stage least squares estimates and subjected to instrument for public investment, two variables are various sensitivity tests. The calibrations show a strong estimated: the marginal increase in private investment degree of complementarity between public and private induced by a change in public investment (β), and, capital. However, while the estimates are positive, the the marginal increase in output induced by a change analysis produces large standard errors. The 95 percent in public investment (βy). Data on predicted disburse- confidence intervals are large in both cases. For that ments of loans from official creditors is taken directly reason, it is important to consider a wider range of the from Kraay (2013), while GDP and investment data parameter estimates in the calibration of the model is from the World Bank and the IMF, respectively, and beyond the point estimates. originally based on national sources. The estimated marginal returns of public Ethiopia stands out among low income coun- investment in Ethiopia are among the lowest tries for having a relatively high public investment observed in low income countries. We define rate and a relatively low private investment rate. ‘the excess return’ as the marginal impact of public Figure 7.1 plots the average investment ratios for investment on output less the sum of the interest the 1980–2012 period. Public investment rates vary rate and the rate of depreciation. In this exercise, widely across countries ranging from as low as 2.5 percent of GDP in Cameroon to nearly 15 percent 53 This assumption is appropriate for most countries in the sample, in Guyana. There is also a weak correlation across although it is somewhat less appropriate in the case of Ethiopia. 94 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE 7.1: Estimated Excess Returns of Public Investment λ (ratio λ (ratio of private of private and public Excess and public Excess Rank Country investment) returns (%) Rank Country investment) returns (%) 1 Cameroon 7.31 1074.88 2 Sri Lanka 5.39 506.36 21 Mali 2.22 47.16 3 Armenia 5.39 505.51 22 Cambodia 1.96 31.51 4 Sudan 5.13 448.69 23 Senegal 1.87 26.78 5 Moldova 4.91 401.68 24 Benin 1.78 22.32 6 Nicaragua 4.24 277.37 25 Burkina Faso 1.77 21.83 7 Burundi 4.09 253.29 26 Togo 1.75 20.96 8 Kenya 3.85 216.11 27 Bolivia 1.73 19.79 9 Uzbekistan 3.75 202.69 28 Niger 1.68 17.61 10 Tanzania 3.70 194.92 29 Sierra Leone 1.43 7.71 11 Honduras 3.43 160.52 30 Djibouti 1.39 6.38 12 Congo 3.32 147.06 31 Côte d’Ivoire 1.38 5.83 13 Mauritania 3.32 146.60 32 Rwanda 1.35 5.00 14 Guinea 3.00 112.08 33 Madagascar 1.19 0.26 15 Uganda 2.95 107.07 34 Malawi 1.17 –0.36 16 Nepal 2.90 102.38 35 Mozambique 1.06 –3.11 17 Pakistan 2.85 97.66 36 CAR 0.99 –4.64 18 Ghana 2.70 83.42 37 Ethiopia 0.92 –5.93 19 Yemen 2.61 76.37 38 Comoros 0.88 –6.59 20 Bangladesh 2.33 54.61 39 Guyana 0.78 –8.28 Note: β =2 and βy=1.5 World Bank data for public and private investment. Source: Eden and Kraay (2014). a 3 percent interest rate and a 10 percent deprecia- percent relative to the current size of the public tion rate is used. For most low income countries in capital stock. This result is derived by comparing the 39 country sample, the excess returns to public the optimal ratio of private to public capital of 1.18 investment fall in the range between 5 and 170 per- derived in Annex A7.1.3 with alternative estimates cent, as illustrated in Table 7.1. However, for some of Ethiopia’s actual ratio of private to public capital countries that already have high public investment from two different data sources. In the first approach, rates, the return to further investment is below the World Bank data for public and private investment is world interest rate (implying negative excess returns). used as a proxy resulting in an actual private to public The excess return public investment in Ethiopia is capital ratio of 0.92 (Table 7.1). Using the formulas estimated at –5.93 percent and the ratio of private of the theoretical model, it can be concluded that the to public investment is 0.92. public capital stock is 12 percent too high in Ethiopia. The empirical analysis further indicates that The second approach uses the Penn World Tables Ethiopia’s growth performance could benefit (PWT) estimates of public and private capital stocks from increasing its private capital stock by 12–20 for Ethiopia and yields a private to public capital ratio Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 95 FIGURE 7.2: Ethiopia: Public and Private Investment and their Returns 1. Public and Private Investment (% of GDP) 2. Estimated Marginal Product of Capital (%) 0.25 0.50 0.45 0.20 0.40 0.35 0.15 0.30 0.25 0.10 0.20 0.15 0.05 0.10 0.05 0 0 1985 1990 1995 2000 2005 2010 1985 1990 1995 2000 2005 2010 Public investment Private investment Note: Figure 7.2.1 uses World Bank data. Figure 7.2.1 uses PWT data. Source: Eden and Kraay (2014). of 0.75, which (using the relevant formula) suggest (2015b) shows that if the rate of return on private that the stock of public capital is 20 percent ‘too high’. capital is greater than 3 percent, then the marginal The results reflect the fact that the marginal return to public investment is lower than the mar- product of private capital is substantially higher ginal return to private investment. The approach than public capital in Ethiopia. Figure 7.2 supports builds on the methodology developed in Caselli and the intuition behind these results. Since the early Feyrer (2007), which decomposes national income 2000s, the public investment rate in Ethiopia has into labor and capital income. Given an assumption soared, while private investment gradually declined on the private rate of return to investment, capital (Figure 7.2.1). Using the estimated production func- income can be decomposed into income attributed tion, the marginal product of public and private capi- to private capital and income attributed to public tal can be computed using PWT data. The marginal capital. A higher assumed rate of return on private product reflects the effect on production from a one capital implies a lower estimated rate of return to dollar increase in either form of capital expressed in public capital. The advantage of this approach is that percent. Figure 7.2.2 reveals that marginal products it measures the returns to public investment without of public and private capital were roughly equalized relying on cross-country analysis. The disadvantage is in the 1987–2003 period, implying that an adequate that the results are sensitive to assumptions regarding balance was struck between public and private invest- the labor income share, which is assumed to be 0.67 ment. Starting in 2004, as the Ethiopian economy (as in countries comparable to Ethiopia). took off, the marginal product of private capital increased substantially, while the marginal product 7.3 Firm-level Constraints: of public capital continued to decline. In 2011, the Infrastructure or Credit? marginal product of private investment was 22.5 percent compared to the marginal product of public Firms depend on a range of high quality inputs to investment of 7.5 percent. be competitive. Macroeconomic and political stabil- Similar results can be obtained from a meth- ity together with a conducive business environment odology that relies only on Ethiopia data. Eden are some of the basic ingredients. Firms also become 96 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE 7.2: Most Binding Constraints to Doing Business in Ethiopia, Various Rankings Global Consultations Competitiveness Large and Enterprise on National Doing Index 2014– Non-Farm Medium Scale Survey Business Agenda Business 2015 Enterprises Manufacturing 2011 2015 2015 Constraints 2013/14 2012/13 1. Credit Taxes Starting a business Government Access to markets Raw materials Bureaucracy 2. Land Credit Credit Foreign exchange Credit Access to markets 3. Energy Land Trade logistics Credit Trade logistics Credit 4. Taxes Energy Protecting minority Corruption Taxes Energy investors 5. Trade logistics Unfair competition Registering Energy property Source: World Bank Doing Business Report (2016); Global Competitiveness Report (2014 and 2015); National Business Agenda (2014); World Bank Enterprise Survey (2011); CSA LMSMI (2012/13); GHS (2013/14). Note: Language has been harmonized across surveys for ease of compa- rability. Credit in bold. Infrastructure underlined. more productive with higher quality infrastructure existing distribution network and the establishment such as roads, energy, telecom and water. In addition, of dedicated power lines to industrial parks. firms rely on a range of financial services, including In Ethiopia, firms that are fully credit con- access to credit, foreign exchange and insurance. They strained exhibit poorer performance and productiv- also need a range of business services (accounting, ity. According to World Bank (2014), firms in Ethiopia auditing, financial advice) to thrive. At any point in are more likely to be fully credit constrained than global time, one or several of these inputs will be the bind- comparators, including SSA countries. Nearly half of ing constraint that prevents a firm from growing or firms in Ethiopia are fully credit constrained.54 For firms, being competitive. being credit constrained means poorer performance Ethiopian firms appear to be more constrained and less productivity. In Ethiopia, a credit constrained in terms of access to credit compared to infra- firm has 15 percentage points lower sales growth, 5 structure. Table 7.2 draws upon six different surveys percentage points lower employment growth, and 11 which shed light on the constraints to doing busi- percentage points lower labor productivity growth than ness in Ethiopia from the perspective of firms. It is firms who are not credit constrained. Instead, invest- noteworthy that access to credit is mentioned as a ment decisions of manufacturing and services firms in greater concern or obstacle for doing business than Ethiopia are heavily dependent on cash flows. infrastructure (energy and trade logistics) across all This is indicative that the economy would ben- six surveys. Additional infrastructure investment may efit from a shift of domestic credit towards private also only address firm needs partially. Good trade firms. If the ultimate purpose of government policy logistics outcomes is a function of ‘hardware’ (roads and rail), but also importantly of ‘software’ (e.g. cus- 54 Fully credit constrained firms are those without external financing and toms procedures). Reliable energy supply for firms which were either rejected for a loan or did not apply even though they needed additional capital. It should be noted that credit constraints can depends not just on total energy generation capac- be a function of both lack of overall credit in the system, and, a reflection ity, but also on investments and rehabilitation of the of firm characteristics (some firms are not credit worthy). Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 97 is to enhance the productivity of private firms, then nor does it require liberalization of the capital account. it is important to understand what the firm level con- It is also noted that the two policy reforms analyzed straints are. If firms really need credit more than access here can be considered as pure extremes on a contin- to new roads or better telecommunication to grow and uum of policies and that they can be mixed as needed. prosper, then government policy would need to sup- A shift towards more private sector credit within the port the alleviation of the credit constraint at the firm existing financial repression model would see public level. Since public infrastructure investment is partially investment fall by a similar magnitude. The deposit financed via the same domestic savings pool, it is clear and lending rates of banks would remain unchanged. that infrastructure financing competes directly with For simplicity of exposition it may be helpful to think the financing of private investment projects. In addi- of savings as being relatively unchanged, even if reality tion, there may be other factors affecting firm credit can be more complex.56 constraints, such as high collateral requirements, Financial repression with a strong private sector which the government can also seek to alleviate. emphasis is a model that was effectively practiced by South Korea. Like Ethiopia, the South Korean 7.4  Domestic Finance Reform government intervened extensively in the pricing and allocation of credit. Unlike Ethiopia, the Koreans The trade-off between financing public infrastruc- directed the bulk of the credit towards priority private ture versus private projects arises from the way in sector activities. ‘Specifically, it ensured that priority which the Ethiopian financial sector is designed. sectors, mainly export-oriented industry such as steel, As explained in Chapter 2, key characteristics include: electronics, ship-building, automobile manufacturing (1) below market-clearing interest rates; (2) market etc., received preferential treatment as far as access to dominance by a state-owned bank, Commercial Bank inexpensive bank credit was concerned’ (Demetriades of Ethiopia (CBE), which effectuates a policy of giv- and Luntel, 2001). Private sector credit to GDP aver- ing public investment projects the funding priority. aged around 30 percent in Korea in the 1970s com- In the economic literature, such a system is referred pared to around 20 percent in Ethiopia today (Nguyen to as ‘financial repression’ (McKinnon, 1973; Shaw, et al (2015). While Ethiopia also favors some priority 1973). The literature itself is divided on to whether private sector activities, particularly in manufacturing, financial repression is good or bad for economic the bulk of total domestic credit is currently directed growth as discussed in Box 7.1. In addition, it is towards public infrastructure projects. To truly emu- important to note that the capital account is closed late Korea in this aspect, Ethiopia would need to shift and that there are no foreign banks operating in the more of credit to priority private sectors. Ethiopian financial sector. Introducing a more liberalized financial sec- In this section we consider two policy reforms tor with market-determined interest rate would be that could help alleviate the private sector credit constraint in Ethiopia. One policy reform would be 55 As more domestic credit is directed towards the private sector, it is to continue the existing system of financial repression, particularly important to serve the small and medium enterprise (SME) but to direct more credit towards private firms at the segment. Ethiopia is characterized by a ‘missing middle’ phenomenon, whereby small enterprises are more credit constrained than either micro expense of public infrastructure projects.55 Another or medium/large enterprises. This represents a key challenge because policy reform involves a gradual move towards a more typically young firms are a great source of job creation but this trend is not seen in Ethiopia, where more established firms dominate the net job liberalized financial system in which interest rates creation, suggesting that there is a lack of competitiveness and innovation reflect the demand and supply for savings/credit. The in the private sector (See World Bank, 2014, for details). 56 In the theoretical model developed later, savings may change even if latter approach does not necessarily imply opening up the interest rate does not because future consumption depends on the the domestic banking sector to foreign competition mix of public and private investment. 98 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT BOX 7.1: Literature Review: Does Financial Repression Inhibit or Facilitate Economic Growth? Economists are divided on the impact of repressive financial policies on economic growth. A large number of studies identify possible mechanisms through which financial liberalization promotes growth, including facilitating financial development, improving allocative efficiency, inducing technological progress and enhancing financial stability (Shaw, 1973; Levine, Loayza and Beck, 2000). Many empirical analyses also confirm this positive correlation (Levine, 2005; Trew, 2006). Other studies, however, cast doubts on this relationship. Kose et al. (2009) find no clear-cut relationship between financial globalization and economic growth in a global dataset of emerging market economies. Stiglitz (2000) attributes increasing frequency of financial crisis during past decades to financial liberalization in the developing world. He argues that developing countries might be more able to manage money supply and financial stability under repressive financial policies. While repressive financial policies would reduce economic efficiency, they might enable the authorities to better deal with problems of market failure and financial risks. This lead some authors to argue that this is essentially an empirical question, influenced by a list of factors such as conditions of financial institutions, markets, regulators and the government. Case study analysis, especially of East Asian countries tend to be supportive some financial repression, especially at the early stages of development. In China, Huang and Wang (2011) find empirical evidence that repressive polices helped economic growth in the 1980s and 1990s thanks to the prudent liberalization approach. However, the effect turned negative in the 2000s as lending to the state sector, interest rate regulation and capital account controls constrained growth. Demetriades and Luintel (2001) develop a theoretical model which predicts a positive association between financial development and the degree of state control over the banking system and mild repression of lending rates. They use data from Korea to derive empirical findings that are consistent with this theoretical prediction. Ang (2009) studies the links between financial policies and private investment. In the case of Malaysia, high reserve and liquidity requirements exerted a positive influence on private investment. However, in the case of India, the effect was negative. Recent cross-country panel regression results are largely supportive of financial sector reforms. A large body of literature suggests that a well-developed financial sector promotes economic growth (e.g. Levine, 2005). However, relatively few studies tries to assess the impact of financial sector reforms on economic growth. Bakaert, Havery, and Lundblad (2005) finds that foreign equity ownership increases growth. Quinn and Toyoda (2008) document that capital account liberalization is positively associated with growth. Prati et al. (2013) finds strong positive effects of financial sector reforms on growth. Finally, Cristiansen et al. (2013) find that domestic financial reforms are robustly associated with growth. Some authors argue that a time series approach is more fruitful than a cross-section approach, though the use of instruments could potentially overcome this methodological criticism. Arestis and Demetriades (1997) worry that the question of causality cannot be satisfactorily addressed in a cross-section framework. Their criticism, in turn, has given rise to the series of cases studies mentioned above. They show uni-directional causality in Germany from financial development to real GDP , but insufficient evidence of such effects in the US, with abundant evidence of reverse causality. While modern econometrics has developed in recent decades, the problem of causality (or more generally, endogenity) continues to be a challenge in this sub-set of the cross-section / panel approach in this literature, which implies that it often refers to association rather than causality between variables. The work by Prati et al. (2013) represents an attempt to overcome this challenge as the authors use the Difference GMM estimator which relies on internal instruments in the form of lagged dependent variables. Source: Author’s compilation. expected to induce higher savings. Through a grad- It would no longer be possible for the government to ual increase in the deposit interest rate, households direct credit to the public infrastructure sector (neither and firms would find it relatively more attractive to via loans nor bond purchases). use formal deposit accounts for their informal savings. The exact consequences of these two reforms The expected result would therefore be an increase are hard to predict with accuracy. Under interest in deposits. The ensuing competition between banks rate liberalization, we would expect private invest- would drive up the deposit rates and attract savings. ment to rise. The deposit rate would increase, attract The magnitude of this increase depends on the elastic- more savings and there would be no targets for public ity of domestic savings with respect to the deposit rate. infrastructure investment. Higher interest rates will Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 99 TABLE 7.3: Predicted Effects of Financial Sector Reforms (1) Financial repression with more (2) Predicted Effect private credit Interest rate liberalization Private investment + + Public investment − ? Total investment 0 + Savings 0 + Financial repression revenue − −− Tax revenue from private activity + + Rents of savers 0 + Public debt sustainability 0 − improve the allocation of capital within the private reform, government would maintain some implicit sector as private projects with low returns will cease revenue as interest rates are unchanged. However, to demand credit, freeing up additional credit for implicit revenues would fall as less public projects are more productive projects. What would happen to financed. Under the liberalization approach, govern- public investment under a more liberal model? Given ment financial repression revenues would eventually higher borrowing cost, it is advisable to reduce pub- disappear as interest rates approach their market-based lic investment and finance a greater share of it from levels. Finally, since financial repression involves a other sources, such as higher tax revenues. Finally, it transfer of resources from savers to borrowers the is expected that total investment would rise given that effects differ under each reform. Financial repression total savings have increased. Table 7.3 summarizes the continues to involve a resource transfer from savers, predicted effects. but now directed to private firms instead of govern- The two policy reforms would have different ment. Interest rate liberalization involves a transfer implications for government finances and the from government back to savers. surplus of savers. To begin with, it is important to The effects on the sustainably of public debt realize that financial repression yields implicit govern- can be challenging under a more liberalized sys- ment revenues, as documented by Giovanni and de tem, while continued financial repression keeps Melo (1993). In Ethiopia, the revenue arises because this challenge at bay. It is important to be aware government infrastructure spending can be financed that current public debt levels are deemed sustainable at below-market interest rates giving rise to interest under the assumption that real interest rates do not rise payment savings. Moreover, if credit is rationed in (IMF and World Bank, 2015). However, since inter- favor of infrastructure projects then government gets est rate liberalization implies a rise in the real interest access to additional credit which it would not have had rates, it is clear that there would be concerns about access to in a market-based system. On the other hand, whether public debt can remain sustainable under a both policy reforms support a much more dynamic more market-based system of setting interest rates. private sector which would ultimately enhance the tax DSA simulations suggest that an increase in the base and yield higher tax revenues. Which of the two real interest rate by 7 percent would make public effects dominate is ultimately an empirical question. debt unsustainable. The current baseline assumption In the ‘financial repression with more private credit’ of the DSA is a real interest rate of –5.7 percent. If the 100 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT real interest rate increases to a positive level of 1.3 per- 2, Ethiopia has experienced a demonetization trend cent then there would be a substantial and protracted over the past decade as reflected by a declining share breach of the PV Debt-to-GDP ratio in the baseline of total credit to GDP. Given Ethiopia’s preference scenario. The associated increase in interest payments for financial repression, the less radical reform may be for the Federal Government and SOEs would be from to maintain this system, but to follow South Korea’s 0.5 percent of GDP to 3.4 percent of GDP. footsteps and direct the bulk of the credit to the pri- Given that each policy reform has pros and vate sector. cons, neither of them is superior to the other. Policy makers that weigh in concerns about debt sus- 7.5 Complementary Infrastructure tainability and find a gradualist approach to reform Financing Options appealing would find financial repression with more private credit most attractive. On the other hand, Continued infrastructure development remains one policy makers concerned about Ethiopia’s demoneti- of Ethiopia’s best strategies to sustain growth, but zation trends and who want to see a strong expansion the current financing model may not be sustain- of credit to the private sector would find liberalization able. Infrastructure was the most important driver more beneficial. If so, such an option would need to of economic growth during the growth acceleration. be combined with additional taxation of the private This is because the economic returns to infrastructure sector which would help address implicit govern- are high and the physical infrastructure expansion in ment revenue losses and the challenges of public debt Ethiopia was substantial. But infrastructure expansion sustainability. was financed via a range of mechanisms that will begin In theory, the policy choice should be informed to show their limits in the future. by two criteria: the relative returns of public and Low domestic resources mobilization, includ- private investment, and, the savings rate. Box 7.2 ing savings and tax revenue is a chief vulnerability. summarizes a simple partial equilibrium model devel- To overcome it and deliver high public infrastructure oped by Eden (2015) specifically for the purposes of investment in the past, policy makers engaged in a this study. The model provides a useful framework series of creative financing mechanism. Going for- within which to analyze Ethiopia’s current situation. ward, inherit policy trade-offs will eventually catch In particular, it becomes clear that financial repression up with this strategy. Public and external indebted- with more private emphasis becomes attractive in situ- ness are gradually rising as is the cost of financing and ations where the marginal return to private investment risks of debt distress. The lack of access to credit and is much higher than the marginal return to public foreign exchange of the private sector holds back an investment. Moreover, if the saving rate of the coun- important driver of growth. An overvalued exchange try is quite low, then the model suggests that welfare rate hurts external competiveness. At some point one would enhance by increasing the deposit rate towards of these constraints will become binding and limit more market-determined levels as in the liberalization Ethiopia’s ability to deliver public infrastructure with reform. Ultimately, which of the two constraints are the current model. Going forward, Ethiopia needs more binding would be an empirical question. more infrastructure, but it would need new mecha- The available empirical evidence suggests nisms to finance it. that Ethiopia has a challenge in both dimensions, In this section we briefly review the range of implying that both types of reform would enhance alternative infrastructure financing options avail- welfare. First, as shown in Section 7.2, marginal able to Ethiopian policy makers. In doing so we returns to private investment appear higher than those distinguish between policy proposals that are broadly of public investment. Second, as shown in Chapter in line with existing government strategy and thinking Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 101 BOX 7.2: A Theoretical Model of Financial Repression and Interest Rate liberalizationa A theoretical framework was developed to support the analysis of Ethiopia’s financial sector reform options. This approach has the advantage of ensuring logical consistency of argument, clarifying the underlying assumptions while also yielding additional insights that may not be immediately intuitive. To illustrate the key ideas, we utilize the two period version of a more general model developed by Eden (2015) for this report. The main features of the model are described below. The model has two periods and four agents: Households, firms, government and the government bank. Output depends positively on private capital and infrastructure (as complementary inputs) and exhibits diminishing returns. Private capital is produced by firms. They borrow from the government bank in the first period to finance capital in the second period and maximize profits given the borrowing interest rate and the returns to capital (which depend positively on infrastructure). Supply of savings. Households receive income from firms and pay taxes. Utility is derived from consumption in both periods and is maximized taking into account the subjective discount rate and the deposit interest rate. In equilibrium, household savings depend positively on the deposit rate. The government has four policy instruments: infrastructure, taxes, the policy deposit rate, and, the lending rate. The government objective is to maximize household welfare. It collects lump sum taxes in both periods and there is a cost associated with collecting taxes. The model has three interest rates. The policy rate is an upper bound on the deposit rate faced by households. The deposit rate is the return on household saving and represent the borrowing cost to the government bank. Under financial repression, the policy rate is equal to the deposit rate. If there is no financial repression then the policy rate can be higher than the decentralized equilibrium rate. The borrowing rate is what firms pay to borrow and it represents a revenue for the government bank. The spread is equal to the borrowing rate less the deposit rate and is a net revenue for government. The optimal policy consists of some financial repression because this reduces the costs associated with collecting taxes. The lower deposit interest rate (equal to the policy rate) associated with government debt allows the government to economize on tax collection costs, in both periods. In optimum, the marginal return to private capital is equal to the returns of infrastructure, net of the marginal costs associated with raising taxes (in period 2). The model abstracts away from costs associated with the collection of private debt. If these costs are equal to tax collection costs, then it is optimal to equalize the returns to government and private capital, as in Eden and Kraay (2014ab). Optimality requires that the marginal costs of taxation are also equalized across periods. In optimum, the government sets a positive interest spread. Given the depressed deposit rate, the government can either ration credit to the private sector or set a spread so that the market for private credit clears. The latter alternative is superior because the spread generates revenues for the government without further distorting private investment. This allows the government to economize on costs associated with tax collection. Furthermore, the price mechanism guarantees that private credit is allocated to the private projects with the highest return. Credit rationing, on the other hand, requires some guess work regarding where are the highest-return projects: since the interest rate is depressed, inefficient projects may find it optimal to request financing. At a higher borrowing rate, only the more productive projects will be profitable. Policy Scenario 1 (financial repression with more private credit) can be illustrated in the model by simulating alternative values of the government infrastructure variable. The properties of the theoretical model can be illustrated by assuming plausible functional forms, including Cobb Douglass production and utility functions and a quadratic tax collection cost function. Figure 1 illustrates the effect of alternative values of government infrastructure (kg) on other variables in the model. There is an optimum level of government infrastructure, kg*, namely the level that maximizes welfare. For kg < kg* there is insufficient crowding-in of infrastructure and production and consumption is too low. For kg > kg* infrastructure crowd-out private capital via the credit markets and this lowers production and consumption as well. The following additional results hold: First, the future tax burden is rising in government infrastructure. Second, any decrease in government infrastructure will be compensated by a corresponding increase in private capital. Third, the deposit interest rate is rising in infrastructure up to the optimum, but unaffected thereafter. An increase in private credit would be welfare enhancing if there is currently too much government infrastructure. In Figure 1, this situation can be illustrated by considering a situation where we are at a point to the right of optimum and reduce the level of government infrastructure investment. As also illustrate, this would result in an increase in private capital. As long as government infrastructure is not reduced too much, welfare is enhanced as more private capital can be invested. At the margin, additional private investment boosts production and consumption more than the resulting loss from lower infrastructure investment. (continued on next page) 102 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT BOX 7.2: A Theoretical Model of Financial Repression and Interest rate liberalizationa (continued) FIGURE 7.3: Numerical Simulations with Alternative Values of Government Infrastructure Welfare Interest rate –1.55 0.15 0.10 –1.60 0.05 –1.65 0 Welfare r –1.70 –0.05 –0.10 –1.75 –0.15 –1.80 –0.20 0 0.05 0.10 0.15 0.20 0 0.05 0.10 0.15 0.20 kg kg Future tax burden Borrowing constraint 0.15 10 8 0.10 MPK – (1 +rf) 6 0.05 T 4 0 2 –0.05 0 0 0.05 0.10 0.15 0.20 0 0.05 0.10 0.15 0.20 kg kg Private capital: second period Output: second period 0.04 0.44 0.43 0.03 0.42 0.41 0.02 kp y 0.40 0.01 0.39 0.38 0 0.37 0 0.05 0.10 0.15 0.20 0 0.05 0.10 0.15 0.20 kg kg Consumption: first period Consumption: second period 0.48 0.44 0.43 0.46 0.42 0.44 0.41 c c 0.42 0.40 0.39 0.40 0.38 0.38 0.37 0 0.05 0.10 0.15 0.20 0 0.05 0.10 0.15 0.20 kg kg (continued on next page) Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 103 BOX 7.2: A Theoretical Model of Financial Repression and Interest rate liberalizationa (continued) FIGURE 7.4: Numerical Simulations with Alternative Values of the Deposit Rate Welfare Interest rate –1.5848 0.115 –1.5849 0.110 –1.5850 0.105 Welfare –1.5851 0.100 r –1.5852 0.095 –1.5853 0.090 –1.5854 0.085 0.08 0.09 0.10 0.11 0.12 0.13 0.08 0.09 0.10 0.11 0.12 0.13 Policy rate Policy rate Future tax burden Borrowing constraint 0.0415 0.10 0.0410 0.08 0.0405 MPK – (1 +rf) 0.06 T 0.0400 0.04 0.0395 0.02 0.0390 0 0.08 0.09 0.10 0.11 0.12 0.13 0.08 0.09 0.10 0.11 0.12 0.13 Policy rate Policy rate Private capital: second period Output: second period –1.5848 0.115 –1.5849 0.110 –1.5850 0.105 –1.5851 0.100 kp y –1.5852 0.095 –1.5853 0.090 –1.5854 0.085 0.08 0.09 0.10 0.11 0.12 0.13 0.08 0.09 0.10 0.11 0.12 0.13 Policy rate Policy rate Consumption: first period Consumption: second period 0.0415 0.10 0.0410 0.08 0.0405 0.06 c c 0.0400 0.04 0.0395 0.02 0.0390 0 0.08 0.09 0.10 0.11 0.12 0.13 0.08 0.09 0.10 0.11 0.12 0.13 Policy rate Policy rate (continued on next page) 104 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT BOX 7.2: A Theoretical Model of Financial Repression and Interest rate liberalizationa (continued) Policy Scenario 2 (interest rate liberalization) can be illustrated in the model by simulating alternative values of the deposit rate. Recall that the deposit rate is the return on household saving and represent the borrowing cost to the government bank. There is an optimum level of the deposit rate, r*, namely the level that maximizes welfare. Consider very low levels of the deposit rate, say zero. As the policy rate increases so does the return to household savings. This encourages households to save rather than consume and the total pool of savings increases in the economy. However, since the government must pay the deposit rate to borrow from households, the borrowing cost to finance government infrastructure is rising. The first effect continues to dominate the second effect until welfare is maximized at the point indicated in the welfare graph below. Beyond this point, the deposit rate becomes ineffective as a policy tool as further increases in the rate are matched by additional savings which ultimately drives down the decentralized equilibrium rate, r, to its maximum level. The following additional results hold: First, the future tax burden is rising in the deposit rate because higher taxes compensate for the increased borrowing cost of government. Second, private investment is rising in the deposit rate as more savings is made available to finance it. Interest rate liberalization will be welfare enhancing if the deposit rate is currently too low. The larger deposit rate encourages household savings and enhances the amount of private investment that can be financed. However, government revenues will decline. The government bank earns a revenue from the difference between the borrowing rate and the deposit rate. As the deposit rate rises, the spread diminishes, and eventually disappears. a This Box summarizes the theoretical model developed by Eden (2015) for the purposes of this study. and policy options which would require a change in If Ethiopia decides to take forward additional trade policy and mind set. Table 7.4 refers. reforms, then such revenues would gradually diminish. Ethiopia’s tax revenue-to-GDP ratio is low In light of the country’s high appetite for public infra- compared to peers and there is substantial scope for structure investment, it is hard to justify the current low raising it further. At 14 percent of GDP, tax revenues levels of revenues. Further analysis would be needed to in Ethiopia are on the low side. The country urgently identify concrete policy recommendations. That said, needs a revenue-enhancing tax reform which broadens a good place to start could be through an examination the base and increases the tax rates. Consistent with and evaluation of existing tax incentives and subsidies, Ethiopia’s lagging performance in terms of reforms, the currently resulting in forgone tax revenues to the tune last major tax reform can be traced back to 2002/03 of 4 percent of GDP. Some of these so-called ‘tax expen- when the sales tax was replaced by the Value Added ditures’ may provide good value for money, while the Tax. In addition to concerns about low levels of tax costs of others may outweigh their benefits. collection, one may add the heavy reliance on foreign Increased involvement of the private sector in taxes, which account for almost a third of tax revenues. infrastructure provision and maintenance can help TABLE 7.4: Alternative Infrastructure Financing Options Consistent with current strategy and thinking Options that would require a change in policy 1. Raising tax revenues 4. Increasing domestic savings and developing capital 2. Increasing private sector financing of infrastructure invest- markets ments and maintenance 5. More selectivity and prioritization of investments 3. Improving public investment management 6. Securitization of infrastructure assets 7. Improved pricing of infrastructure services, such as electricity Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 105 reduce financing requirements of the public sector. countries, and must thus be interpreted with a great Ethiopia has made recent progress in this direction. deal of caution and should not be a cause for com- The most prominent example is the 1,000MW, US$4 placency. To illustrate, Ethiopia continues to register billion geothermal energy project at Corbetti, where significant time and cost overruns in some public a private foreign investor consortium will sell energy infrastructure projects (see World Bank, 2013). A to the national grid according to a Power Purchasing detailed country study of Ethiopia’s PIM performance Agreement (PPA). There have also been discussions is currently not available. In all cases, further prog- about an oil pipeline project potentially financed ress in all four dimensions of Ethiopia’s PIM capacity by Black Rhino, although agreements are yet to be would undoubtedly enhance the positive economic signed. Aside from direct provision of infrastructure, returns expected from publically financed projects. there are also recent examples of private user contribu- (See World Bank 2015 for more details on this topic). tions in road projects, such as the Addis-Adama Toll Further resources could be raised through Express Way. Ethiopia currently does not have a Public domestic savings mobilization and the eventual Private Partnership (PPP) Framework which prevents establishment of capital markets. The government a holistic approach to addressing the challenge.57 An has been actively aiming to raise domestic savings, important lesson from Corbetti is that ad hoc negoti- among others through bank branch expansions. ated deals are not the way to go. PPPs or pure private As documented in the Second Ethiopia Economic solutions are important alternatives to public provi- Update (World Bank, 2013), this policy has indeed sion of infrastructure that must be kept in mind. If had a demonstrable effect on domestic savings in the private finance is going to play a significant role then formal banking system. However, as shown in the the government needs to be proactive and systematic same analysis, a key determinant of domestic savings in its approach. (Interested readers can consult IMF is the real deposit interest rate. Since this rate is cur- (2014) for further details). rently negative, households have strong incentives Improved public investment management can to channel monetary savings into informal savings help ensure that Ethiopia gets as much infrastruc- mechanisms. A negative real interest rate is also a major ture for the public money that it spends. Since obstacle for the development of a secondary market for weakness in public investment management can treasury bills as investors would not earn a sufficient negate the core argument that impressive rates of pub- return for voluntary purchase of such assets. On the lic investment are necessary for a country to sustain other hand, the negative real interest rate is a part of rapid economic growth, attention to the processes an overall financial repression strategy which yields that govern project selection and management is criti- other important benefits for government, including cal. Encouragingly, Public Investment Management the cheap access to finance for public infrastructure (PIM) in Ethiopia is better than expected given its investment. level of development. An international comparison of Selectivity and prioritization in public invest- PIM capacity across a sample of 71 poor and middle ments. In theory, the Government should finance the income countries conducted by the IMF and World projects with the highest expected economic return, Bank, places Ethiopia at the median, while it is the but in practice such calculations are seldom available. eleventh poorest countries in the world (Dabla-Norris Based on the limited information available, most et al. 2010). Of the four dimensions measured in the public investment projects would potentially have an study, Ethiopia scores above the median in ‘manage- important positive long term impact on exports and ment’ and ‘appraisal’, at the median in ‘evaluation’ and below the median in ‘selection’. On the other 57 The Government, together with the African Development Bank, is in hand, such indices may lack precision about individual the process of developing a PPP Framework and a PPP Unit. 106 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT growth. However, not all projects (whether infrastruc- investments. Moreover, raising tariffs would be a pro- ture or productive) have an equally compelling merit gressive policy, as it is the better-off households that of financing. Thus, to free overall resources for prior- have access to electricity and benefit from the implicit ity infrastructure investment, the government could subsidy currently in place. (See World Bank 2015ab reduce financing of marginal projects that do not have for more details). a demonstrable strong economic rationale. In summary, this chapter has argued that to Securitization of infrastructure assets is sustain high growth, Ethiopia needs to explore ways another potential source of public investment of financing private investment while also finding financing. Ethiopia has a series of prominent, success- complementary ways of financing infrastructure. ful, and profitable State Owned Enterprises, including This is because both infrastructure investment and Ethiopian Airlines, Ethio Telecom and Commercial private investment are needed, as the experience of Bank of Ethiopia. These enterprises are currently 100 high-growth economies show. Ethiopia’s financing percent owned by Government. Securitization refers choice currently has a clear bias in favor of public to a process whereby a small share of these assets infrastructure investment. Infrastructure investment were sold to the general public. This approach has enhances the productivity of the private sector, but been successfully used in countries such as China only when lack of infrastructure itself is constraining and Colombia. growth. Ethiopian firms appear to be more constrained A final proposal include the cost-based pric- in credit and empirical estimates suggest that, at the ing of infrastructure services, such as electricity. margin, credit could yield better growth returns in Households and firms currently pay energy tariffs that the private sector. This suggests some reduction in are below the cost of providing such services. At the public infrastructure financing would be beneficial for same time, Ethiopia is investing billions of dollars in long-term growth. To ensure continued support for new energy generation. By charging more for energy infrastructure finance, this chapter presented a menu services, consumers could help finance such energy of policy options for consideration. Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 107 Annex 7.1 Public and Private capital stocks: in the constrained case, aggregate capital Investment Model: The Model should equal I, and the marginal returns to investment may exceed the world interest rate. In contrast, in the Consider the following problem, in which the gov- unconstrained case, the levels of investment are deter- ernment takes the rental rate of capital as given and mined by the condition that the marginal return to chooses public and private capital stocks, subject to a (both types of ) investment is equated with the world constraint on the total capital stock: interest rate. In both cases, the marginal products of public and private capital should be equalized, which max F ( kg , kp ) − r ( kg + kp ) s.t . kg + kp = I is the only condition used for the analysis that fol- kg ,k p lows. Consider next the following functional form (Constant Elasticity of Substitution, CES): Where F is a strictly positive, increasing production function, that has decreasing returns in both argu- ∝ ments, is the per capita public capital stock, is the per F ( kg , k p ) = A ( γ k g σ + (1 − γ ) kg ) σ σ capita private capital stock, is the interest rate, and is the constraint on the total capital stock. where 0 ≤ γ ≤ 1 and σ ≤ 1 and A represents aggregate This simplified problem captures a static repre- productivity. This functional form is quite flexible: sentation of the dynamic problem in Eden and Kraay depending on parameters, public and private capital (2014), in which the government maximizes welfare can be substitutes or compliments, and have differ- subject to a binding aggregate credit constraint. (In ent levels of relative productivity. As special cases, this the dynamic setting, is pinned down as the sum of functional form nests the cases of perfect substitutes depreciated capital stocks and availability of credit). (σ = 1), Cobb-Douglas (σ = 0), and perfect comple- Substituting in the constraint, the optimization prob- ments (also known as Leontief, and given by σ = – ∞). lem can be rewritten as: This functional form implies that the optimality con- dition is given by: ( ) 1 ∗ kp 1− γ 1−σ λ = ∗= ∗ kg γ The first order conditions of this problem yields: Here, λ* is the optimal ratio of public and private ∂F ∂F capital. It depends both on the elasticity of substi- = ∂kg ∂kp tution between public and private capital, σ, and on the relative productivity parameters γ and 1 – γ. In other words, the marginal product of public Thus, to determine the optimal allocation of credit capital equals the marginal product of private capital between public and private capital, a quantitative and the levels of public and private capital are deter- sense of γ and σ is needed. It is useful to note that mined by the aggregate credit constraint. λ* is the optimal ratio of public and private capital, It is worth noting that the optimality condition regardless of the extent to which the credit constraint does not depend on the assumption of a binding aggre- is binding. However, in general, if the constraint is gate credit constraint, as a similar optimality condition binding, the aggregate capital stock is too low, and can be derived without this assumption (see Eden and there may be positive excess returns to both types of Kraay (2014) for details). The difference between the capital (marginal output effect less the interest rate two cases is in the implications for the levels of the and depreciation). 108 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Annex 7.2 Public and Private Investment Model: Empirical Results TABLE A7.2.1: The Response of Private Investment to Public Investment All IDA Countries IDA Countries in Sub-Saharan Africa Control Control Excluding for Lagged Excluding for Lagged All Influential Dependent All Influential Dependent Observations Observations Variable Observations Observations Variable Panel A: Ordinary Least Squares (Dependent Variable is Change in Private Investment) Change in Government –0.0858 –0.113 –0.0780 –0.0900 –0.108 –0.0778 Investment (0.125) (0.116) (0.126) (0.167) (0.148) (0.166) Panel B: Two-Stage Least Squares (Dependent Variable is Change in Private Investment) Change in Government 1.881* 2.298* 1.892* 1.891 1.668 1.939 Investment (1.066) (1.297) (1.059) (1.526) (1.204) (1.557) Panel C: First-Stage Regression (Dependent Variable is Change in Government Investment) Change in Predicted 0.290*** 0.270*** 0.290*** 0.248** 0.326*** 0.246** Disbursements (0.0778) (0.0750) (0.0775) (0.0918) (0.0732) (0.0915) First-Stage F-Statistic 13.94 12.98 14.00 7.30 19.88 7.24 Weak Instrument Consistent 95% [0.583, [0.772, [ 0.594, [0.137, [0.292, [0.180, Confidence Interval 4.863] 6.275] 4.878] 9.526] 5.219] 9.854] for β Number of Observations 916 908 916 611 607 611 Notes: This table reports the results from a series of regressions of changes in private investment on changes in public investment, with country and year fixed effects. All changes are in constant local currency units and are scaled by lagged GDP . The sample consists of IDA-eligible countries (first three columns) and IDA-eligible countries in Africa (second three columns). Panel A reports OLS estimates, Panel B reports 2SLS estimates, and Panel C reports the corresponding first-stage regressions. Weak instrument-consistent confidence intervals are based on the Moreira Likelihood Ra- tio statistic. Changes in predicted disbursements on loans from official creditors are used as an instrument for changes in government investment in Panel B. Heteroskedasticity-consistent standard errors clustered at the country level are indicated in parentheses. * (**) (***) indicates significance at the 10 (5) (1) percent level. Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 109 TABLE A7.2.2: The Response of Output to Public Investment All IDA Countries IDA Countries in Sub-Saharan Africa Control Control Excluding for Lagged Excluding for Lagged All Influential Dependent All Influential Dependent Observations Observations Variable Observations Observations Variable Panel A: Ordinary Least Squares (Dependent Variable is Change in Output) Change in Government 0.190* 0.148* 0.0911 0.184 0.127 0.0982 Investment (0.0994) (0.0757) (0.0830) (0.122) (0.0928) (0.107) Panel B: Two-Stage Least Squares (Dependent Variable is Change in Output) Change in Government 1.418* 1.248 1.267 2.580** 1.684* 2.439* Investment (0.797) (0.819) (0.783) (1.177) (0.831) (1.305) Panel C: First-Stage Regression (Dependent Variable is Change in Government Investment) Change in Predicted 0.290*** 0.255*** 0.269*** 0.248** 0.308*** 0.211** Disbursements (0.0778) (0.0720) (0.0786) (0.0918) (0.0663) (0.0967) First-Stage F-Statistic 13.94 12.57 11.71 7.3 21.59 4.74 Weak Instrument Consistent 95% [–0.046, [–0.336, [–0.293, [0.346, [0.109, [–0.168, Confidence Interval 3.933] 4.160] 4.035] 11.943] 5.713] 25.653] for β Number of Observations 916 907 916 611 606 611 Notes: This table reports the results from a series of regressions of changes in real GDP on changes in public investment, with country and year fixed effects. All changes are in constant local currency units and are scaled by lagged GDP . The sample consists of IDA-eligible countries (first three columns) and IDA-eligible countries in Africa (second three columns). Panel A reports OLS estimates, Panel B reports 2SLS estimates, and Panel C reports the corresponding first-stage regressions. Weak instrument-consistent confidence intervals are based on the Moreira Likelihood Ra- tio statistic. Changes in predicted disbursements on loans from official creditors are used as an instrument for changes in government investment in Panel B. Heteroskedasticity-consistent standard errors clustered at the country level are indicated in parentheses. * (**) (***) indicates significance at the 10 (5) (1) percent level. 110 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Annex 7.3 Public and Private Annex 2. What do these results imply for the optimal Investment Model: Calibration level of the public capital stock in Ethiopia? The first approach using World Bank invest- Table A7.3.1 presents the calibrated values of the ment data indicates that Ethiopia may benefit from parameters of the CES production function (σ and reducing its public capital stock by 12 percent. The γ,) as a function of the empirical estimates of the point estimate in Table A7.3.1 suggests that optimal response of private investment to public investment ratio of public and private capital is given by λ* = 1.18. (β) and the response of output to public investment By comparison, the average ratio of private and pub- (βy). The excess returns to public investment for λ = 3 lic investment in Ethiopia is λ = 0.92. As a first step, and λ =1 implied by this calibration are calculated it is assumed that average ratio of public and private using r*+ ∂ = 0.13 under the assumption that credit investment roughly corresponds to the ratio of capital constraints are not binding. The last two columns stocks. The observed ratio λ = 0.92 is a little bit low present the threshold level of compared to the estimated optimal ratio (1.18): k*p λ∗ = , k*g above which there are positive excess returns to public investment, and the percent of countries in the sample for which λ is above the threshold. The bolded row rep- resents the calibrated values for the point estimates of An alternative approach using estimated capital β = 2 and βy = 1.5 derived from the regression model in stock indicates that public capital stock in Ethiopia TABLE A7.3.1: Calibration Results Empirical estimates Calibrated parameters Excess return to public investment β βy σ γ λ=3 λ=1 λ* Countries (%) Private inv. Output Degree of Relative response response substitutability productivity Optimal ratio Optimal ratio Optimal ratio Optimal ratio 1 0.5 –0.35 0.4 0.25 –0.04 1.35 82% 1.5 0.5 –1.19 0.17 0.17 –0.10 2.06 54% 1.5 1 –0.67 0.49 0.66 –0.01 1.02 90% 1.5 1.5 –0.54 0.65 1.17 0.11 0.67 100% 1.5 2 –0.48 0.73 1.7 0.22 0.51 100% 2 0.5 –3.18 0.02 0.11 –0.13 2.54 49% 2 1 –1.72 0.22 0.6 –0.09 1.59 72% 2 1.5 –1.43 0.4 1.11 –0.04 1.18 85% 2 2 –1.32 0.5 1.5 0 1.00 90% 2.5 0.5 –10.48 0.00 0.05 –0.13 2.92 38% 2.5 1 –4.96 0.007 0.54 –0.13 2.30 51% 2.5 1.5 –4.13 0.03 1.05 –0.13 1.97 54% 2.5 2 –3.8 0.06 1.56 –0.12 1.77 54% Ethiopia’s Financing Choice: Public Infrastructure or Private Investment? 111 should be about 20 percent lower. One limitation of of the aggregate capital stock reported in PWT. Public the previous approach is that 85 percent of the coun- and private capital stocks are then cumulated forward tries in the sample exhibit a ratio of private and public using the investment rates Figure 2.1 to arrive at a capital which is too high relative to the optimum. The series of public and private capital stocks. The esti- estimate based on average investment rates is imperfect, mated production function is then used to compute because the ratio of capital stocks depends both on the the marginal returns to public and private capital over timing on investment and on the initial capital stocks. the past two decades. The estimates suggest that, in An alternative, and possibly more accurate, measure of 2011, Ethiopia’s ratio of private and public capital the ratio of public and private capital stocks in Ethiopia was λ = 0.75. Using the formula above, this estimate is therefore computed by complementing data on gov- suggests that the public capital stock should be about ernment and private investment with estimated capital 20 percent lower. Given the short length of the series, stocks from the Penn World Tables (PWT). The series the estimates of the capital stocks may be sensitive to are computed using a 3 percent annual depreciation the specification of initial conditions. To address this rate, which is significantly lower than the 10 percent concern, the analysis was repeated under the alterna- used in Eden and Kraay (2014). This difference is not tive assumption that the initial public capital stock important for the calibration of the parameters σ and in 1987 is zero. It is noted that a lower initial public γ, but is important when computing excess returns. To capital stock implies higher returns to public invest- compute public and private capital stocks in Ethiopia, ment. Even under this highly conservative assumption, it is assumed that the initial public capital stock in roughly the same gap in marginal products is derived 1987 (when the investment data start) is equal to half for 2011, suggesting a fairly robust conclusion. 113 GROWTH AND STRUCTURAL REFORMS58 8 Ethiopia’s reform effort was substantial in the Using illustrative cross-country simulations at the macro and firm levels, this chapter shows that even modest 1990’s, but few reforms took place in the 2000s. structural reforms may yield substantial growth pay-offs The period immediately following the overthrow for the country. In terms of reform sequencing, Ethiopia of the communist Derg in 1991 was characterized has already followed international best practice through its ‘trade-first’ approach, although it has proceeded relatively by deep and wide-ranging reforms, including trade, slowly. Recommended next steps include the completion of agriculture, exchange rate, banking, taxation and trade reforms by also opening up services followed by a privatization. However, reform effort slowed down liberalization of the domestic financial sector while being cognizant of potential risks to reform efforts. Such reforms towards the end 1990s and has not picked up since. may be important contributors to Ethiopia’s future growth Although regulatory improvements were made to given their effect of improving the efficiency of resources domestic competition, investment, customs and busi- allocation and enhancing productivity. ness licensing/registration, reforms were fewer and less deep in the 2000s. Indeed, the absence of concurrent major structural reforms is a curious characteristic of 8.1 Introduction the growth acceleration episode. This experience raises an important question Economists have long debated whether reforms pro- about what the relationship between growth and mote growth. The discussion between proponents and reform is for Ethiopia. Arguably, the market-oriented opponents of the Washington Consensus is a classical reforms of the 1990s provided a necessary founda- illustration hereof. Because economic theory does not tion for the subsequent economic take-off in 2004. reach clear conclusions on the conditions that best sup- As documented in World Bank (2007), the changes port income catch-up, researchers have sought to draw les- that occurred since 1992 constituted the first stages sons from the experience of a broad segment of countries of a major, and potentially, long lasting transition to using cross-country, industry, and firm level evidence. institutional arrangements which were much more There is a growing consensus that both macro- conducive to the pursuit of long-term prosperity than and microeconomic reforms can lead to improve- earlier conditions. In other words, growth was not ments in resource allocation, productivity, and necessarily ‘reform-less’. At the same time, contin- growth. In particular, higher quality and quantity of ued or concurrent economic reforms could not have infrastructure and human capital, trade openness, effi- been a major driver of sustained growth in the 2000s cient and well-developed financial systems, appropri- precisely because the reform effort slowed down at ate tax and expenditure policies, and sound economic the same time. institutions (e.g. strong rule of law, and avoidance of Does this mean that Ethiopia does not need overly stringent regulation of product and labor mar- to reform to grow fast? In this chapter we argue kets) that promote competition, facilitate entry and that it could be erroneous and complacent to reach exit, and encourage entrepreneurship and innovation have been variously found to increase productivity 58 This chapter draws upon the background papers prepared by Haile growth (Dabla-Noris et al. 2014). (2015) and Hollweg, Rojas and Varela (2015). 114 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 8.1: Ethiopia: Structural Reform Indices, 1973–2005 1. Real Sector 2. Financial Sector 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 Trade Current account Agriculture Domestic finance Banking Telecom Electricity Securities Capital account Source: Based on data from Prati et al. (2013). Note: Values cannot be compared across indices (see Box 8.1). such a conclusion. Ethiopia was able to sustain high 3 evaluates its impact on economic growth using growth in the past because of earlier reforms and, macro level data while Section 4 estimates the as documented in previous chapters, because it ‘got impact on firm productivity. Section 5 summarizes infrastructure right’ at the early stage of development. international best practice on reform sequencing However, as shown in this chapter, even for countries and relates it to Ethiopia’s experience. Section 6 such as Ethiopia, there are substantial growth benefits highlights potential risks of reform. Section 7 con- to further reforms. This implies that structural reforms cludes by sketching the broad direction of future offer a growth potential, an opportunity that Ethiopia reform efforts in Ethiopia. Finally, Section 8 sets up can tap into when needed. If growth slows down, as a framework for monitoring the sustainability of we have argued in Chapter 6 is a real possibility, reig- Ethiopia’s growth model. niting the structural reform agenda may well become necessary to sustain high growth. Trends and Status in Structural 8.2  This chapter seeks to address the following Reforms questions: How has Ethiopia’s reform effort evolved over time and how does it compare to other countries? Structural reform indices illustrate that Ethiopia’s What are the potential growth benefits to reform- reform effort accelerated since 1991 and slowed ing? Which reforms are most potent? How does the down in the late 1990s. Figure 8.1.1 illustrates this country’s nascent reform sequencing compare with point for the real sector and Figure 8.1.2 for the international best practice? What are the risks of financial sector. Box 8.1 contains details on data and reform? Are there alternatives to the ‘trodden path of definitions. Prior to 1991, there were few reforms, as reform’? What would be the logical next reform step illustrated by relatively low levels of the reform indi- for the country? ces. After 1991, most indices exhibit a rising trend, This chapter is structured as follows: Section indicating reform progress. From the late 1990s 2 examines Ethiopia’s structural reform status and onwards, most indices exhibit a flat trend, suggest- trends, and compares with peers. Sections 3 and 4 ing that reforms came to a halt. Trade is an impor- examine the economic impact of reforms. Section tant exception to this overall trend as tariffs were Growth and Structural Reforms58 115 BOX 8.1: Structural Reform Indices: Data and Definitions The data on structural reforms cover more than ninety developed and developing countries, and span a fairly long period of time, namely 1973 to 2006. These data are taken from Prati et al. (2013), which are in turn based on databases constructed by Abiad et al. (2008) and Ostry et al. (2009). The indices of structural reforms are categorized into two groups: real sector and financial sector reforms. The reform indicators in the real sector comprise openness to international trade and domestic product market liberalization. Openness to international trade is measured along two dimensions: average tariff rates and restrictions (or lack thereof) on current account transactions (including payments and receipts on exports and imports of goods and services). The average tariff index takes the value 0 if average tariff rates are 60 percent or higher, the value 1 if tariff rates are zero, and varies linearly for intermediate tariff rates. The degree of reforms in the product market is captured by two different indices. The first corresponds to the agricultural sector and measures the extent of state intervention (i.e. the presence of export marketing boards and price controls) in the market for the country’s main agricultural export commodity. The second captures the degree of liberalization (i.e. the extent of competition and regulatory quality) in the networks (telecommunications and electricity) sector. Financial sector reforms are captured by two indices measuring domestic finance and capital account liberalization. The indicator of domestic financial reform is derived as the average of six sub-indices. Five of them involve the banking system: (i) credit controls, such as subsidized lending and direct credit; (ii) interest rate controls, such as floors or ceilings; (iii) competition restrictions, such as entry barriers and limits on branches; (iv) the degree of state ownership; and (v) the quality of banking supervision and regulation. The sixth sub-index focuses on the securities markets and measures the degree of legal restrictions on the development of domestic bond and equity markets, and the existence of independent regulators. The index of capital account liberalization measures the intensity of restrictions on financial transactions for residents and nonresidents, as well as the use of multiple exchange rates. We use both the aggregate indicator of capital account reforms and two sub-indices of external capital account openness for resident and nonresident. The two sub-indices capture the degree of legal restrictions on residents’ versus nonresidents’ ability to move capital in and out the country. All reform indices range between 0 and 1: the higher the rating, the greater the degree of liberalization. We note that variations in the values of each index over time and across countries reflect differences in the absolute degree of economic liberalization within each sector. As the indices were constructed using different methodologies, quantitative differences in the values of the indices across sectors do not provide an exact measure of whether one sector is more liberalized than another. Source: Prati et al. (2013). reduced throughout the period of analysis. Electricity Ethiopia maintains a high level of regulatory is another exception as no reforms were observed restrictiveness towards foreign services providers. throughout the 1993–2006 period. The World Bank Services Trade Restrictions Database Ethiopia lags behind in most dimensions of shows that Ethiopia exhibits high levels of restric- reform, especially in domestic finance, the current tiveness across all modes of supply, scoring higher account and the capital account. Figure 8.2 com- than a relevant set of comparator countries in each pares Ethiopia’s reform experience with the averages (Figure 8.3). Mode 3, or commercial presence, is in for Sub-Saharan Africa (SSA), Low Income Countries fact the most restrictive form for foreign providers to (LIC) and Lower Middle Income Countries (LMIC). supply services in Ethiopia. Ethiopia is substantially It draws upon the reform indices of Prati et al. more restrictive than its comparators in mode 1 (cross- (2013) described in Box 8.1. Ethiopia has done well border supply), mode 2 (consumption abroad) and in reducing tariffs, implying that its trade reform index mode 3, but only slightly more restrictive than com- is comparable to peers. The reform gaps in domestic parators in mode 4 (movement of natural persons), finance, the current account and the capital account traditionally the most protected mode of supply. are substantial. Gaps in agriculture reform and net- Ethiopia also exhibits high restrictiveness when works are less pronounced. considering five key services sectors—financial, 116 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT FIGURE 8.2: Structural Reform Indices by Country and Income Groups 1. Trade Index 2. Current Account Index 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 3. Agriculture Index 4. Networks Index 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 5. Capital Account Index 6. Domestic Finance Index 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 Ethiopia SSA LIC LMIC Source: Based on data from Prati et al. (2013). Growth and Structural Reforms58 117 FIGURE 8.3: Services Trade Restrictiveness Index by Sector and Mode 100 80 60 40 20 0 China Ethiopia Kenya Korea Rwanda Tanzania Uganda Vietnam Zambia Overall Mode 1 Mode 3 Mode 4 100 80 60 40 20 0 China Ethiopia Kenya Korea Rwanda Tanzania Uganda Vietnam Zambia Overall Financial Professional Retail Telecommunication Transportation Source: World Bank Services Trade Restrictions Database. professional, retail, telecommunications and trans- including Sub-Saharan African, low income, and lower port services. As illustrated in Figure 8.3 Ethiopia is middle income countries. We also include selected completely closed in retail and telecommunications, and countries for illustrative purposes. This includes almost entirely closed (with a score above 75 percent) in Uganda and Tanzania as examples of structural peers, transportation, professional, and financial services. and Ghana and Sri Lanka as examples of aspirational peers. The main results are summarized below and The Potential Growth Impact of 8.3  explained in detail subsequently. Annex 8.1 details Reforms the methodology and Annex 8.2 summarizes the robustness checks. Overview Overall, the results indicate that even modest reforms that close gaps with Sub-Saharan Africa What would be the potential impact on growth peers would potentially have considerable impact of structural reforms in Ethiopia? To address this on GDP per capita growth. In particular, the largest question, we perform a benchmarking exercise based potential growth payoffs could be reaped from closing on the cross-country growth regression model in Prati gaps in domestic financial reforms, the current account et al. (2013), using the data set described in Box 8.1. and the capital account. Effectively, we simulate the growth effect of closing We acknowledge upfront that these results are Ethiopia’s reform gap relative to relevant peer groups, only indicative. The results shed light on the potential 118 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT growth payoffs that reforms could deliver but do not the growth effect of liberalizing trade: for every unit constitute a comprehensive appraisal of reforms that of increase in the trade reform index, the real GDP have actually been introduced. Among other shortcom- per capita growth rate would increase by 0.019 per- ings, the exercise does not control for external factors centage points the subsequent year. This is a dynamic that may affect the link between structural reforms and effect, so a similar growth boost would take place the growth. It focuses exclusively on the impact on economic subsequent year at a diminishing rate until the growth growth, thereby ignoring other important dimensions. effect eventually vanishes. This cumulates into long Finally, by focusing on ‘average effects’ rather than sta- term effect, which we examine later. tistical tail events the results also tend to overlook the The potential impact of structural reforms on potential risks of reform, as discussed later the chapter. GDP per capita growth is estimated based on the reform gaps for the period 2000–2006. Because Detailed Results and Discussion the averages for the period 1973–2006 are likely to smooth out significant fluctuations in the reform Owing to the slowdown in reform effort in the indices over time, the discussion focuses on the reform late 1990s, Ethiopia’s reform gap with peers has gaps for the more recent period, which better reflects increased. Table 8.1 presents the average values of the the countries’ recent reform trends. Although the structural reform indices for Ethiopia and comparator reform gaps for the most recent period, namely 2006, groups and countries over the periods 1973–2006 and would be more relevant, in some cases we might end 2000–06. Prior to 2000, Ethiopia lagged behind all up capturing anomalies. peer groups, except in trade reform. After 2000, the The results suggest that closing Ethiopia’s gap increased because many of the comparator coun- reform gaps would generally be associated with tries and groups, unlike Ethiopia, implemented con- significant increases in economic growth. Table 8.3 siderable reforms. Reform gaps for Ethiopia are most presents the results of the benchmarking exercise. We substantial in the capital account, domestic finance, illustrate the results by using the trade liberalization electricity, and the current account and smallest in example. Suppose Ethiopia were to close the gap with trade. This trend is also visible in Figure 8.2. the Sub-Saharan African average in this dimension. The relationship between economic growth The reform gap is calculated as 0.738–0.672= 0.066, and reform is derived using a cross-country growth as per Table 8.1. The resulting growth effect, in turn, regression model. Table 8.2 presents the results of an is derived by multiplying the reform gap with the ordinary least squares regression (OLS) in which real coefficient derived in Table 8.2, i.e. 0.066*0.019 = GDP per capita growth is regressed individually on the 0.13%. In other words, Ethiopia’s per capita growth values of real GDP (in logs) and the respective reform rate would be 0.13 percentage points higher in the first indicator (both lagged) using country and year fixed year after the reform if it could close the trade reform effects.59 The results are similar to those presented in gap with SSA. This growth effect would persist for a Prati et al. (2013), except we re-estimated the origi- while, but gradually decline over time. nal model by disaggregating the network index into The most substantial growth impacts would be electricity and telecommunications to gain further realized by reforming the domestic financial sec- insights about these sub-sectors. tor, the current account and the capital account. The baseline regression results conform to our Generally speaking, the growth effects depend on expectations. All coefficients are positive and statisti- cally significant at conventional critical values, except 59 There are no major differences in results if the regression is estimated the networks index (including electricity and telecom- using a multivariate approach, i.e. considering all reforms simultaneously. munications sectors). To illustrate the results, consider See Haile (2015) for details. Growth and Structural Reforms58 119 TABLE 8.1: Average Values of Structural Reform Indices 2000–2006 Structural reforms Ethiopia Uganda Tanzania Sri Lanka Ghana SSA LIC LMIC Real sectors Trade 0.672 0.778 0.746 0.828 0.764 0.738 0.750 0.755 Current account 0.438 1.000 0.750 0.698 0.646 0.683 0.675 0.664 Agriculture 0.333 1.000 0.667 0.667 0.333 0.464 0.413 0.519 Network 0.091 0.545 0.212 0.212 0.636 0.192 0.158 0.310 Electricity 0.000 0.600 0.033 0.267 0.800 0.119 0.106 0.334 Telecommunications 0.167 0.500 0.361 0.167 0.500 0.254 0.203 0.280 Financial sectors Domestic finance 0.370 0.685 0.815 0.713 0.537 0.671 0.606 0.581 Banking 0.333 0.756 0.867 0.722 0.511 0.709 0.636 0.594 Securities 0.556 0.333 0.556 0.667 0.667 0.483 0.461 0.519 Capital account (CA) 0.250 1.000 0.375 0.500 0.375 0.586 0.512 0.602 CA (resident) 0.250 1.000 0.250 0.250 0.250 0.551 0.461 0.552 CA (non-resident) 0.250 1.000 0.500 0.750 0.500 0.638 0.547 0.689 1973–2006 Real sectors Trade 0.507 0.459 0.507 0.459 0.507 0.459 0.507 0.459 Current account 0.250 0.544 0.250 0.544 0.250 0.544 0.250 0.544 Agriculture 0.146 0.364 0.146 0.364 0.146 0.364 0.146 0.364 Network 0.028 0.124 0.028 0.124 0.028 0.124 0.028 0.124 Electricity 0.000 0.127 0.000 0.127 0.000 0.127 0.000 0.127 Telecommunications 0.052 0.121 0.052 0.121 0.052 0.121 0.052 0.121 Financial sectors Domestic finance 0.125 0.347 0.125 0.347 0.125 0.347 0.125 0.347 Banking 0.123 0.390 0.123 0.390 0.123 0.390 0.123 0.390 Securities 0.135 0.131 0.135 0.131 0.135 0.131 0.135 0.131 Capital account (CA) 0.195 0.492 0.195 0.492 0.195 0.492 0.195 0.492 CA (resident) 0.250 0.477 0.250 0.477 0.250 0.477 0.250 0.477 CA (non-resident) 0.141 0.508 0.141 0.508 0.141 0.508 0.141 0.508 Source: Staff estimates based on data from Prati et al. (2013). two factors: First, the size of the reform gap: the country-specific, while the second factor is an average larger the gap, the larger the growth effect of clos- effect estimated for all countries in the sample.60 To ing it (Table 8.1). Second, the size of the general illustrate, additional domestic financial liberalization impact of the reform in question with respect to growth, as reflected by the size of the estimated coef- 60 Prati et al. (2013) and Haile (2015) show how the estimated coefficients ficient (Table 8.2). We note that the first factor is are broadly similar across income groups. 120 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE 8.2: Baseline Growth Regressions (Dependent variable: Real GDP Per Capita, Growth Rate) Regressions (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Real sectors Trade 0.019 (t–1) (1.9)* Current account 0.033 (t–1) (4.1)*** Agriculture 0.018 (t–1) (2.3)** Network 0.004 (t–1) (0.4) Electricity 0.007 (t–1) (1.41) Telecom 0.001 (t–1) (0.43) Financial sector Domestic finan. 0.064 (t–1) (4.6)*** Banking 0.050 (t–1) (4.2)*** Securities 0.037 (t–1) (4.6)*** Capital account 0.021 (t–1) (2.3)** Capitalresident 0.015 (t–1) (2.14)** Capitalnonresident 0.016 (t–1) (2.00)** Log (GDP per –0.048 –0.051 –0.036 –0.045 –0.042 –0.041 –0.047 –0.051 –0.049 –0.051 –0.045 –0.044 capita) (t–1) (6.9)* (5.7)*** (5.1)*** (5.6)*** (5.3)*** (5.1)*** (5.9)*** (5.7)*** (5.4)*** (5.7)*** (5.6)*** (5.7)*** Observations 3,418 3,530 3,390 3,796 2,653 2,653 2,653 3,530 3,556 3,530 3,846 3,814 R-squared 0.19 0.14 0.17 0.15 0.20 0.19 0.20 0.14 0.14 0.14 0.15 0.15 Source: Prati et al. (2013) and staff estimates. Note: t-values (computed based on robust standard errors clustered at country level) in parentheses. All specifications are estimated by OLS and include country and year fixed effects. The regressions include only one indicator of structural reform at a time. Annual data over 1973–2006 when available. GDP in real terms and PPP adjusted. ***, **, * indicate statistical significance at the one, five, and ten percent. in Ethiopia, up to a point where it reaches the average The potential long-run effects of reforms are SSA country level, would increase the real per capita substantial. These effects can be estimated effectively as GDP rate by 1.92 percentage points the first year after the cumulative sum of the gradually declining short-run reform with marginally declining rates in subsequent effects. The results are reported in Table 8.4. If we focus years. We defined financial liberalization broadly here exclusively on the effect of Ethiopia catching up with the as a higher index value in the six dimensions of finan- Sub-Saharan Africa average, the results would be as fol- cial sector reform defined in Box 8.1 lows: In the case of domestic financial reforms, Ethiopia’s Growth and Structural Reforms58 121 TABLE 8.3: Coefficient Estimates and Potential Growth Impact of Reforms Coefficient Predicted effect on Ethiopia’s real GDP per capita, growth rate (%) Structural reforms estimates* Uganda Tanzania Sri Lanka Ghana SSA LIC LMIC Real sectors Trade 0.019 0.20 0.14 0.30 0.17 0.13 0.15 0.16 Current account 0.033 1.86 1.03 0.86 0.69 0.81 0.78 0.75 Agriculture 0.018 1.20 0.60 0.60 0.00 0.23 0.14 0.33 Network 0.004 0.18 0.05 0.05 0.22 0.04 0.03 0.09 Electricity 0.007 0.42 0.02 0.19 0.56 0.08 0.07 0.23 Telecommunications 0.001 0.03 0.02 0.00 0.03 0.01 0.004 0.01 Financial sectors Domestic finance 0.064 2.01 2.84 2.19 1.07 1.92 1.51 1.35 Banking 0.050 2.11 2.67 1.94 0.89 1.88 1.51 1.30 Securities 0.037 –0.82 0.00 0.41 0.41 –0.27 –0.35 –0.14 Capital account 0.021 1.58 0.26 0.53 0.26 0.71 0.55 0.74 Capital (resident) 0.015 1.13 0.00 0.00 0.00 0.45 0.32 0.45 Capital (nonresidents) 0.016 1.20 0.40 0.80 0.40 0.62 0.48 0.70 Source: Author’s computation based on data from Prati et al. (2013). TABLE 8.4: Coefficient estimates and Potential Long-run Growth Impact of Reforms Long-run Predicted effect on Ethiopia’s real GDP per capita, growth rate (%) Structural reforms multiplier* Uganda Tanzania Sri Lanka Ghana SSA LIC LMI Real sectors Trade 0.396 4.20 2.91 6.18 3.64 2.61 3.06 3.28 Current account 0.647 36.40 20.22 16.85 13.48 15.85 15.35 14.66 Agriculture 0.500 33.33 16.67 16.67 0.00 6.52 4.00 9.26 Network 0.089 4.04 1.08 1.08 4.85 0.90 0.59 1.95 Electricity 0.156 9.33 0.52 4.15 12.44 1.85 1.65 5.20 Telecommunications 0.023 0.76 0.44 0.00 0.76 0.20 0.082 0.26 Financial sectors Domestic finance 1.524 47.97 67.72 52.20 25.40 45.80 35.98 32.14 Banking 1.220 51.49 65.04 47.43 21.68 45.76 36.86 31.77 Securities 0.787 –17.49 0.00 8.75 8.75 –5.72 –7.44 –2.92 Capital account (CA) 0.412 30.88 5.15 10.29 5.15 13.84 10.77 14.51 CA (resident) 0.306 22.96 0.00 0.00 0.00 9.21 6.45 9.25 CA (nonresidents) 0.314 23.53 7.84 15.69 7.84 12.18 9.31 13.78 Source: Author’s computation based on data from Prati et al. (2013). Note: SSA, Sub-Saharan Africa; LI, Low-income countries; LMI, Lower-middle-income countries. *Coefficient estimates obtained through a simple manipulation of the coefficient estimates in Tables 2 and 3. The figures in the last seven columns are in percentages and represent the growth payoffs from closing the reform gaps between Ethiopia and the respective benchmark country in the second row. † Potential growth payoffs from getting closer to the technology frontier for the period 2000–2006. 122 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT real GDP per capita would be 46 percent higher in as a reflection of moderate growth coefficient the long run.61 Current account reform, to the level (0.018) and reform gaps. The long term effect observed for the average SSA country, would yield a real would be a real GDP level that would be 6 per- GDP per capita for Ethiopia that is 16 percent higher. cent higher. Agriculture reform would involve a For capital account reform, the effect is 14 percent. reduction of state interventions the market for the In summary, we get the following results country’s main agricultural export commodity. (ranked in order of importance):  The estimated electricity and telecom reforms yield negligible growth gains (see below). This  Domestic financial reforms would yield the is primarily the result of very small growth coef- most substantial growth pay-off. This type of ficients. Electricity and telecom reforms refer to reform is generally the most potent in terms of increasing the extent of competition and improv- growth (coefficient: 0.064) and the reform gap in ing the regulatory quality in these sectors. this sector is the second largest that Ethiopia is facing. To illustrate, if Ethiopia were to catch up The lack of reform data for the recent period with the average Sub-Saharan country in terms since 2006 is not necessarily a shortcoming for the of financial liberalization, its per capita growth findings as reform gaps and coefficients are unlikely rate could be boosted by 1.9 percentage points in to have changed significantly: First, Ethiopia’s reform the first year after the reform. This effect pertains gap may arguably even have widened since 2005. Even primarily to the banking sector. if we don’t have data for this period, a qualitative assess-  Current account reform has the second largest ment strongly suggests that there have been no major growth effect. Current account reform has the second reforms in Ethiopia over this period. Moreover, given largest growth coefficient (0.033) and Ethiopia’s gap the historical trend for comparator countries, it is rea- is sizable. Catching up with the SSA average would sonable to expect that these countries may have main- potentially add 0.8 percentage points to Ethiopia’s tained or continued their reform efforts and this would annual per capita growth rate. Current account also be consistent with the available qualitative evidence. reform would involve a reduction of restrictions Second, the coefficient estimates for reform are quite on payments and receipts on exports and imports robust, implying that they are not sensitive to the omis- of goods and services. Examples on such restric- sion or inclusion of additional observations. In particu- tions include rules associated with import permits. lar, the robustness checks shown in the Annex show  An opening of the capital account has a growth that the potential impact of reforms for the sub-periods potential comparable to that of a current 1973–1989 and 1990–2006 are generally consistent account reform. This is a reflection of the fact with each other, which might suggest that expanding that Ethiopia lags substantially behind in this area the sample period with few more observations would (largest gap observed) even if the growth coeffi- not have a substantial impact on the empirical estimates. cient is not particularly large (0.021). Closing the Growth benefits to telecom and electricity capital account gap with the Sub-Saharan Africa reforms may be positive for Ethiopia, even if the average yields an additional growth rate of 0.7 results do not show this. A notable exception to percentage points, in the short term. This would the above explanation about the robustness of coef- involve a reduction of restrictions on financial ficient estimates to new data is telecom and electricity transactions for residents and non-residents.  The growth payoffs from agriculture reform 61 The long term effect is estimated over a 33 year period (1973–2006). are modest. In agriculture, growth could be This is potentially indicative that the long run effect may be around 0.2 percentage higher by catching up with SSA thirty years. Growth and Structural Reforms58 123 TABLE 8.5: Reforms, Growth, and Distance to the Technology Frontier Coefficient Predicted effect on Ethiopia’s GDP estimates* per capita growth (%) Growth payoff from First Second Reform getting closer to the Structural reforms quartile quartile gap First quartile Second quartile technology frontier Real sectors Trade 0.041 0.027 0.156 0.641 0.422 –0.219 Current account 0.028 0.054 0.260 0.729 1.406 0.677 Agriculture 0.010 0.041 0.333 0.333 1.367 1.033 Network 0.025 0.006 0.121 0.303 0.073 –0.230 Electricity 0.022 0.015 0.267 0.575 0.402 –0.173 Telecommunications 0.013 –0.008 0.000 0.000 0.000 0.000 Financial sectors Domestic finance 0.026 0.109 0.343 0.891 3.734 2.844 Banking 0.017 0.083 0.389 0.661 3.228 2.567 Securities 0.030 0.075 0.111 0.333 0.833 0.500 Capital account 0.006 0.038 0.250 0.150 0.950 0.800 Capital (resident) 0.016 0.022 0.000 0.000 0.000 0.000 Capital (nonresidents) –0.008 0.030 0.500 –0.400 1.500 1.900 Source: Author’s computation based on data from Prati et al. (2013). reforms. Since such reforms are relatively recent, it is To illustrate this general result, we simulate possible that their growth effect is not accurately esti- the growth impact for Ethiopia if it were to move mated (see Prati et al., 2013 for details). Most likely, closer to the technology frontier. To test whether the the growth effect would be larger than the baseline simulated growth impacts of reforms in Ethiopia are estimates suggest. Since Ethiopia has not reformed affected by the country’s distance to the technology these two sectors, it can be conjectured that closing frontier, the countries in our sample are first assigned the reform gap with other countries could also yield to different quartiles depending on their respective growth benefits for the country. distances to the technological frontier.62 The regression The growth effect of reform generally increases model is then estimated for each quartile. Ethiopia as countries develop. Prati et al. (2013) find that both is currently in the first quartile and we estimate the real and financial sector reforms are positively associ- hypothetical effect if it were in the second quartile. ated with higher growth. However, their results also If Ethiopia were more developed, predicted suggest that reforms are more effective when markets growth payoffs would be even higher. Table 8.5 and institutions are not at their infancy but at a some- what more advanced stage in their process of develop- 62 Following Prati et al. (2013), we use the ratio of each country’s per ment. A similar result is reached by Christiansen et al. capita GDP to that of the United States as a proxy for its distance to (2013), who find that financial and trade reforms are the technology frontier in a given year. Note that the regression model estimated here includes a one-year lag of the ratio of each country’s robustly associated with economic growth, but only GDP per capita to that of the United States instead of per capita income in middle income countries. lagged one period. 124 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT presents the growth payoffs from getting closer to reinforcing, the full potential of each not being real- the technology frontier. If Ethiopia had the same dis- ized without adequate openness in the other (Duggan tance to the technology frontier as the countries in et al., 2013). Increased openness in the service sector the second quartile, closing gaps in current account not only implies increased foreign presence, but more and agriculture reforms would yield an additional broadly, it implies encouraging entry and inducing 0.68 and 1.03 percentage points increase in GDP per increased competition between foreign and domestic capita growth, respectively. Similarly, closing the gaps providers alike. in domestic finance and capital account liberalizations Services liberalization differs from goods trade would increase income per capita growth rate by an liberalization in its effects on domestic activity in additional 2.84 and 0.80 percentage points. This is the import-competing sector. Indeed, as pointed out not, however, the case for openness to international by Mattoo et al. (2006), for the case of services, liber- trade (as measured by average tariff rates) and reforms alization implies increased scale of domestic activity of the networks sector, the growth effects of which are in import competing sectors because foreign factors larger for countries in the first quartile (the farthest tend to locate domestically or domestic competition from the technology frontier) than for those in the increases by more effective regulation. These dynamics second quartile. In sum, the results suggest that addi- of competition will lead to better and more reliable tional domestic financial liberalization would provide provision of existing services, new varieties of services, the largest growth gain, followed by agriculture sector and competitive pricing. reforms, if Ethiopia moved closer to the technology The benefits of a more competitive services sec- frontier. tor typically have economy-wide effects.63 In fact, This simulation underscores a fundamental the inefficient supply of services inputs acts as a tax point: Even if growth payoffs are higher at higher on production of goods that use these services. Firms levels of development, they are sufficiently high at in any industry rely on financial services, on energy, Ethiopia’s current level of development. In other on telecommunications, on transport, on professional words, there is sufficient incentive even for low income services, etc. An efficient and well-regulated financial countries, such as Ethiopia, to initiate reform efforts sector is necessary to transform saving to investment at an early stage of development. efficiently, to ensure that resources are deployed where they have the highest returns. Improved efficiency in Productivity Impacts of Services 8.4  telecoms generates economy-wide effects as they are Trade Liberalization crucial for the dissemination of knowledge. Transport services contribute to an efficient distribution of goods In this section, we consider the microeconomic within a country and beyond. Professional accounting impact of services trade reform. Section 3 made services, for example, are key in reducing transaction use of the cross-country structural reform data set by costs—one of most significant impediments to growth Prati et al. (2013) to estimate the economic growth in Africa (Collier and Gunning, 1999). However, impact of reform at the macro level. In this section we access to good quality, reliable services varies substan- make use of the World Bank Enterprise Survey data tially across SSA and even across regions in Ethiopia. to estimate the economic impact of services sector A vast literature documents the links between liberalization on firm productivity. services sector reform and economic performance. Openness in the services sector is part and The existing work linking service sector reform and parcel of a comprehensive trade policy reform package. The benefits of opening up the services 63 A vast literature documents the links between services sector reform and goods markets are ‘multiplicative’ or mutually and economic performance. (See Hollweg et al. 2015 for details). Growth and Structural Reforms58 125 performance focuses on different channels: (i) services systematic econometric and case-study evidence that reform and economy wide gains, (ii) services reform point to three mechanisms at work when services and services sector performance, (iii) services reform trade is liberalized. First, services product variety and manufacturing export competitiveness, and increases. Second, quality improves. Third, services (iv) services reform and manufacturing productivity. input prices fall. These mechanisms are in turn asso- (See Hollweg et al. 2015 for details). ciated with improved firms’ productivity and export International experience suggest that the effects competitiveness. of liberalizing services trade on firms’ efficiency are We exploit cross-country data on industrial sizable. Because burdensome regulations and restric- dynamics (including Ethiopia), to gauge the cost tiveness to services trade affect firms’ input choices, in terms of productivity of poor service provision. they are also typically associated with productivity Table 8.6 summarizes the results and Annex 7.3 details costs for firms in downstream sectors. A more open the methodology. We find that productivity perfor- regime to services trade brings about increased FDI mance is heterogeneous across firms. As typically in the services sector (currently almost negligible found in the literature, firms that are more integrated in Ethiopia) that increases competition and leads in the global marketplace show productivity premia to improved performance. This is backed both by when compared to others in the same country, same TABLE 8.6: Labor Productivity Determinants Based on Perception of Services’ Performance Regional Regional Regional Regional Regional Regional Regional Regional Dep. Var. Labor Average Average Average Average Average Average Average Average Productivity All firms All firms All firms All firms Africa Africa Africa Africa Exporter 0.113*** 0.113*** 0.113*** 0.115*** 0.172*** 0.169*** 0.170*** 0.151** (0.017) (0.017) (0.017) (0.017) (0.058) (0.058) (0.058) (0.066) Firm size 0.101*** 0.102*** 0.102*** 0.104*** 0.096*** 0.101*** 0.100*** 0.088*** (0.009) (0.009) (0.009) (0.009) (0.029) (0.029) (0.029) (0.032) Firm age –7.9e–5 –7.5e–5 –8.2e–5 –0.0002 0.004*** 0.004*** 0.004*** 0.005*** (0.0004) (0.0004) (0.0004) (0.0004) (0.001) (0.001) (0.001) (0.001) Finance Obstacle –0.07*** –0.165* (0.03) (0.09) Transportation Obstacle   –0.08***   –0.03   (0.03)   (0.128) Electricity Obstacle   –0.036   –0.03   (0.025)   (0.09) Telecommunications   –0.08***   –0.113 Obstacle   (0.024)   (0.117) Constant 1.983*** 1.985*** 1.934*** 1.927*** 2.232*** 1.939*** 1.971*** 2.031*** (0.175) (0.174) (0.180) (0.169) (0.270) (0.308) (0.324) (0.253) Sector dummies Yes Yes Yes Yes Yes Yes Yes Yes Country-year dummies Yes Yes Yes Yes Yes Yes Yes Yes Observations 41,456 41,456 41,456 39,169 6,596 6,596 6,596 5,254 R-squared 0.172 0.172 0.172 0.176 0.195 0.194 0.194 0.218 Source: Hollweg et al. (2015). Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 126 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT sector and same size class. For example, exporters in percent, keeping all else equal. Similarly, if electricity the sample are 12 percent more productive than non- conditions were to also match those of Rwanda, the exporters, when looking across the 127 countries in labor productivity gains would be close to a 2.2 per- the sample and focusing on labor productivity (their cent. Finally, matching China’s transportation services TFP is 17.5 percent higher), and about 18.5 percent would imply productivity gains of 4.2 percent.64 more productive when focusing on African firms only (their TFP is 22 percent higher). Larger firms are 10 Reform Sequencing: Best Practice 8.5  percent more productive than medium-size firms, and and Ethiopia’s Experience this premium is homogeneous for all firms African firms. The premium on old firms is only significant The analysis thus far has revealed the potentially for African firms and suggests that each extra year of substantial economic benefits to liberalization, experience is associated with half a percentage point but is silent about reform sequencing. Economic extra of labor productivity. benefits seem particularly large for domestic finance, Firms’ productivity is affected by poor services current and capital account reform (section 4) as well provision. Evidence suggests that the quality of finan- as reforms in telecommunications and transport ser- cial, transport, electricity and telecommunications vices (section 5). If policy makers were interested in services provided to downstream firms, measured pursuing reform, what would be the best place to start? through firm’s perceptions affects firm’s performance We address this question by examining the normative significantly, both from a statistical point of view and guidance derived from theoretical and empirical work. from an economic point of view. The finding is rela- According to ‘interest group theory’ countries tively robust when using alternative measures of per- should liberalize trade and the capital account formance (labor productivity or TFP). When focusing simultaneously and prior to liberalizing the finan- on African firms only, the negative effect of poor cial sector. Rajan and Zingales (2003) argue that services provision on performance is blurrier, likely incumbent firms in a closed economy benefit from due to the smaller sample of firms, but still negative. lack of financial development because it denies poten- There are considerable gains to be reaped from tial new competitors the financial resources to enter services sector liberalization in Ethiopia, especially the market. Liberalizing trade and opening the capi- in credit access, energy and transport services. The tal account also disturbs the status quo by exposing largest effects of services provision on firms’ perfor- incumbents to external competition and by allowing mance are found on these two services. Slightly below domestic entrants to tap international capital markets. is the estimated effects of access to finance services, Trade openness would expose incumbent firms to while the estimated effect of electricity services is more competition and induce them to tap domestic slightly less than half the size of the average of the financial markets more to survive, leading to more previous effects. For electricity, the effect is not well- financial repression. That would be the case unless determined. To get a sense of the economic size of capital markets were opened simultaneously, thereby the effect, we combine the estimated coefficients for allowing incumbents to tap international markets. If each of these services with firms’ perceptions about so, liberalization of the domestic financial sector would the quality of their provision in Ethiopia, and simulate face less political opposition. the productivity-impact of reforms that would make firms perceive the same level of service provided in 64 These effects are calculated as the product of the change in the percep- comparator countries. For example, if Ethiopia’s access tion indicator needed to match the comparator’s perception about the given service and the estimated coefficient of the effect of the perception to finance conditions were to match those of Rwanda, about the given service on firms’ performance. This is essentially similar then firm labor productivity would increase by 4.3 to the benchmarking approached adopted in Section 4. Growth and Structural Reforms58 127 TABLE 8.7: International Best Practice Guidance on Reform Sequencing Method Political Economy Economic Theory Empirical Analysis Empirical Analysis Empirical Analysis Metric Reform progress Efficiency/growth Growth Macro stability As observed Approach Normative Normative Normative Normative Positive Step 1 Trade and Capital Trade Trade Trade Step 2 Financial Financial Financial, Capital Financial Financial/Capital Step 3 Capital Capital Source Rajan & Zingales McKinnon (1973, IMF (2007, 2008) IMF (2008) IMF (2008) (1993) 1991) Source: Own Elaboration. Note: ‘Normative’ refers to guidance as to the reform sequence countries should follow while ‘positive’ is an analysis of what reform sequence countries actually pursue in practice. A more popular strand in the literature, the outcome than the reverse sequence. A ‘trade first’ strat- normative order of economic liberalization, offers egy is also better for growth than a ‘big bang’ approach additional insights. McKinnon (1973, 1991) effec- of liberalizing all sectors at once. While there are no tively argues for the following sequencing of reforms: clear growth effects of alternative sequences, addi- (1) trade liberalization, (2) domestic financial sector, tional empirical results can be derived with respect to (3) opening the capital account.65 If trade liberalization macroeconomic stability: a liberalized financial sector preceded capital account opening then the resulting enjoys lower macroeconomic volatility and experience capital inflows would undermine competitiveness lower incidences of sudden stops. Moreover, volatility through real exchange rate appreciation while also and crisis risks in capital open economies are higher resulting in capital flight. The domestic financial when domestic financial sector liberalization is low, sector should also be liberalized before the external suggesting that the financial sector should be liberal- capital account, McKinnon argues. If not, capital ized before the capital account (IMF, 2008). inflows would lead to over-borrowing in foreign cur- Ethiopia’s reform sequence thus far has been rency, which a dysfunctional domestic financial sector consistent with international best practice even if would misallocate, and capital outflows could erode the country has been a slow reformer. Table 8.7 sum- the domestic deposit base. marizes this guidance based on alternative approaches. Empirical analyses is generally quite support- Accordingly, trade liberalization was the logical first ive of McKinnon’s theory. The data shows that most step for Ethiopia. countries tend to liberalize trade as their first reform step. As a second step, countries either liberalize the 8.6  Reform Risks domestic financial sector or the capital account in about equal frequency with a weak statistical tendency Although there are economic benefits to reforms as for financial reforms to come first (Hauner and Prati, well as an emerging consensus about their sequenc- 2008). What is the impact on growth of alternative ing, policy makers are often concerned about risks. reform sequences? The trade-first sequence is generally This section briefly discusses some of the reasons why good for growth as a liberal trade regime is involved policy makers may hesitate to pursue reforms. in both igniting growth and sustaining it. There is also empirical support for liberalizing trade before the 65 Note that only ‘trade liberalization first’ would be consistent with capital account as this yields a more favorable growth Rajan and Zingales (1993). 128 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT In general, resistance to reforms can be attrib- pursuing financial liberalization. Indeed, countries uted to a number of factors inherent to the reform that strengthened their supervisory and regulatory process itself. This includes: uncertainty about the frameworks prior to the introduction of liberalization benefits of reform; implementation costs of reforms fared better than those that liberalized first. The inad- are usually paid upfront while benefits take time to equate sequencing of reform measures had a lot to do materialize, and; some regulations tend to create with the disappointing results of liberalization in some rents which are shared among the beneficiaries of the countries. In the case of Ethiopia, it is often argued regulation.  To implement reforms, the government that the ability of the National Bank of Ethiopia also needs to create consensus among the different (NBE) to regulate foreign banks and other financial actors of the economy. Finally, reforms suffer from institutions is limited. NBE may not be familiar with ‘collective action’ challenges as their cost is concen- the services and products that foreign banks will trated on relatively small and well-defined interest bring with them and will therefore face difficulties in groups, while the benefits tend to be thinly spread effectively supervising (partly or fully) foreign owned over a much larger and less organized population financial services providers (BKP, 2007). (Olson, 1965). In Ethiopia, potential reforms in the services International experience with pursuing struc- sector would have negative implications for some tural economic reforms also seem to be somewhat SOEs and positive implications for consumers. mixed, even if average long term net benefits are Indeed, this is the expression of ‘costs to a few and positive across countries. One way of interpreting benefits to many’, mentioned above. The potential the empirical results presented in the previous sections introduction of more foreign competition in tele- is to think of them as the average, long term effects of com, trade logistics or finance would reduce the reform. Put differently, short term net benefits may be market shares and profits of Ethio Telecom, ESLSE negative for some countries and not all countries nec- and CBE. At the same time, there would be benefits essarily achieve net benefits in the long term. Similarly, to millions of consumers in the form of lower prices the emerging consensus around reform sequencing has and higher quality services. Government need not developed on the basis of country experience and the necessarily lose revenue as a result. In fact, it may earn lessons from the success and failures of early reformers. more revenues from taxation and license fees of new To illustrate, although financial liberalization reforms entrants. Moreover, by preparing suitable regulation, in Latin America in the 1980s were ‘particularly pain- Government can ensure that new entrants contribute ful’ (Arestis and Demetriades, 1997), they eventu- substantially and meaningfully to social objectives ally paved the way for robust financial institutions such as serving rural or marginal customers. Similarly, that successfully intermediated savings to support there are many nuances to reform and designs can be growth and were sufficiently robust to withstand the identified that mitigate some of the risks. For instance, 2008/09 global financial crisis. Similarly, the 1997–99 rather than posing a competitive threat to the domestic East Asian financial crisis is often cited as the prime banking industry, regulation can be introduced that example for why countries need to be careful not to facilitates joint ventures between foreign and domestic liberalize the capital account ‘prematurely’ (Stiglitz, banks with benefits to both sides. 2000). While these experiences offer benefits to late- Further detailed, sector-by-sector analysis comers of reform, there is no guarantee that reforms would be needed if the Government were to be will automatically bring strong positive net benefits interested in re-initiating the reform agenda. to a country such as Ethiopia. Reforms in finance, telecom and trade logistics share Putting in place an effective regulatory and common features, but obviously have their idiosyn- supervisory framework is a pre-requisite to cratic features. Moreover, detailed studies already exist Growth and Structural Reforms58 129 FIGURE 8.4: Structural Reform Indices for East Asian Countries 1. South Korea 2. Taiwan 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 3. China 4. Malaysia 1.0 1.0 0.9 0.9 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 Agriculture Networks Trade Domestic finance Capital account Current account Source: Own calculation’s based on Prati et al. (2013). for several of these sectors, although they may need 8.7 Quo Vadis Ethiopia? to be updated (see BKP 2007 for an example in the financial sector). Such studies highlight the complexity A good place to start embarking on reforms would and many nuances that need to be taken into account be to complete the process of trade liberalization by by policy makers. While the ‘devil is in the detail’, it also opening the services sectors to foreign firms. is clear from their analysis that adequate reforms can The analysis presented in Section 8.4 suggested that be designed for a country such as Ethiopia in a way trade logistics and telecom reforms could be particu- that maximizes potential benefits and minimize costs larly beneficial to Ethiopian manufacturing firms. and risks. Domestic financial liberalization could be a useful In sum, Ethiopia would need to proceed subsequent step of the reform process, according to thoughtfully and strategically if it is to reap the international best practice. potential economic benefits of structural reforms. East Asian developmental states have also That said, the evidence and country experience is gradually moved towards market liberalization, suggestive that the benefits of reform often outweigh although they did so at a slower pace and at a their costs, but this is not to say that there are no risks later stage of development. As discussed in Chapter involved nor that benefits are guaranteed. 2, Ethiopia aims to pursue a developmental state 130 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT approach inspired by the experiences of the East useful to keep an eye on when ‘bad times’ may be Asian Tigers. This model involves considerable state approaching. To that effect, in the subsequent and intervention in the allocation of resources. Structural last section of the report, we propose a series of indica- reforms often, but not always, implies the introduc- tors that would be worth monitoring going forward. tion of market-based mechanisms for resource allo- These indicators were designed to capture the many cation given that these are typically most efficient. trade-offs that are embedded in the current growth This raises the question, then, of how the East Asia strategy. At some point, we argue, the costs of pursu- developmental states have progressed in terms of their ing the current policy would outweigh its benefits. structural reforms. Countries such as Korea, Malaysia E.g. the loss of competitiveness associated with an and Taiwan, score quite highly in the reform indices overvalued exchange rate would outweigh the benefits developed by Prati et al. (2013). As illustrated in in the form of cheaper public capital imports. Policy Figure 8.4, these countries tend to exhibit rising trends makers would do well to anticipate such developments in all reform indices, reflecting increased liberalization and act in time. efforts.66 China, however, does not appear to be a big What would be the leading indicators that the reformer in this data set though this is in contrast with current growth model is no longer sustainable? received wisdom that sustained economic reforms Even if Ethiopia has found a strong blue print to were key to China’s double digit growth acceleration deliver economic growth, this blue print will have that lasted for three decades (e.g. Knight and Ding, to evolve over time and adjust to circumstances. 2012). One potential explanation could be that sev- Indeed, what distinguishes successful countries from eral of China’s reforms took place after the data set unsuccessful ones is their ability to adjust and adapt ends in 2006. In all cases, the resounding impression to evolving circumstances. How would policy mak- is that all countries seem to move towards gradually ers in Ethiopia know that it is time to make such an more liberalized markets, including the East Asian adjustment? Ideally, adjustments would be made prior developmental states. to the data showing a growth deceleration. In other In light hereof, Ethiopia would eventually be words, we need not wait for the growth deceleration expected to re-initiate its structural reform agenda. to come before policy action is taken. Ethiopia has been delaying a major reform effort for The need to monitor the performance of years, namely the process of joining the WTO and ini- Ethiopia’s growth model arises from the presence tiating the process of liberalizing its services sector. Our of a number of policy trade-offs. Put differently, the analysis of normative reform sequencing and country model has both benefits and costs. As long as growth experiences leads us to conclude that i is exactly the is promoted in a sustainable manner, the net benefits right next step for Ethiopia to accede WTO and to prevail. However, at some point the costs may be make a credible services offer which is acceptable to higher than the benefits and then it is time to change WTO members. In doing so, Ethiopia would fol- gear. For instance, publically financed infrastructure low the trodden path of reform also followed by East provision yields clear growth benefits, but has costs Asian countries, which starts with trade liberalization in terms of rising domestic and external debt as well and gradually embraces financial sector liberalization. as financial crowding out of the private sector. We consider each of these in turn as we propose a list of Growth Model Sustainability 8.8  indicators to monitor (Table 8.8): Monitoring As policy makers are unlikely to initiate struc- 66 The introduction of capital account restrictions in Malaysia in the tural economic reforms in ‘good times’, it may be mid-1990s is the only notable exception to this trends. Growth and Structural Reforms58 131 TABLE 8.8: Ethiopia Growth Model Sustainability Indicators Main Indicator Supplementary Indicators Source 1. External Debt Sustainability External risk of debt distress IMF-WB DSA External debt to export ratio International bond markets Sovereign risk premium Moody’s, S&P , Fitch Sovereign credit risk rating 2. Domestic public debt sustainability Debt-to-GDP sensitivity analysis IMF-WB DSA Nominal interest rate NBE Inflation CSA 3. External competitiveness Real effective exchange rate IMF Exports NBE Trade and services balance NBE Trade logistics performance WB FDI NBE 4. Private sector credit/forex shortage Total outstanding private credit NBE Black market premium NBE Anecdotal evidence of shortage Marginal returns to public and 5.  Public and private investment MOFED (National Accounts) private investment Firm surveys WB Doing Business, WB Enterprise Surveys, Global Competitiveness Index, National Dialogue 6. Marginal cost of financing Terms of external non-concessional loans MOFED (Debt Directorate) 7. Inflation Consumer prices CSA Producer prices CSA Nominal wages CSA 8. Government recurrent spending Capital and recurrent spending MOFED Operations and maintenance Public wage bill (real terms) 9. Domestic resources mobilization M2/GDP NBE Domestic credit to GDP NBE Domestic revenue to GDP MOFED 10. Corruption Corruption perception indices World Governance Indicators Anecdotal evidence Transparency International Mo Ibrahim i. External Public Debt Sustainability. The annual At the same time export growth started to slow joint IMF-World Bank Debt Sustainability down substantially. As a consequence, the risk Assessment (DSA) offers insights into this indi- rating was labelled ‘low on the cusp to moder- cator. Following HIPC and MDRI debt relief, ate’ in 2014, and in 2015, the risk rating was Ethiopia’s debt burden diminished considerably. downgraded to ‘moderate’. The rising imbalance To illustrate, the external risk of debt distress rat- between the level of external debt and the poor ing was classified as ‘low’ in the years 2011–13. export performance is a potential vulnerability However, in recent years, the authorities have that deserves careful monitoring, especially the increasingly resorted to external non-concessional Present Value (PV) of external debt-to-exports borrowing to finance infrastructure investments. indicator. 132 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT ii. Domestic Public Debt Sustainability. The Joint the government can keep an eye on the degree IMF-Bank DSA also provides a good overview of crowding out in the financial markets. The of public debt sustainability. While public debt shortage of forex is fluctuating over the year and is currently sustainable, this assessment is based the best two indicators would be: (a) the forex on nominal interest rates on public sector bor- black market premium, and; (b) evidence from rowing remaining significantly below inflation. firms expressing that they lack forex or credit With inflation projected to remain at a single- (see below). digit level, under current policies interest rates v. Marginal returns to public and private invest- on SOEs’ domestic borrowing would be at nega- ment. In an optimum, we would want to make tive real interest rate. Ethiopia’s relatively benign sure that the marginal return to private invest- public sector debt outlook hinges strongly on ment is equal to the marginal return to public the continuation of these financing conditions. investment (Eden and Kray, 2014). As public If the actual cost of borrowing were to rise above infrastructure investment increases, it is reasonable inflation, the debt indicators would worsen or to expect that their marginal benefits decline. But fiscal adjustment would be required to maintain how would we know this in practice? A key prob- sustainability. lem is that two effects are conflated: (1) the short- iii. External competitiveness and the real exchange term economic activity effect affecting aggregate rate. As explained earlier, Ethiopia has a policy demand in any given year; (2) the long-term effect preference for maintaining a strong real exchange of enhanced private sector productivity because rate. One important benefit hereof is that public of better infrastructure. It’s the second effect we capital imports are cheapened. The drawback of a are interested in, but we don’t have any hard data. strong currency, however, is that it affects export On the flip side we are interested in the marginal performance and the trade balance. Exports of return to private investment. Again, in the absence goods and services have performed poorly over of hard data, we could analyze the constraints the past 3 years (FY13-FY15) hardly register- of doing business to gauge whether credit and ing any growth compared with a normal annual forex is mention as bigger problems than public growth rate of 20 percent. While the main culprit infrastructure shortages. There are four sources is lower international commodity prices, the real with alternative frequencies: DB, CGI, national exchange rate also has an important impact on consultations, and Enterprise survey data. exports. Haile (2015) estimates that a 1 percent vi. Marginal cost of financing. In optimum, public real devaluation would increase exports by half a investment should be financed as long as the mar- percentage points. The effect is higher for manu- ginal benefits equal the cost of financing (Eden facturing 1 percentage point than for agriculture and Kraay, 2014). In the absence of reliable infor- 1/3 percentage points. Obviously, there are many mation about the benefits of public investment, other ways to promote exports (see World Bank, policy makers could monitor the external cost of 2013), but the bottom line is that exports, trade financing. External non-concessional financing balance, the real exchange rate, the black market is arguably a good indicator of the marginal cost premium are indicators to watch. of financing. Arguably, the probability that the iv. Private sector credit and forex shortage. In light optimality condition holds is declining in the of the rationing of credit and foreign exchange in marginal cost of financing. In other words, the favor of public infrastructure projects, it is clear risks that public projects receive financing when that the private sector is constrained. By monitor- they should not have been financed increases as ing the total outstanding credit by major sector, the marginal financing cost increases. Growth and Structural Reforms58 133 vii. Inflation. There are several reasons to consider this re-built because they have not been properly indicator. From a macroeconomic perspective, it maintained. is clear that Ethiopia’s growth model is supported ix. Domestic resources mobilization (savings and by expansionary fiscal and monetary policies. taxes). Ethiopia maintains a low deposit real Moreover, economic activity is very high. In this interest rate. This may explain the demonetiza- environment, we would also expect relatively high tion trend observed over the past decade where inflation rates. A model based on high public broad money and total outstanding credit have investment may also run into challenges associated declined substantially as a share of GDP. In such with absorptive capacity constraints. High demand an environment there is less money for both pub- for construction services and material may drive lic and private investment. In addition, govern- up domestic prices and wages and hence induce ment revenue as a share of GDP has remained at inflation. High inflation has a number of economic low levels, although government has been able disadvantages and is also a variable that directly to finance public investment through implicit affects the well-being of the population. revenues arising from seignorage and financial viii. Recurrent government spending. Ethiopia repression. Raising savings rates and high tax benefitted from keeping government consump- revenues are critical to support the sustainability tion low as this created fiscal space to finance of the growth model. public investment. What government needs to x. Governance and corruption. Ethiopia does rela- make sure, however, is that public employment tively well on indicators of corruption, including is adequately remunerated. While studies suggest the World Wide Governance indicators. However, that real civil servant salaries have declined over since construction is an activity which is inher- the decade (World Bank, 2015d) they also suggest ently associated with corruption and Ethiopia is that public employment is a good proposition for constructing more than even, it is reasonable to most urban workers (World Bank, 2015a). Still, assume that challenges of corruption are rising. experience from other countries suggest that cor- Indeed, this trend is consistent with anecdotal evi- ruptive practices are associated with low public dence. Corruption is a hard indicator to monitor sector remuneration. In addition, it is critical that in practice. In addition to relying on international government sets enough money aside for opera- indices, such as Transparency International and tions and maintenance. O&M helps preserve the Mo Ibrahim indices, one would also need to rely value of the initial investment and it can become on records of the Anti-Government Commission very costly if infrastructure facilities need to be as well as anecdotal and evidence. 134 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Annex 8.1 Methodology for Estimating The analysis is based on data for more than ninety Growth Impact of Reforms developed and developing countries covering the period 1973–2006. We use the cross-country growth regression model The simulated growth impact of alternative in Prati et al. (2013). It examines the association values of the reform indices corresponding to the between structural reforms and economic growth and benchmark countries can be computed based on takes the following form: the following equation: ∆yi ,t = β0 yi ,t −1 + β1reformi ,t −1 + θ Xi ,t −1 + ηi + ν i + εit (1) where the subscripts i and t denote country and year respectively; yi,t is the log of real GDP per capita of country i at time t, hence the difference ∆yi,t repre- where is the annual growth rate of GDP per capita sents the annual growth rate of per capita income; yi,t–1 obtained based on the average value of the benchmark stands for one-period lagged GDP per capita, with β0 country’s reform index, with bc denoting benchmark measuring the speed of convergence in income per country; represents the annual GDP per capita capita across countries; reformi,t–1 represents the indi- growth derived using the average value of the reform cators of structural reforms in the real and financial index for Ethiopia; and stand for sectors; β1 is the key parameter of interest and cap- the average values of the reform indices for Ethiopia tures the effect of reforms on GDP per capita growth; and the benchmark country, respectively; reform gap Xi,t is a vector comprising a set of control variables is thus the gap in the average levels of sectoral reforms that may affect both ∆yi,t and reformi,t–1; ηi is a set between the comparator country and Ethiopia; growth of country fixed effects accounting for unobserved payoff is the change in GDP per capita growth asso- country-specific and time-invariant factors (such as ciated with closing the aforementioned reform gap; geographical location, historical legacies, and legal and the remaining variables are as defined previously. origins) that may have significant bearing on both Equation (2) shows that ceteris paribus the growth pay- the introduction of reforms and economic growth; off from closing the reform gaps with the benchmark υi is a vector of time dummies capturing universal country is given by the product of the reform gap and time trends; and finally, represents the error term. the estimated coefficient δ1. Growth and Structural Reforms58 135 Annex 8.2 Robustness Checks sectors) turns out to be negative, which seems coun- terintuitive, albeit still statistically insignificant. In this section, we run a wide array of robustness Another cause for concern is that the empirical checks to test the validity of our baseline findings. results on the reform-growth nexus are heteroge- The sensitivity analysis is performed based on the neous across different time periods. Thus, Annex original model of Prati et al. (2013), who also con- Tables 3 and 4 report the predicted growth effects duct several econometric exercises to investigate the for the periods 1973–89 and 1990–2006, which sup- robustness of their main results. posedly represent homogenous periods as most of the We first re-estimate the baseline regressions countries in our sample implemented major structural using the GMM estimator proposed by Arellano reforms in the 1990s and early 2000s. Overall, the and Bond (1991).67 This more or less circumvents main results for the full sample generally hold for the the inconsistency of fixed-effect OLS estimates arising subsamples as well. from the correlation between the lagged dependent Finally, we conduced some sensitivity tests (not variable and the lagged error term.68 The simulation reported here). These indicate that the predicted results based on the GMM estimates are reported in growth impacts of reforms based on coefficient esti- Annex Table A.1.69 We note that the potential growth mates from a regression that includes only emerging effects of closing reform gaps are qualitatively similar and developing economies are generally consistent to the baseline outcomes, although the latter are in with our baseline findings. Further, the main findings most cases quantitatively larger. This is to be expected generally prove robust to using three- and five-year as the GMM regressions, unlike the baseline specifica- interval data instead of employing annual data, which tions which examine the association between reforms are often considered to be prone to measurement error. and GDP per capita growth, look at the effects of reforms on GDP per capita. Next we check that omitted variables are not 67 Note in passing, however, that the Arellano and Bond (1991) first- differenced GMM estimator may suffer from large finite-sample biases biasing our results. We compute the predicted and poor precision when the time series are persistent. In such cases, growth effects based on a regression that includes the lagged levels of the series are weakly correlated with the lagged first differences, thereby making the instruments for the first-differenced additional set of time-varying control variables: politi- equations weak (Blundell and Bond, 1998). cal institutions (proxied by the Polity IV indicator of 68 For fixed N, OLS estimates are consistent only for T ➝ ∞. Although the number of time periods in our sample is not too small, the persistence democracy), terms of trade, and tertiary educational of the lagged dependent variable can still render fixed-OLS estimates attainment. Annex Table A2 shows that the results inconsistent (Wooldridge, 2010). 69 Note that Equation (1) was rearranged before applying the GMM are more or less in line with our previous findings. estimator. More specifically, the baseline specification is rewritten as However, the sign of the coefficient for reforms in a dynamic model where the lagged dependent variable appears in the right-hand-side of the equation: the networks sector (as well as the separate indices for reforms of the electricity and telecommunications yi,t = a0yi,t−1 + a1reformi,t−1 + θXi,t−1 + ηi + vi + εit, where a0= 1 + β0. 136 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT TABLE A8.2.1: GMM Regression: Coefficient Estimates and Growth Impact of Reforms Coefficient Predicted effect on real GDP per capita, growth rate (%) Structural Reforms estimates* Uganda Tanzania Sri Lanka Ghana SSA LIC LMI Real sectors Trade 0.031 0.33 0.23 0.48 0.29 0.26 0.20 0.24 Current account 0.054 3.04 1.69 1.41 1.13 1.22 1.32 1.28 Agriculture 0.055 3.67 1.83 1.83 0.00 1.02 0.72 0.44 Network 0.001 0.05 0.01 0.01 0.05 0.02 0.01 0.01 Electricity 0.018 1.07 0.06 0.47 1.42 0.59 0.21 0.19 Telecom –0.014 –0.48 –0.28 0.00 –0.48 –0.16 –0.13 –0.05 Financial sectors Domestic finance 0.116 3.65 5.16 3.97 1.93 2.45 3.49 2.74 Banking 0.095 4.01 5.07 3.69 1.69 2.47 3.56 2.87 Securities 0.07 –1.56 0.00 0.78 0.78 –0.26 –0.51 –0.66 Capital account (CA) 0.037 2.78 0.46 0.93 0.46 1.30 1.24 0.97 CA (resident) 0.034 2.55 0.00 0.00 0.00 1.03 1.02 0.72 CA (nonresident) 0.024 1.80 0.60 1.20 0.60 1.05 0.93 0.71 Source: Author’s computation based on data from Prati et al. (2013). Note: Coefficient estimates obtained from regressing GDP per capita growth on lagged GDP per capita and the reform indices one at a time. All estimated coefficients are based on the Arellano and Bond (1991) first-differenced GMM estimator. Two and more lags of the independent vari- ables are used as instruments. The figures in the last seven columns are in percentage points and represent the growth payoffs from closing the reform gaps between Ethiopia and the benchmark country in the second row. TABLE A8.2.2: Control Variable Check: Coefficient Estimates and Growth Impact Structural Coefficient Predicted effect on real GDP per capita, growth rate (%) Reforms estimates* Uganda Tanzania Sri Lanka Ghana SSA LI LMI Real sectors Trade 0.021 0.22 0.15 0.33 0.19 0.17 0.14 0.16 Current account 0.034 1.91 1.06 0.89 0.71 0.77 0.83 0.81 Agriculture 0.024 1.60 0.80 0.80 0.00 0.44 0.31 0.19 Network –0.009 –0.41 –0.11 –0.11 –0.49 –0.20 –0.09 –0.06 Electricity –0.001 –0.08 0.00 –0.04 –0.11 –0.05 –0.02 –0.02 Telecom –0.013 –0.42 –0.25 0.00 –0.42 –0.14 –0.11 –0.05 Financial sectors Domestic finance 0.060 1.89 2.67 2.06 1.00 1.27 1.80 1.42 Banking 0.046 1.94 2.45 1.79 0.82 1.20 1.73 1.39 Securities 0.035 –0.78 0.00 0.39 0.39 –0.13 –0.25 –0.33 Capital account (CA) 0.022 1.65 0.28 0.55 0.28 0.78 0.74 0.58 CA (resident) 0.017 1.28 0.00 0.00 0.00 0.51 0.51 0.36 CA (nonresident) 0.016 1.20 0.40 0.80 0.40 0.70 0.62 0.48 Source: Author’s computation based on data from Prati et al. (2013). *Coefficient estimates obtained from regressing GDP per capita growth on lagged GDP per capita and the reform indices one at a time. The figures in the last seven columns are in percentage points and represent the growth payoffs from closing reform gaps between Ethiopia and the respective benchmark countries in the second row. Growth and Structural Reforms58 137 TABLE A8.2.3: Coefficient Estimates and Potential Growth Impact (1973–1989) Structural Coefficient Predicted effect on real GDP per capita, growth rate (%) reforms estimates* Uganda Tanzania Sri Lanka Ghana SSA LIC LMIC Real sectors Trade 0.022 0.23 0.16 0.34 0.20 0.15 0.17 0.18 Current account 0.051 2.87 1.59 1.33 1.06 1.25 1.21 1.16 Agriculture 0.002 0.13 0.07 0.07 0.00 0.03 0.02 0.04 Network 0.102 4.64 1.24 1.24 5.56 1.03 0.68 2.24 Electricity 0.065 3.91 0.22 1.74 5.22 0.78 0.69 2.18 Telecom 0.062 2.07 1.21 0.00 2.07 0.54 0.22 0.70 Financial sectors Domestic finance 0.053 1.67 2.36 1.82 0.88 1.59 1.25 1.12 Banking 0.043 1.82 2.29 1.67 0.76 1.61 1.30 1.12 Securities 0.032 –0.71 0.00 0.36 0.36 –0.23 –0.30 –0.12 Capital acct. (CA) 0.054 4.05 0.68 1.35 0.68 1.81 1.41 1.90 CA (resident) 0.038 2.85 0.00 0.00 0.00 1.14 0.80 1.15 CA (nonresident) 0.037 2.78 0.93 1.85 0.93 1.44 1.10 1.63 Source: Author’s computation based on data from Prati et al. (2013). Notes: SSA, Sub-Saharan Africa; LIC, Low-income countries; LMIC, Lower-middle-income countries. *Coefficient estimates obtained from regressing GDP per capita growth on lagged GDP per capita and the reforms indices one at a time. The figures in the last seven columns are in percentage points and represent the growth payoffs from closing the reform gaps between Ethiopia and the respective benchmark countries in the second row. TABLE A8.2.4: Coefficient Estimates and Potential Growth Impact (1990–2006) Structural Coefficient Predicted effect on real GDP per capita, growth rate (%) reforms estimates* Uganda Tanzania Sri Lanka Ghana SSA LIC LMIC Real sectors Trade 0.022 0.19 0.13 0.28 0.17 0.12 0.14 0.15 Current account 0.051 2.19 1.22 1.02 0.81 0.96 0.93 0.88 Agriculture 0.002 3.13 1.57 1.57 0.00 0.61 0.38 0.87 Network 0.102 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Electricity 0.065 0.39 0.02 0.17 0.52 0.08 0.07 0.22 Telecom 0.062 –0.19 –0.11 0.00 –0.19 –0.05 –0.02 –0.07 Financial sectors Domestic finance 0.053 3.90 5.51 4.25 2.07 3.73 2.93 2.62 Banking 0.043 4.10 5.17 3.77 1.72 3.64 2.93 2.53 Securities 0.032 –1.40 0.00 0.70 0.70 –0.46 –0.60 –0.23 Capital acct. (CA) 0.054 1.28 0.21 0.43 0.21 0.57 0.44 0.60 CA (resident) 0.038 1.35 0.00 0.00 0.00 0.54 0.38 0.54 CA (nonresident) 0.037 0.15 0.05 0.10 0.05 0.08 0.06 0.09 Source: Author’s computation based on data from Prati et al. (2013). Notes: SSA, Sub-Saharan Africa; LIC, Low-income countries; LMIC, Lower-middle-income countries. *Coefficient estimates obtained from regress- ing GDP per capita growth on lagged GDP per capita and the reform indices one at a time. The figures in the last seven columns are in percentage points and represent the growth payoffs from closing the reform gaps between Ethiopia and the respective benchmark countries in the second row. 138 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Annex 8.3 Estimating the Impact of performance of services, controlling for factors Services Inputs Quality on Firms’ relevant for firm performance. Factors typically Productivity identified in the literature include firm’s export status, firm’s size, and firm’s age. In addition, we control for How does access to quality services inputs affect country-year fixed effects, to eliminate the potential firms’ performance? Answering this question requires of distortions due to changes in the relative values access to a dataset on the basis of which we can obtain of the different currencies in which output, wages, comparable measures of quality services input provi- intermediates and capital stock are expressed and to sion and of firms’ performance. Then, it is necessary eliminate the effect of country-year unobservables to test whether a systematic relationship exists between that may affect both productivity and the percep- the two. Our approach follows that of Arnold, Mattoo tion of services’ quality, as well as sector fixed effects and Narciso (2006). to control for time-invariant and sector-specific The dataset comes from the World Bank unobservables. Enterprise Surveys. Data from these surveys are avail- Concerns about endogeneity arise because it able for a cross-section of firms from 188 country-year is possible that poor performance affects firms’ combinations (127 countries are surveyed, with some perceptions about the obstacles that services input countries being surveyed in more than one year, of provision represent. This would imply a bias upwards which 42 are African countries (including Ethiopia)). in the coefficient linking services performance with The surveys were undertaken between 2006 and 2013. productivity. This makes a specification that links The measure of firm performance chosen is pro- firm-level perceptions of services quality with firm- ductivity. We use three alternative measures: (i) labor level productivity inappropriate. Our strategy, fol- productivity (the ratio of output to total labor costs), lowing Arnold et al (2006) consists in aggregating (ii) total factor productivity (TFP) estimated in two the individual firm’s responses to the services-related as a residual of a Cobb-Douglas production function, questions on the right hand side at the regional level, with output as a function of the capital stock, labor within each country. This reduces the influence that and intermediate inputs; and (iii) TFP estimated as a an individual firm’s performance has on the regressor. residual from a translog specification in which output In addition, it is likely to better summarize the quality is expressed as a function of the capital stock, labor, provision of services in a given region. intermediate inputs and their squared terms, and their The chosen specification is as follows: cross-products. The performance of services sectors is also µi = αct + γ s + β ServPerformancer + π Xi + εi (1) obtained from the Enterprise Surveys. We used sub- jective measures of local services performance, which where µ is the indicator of productivity (labor produc- are firms’ valuations as to how much of a constraint tivity, residual from Cobb Douglas or residual from they consider electricity, telecommunications, trans- translog), α is a country-year fixed effect, is a sector port, and access to finance for their businesses. Firms fixed effect, ServPerformance is a vector of perception are asked to select, on a scale from 0 to 4, whether based indicators of obstacles represented by access to they consider each of these dimensions to be not an finance, electricity, transport, and telecommunica- obstacle for their operations (0), a minor obstacle (1), tions, that vary at the regional level, X is a vector of a moderate obstacle (2), major obstacle (3) and severe controls varying at the firm level, and ε is an error obstacle (4). term assumed orthogonal to the regressors. The empirical strategy consists in regressing We focus on the impact on domestic firms, the measure of productivity on measures of the so all regressions are estimated on a sample of Growth and Structural Reforms58 139 domestic-owned firms or firms with less than 10% of African countries, including Ethiopia (42). Specific of foreign ownership. Results are presented for the models for Ethiopia were not fitted due to the insuf- entire cross-section of countries (188), and for the set ficient size of the sample. 141 REFERENCES Abiad, A., Detragiache, E., Tressel, T., 2008, ‘A New Uganda. Journal of International Development, Database of Financial Reforms’. IMF Working (8), 415–444. Paper 08/266. International Monetary Fund, Araujo, J., M. Brueckner, M. Clavijo, E. Vostroknutova, Washington, DC. and K. Wacker, 2014, ‘Benchmarking the Abramovitz, M., 1986, ‘Catching-Up, Forging Ahead Determinants of Economic Growth in Latin and Falling Behind, The Journal of Economic America and the Caribbean’, World Bank, History 46(2): 385–406. Washington DC. Addison, T. and A. Geda, 2001, ‘Ethiopia’s New Arellano, M. and O. Bover, 1995, ‘Another look at Financial Sector and its Regulation’, WIDER the instrumental variable estimation of error- Discussion Paper 2001/55. components models.’ Journal of Econometrics Afonso, A. and D. Furceri, 2010, ‘Government 68(1): 29–51 size, volatility and economic growth.’ European Arestis P. and P. Demetriades, 1997, ‘Financial Journal of Political Economy 26(4): 517–532 Development and Economic Growth: Assessing Aiyagari, S.R., L.J. Christiano and M. Eichenbaum, the Evidence’, The Economic Journal 107 (May): 1992, ‘The output, employment, and interest rate 783–799. effect of government consumption.’ Journal of Arezki, R. and M. Brueckner, 2012, ‘Commodity Monetary Economics 30: 73–86 Windfalls, Democracy, and External Debt.’ Alemu, D., Rashid, S., Tripp, R. (2010). Seed sys- Economic Journal 122: 848–866. tem potential in Ethiopia: Constraints and Asfaw, A., Admassie, A. (2004). The role of education opportunities for enhancing the seed sector. on the adoption of chemical fertiliser under dif- IFPRI. Downloaded from: http://www.ata.gov. ferent socioeconomic environments in Ethiopia. et/wp-content/uploads/Ethiopia-Seed-System- Agricultural Economics, 30(3), 215–228. Diagnostic-July-2010.pdf ATA-MoA (Agricultural Transformation Agency – Allaro, H. B., 2012, ‘The Effect of an Export-Led Ministry of Agriculture) (2014). Seed system Growth Strategy on the Ethiopian Economy’, Development Strategy, downloaded from: http:// American Journal of Economics 2(3): 50–56. www.ata.gov.et/wp-content/uploads/20130415_ Alston, J. M., Chan-Kang, C., Marra, M. C., Pardey, Overall-seed-system-strategy_vFINALDRAFT_ P. G., & Wyatt, T. J. (2000). A meta-analysis of Offical-Release_final.pdf rates of return to agricultural R&D: Ex pede Bachewe, F., Berhane, G., Minten, B., and Taffesse, Herculem? (Vol. 113). Intl Food Policy Res Inst. A.S. (2015). ‘Agricultural Growth in Ethiopia Ang, J., 2009, ‘Private Investment and Financial (2004–2014): Sector Policies in India and Malaysia’, World Evidence and Drivers’, International Food Policy Development 37(7): 1261–73. Research Institute (IFRPI), Ethiopia Strategy Appleton, S. M., & Balihuta, A. (1996). Education Support Program (ESSP), Background Paper and agricultural productivity: evidence from Prepared for this Report. 142 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Bakaert, G., C. Harvey, and C. Lundblad, 2005, ‘Does Caselli, F. and J. Feyrer, 2007, ‘The Marginal Product Financial Liberalization Spur Growth?’, Journal of Capital’, The Quarterly Journal of Economics: of Financial Economics 77: 3–55. 122(2):535–568. Baldwin, R., 2011, ‘Trade and Industrialization after Central Statistical Agency of the Federal Democratic Globalization’s 2nd Unbundling: How Building Republic of Ethiopia. Various years. Agricultural and Joining a Supply Chain are Different and Sample Survey: Volume I-Report on Area and Why it Matters’, NBER Working Paper 17716. Production of Crops (Private Peasant Holdings, Baldwin, R., P. Martin, and G. Ottaviano, 2011, ‘Global Meher Season), (Central Income Divergence, Trade and Industrialization: Christiansen, L., Schindler, M., and Tressel, T., The Geography of Growth Take-Offs’, NBER 2013, ‘Growth and Structural Reforms: A New Working Paper 6458. Assessment.’ Journal of International Economics, Bazzi, S. and M. Clemens, 2013, ‘Blunt Instruments: 89: 347–356. Avoiding Common Pitfalls in Identifying Collier, P., 2006, ‘Africa and Growth’, Federal Reserve the Causes of Economic Growth’, American Bank of Kansas City, Fall, pages 18–21. Economic Journal: Macroeconomics 5(2): Comin D. and B. Hobijn, 2010, ‘Technology 152–186. Diffusion and Postwar Growth’, NBER Working Benin, S. (2014). Implementation performance and Paper 16378. progress towards core CAADP targets, presen- Commission on Growth and Development, 2008, tation at the 2014 Resakks annual conference, ‘The Growth Report – Strategies for Sustained October 8–10, Addis Ababa. Growth and Inclusive Development’, World Benson, T., Spielman, D., Kasa, L. (2014). Direct Bank, Washington. D.C. seed marketing program in Ethiopia in 2013: An Dabla-Norris, E., J. Brumby, A. Kyobe, Z. Mills, operational evaluation to guide seed sector reform, and C. Papageorgiou, 2011, ‘Investing in Public IFPRI Discussion Paper 01350, Washington DC. Investment: An Index of Public Investment Berhane, G., Woldu, T.A., Taffesse, A.S., Tadesse, Efficiency’, IMF Working Paper 11/37, F. (forthcoming). “Understanding Ethiopia’s Washington D.C. Extension System”. ESSP Discussion Paper. Dabla-Noris, E., Ho, G., Kochhar, K., Kyobe, A., Bigsten, A., M. Gebreeysus, and M. Soderbom, Tchaidze, R., 2014, ‘Anchoring Growth: The 2009, ‘Gradual Trade Liberalization and Firm Importance of Productivity-Enhancing Reforms Performance in Ethiopia’, Center for the Study in Emerging Market and Developing Economies.’ of African Economies Working Paper 2009–21. Journal of International Commerce, Economics, BKP Development and Economisti Associati Srl, and Policy 5(2). 2007, ‘The Impact of WTO Accession on the Davis, K., Swanson, B., Amudavi, D., Mekonnen, Financial Services Sector of Ethiopia – A Study D.A., Flohrs, A., Riese, J., Lamb, C., Zerfu, Commissioned by the Ministry of Trade and E. (2010). In-depth Assessment of the Public Industry’. Agricultural Extension System of Ethiopia and Blundell, R., and S. Bond, 1998, ‘Initial conditions Recommendations for Improvement. IFPRI and moment restrictions in dynamic panel data Discussion Paper 01041. models.’ Journal of Econometrics 87: 115–43. Deininger, K., Ayalew, D., Alemu, T. (2011). Brueckner, M., 2013, ‘Background Paper for Latin Impacts of Land Certification on Tenure Security, America and the Caribbean’s Study on the Investment, and Land Market Participation: Determinants of Economic Growth: Econometric Evidence from Ethiopia. Land Economics, Analysis’, World Bank, Washington D.C. 87(2):312–334 References 143 Deininger, K., Ayalew, D., Holden, S., Zevenbergen, Easterly, B., 2002, ‘Growth in Ethiopia: Retrospect J. (2008). Rural Land Certification in Ethiopia: and Prospect’, unpublished manuscript prepared Process, Initial Impact, and Implications for for the World Bank. Other African Countries. World Development, Eden, M. and A. Kraay, 2014a, ‘Crowding in and 36(10):1786–1812 the Returns to Government Investment in Low- Demetriades, P. and K. Luintel, 2001, ‘Financial income Countries’, World Bank Policy Research restraints in the South Korean miracle’, Journal Working Paper 6781. of Development Economics, 64: 459–479. Eden, M. and A. Kraay, 2014b, ‘Crowding in and the Dercon, S., & Christiaensen, L. (2011). Consumption Returns to Government Investment in Low-income risk, technology adoption and poverty traps: Countries: Policy Implications for Ethiopia’, unpub- Evidence from Ethiopia. Journal of development lished manuscript prepared for the World Bank. economics, 96(2), 159–173. Eden, M., 2015a, ‘Is Financial Repression Part of An Dercon, S., Gilligan, D. O., Hoddinott, J., Optimal Policy?’, Background note prepared for Woldehanna, T. (2009). The impact of agricul- this report. tural extension and roads on poverty and con- Eden, M. 2015b, ‘The Rate of Return to Government sumption growth in fifteen Ethiopian villages. Capital in Ethiopia’, Background note prepared American Journal of Agricultural Economics, for this report. 91(4), 1007–1021. Endale, K. (2011). Fertilizer Consumption and Dinh H., V. Palmade, V. Chandra, and F. Cossar, Agricultural Productivity in Ethiopia. Working 2012, ‘Light Manufacturing in Africa: Targeted Paper 003. Addis Ababa: Ethiopia Development Policies to Enhance Private Investment and Create Research Institute. Jobs’, World Bank. EPPCF, ‘Ethiopia’s National Business Agenda’, 2014, Dollar, D. and A. Kraay, 2003, ‘Institutions, trade, Addis Ababa. and growth,’ Journal of Monetary Economics 50 Evenson, R.E., Gollin, D. (2003). Assessing the impact (1): 133–162. of the Green Revolution. Science, 300:758–762. Doornik, J. A., Hendry, D. F., and Pretis, F. (2013). FAO (2012). Current world fertilizer trends and out- Step-Indicator Saturation. Discussion Paper No. look to 2016. Food and Agricultural Organization 658, Department of Economics, University of of the United Nations. Rome. Oxford.  Galor, O. and D. Weil, ‘The Gender Gap, Fertility Dorosh, P., Wang, H. G., You, L., & Schmidt, E. and Growth’, American Economic Review 86(3): (2012). Road connectivity, population, and crop 374–87. production in Sub-Saharan Africa. Agricultural Gebregziabher, F. (2014). The long-run macroeco- Economics, 43(1), 89–103. nomic effects of aid and disaggregated aid in Duggan, Victor and Rahardja, Sjamsu and Varela, Ethiopia. Journal of International Development, Gonzalo J., Service Sector Reform and 26(4), 520–540. Manufacturing Productivity: Evidence from Gerschenkron, A., 1962, ‘Economic Backwardness in Indonesia (2013). World Bank Policy Research Historical Perspective, Belknap Press of Harvard Working Paper No. 6349. University Press. Durlauf, S., 2009, ‘The rise and fall of cross-country Ghani E. (ed.), ‘The Service Revolution in South Asia’, growth regressions’, in: Boianovsky and Hoover Oxford University Press. (eds.): Robert Solow and the development of Ghani, E. and S. O’Connell, 2014, ‘Can Service Be growth economics, Duke University Press: a Growth Escalator in Low Income Countries?’, 315–333 World Bank Policy Research Working Paper 6971. 144 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Ghani, E., 2012. ‘Service with a Smile’, Economic Hirvonen, K., Hoddinott, J. (2014). Agricultural Premise 96: 1–6, November. production and children’s diets: Evidence from Ghebru, H., Holden, S. (2013). Links between rural Ethiopia. Ethiopia Strategy Support Program tenure security and food security: Evidence Discussion Paper 69. Addis Ababa. from Ethiopia: IFPRI discussion papers 1288, Hoddinott, J., Headey, D., Dereje, M. (2013). Cows, International Food Policy Research Institute missing milk markets and nutrition in rural (IFPRI) Ethiopia. Ethiopia Strategy Support Program Ghebru, H., Koru, B., Taffesse, A.S. (2015). Further Discussion Paper 63. Addis Ababa. formalization or have we had enough? Land pol- Holden, S., Deininger, K., Ghebru, H. (2007). icy, agricultural transformation and smallholder Impacts of Low-Cost Land Certification on farmers in Ethiopia, ESSP Discussion paper, Investment and Productivity. American Journal forthcoming. of Agricultural Economics, 91(2):359–373 Growth and Development Centre, Research Hollweg, C., E. Rojas, and G. Varela, 2015, Memorandum 149. ‘Integration of Services and Manufacturing in Hagos, F., Makombe, G., Namara, R. E., Awulachew, Ethiopia’, Background Paper for this Report, S. B. (2009). Importance of irrigated agriculture World Bank. to the Ethiopian economy: Capturing the direct Hostland, M. and M. Giugale, ‘Africa’s Macroeconomic net benefits of irrigation. Colombo, Sri Lanka: Story’, World Bank Policy Research Working International Water Management Institute. 37p. Paper 6635. (IWMI Research Report 128) Huang, Y. and X. Wang, 2011, ‘Does Financial Haile, F., 2015, ‘Structural Reforms and Growth in Repression Inhibit or Facilitate Economic Ethiopia: A Benchmarking Exercise. Mimeo. Growth? A Case Study of Hauner D. and A. Prati, 2008, ‘Openness and Chinese Reform Experience’, Oxford Bulletin of Domestic Financial Liberalization: Which Comes Economics and Statistics 73(6): 833–855. First’, On the Causes and Consequences of Huffman, W. E. (2001). Human capital: Education Structural Reforms. Mimeo. and agriculture. Handbook of agricultural eco- Hausmann, R., 2001, ‘Prisoners of Geography’, nomics, 1, 333–381. Foreign Policy 122:44. IFDC (2012). “Ethiopia Fertilizer Assessment”. Hausmann, R., D. Rodrik and A. Velasco, 2005, International Monetary Fund, 2008, ‘Structural ‘Growth Diagnostics’, Working Paper. Reforms and Economic Performance in Advanced Hausmann, R., C. Hidalgo, S. Bustos, M. Coscia, S. and Developing Countries’, IMF Research Chung, J. Jimenez, A. Simoes, and M. Yildrim, Department. Washington D. C. 2014, ‘The Atlas of Economic Complexity: International Monetary Fund, 2013, ‘Ethiopia: 2013 Mapping Paths to Prosperity’. Cambridge, Article IV Consultation’, Washington D.C. Massachusetts: Harvard University, Center for International Monetary Fund, 2013, ‘Regional International Development. Economic Outlook – Sub-Saharan Africa: Hausmann, R., L. Pritchett and D. Rodrik, 2004, Keeping the Pace’, October, Washington D.C. ‘Growth Accelerations’, NBER Working Paper International Monetary Fund, 2014, ‘Ethiopia: 2014 10566. Article IV Consultation’, Washington D.C. Headey, D., Dereje, M., Taffesse, A. S. (2014). Land International Monetary Fund, 2014, World constraints and agricultural intensification in Economic Outlook, ‘Chapter 3: Is it Time for an Ethiopia: A village-level analysis of high-potential Infrastructure Push? The Macroeconomic Effects areas. Food Policy, 48, 129–141. of Public Investment’, Washington D.C. References 145 Jack, B.K. (2011). Constraints on the adoption of Lewis, Arthur. 1954. “Economic Development with agricultural technologies in developing coun- Unlimited Supplies of Labour.” The Manchester tries. White paper, Agricultural Technology School. Volume 22, Issue 2, pages 139–191. Adoption Initiative, J-PAL (MIT) and CEGA Li, J., Gorham, R., Nuru, M. (2011). Road (UC Berkeley). Development, Agglomeration and Agricultural Jamison, D. T., Lau, L. J. (1982). Farmer education Growth in Ethiopia: an Empirical Analysis. and farm efficiency. Johns Hopkins University Manuscript. Press. Lin, J., 2011, ‘New Structural Economics: A Kaufmann, D., A. Kraay and M. Mastruzzi, Framework for Rethinking Development,’ World 2010, ‘The Worldwide Governance Indicators: Bank Policy Research Working Paper 5197. Methodology and Analytical Issues.’ World Bank Loayza, N., P. Fajnzylber, and C. Calderón, 2005, Policy Research Working Paper No. 5430 ‘Economic Growth in Latin America and the Kedir, M., Schmidt, E., Tilahun, H. (2015). Caribbean: Stylized Facts, Explanations, and Production Patterns and Fertilizer Use. Mimeo. Forecasts’. World Bank, Washington D.C. Knight, J., Weir, S., Woldehanna, T. (2003). The role Martins, P., 2014, ‘Structural change in Ethiopia – of education in facilitating risk-taking and innova- an employment perspective’, World Bank Policy tion in agriculture. The Journal of Development Research Working Paper 6749. Studies, 39(6), 1–22. Martins, P., 2015, ‘Structural change in Ethiopia—an Knight, J., and S. Ding, 2012, ‘China’s Remarkable update using 2013 LFS data’, Background Paper Economic Growth’, Oxford University Press. prepared for this report. Kose, M. A., Prasad, E., Rogoff, K. and Wei, S. J., Mattoo, A., R. Rathindran, A. Subramanian (2006), 2009, ‘Financial globalization: a reappraisal’, IMF ‘Measuring Services Trade Liberalization and Its Staff Papers, Vol. 56, pp. 8–62. Impact on Economic Growth: An Illustration’, Kraay, A., in progress. ‘Weak Internal Instruments in Journal of Economic Integration 21(1):64–98. Dynamic Panel Growth Regressions: A Comment Maxwell, D., Majid, N., Stobaugh, H., Kim, J. J., on “Redistribution, Inequality, and Growth”.’ Lauer, J., Paul, E. (2014). Lessons learned from World Bank mimeo the Somalia famine and the greater Horn of Krishnan, P. and Patnam, M. (2014). Neighbors and Africa 2011–2012: Desk Review of Literature. extension agents in Ethiopia: Who matters more Feinstein International Center, Tufts University: for technology adoption?. American Journal of Medford, USA. Agricultural Economics, 96(1), 308–327. McKinnon, R., 1973, Money and Capital in Economic Kumar, N., Quisumbing, A. R. (2015). Policy reform Development, The Brookings Institution, toward gender equality in Ethiopia: Little by little Washington, D.C. the egg begins to walk. World Development, 67, McMillan, M. and D. Rodrik, 2011, ‘Globalization, 406–423. Structural Change, and Productivity Growth’, Levine, R., 2005, ‘Finance and growth: theory, mecha- NBER Working Paper 17143, National Bureau nisms and evidence’, in Aghion P. and Durlauf of Economic Research. S. N. (eds), Handbook of Economic Growth, Melesse, M.B., Bulte, E. (2015). Does land registra- Amsterdam: North-Holland. tion and certification boost farm productivity? Levine, R., Loayza, N. and Beck, T., 2000, ‘Financial Evidence from Ethiopia, Agricultural Economics, intermediation and growth: causality and causes’, forthcoming. Journal of Monetary Economics, Vol. 46, pp. Merotto, D., and H. Dogo, ‘Similar but Different: 31–77. Explaining Ethiopia’s Growth Take-off – What 146 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT Caused it and Is it Sustainable?’, unpublished Parente, S. and E. Prescott, 1999, ‘Barriers to Riches’, manuscript. Third Walras-Pareto Lecture, University of Minten, B., Koru, B., Stifel, D. (2013). The Lausanne. (http://www.sfu.ca/~dandolfa/barri- last mile(s) in modern input distribution: erstoriches.pdf ) Pricing, profitability, and adoption, Agricultural Parente, S. and E. Prescott, 2003, ‘A Unified Theory Economics, 44: 1–18 of the Evolution of International Income Minten, B., Tamiru, S., Engeda, E., Kuma, T. (2013). Levels’ (http://www.minneapolisfed.org/research/ Ethiopia’s value chains on the move: The case of prescott/papers/Unified_Theory.pdf ). teff. ESSP II Working Paper 52, International Polity IV, 2012, ‘Political Regime Characteristics Food Policy Research Institute/Ethiopia Strategy and Transitions.’ Online Database. http://www. Support Program II, Addis Ababa, Ethiopia. systemicpeace.org/polity/polity4.htm Minten, B., Tamiru, S., Stifel, D. (2014). Structural Prati, A., M. Onorato, and C. Papageorgiuo, 2013, transformation in cereal markets in Ethiopia, ‘Which reforms work and under what institu- Journal of Development Studies, 50(5): 611–629 tional environment? Evidence from a new data set Moller, L. and K. Wacker, 2015, ‘Explaining Ethiopia’s on structural reforms’, The Review of Economics Growth Acceleration – Insights from a Cross- and Statistics 95(3): 946–968. Country Regression Model’, World Bank Policy Pritchett, L., 2001, ‘Where Has All the Education Research Working Paper 7292. Gone?’ World Bank Economic Review 15(3): Morris, M. L. (2007). Fertilizer use in African agricul- 367–391 ture: Lessons learned and good practice guidelines. Private Sector Development Hub, 2010, ‘A Study on World Bank Publications. the Determination of the Share of Private Sector Mulat, Y.S. (2011). The status of irrigated agriculture in GDP’, Addis in Ethiopia. Research report submitted to the Ababa. Ethiopian Development Research Institute. Quinn, D. and A. Toyoda, 2008, ‘Does Capital National Bank of Ethiopia (NBE) (2014). Annual Account Liberalization Lead to Growth?’, Review report 2013–2014. Addis Ababa. of Financial Studies 1403–49. Nickel, 1981, ‘Biases in dynamic models with fixed Ragasa, C., Berhane, G., Tadesse, F., Taffesse, A. S. effects’, Econometrica 49: 1417–1426. (2013). Gender differences in access to extension Nin-Pratt, Alejandro (2015). Inputs, Productivity, and services and agricultural productivity. The Journal Agricultural Growth in Africa South of the Sahara. of Agricultural Education and Extension, 19(5), IFPRI Discussion Paper 01432, Washington D.C. 437–468. Ngyen H., L. Ngyen and M. Geiger (2015), ‘Ethiopia Rajan, R. and L. Zingales, 1998, ‘Financial Dependence and the East Asian Model’, Processed. and Growth’, American Economic Review 88(3): Ogundari, K. (2014). The paradigm of agricultural 559–86. efficiency and its implications on food security Rashid, S., Tefera, N., Minot, N., Ayele, G. (2013). in Africa: What does meta-analysis reveal? World Can modern input use be promoted without Development, 64: 690–702 subsidies? An analysis of fertilizer in Ethiopia. Ostry, J., Prati, A., and Spilimbergo, A., 2009, Agricultural Economics, 44(6), 595–611. ‘Structural Reforms and Economic Performance Reimers, M. and S. Klasen, 2013, ‘Revisiting the in Advanced and Developing Countries.’ IMF Role of Education for Agricultural Productivity’, Occasional Paper 268. International Monetary American Journal of Agricultural Economics Fund, Washington, DC. 95(1): 131–152. References 147 Reinhart, C. and K. Rogoff, 2011, ‘From Financial in Ethiopia: The search for appropriate public and Crash to Debt Crisis.’ American Economic private roles. Food Policy, 35(3), 185–194. Review 101: 1676–1706. Spielman, D., Kelemwork, D., Alemu, D. (2012). Robinson, S., Strzepek, K., Cervigni, R. (2013). Seed, fertilizer, and agricultural extension in The cost of adapting to climate change in Ethiopia. In In P. Dorosh & S. Rashid (Eds.), Ethiopia: Sector-wise and macro-economic esti- Food and agriculture in Ethiopia: Progress and mates. Ethiopia Strategy Support Program II policy challenges, 132–183. Philadelphia: IFPRI/ (ESSP) working Paper, 53. Pennsylvania University Press. Rodrik, D., 2008, ‘Notes on the Ethiopian economic Spielman, D., Mekonnen, D. (2013). Transforming situation’, Unpublished manuscript, December demand assessment and supply responses in 22. Ethiopia’s seed system and market, mimeo. Rodrik, D., 2013a, ‘The Past, Present and Future Statistical Agency, Addis Ababa, 2004 to 2015). of Economic Growth’, Global Citizen Stiefel, D., Minten, B. (2015). Market access, welfare, Foundation,Working Paper 1. and nutrition: Evidence from Ethiopia. ESSP Rodrik, D., 2013b, ‘Structural Change Fundamentals Working Paper. and Growth – An overview’. Stiglitz, J. E., 2000, ‘Capital market liberaliza- Rodrik, D., 2013c, ‘Unconditional Convergence in tion, economic growth and instability’, World Manufacturing’, Quarterly Journal of Economics Development, Vol. 28, pp. 1075–1086 128(1): 165–204. Summers, L. and L. Pritchett, 2014, ‘Asiaphoria meets Rodrik, D., 2014, ‘An African Growth Miracle?’, Regression to the Mean’, NBER Working Paper Institute for Advanced Study, Princeton 20573 Rodrik, D., 2015, ‘Premature Deindustrialization’ Tadesse, G., Bahiigwa, G. (2015). Mobile phones and NBER Working Paper No. 20935, February farmers’ marketing decisions in Ethiopia, World 2015. Development, 68:296–307. Ruttan, V. W., Hayami, Y, 1984, Toward a theory Taffesse, A.S., Yimer, F., Dereje, M. (2015). The of induced institutional innovation. Journal of impact of the PSNP-created community assets Development Studies, 20(4), 203–223. on crop yield growth, ESSP working paper, Sachs, J., A. Mellinger and J. Gallup, 2001, ‘The forthcoming. Geography of Poverty and Wealth’, Scientific The Economist, 2015, ‘Global Manufacturing: Made American 284(3): 70–75 in China?’, March 14 (Editorial). Schultz, T. W. 1951. A Framework for Land Timmer, M., G. de Vries and K. de Vries (2014). Economics. The Long View. Journal of Farm Patterns of Structural Change in Developing Economics, 204–215. Countries. Groningen Shaw, A. S., 1973, Financial Deepening in Economic Trew, A., 2006, ‘Finance and growth: a critical survey’, Development, Oxford University Press, New The Economic Record, Vol. 82, pp. 481–490. York. USAID, 2015, ‘Finding the Binding Constraints to Shiferaw, A., M. Söderbom, E. Siba and G. Alemu, Sustained and Inclusive Growth in Ethiopia’, 2013, ‘Road Infrastructure and Enterprise unpublished manuscript. Dynamics in Ethiopia.’ College of William and von Braun, J., Olofinbiyi, T. (2007). Famine and Mary Working Paper 128. food security in Ethiopia. Case study #7–4 of the Spielman, D. J., Byerlee, D., Alemu, D., Kelemework, program: “Food policy for developing countries: D. (2010). Policies to promote cereal intensification The role of government in the global food system” 148 ETHIOPIA’S GREAT RUN – THE GROWTH ACCELERATION AND HOW TO PACE IT 2007, edited by Pinstrup-Andersen, P., and F. World Bank, 2014a, ‘Ethiopia: Poverty Assessment’, Cheng. Ithaca, NY: Cornell University. World Bank, Washington D.C. von Braun, J., Teklu, T., Webb, P. (1998). Famine World Bank, 2014b, ‘SME Finance in Ethiopia: in Africa. Causes, Responses and Prevention. Addressing the Missing Middle Challenge’, World International Food Policy Research Institute (IFPRI). Bank, Washington, Wacziarg, R. and K. Welch, 2008, ‘Trade Liberalization World Bank, 2014c, ‘3rd Ethiopian Economic and Growth: New Evidence’, World Bank Update: Strengthening Export Performance Economic Review 22(2): 187–231. through Improved Competitiveness’, World Bank, Weir, S., 1999, ‘The Effects of Education on Farmer Washington D.C. Productivity in Rural Ethiopia’, CSAE Working World Bank, 2014d, ‘Africa’s Pulse’, Volume 10 Paper 99–7. (October 2014), Office of the Chief Economist Weir, S., Knight, J., 2004, ‘Externality effects of edu- for the Africa Region. cation: dynamics of the adoption and diffusion World Bank, 2014e, ‘People-Powered Growth: The of an innovation in rural Ethiopia’, Economic Demographic Dividend and the Future African Development and Cultural Change 53(1): Lions’, Washington D.C. 93–113. World Bank, 2015a, ‘Lessons from Urban Labor Wooldridge, J., 2010, ‘Econometric Analysis of Cross- Markets in Ethiopia’, World Bank, Washington, Section and Panel Data’, MIT Press, Cambridge, D.C. MA. World Bank, 2015b, ‘Doing Business 2015: Going World Bank, 2007, ‘Country Economic Memorandum: Beyond Efficiency’, World Bank, Washington, Ethiopia Accelerating Equitable Growth’, D.C. Washington D.C. World Bank, 2015c, ‘The Power of Public Investment World Bank, 2011, ‘Ethiopia Enterprise Survey, Management: Transforming Resources into Assets 2011’, Washington, DC. for Growth’, Washington D.C. World Bank, 2012a, ‘Ethiopia Economic Update I: World Bank, 2015d, ‘4th Economic Update: Overcoming Inflation, Raising Competiveness’, Overcoming Constraints in the Manufacturing Washington, D.C. Sector’, World Bank, Washington, D.C. World Bank, 2012b, ‘Secondary Education in Ethiopia World Development Indicators (WDI), 2013, Online – Supporting Growth and Transformation’, Database. http://data.worldbank.org/indicator, Washington D.C. World Bank. World Bank, 2013, ‘Ethiopia Economic Update II: Yusuf, S., 2014, ‘Sustaining Ethiopia’s Growth Laying the Foundation for Achieving Middle Acceleration’, Background Paper prepared for Income Status’, World Bank, Washington, DC. this report. The World Bank 1818 H Street, NW Washington, DC 20433 Phone: (202) 473-1000 Fax: (202) 477-6391