WPS6327 Policy Research Working Paper 6327 Does Trade Reduce Poverty? A View from Africa Maëlan Le Goff Raju Jan Singh The World Bank Africa Region Poverty Reduction and Economic Management Unit January 2013 Policy Research Working Paper 6327 Abstract Although trade liberalization is being actively promoted openness is reflected in the available empirical evidence. as a key component in development strategies, This paper examines how the effect of trade openness on theoretically, the impact of trade openness on poverty poverty may depend on complementary reforms that help reduction is ambiguous. A more liberalized trade regime a country take advantage of international competition. is argued to change relative factor prices in favor of the Using a non-linear regression specification that interacts more abundant factor. If poverty and relative low income a proxy of trade openness with proxies of various country stem from abundance of labor, greater trade openness structural specificities and a panel of 30 African countries should lead to higher labor prices and a decrease in over the period 1981–2010, the analysis finds that trade poverty. However, should the re-allocation of factors be openness tends to reduce poverty in countries where hampered, the expected benefits from freer trade may not financial sectors are deep, education levels high and materialize. The theoretical ambiguity on the effects of governance strong. This paper is a product of the Poverty Reduction and Economic Management Unit, Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at rsingh9@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team DOES TRADE REDUCE POVERTY? A VIEW FROM AFRICA 1 Maëlan Le Goff Raju Jan Singh JEL Classification Numbers: F02, F15, O11, O55 Keywords: Trade, poverty, governance, Africa Sector Board: EPOL Author’s E-mail Address: rsingh9@worldbank.org; maelan.legoff@cepii.fr 1 We would like to thank Richard Baldwin, Daniel Lederman and Sherry Stephensen for their helpful comments. We would also like to thank participants at the Center for the Study of African Economies Conference, March 18-20, 2012, Oxford and at the Poverty Reduction, Equity, and Growth Network’s Conference, September 6-7, 2012, Dakar for suggestions on earlier versions. The usual disclaimer applies. Maëlan Le Goff is affiliated to the CEPII (Centre d’Etudes Prospectives et d’informations Internationales), Raju Jan Singh is with the World Bank. 2 I. INTRODUCTION Most economists accept that, in the long run, open economies fair better in aggregate than do closed ones, and that relatively open policies contribute significantly to development. Many commentators fear, however, that in the shorter run, one of the steps towards openness – trade liberalization – harms the poorer actors in the economy, and that even in the longer run successful open regimes may leave some people behind in poverty. Africa remains the poorest continent of the world. Yet, at the same time, African countries have experienced significant improvements in trade liberalization. It seems that the large gains expected from opening up to international economic forces have, to date, been limited in Africa, especially for poor people. While the traditional trade theory predicts welfare gains from openness at the country level through specialization, investment in innovation, productivity improvement, or a better resource allocation, the theoretical impact of trade on the poor remains uncertain. Besides, empirical results do not converge on this point and it seems that developing countries are not equally able to make use of the opportunities arising out of increased access to markets in the developed world. The contribution of this paper lies in providing new cross-country empirical evidence focused on Africa on how the poverty reduction effect of greater trade openness depends on a variety of structural characteristics, including some that are subject to reform. Using a panel of African countries over the period 1981-2010 and testing for non-linearities in the trade-poverty relationship, this paper explores the empirical link between trade openness and poverty. Its results uncover an interesting pattern of reform complementarity: trade openness tends to reduce poverty in countries as their financial sector grows deeper, their education level higher and their governance stronger. Our concern is with poverty, not inequality. Since trade liberalization tends to increase the opportunities for economic activity, it can very easily widen income inequality while at the same time reduce poverty. Consequently, statements about its effects on inequality cannot be translated directly into statements about its impact on absolute poverty. There may be sound positive and normative reasons for interest in inequality, but they are not the concerns of this paper. The rest of the paper is organized as follows. Section II reviews the impact of trade on poverty by focusing on the transmission channels and the conditional variables that can influence the trade-poverty relationship. Section III sets up the empirical analysis of the non-linear trade-poverty relationship. Conclusions are presented in Section IV. 3 II. TRADE OPENNESS AND POVERTY: DISCUSSION OF THE LITERATURE Bhagwati and Srinivasan (2002) distinguish two broad strands of argumentation when discussing the effects of freer trade on poverty: static and dynamic. In the former case, we examine how freer trade effects poverty taking resources and technology as given. In the latter case, we consider growth effects and the evolution of poverty over time. 2 Following the static approach, the Stolper-Samuelson theorem, in its simplest form, suggests that the abundant factor should see an increase in its real income when a country opens up to trade. If the abundant factor in developing countries is unskilled labor, then this framework suggests that the poor (unskilled) in developing countries have the most to gain from trade. Krueger (1983) has used this insight to argue that trade reforms in developing countries should be pro-poor, since these countries are most likely to have a comparative advantage in producing goods requiring unskilled labor. However, for comparative advantage to increase the incomes of the unskilled, they need to be able to move out of contracting sectors and into expanding ones. Davis and Mishra (2006), Goh and Javorcik (2006) and Topalova (2006) suggest that labor in the real world may not be as mobile: there are too many barriers to entry and exit for firms, and too many barriers to labor mobility for workers. Focusing on urban Colombia, Attanasio, Goldberg and Pavenik (2004) find for example that the probability of being unemployed is more important in traded-good sectors than in non- traded-good sectors. Trade openness may also increase the size of the informal sector (Goldberg and Pavenik, 2003). Being more exposed to foreign competition, firms may be incited to reduce their costs by hiring temporary workers instead of permanent ones, or even to lay off workers, who may in turn obtain informal jobs. Depending on the wage differences between sectors, this could lead to an increase in poverty. In addition, if the poor are mostly completely unskilled, while it is semi-skilled labor that is on increased demand, poverty will be unaffected – or possibly, worsened. Trade liberalization may even be accompanied by skill-biased technical change, which can mean that skilled labor may benefit relative to unskilled labor. Lower prices for capital goods or increased competition following trade liberalization could encourage firms to import 2 See Winters and McKay (2004) for a detailed discussion on the possible various channels for freer trade to affect poverty. 4 machines and increase their demand for skilled labor (Acemoglu, 2003; Behrman, Birdsall, and Szekely, 2000; Harrison and Hanson, 1999). Furthermore, many developing countries are rich in natural resources. Trade would stimulate this sector rather than labor-intensive ones. From a dynamic perspective, economic growth is key to sustained poverty alleviation and trade liberalization is argued to lead to the needed increases in productivity to sustain growth. Freer trade provides greater incentives for investment, the benefits of scale and competition, limitation on rent-seeking activities favored by trade restrictions and openness to new ideas and innovations (Berg and Krueger, 2003; Grossman and Helpman, 1991; Lucas, 1988). Empirically, cross-county growth regressions have produced mixed evidence. Using either trade shares or indices of trade liberalization, Sachs and Warner (1995), Edwards (1997), Frankel and Romer (1999) Dollar and Kraay (2001) and Lee et al. (2004) find a positive association between liberalization and growth. However, noting that trade liberalization often occurs at the same time than other policy reforms, Rodriguez and Rodrik (2001) have criticized the literature which associates trade openness measured by trade shares with more rapid growth. Considering trade policy measures instead, Harrison (1996), Edwards (1997), and Irwin and Terviö (2002) find a significant negative effect of trade on economic growth, while Vamvakidis (2002) still shows that trade enhances growth. Examining the effect of trade on poverty more directly, Krueger (1983) shows in her case studies that developing countries’ manufactured exports were, indeed, labor-intensive, but that the employment effects of freer trade policies were generally rather limited. A number of cross-country studies on poverty, while not dealing with trade explicitly, incorporate trade openness as a control variable and showed similar results: at best the benefits of greater trade openness seem to have bypassed the poor. Looking at developing countries only, Beck et al. (2007) and Kpodar and Singh (2011) find no effect on the poor. Similarly, Dollar and Kraay (2001) find a lack of any evidence of an impact of openness on the income of the poorest quintile in a sample of advanced and developing economies. By contrast, looking at a sample of developing countries, Guillaumont-Jeanneney and Kpodar (2011) find a negative relationship between trade openness and the income of the poorest quintile. Similarly, Singh and Huang (2011) focusing on a sample of sub-Saharan African countries suggest that greater trade openness increases headcount poverty, widens the poverty gap, and reduces the income of the poorest quintile. 5 This lack of any clear correlation between openness measures and poverty indicators in aggregate could be because there is too much heterogeneity in the effects of trade reforms on the poor. Since poor workers in import-competing sectors lose from reforms, while poor workers in export-oriented sectors gain, it could be that in the aggregate these different effects cancel each other. Similarly, cross-country studies have tended to favor larger samples and focus on developing countries at best. While using a broader sample increases the degrees of freedom, it may also introduce unwanted heterogeneity if factors explaining poverty differ between country groups. Another possibility suggested by Winters and McKay (2004) is that trade liberalization almost certainly requires combination with other appropriate policies. The sort of policies envisaged would be those that encourage investment, allow effective conflict resolution, and promote human-capital accumulation. Linear regression models would not capture such complementary dynamics. A number of recent studies emphasize the importance of complementary policies in determining the benefits or costs of trade reforms for developing countries. For example, Bolaky and Freund (2008) show that trade reforms actually lead to income losses in highly regulated economies. Excessive regulations restrict growth because resources are prevented from moving into the most productive sectors and to the most efficient firms within sectors. Similarly, Sindzingre (2005) suggests that institutions could help explain the heterogeneity in the globalization-poverty relationship. She argues that domestic political structures and institutions (such as oligarchic or predatory regimes) may prevent the poor from benefiting from globalization. More recently, Haltiwanger (2011) and McMillan and Verduzco (2011) argue that benefits of trade depend to a large extent on national institutional settings. The process of trade- induced growth entails a continual reallocation of resources away from less productive activities to more productive ones. Many things can go wrong in this reallocative process if economies are distorted, for instance if transportation or communication infrastructure are not sufficiently developed, if ineffective (or non-existent) competition policy does not prevent large firms from abusing their market power, or if financial markets are not sufficiently developed to fund new and expanding businesses. In such distorted economic environments, there is little chance that the benefits of greater trade openness materialize and – in extreme cases – a “de-coupling� may take place, i.e. cases in which policy reforms induce downsizing and exit of some firms but do not lead to the expansion of other firms. Similarly, reviewing the new wave of research under the International Collaborative Initiative on Trade and Employment, Newfarmer and Sztajerowska (2012) conclude that the benefits of trade do not accrue automatically, and policies that complement trade opening are needed. Policies that embed trade reforms in a context of macroeconomic stability and a sound investment climate on the one hand and, on the other, protection for 6 workers, maintenance of high-quality working conditions, and facilitation of labor transitions, can play an important role in realizing the potential wage, employment, and income gains associated with trade. Finally, Agénor (2004) examines whether the effect of globalization on poverty depends on the degree of liberalization. Introducing a square term of globalization index in his base model, he finds that under a certain level globalization is detrimental for the poor and that beyond this threshold, globalization appears to reduce poverty (illustrating an inverted U- shaped relationship between globalization and poverty). One explanation suggested by Agénor (2004) is that at the first stage, globalization generates a decrease in the output of import-competing sectors, while at the next stage, output increases thanks to the development of the exportable sector induced by greater globalization. Using endogenous threshold regression techniques to investigate the trade-poverty relationship in China, Liang (2006) reaches similar results: globalization promotes poverty-reduction in Chinese provinces only after the economy has reached a certain level of globalization. 7 III. EMPIRICAL ANALYSIS A- Sample Our empirical objective is to examine how the poverty reduction effect of openness may depend on a variety of country characteristics, including some that are subject to reform. For this purpose, we work with pooled cross-country and time-series data for 30 African countries averaged over five-year periods from 1981 to 2010. Summary statistics and the correlation matrix for the variables used in our estimation exercises are provided in Appendix 1. Following the same approach as Chang (2009), we start with a linear regression specification and then extend it to account for interaction terms between an openness measure and proxies for various county characteristics (financial depth, education and governance). These three dimensions (finance, education, and governance) should characterize an economy’s ability to reallocate resources away from the less productive sectors to the more productive ones and, hence, take advantage of the opportunities offered by greater trade openness. A more developed financial sector, as measured by the private sector credit-to- GDP ratio, would allow a faster identification of new and promising sectors and a redirection of credit. A more educated population, as measured by primary completion rates, would be more able to acquire the new skills sought by growing sectors and adjust more rapidly to the new conditions of the labor market. Finally, better governance, as measured by the rule of law, would allow contracts to be made and conflicts to be resolved more easily. B- Model and Definition of Variables Model We use the basic growth–poverty model suggested by Ravallion (1997) and Ravallion and Chen (1997) as well as the frameworks posited by Dollar and Kraay (2001) and Berg and Krueger (2003). Our point of departure can be written as: 𝑃𝑜𝑣𝑒𝑟𝑡𝑦𝑖,𝑡 = 𝜷� 𝑻𝑶𝒊,𝒕 + 𝛽2 𝑋𝑖,𝑡 + 𝜑𝑡 + 𝜇𝑖 + 𝜀𝑖,𝑡 (1) where the subscripts i and t represent country and time period, respectively, Poverty is the log of a poverty indicator, X is the matrix of control variables, TO is a measure of trade openness, 𝜑𝑡 corresponds to time effects, 𝜇𝑖 denotes unobserved country-specific effects, and 𝜀𝑖,𝑡 the error term. 8 Following Chang (2009), we then extend the regression specification by allowing the poverty reduction effect of openness to vary with some country characteristics represented in the control set (financial depth, education, and governance). We do this by interacting the openness measure with each of these conditional variables in turn. The regression equation with the interaction terms is the following: 𝑃𝑜𝑣𝑒𝑟𝑡𝑦𝑖,𝑡 = 𝜷� 𝑻𝑶𝒊,𝒕 + 𝛽2 𝑋𝑖,𝑡 + 𝜷𝟑 𝑻𝑶𝒊,𝒕 × 𝒙𝒊,𝒕 + 𝜑𝑡 + 𝜇𝑖 + 𝜀𝑖,𝑡 (2) where 𝑥𝑖,𝑡 represents one of the control variables. Variables Poverty. Poverty is a complex and multidimensional phenomenon, and there is considerable controversy in the literature about how it should be defined and measured. The majority of the empirical economic literature on poverty tends to adopt on absolute income or consumption metric. Therefore, while recognizing that there are many legitimate approaches to the measurement of poverty, we will follow this approach. In this regard, the most popular indicator is the poverty headcount index which measures the percentage of the population living with consumption or income per person below a certain poverty line. It is a measure of absolute poverty. Another popular measure is the poverty gap, which measures the mean distance below the poverty line as a proportion of the poverty line. As dependent variable, we use the poverty headcount and the poverty gap considering the $1.25 poverty line. Trade openness. Two main categories of trade openness measures can be found in the economic literature. Spilimbergo et al. (1999) distinguish for instance between incidence- based measures of openness, based on tariff data and trade policy, and outcome-based measures of openness, based on trade data. Calderón et al (2005) make a similar distinction between policy or legal measures and outcome or de facto measures of openness. In this paper we focus on the impact of actual globalization on poverty and we have chosen as our dependent variable (Trade openness) an ex-post measurable and objective definition of globalization, namely effective trade openness (measured as the sum of exports and imports as a share of GDP), and not a measure of liberalization policies. Such an indicator is arguably better than legalistic (de jure) measures to the extent that the latter are difficult to summarize in a single indicator (there are multiple tariff rates, multiple non-tariff barriers, multiple trade agreements) and even more difficult to adjust for the strength with which they are enforced. 9 Control variables. We also include a set of control variables that are commonly used in poverty equations: overall income per capita (GDP per capita) to control for economic development, a measure of human capital level (Education); a variable of financial deepening (Private credit/GDP); growth of the consumer price index (Inflation) to control for the macroeconomic environment; and an indicator of institutional quality (Law and order) drawn from the International Country Risk Guide (ICRG) database which measures the strength and impartiality of the legal system, and the popular observance of the law. Data sources are provided in Appendix 3. C- Methodology The poverty regression presented above poses several challenges for estimation. The first is the presence of unobserved period- and country-specific effects. The second challenge is that most explanatory variables could be jointly endogenous with poverty, so we need to control for the biases resulting from simultaneous or reverse causation. For this purpose, we estimate the coefficients of our model by using the System Generalized Method-of- Moment (GMM) estimator developed by Blundell and Bond (1998). The estimator combines two sets of equations. The first set includes first-differenced equations where the right-hand-side variables are instrumented by the levels of the series lagged one period or more. The second set consists of the equations in levels with the right-hand side variables being instrumented by lagged first or higher-order differences. The consistency of the GMM estimators depends on whether lagged values of the explanatory variables are valid instruments in the poverty regression. We examine this issue by considering the Hansen test of over-identifying restrictions. The no rejection of the null hypothesis implies that instrumental variables are not correlated with the residual and are satisfying the orthogonality conditions required. A serial correlation test is also carried out and demonstrates that the errors exhibit no second-order serial correlation. D- Results A look at the plots in Appendix 4 which compare the trade-poverty relationship in the top country group and the bottom country group in terms of financial development, education and governance, 3 suggests that the effect of trade openness on poverty would depend on a variety of country characteristics. Indeed, for each conditional variable considered, the 3 Countries are classified in the top (bottom) group if they belong to the top one-fourth (bottom three- fourths) of a rank distribution given by each conditional variable (financial development, education level and quality of institutions). 10 slope for the relationship between trade openness and poverty appears negative and steeper in the top group than in the bottom group. To control for other poverty determinants and endogeneity issues, we now test our model through an econometric analysis. Results of the basic regression with no interaction term (equation 1) are presented in Tables 1 (for the poverty headcount) and 2 (for the poverty gap), columns 1. The log transformation of all the variables allows us to interpret the coefficients as elasticities. A positive sign in the corresponding coefficient indicates an increase in the poverty headcount (or poverty gap), corresponding to a worsening in the situation of the poor. Overall, the results are consistent with the previous empirical literature. There is a strong negative relationship between income per capita and poverty incidence, suggesting that higher levels of economic development are associated with lower levels of poverty. The proxy for the education level has a negative and significant coefficient, denoting its beneficial impact on poverty reduction. Inflation, financial development, and governance variables do not seem to be robustly linked to poverty. Looking at trade openness, the results are in line with Beck et al. (2007) and Kpodar and Singh (2011): they would suggest that greater trade openness is not significantly associated with either lower or higher levels of poverty. Since in the basic specification only linear effects are allowed, the coefficient associated with trade openness is an average. To uncover what is behind this average, we examine next the influence of some structural country characteristics in the trade-poverty relationship. The results with the interaction terms are presented in Columns 2, 3 and 4 of Tables 1 and 2. An interesting pattern of reform complementarity emerges: the coefficient on the trade openness proxy suggests that greater trade is associated with higher levels of poverty, but the coefficients on the interaction terms indicate that this adverse relation could be reversed if financial development grows deeper, education levels higher, and governance improves. We first test whether the trade-poverty relationship would change with the development of the financial sector (columns 2). The coefficient on the interaction term with the financial depth (as measured by the private sector credit-to-GDP ratio) is negative and significant. These results suggest that a greater openness of African economies to trade is associated with lower levels of poverty when the financial system is more developed. In other words, easier access to cheaper credit may allow the poor to benefit more from trade openness. A threshold of 17.7 percent of GDP is estimated for the development of domestic private credit beyond which the poor can benefit from trade openness, which is far below the average of our sample (21.2 percent of GDP). To illustrate, the Central African Republic, 11 Ghana, Malawi and Uganda are under this threshold while South Africa, Algeria and Tunisia are well above it. 4 Column 3 of Tables 1 and 2 shows the results of the estimations testing the role of human capital in the trade-poverty relationship. The beneficial impact of an increase in trade openness on poverty reduction is larger when the investment in human capital is stronger. We find that when the share of the population over age 15 with completed primary education exceeds 46.7 percent (the average in the sample is 55 percent), trade openness starts being favorable to the poor. In other words, with appropriate learning skills, people are better able to take advantage of the new opportunities offered by trade. Finally, we examine whether the relationship between the openness to trade and poverty may hinge on a country’s institutional environment (columns 4). Results suggest that the negative association between trade openness and poverty could diminish with improvements with respect to law and order, and even reverse. We find that trade openness could be favorable to the poor when institution quality (measured by the Law and order variable) reaches 3.3 (the average of our sample is 2.9). 5 In other words, an environment where laws and contracts are enforced seems to be more favorable for the emergence of new enterprises and the closing of older ones, allowing the economy the needed flexibility to adjust. Robustness tests We conducted several robustness tests: • First, we used an alternative measure of absolute poverty, the poverty gap. The results using the poverty gap confirmed the observations for the poverty headcount ratio (see Table 2); • We also removed outliers. The results reported in Table A.1 and Table A.2 confirm and reinforce our previous observations. 4 See Appendix 6. 5 We also used another ICRG indicator measuring Institutional strength and quality of the bureaucracy (ranging from 0 to 4, with a higher figure indicating a bureaucracy which has the strength and expertise to govern without drastic changes in policy or interruptions in government services), as a proxy of the institutional environment. We found similar results. 12 IV. CONCLUDING REMARKS While trade liberalization is being actively promoted as a key component of development strategies, both theoretically and empirically, its impact on poverty is ambiguous. At best, the benefits of freer trade seem to bypass the poor. By focusing on African countries and taking into account possible non-linearities, this paper attempted to reach more robust results. While on average trade does not seem to be associated with lower poverty, this observation hides important non-linearities and an interesting pattern of policy complementarities. More openness results in a reduction in poverty when the financial sector is deeper, education levels are higher, and governance is stronger. These results are consistent with the recent literature arguing that the benefits of trade are not automatic and that policies to accompany trade opening are needed. These policies would aim at encouraging the financing of new investment, the effective resolution of conflicts, and the ability to adjust and learn new skills. This accompanying policy agenda would allow resources to be reallocated away from less productive activities to more promising ones. Trade liberalization should therefore not be seen in isolation and additional policies will be needed to enhance its impact, including on poverty. This also means that poor policies and institutions, weak human capital and limited financial development, have not only a direct negative effect on countries’ welfare, they also prevent the poor in developing countries from benefiting from the gains of trade liberalization. 13 References Attanasio, O., Goldberg, P. and Pavenik , N. (2004), “Trade Reforms and Wage Inequality in Colombia�, Journal of Development Economics, Vol. 74: 331-366. Acemoglu, D. (2003), “Patterns of Skill Premia�, Review of Economic Studies, Vol. 70: 199- 230. Agénor, P.-R. (2004), “Does Globalization Hurt the Poor?�, International Economics and Economic Policy, Vol. 1 (1): 21–51. Beck, T., Demirguc-Kunt, A. and Levine, R. (2007), “Finance, Inequality and the Poor�, Journal of Economic Growth, Vol. 12: 27-49. Behrman, J., N. Birdsall, M Szekely (2000), “Economic Reform and Wage Differentials in Latin America�, Inter-American Development Bank, Working Paper, No 435. Besley, T., and R. (2003), “Halving Global Poverty�. Journal of Economic Perspectives, Vol. 17 (3): 3–22. Bhagwati, J. and Srinivasan, T. N. (2002), “Trade and Poverty in the Poor Countries�, AEA Papers and Proceedings, Vol. 92(2): 180-183. Blundell, R., and S. Bond (1998), "Initial Conditions and Moment Restrictions in Dynamic Panel Data Models", Journal of Econometrics, Vol. 87(1): 115-143 Bolaky, B. and Freund, C. (2008), “Trade, Regulations, and Income�, Journal of Development Economics, Vol. 87(2): 309-21. Calderón, C., N. Loayza, K. Schmidt-Hebbel (2005), “Does Openness Imply Greater Exposure?� World Bank Policy Research Working Paper, No 3733. Chang, R., L., Kaltani and N., Loayza, N. V., (2009.), “Openness Can Be Good for Growth: The Role of Policy Complementarities�, Journal of Development Economics, Vol. 90(1): 33- 49. Davis, D. and Mishra, P. (2006), “Stopler-Samuelson Is Dead and Other Crimes of Both Theory and Data� in Harrison, A. (ed.), Globalization and Poverty, University of Chicago Press for NBER, Chicago, IL Dollar D. and Kraay, A. (2001), “Trade, Growth and Poverty�, mimeo, Development Research Group, The World Bank, Washington. 14 Edwards, S. (1997), “Trade Policy, Growth, and Income Distribution�, American Economic Review, Vol. 87(2): 205-210. Frankel, J. (1999), "Does Trade Cause Growth?", The American Economic Review, Vol. 89 (3): 379-399 Goh, C. and Javorcik, B.S. (2006), “Trade Protection and Industry Wage Structure in Poland�, in Harrison, A. (ed.), Globalization and Poverty, University of Chicago Press for NBER, Chicago, IL Goldberg, P. and Pavenik, N. (2003), “The Response of the Informal Sector to Trade Liberalization�, Journal of Development Economics, Vol. 72: 463-496. Görg, H. and Strobl, E. (2001), “Relative Wages, Openness and Skill-Biased Technological Change in Ghana�, CREDIT Research Paper, No.01/18, CREDIT, University of Nottingham, Nottingham. Grossman, G. M. and E. Helpman (1991), Innovation and Growth in the Global Economy, Cambridge, MA and London: MIT Press. Guillaumont-Jeanneney, S. and Kpodar, K. (2011), “Financial Development and Poverty Reduction: Can There Be a Benefit Without a Cost?�, Journal of Development Studies, Vol 47(1): 143-163. Haltiwanger, J. (2011), “Globalization and Economic Volatility�, in M. Bacchetta and M. Jansen (eds): Making Globalization Socially Sustainable, ILO and WTO, Geneva: 119- 146. Harrison, A. , "Openness and Growth: A Time-Series, Cross-Country Analysis for Developing Countries", Journal of Development Economics, Vol. 48(2): 419-447. Harrison, A. and Hanson, G. (1999), “Who Gains from Trade Reform? Some Remaining Puzzles�, Journal of Development Economics, Vol. 59: 125-154. Irwin D. and M Terviö (2002), "Does Trade Raise Income?", Journal of International Economics, Vol. 58(1): 1-18. Kpodar. K. and Singh R. (2011), “Does Financial Structure Matter for Poverty? Evidence from Developing Countries�, World Bank Policy Research Working Paper, WPS5915. Krueger, A. (1983), Trade and Employment in Developing Countries: Volume 3: Synthesis and Conclusions, NBER, New York. 15 Krueger, A., and A. Berg (2003), “Trade, Growth, and Poverty: A Selective Survey�, IMF Working Paper, WP/03/30 (Washington: International Monetary Fund). Lee, H., Ricci L. and Rigobon, R. (2004), “Once Again Is Openness Good for Growth ?�, Journal of Development Economics, Vol. 75(2): 451-472. Liang, Z. (2006), “Threshold Estimation on the Globalization-Poverty Nexus: Evidence from China�, Research Paper, UNU-WIDER, United Nations University, No. 2006/57. Lucas, R. E. (1988), “The Mechanics of Economic Development,� Journal of Monetary Economics, Vol. 22: 3–42. McMillan, M. and I., Verduzco (2011), “New Evidence on Trade and Employment: An Overview� in Marion Jansen, Ralf Peters, José Manuel Salazar-Xirinachs (eds.), Trade and Employment: From Myths to Facts, International Labor Organization, Geneva: 23-60. Newfarmer, R. and M. Sztajerowska (2012), “Trade and Employment in a Fast-Changing World� in Douglas Lippoldt (ed.), Policy Priorities for International Trade and Jobs, pp. 7-73, OECD, Paris Ravallion, M. (1997), “Can High-Inequality Developing Countries Escape Absolute Poverty?�, Economics Letters, Vol. 56(1): 51-57. Ravallion, M. and Chen, S. (1997), “What Can New Survey Data Tell Us About Recent Changes in Distribution and Poverty?�, World Bank Economic Review, Vol. 11(2): 357-82. Rodríguez, F. and Rodrik, D. (2001), Trade Policy and Economic Growth: A Skeptics Guide to the Cross-national Evidence� in Bernanke, B. and K. Rogoff (eds.), NBER Macroeconomics Annual 2000, MIT Press, Cambridge, MA, Vol. 15: 261–325. Sachs, J. and W. Andrew (1995), "Economic Reform and the Process of Economic Integration", Brookings Papers of Economic Activity, Vol. 1: 1-95. Sindzingre, A. (2005), “Explaining Threshold Effects of Globalization on Poverty: An Institutional Perspective�, Research Paper, UNU-WIDER, United Nations University, No. 2005/53. Singh, R. and Huang Y. (2011), “Financial Deepening , Property Rights, and Poverty : Evidence from Sub-Saharan Africa�, IMF Working Paper, WP/11/196 (Washington: International Monetary Fund). Spilimbergo A., J. L. Londoño, M. Székely (1999), "Income Distribution, Factor Endowments, and Trade Openness", Journal of Development Economics, Vol. 59 (1): 77-101. 16 Topalova, P. (2006), “Trade Liberalization, Poverty and Inequality: Evidence From Indian Districts� in Harrison, A. (ed.), Globalization and Poverty, University of Chicago Press for NBER, Chicago, IL. Vamvakidis, A. (2002), "How Robust is the Growth-Openness Connection? Historical Evidence", Journal of Economic Growth, Vol. 7 (1): 57-80. Winters A., McCulloch, N. and McKay, A. (2004), “Trade Liberalization and Poverty: The Evidence So Far�, Journal of Economic Literature, Vol XLII: 72-115. 17 Table 1 - Trade Openness and Poverty Incidence in Africa-GMM System Poverty headcount (log) (1) (2) (3) (4) Trade openness (log) 0.390 2.670** 5.037* 2.402** (0.91) (1.99) (1.75) (2.09) GDP per capita (log) -0.888** -0.818** -0.863*** -0.553** (-2.45) (-2.10) (-3.70) (-2.14) Inflation (log) 0.157* 0.184* 0.058 0.119 (1.84) (1.84) (0.58) (1.40) Education (log) -0.778** -0.447 4.884* -0.425 (-2.36) (-1.50) (1.74) (-1.55) Law and order 0.152 0.014 -0.097 2.917** (1.03) (0.09) (-0.70) (2.18) Private Credit/GDP (log) -0.192 3.778* -0.034 -0.276 (-0.81) (1.69) (-0.23) (-1.46) Private Credit/GDP *Trade openness -0.928* (-1.72) Education* Trade openness -1.310* (-1.80) Law and order *Trade openness -0.725** (-2.08) Constant 11.059*** -0.148 -8.339 -0.065 (2.97) (-0.02) (-0.82) (-0.01) Observations 64 64 64 64 Number of countries 30 30 30 30 Sargan/Hansen test 0.59 0.47 0.82 0.57 AR(2) 0.29 0.47 0.64 0.75 Notes: data are averaged over five years. Absolute value of z statistics in parentheses; * significant at 10%; ** significant at 5%; ***significant at 1%. 18 Table 2 - Trade Openness and Poverty Gap in Africa-GMM System Poverty gap (log) (1) (2) (3) (4) Trade openness (log) 0.656 1.310** 7.899* 3.601** (1.09) (1.97) (1.89) (2.19) GDP per capita (log) -1.209** -1.358*** -1.414*** -0.713** (-2.46) (-3.08) (-3.32) (-2.04) Inflation (log) 0.267** 0.078 0.144 0.207 (2.19) (0.73) (1.49) (1.62) Education (log) -1.136** 0.636 7.707* -0.687* (-2.56) (1.33) (1.76) (-1.78) Law and order 0.283 -0.03 0.083 4.332** (1.22) (0.12) (0.52) (2.27) Private Credit/GDP (log) -0.342 3.409** -0.054 -0.394 (-1.05) (2.33) (-0.43) (-1.53) Private Credit/GDP *Trade openness -0.969*** (-2.75) Education* Trade openness -2.055* (-1.86) Law and order *Trade openness -1.069** (-2.15) Constant 12.366** 3.594 -17.184 -0.378 (2.38) (1.29) (-1.13) (-0.50) Observations 64 64 64 64 Number of countries 30 30 30 30 Sargan/Hansen test 0.78 0.49 0.88 0.51 AR(2) 0.37 0.42 0.87 0.90 Notes: data are averaged over five years. Absolute value of z statistics in parentheses; * significant at 10%; ** significant at 5%; ***significant at 1%. Appendix 1 - Summary Statistics and Correlation Matrix Variable Observations Mean Standard Minimum Maximum deviation Poverty incidence 64 42.5 26.3 2.0 86.1 Poverty gap 64 17.6 13.9 0.4 53.1 Trade openness 64 65.8 28.0 19.1 178.3 GDP per capita 64 2255.9 2443.3 295.1 13003.9 Inflation 64 27.4 130.1 0.3 1042.7 Education 64 55.0 22.4 9.6 93.9 Law and order 64 2.9 1.1 1 5.4 Private Credit/GDP 64 21.2 24.9 0.6 151.6 (1) (2) (3) (4) (5) (6) (7) (8) Poverty incidence (1) 1 Poverty gap (2) 0.97 1 Trade openness (3) -0.02 0.03 1 GDP per capita (4) -0.61 -0.55 0.17 1 Inflation (5) 0.09 0.15 0.28 -0.01 1 Education (6) -0.54 -0.49 0.28 0.58 -0.13 1 Law and order (7) -0.18 -0.17 0.06 0.14 -0.15 -024 1 Private Credit/GDP (8) -0.50 -0.47 -0.06 0.56 -0.11 0.42 0.16 1 20 Appendix 2 - List of the Sample Countries (30) Algeria Kenya Angola Liberia Botswana Madagascar Burkina Faso Malawi Cameroon Mali Congo, Dem. Rep. Morocco Congo, Rep. Mozambique Cote d’Ivoire Niger Egypt Arab Rep. Nigeria Ethiopia Senegal Gabon South Africa Gambia Tanzania Ghana Togo Guinea Tunisia Guinea-Bissau Uganda 21 Appendix 3 - Variable Definition and Sources Variables Description Data sources The percentage of the population living below the $1.25/day Poverty incidence international poverty line World Bank Global Poverty Index Poverty gap The average shortfall of the poor Database with respect to the poverty line, multiplied by the headcount ratio Trade openness Sum of exports and imports as a share of GDP Annual percentage change in Inflation consumer prices International Financial Statistics and World Development Indicators Nominal GDP divided by GDP per capita population size Domestic credit to private sector Private Credit/GDP (% of GDP) Primary completion rate: total number of new entrants in the last grade of primary education, Education regardless of age, expressed as UNESCO database percentage of the total population of the theoretical entrance age to the last grade of primary. Strength and impartiality of the legal system, and popular observance of the law. Its International Country Risk Guide (ICRG) values range from 0 to 6, with a Law and order higher figure indicating a better quality and enforcement of the legal system 22 Appendix 4 - Poverty and Trade Openness for Top and Bottom Reformers Trade and poverty depending on financial development (Private credit) Trade and poverty depending on the quality of institutions (Law and order) 4 4 Log of poverty headcount Log of poverty headcount 2 2 0 0 -2 -2 2 3 4 5 6 2 3 4 5 6 Log of trade openness Log of trade openness Bottom group Linear prediction (bottom group) Bottom group Linear prediction (bottom group) Top group Linear prediction (top group) Top group Linear prediction (top group) Trade and poverty depending on education level (Education) 4 Log of poverty headcount 0 -2 2 2 3 4 5 6 Log of trade openness Bottom group Linear prediction (bottom group) Top group Linear prediction (top group) 23 Appendix 5 - Robustness Checks Table A.1- Trade openness and poverty incidence in Africa-GMM system excluding Outliers Poverty headcount (log) (1) (2) (3) (4) Trade openness (log) 0.549* 2.459*** 4.346* 2.324* (1.68) (2.66) (1.93) (1.86) GDP per capita (log) -0.739*** -0.712*** -0.939*** -0.559** (-2.78) (-2.85) (-3.03) (-2.43) Inflation (log) 0.174** 0.233** 0.030 0.162* (1.98) (2.49) (0.27) (1.87) Education (log) -0.704** -0.382 4.446* -0.424* (-2.33) (-1.39) (1.93) (-1.70) Law and order 0.159 0.036 -0.090 2.975** (1.04) (0.24) (-0.66) (2.15) Private Credit/GDP (log) -0.357* 3.211* -0.002 -0.408** (-1.70) (1.96) (-0.02) (-2.10) Private Credit/GDP *Trade openness -0.830** (-1.98) Education* Trade openness -1.203** (-2.07) Law and order *Trade openness -0.724** (-2.04) Constant 9.352*** -0.045 -5.495 0.336 (3.58) (-0.01) (-0.63) (0.06) Observations 61 61 61 61 Number of countries 28 28 28 28 Sargan/Hansen test 0.47 0.47 0.92 0.53 AR(2) 0.27 0.92 0.78 0.71 Notes: data are averaged over five years. Absolute value of z statistics in parentheses; * significant at 10%; ** significant at 5%; ***significant at 1%. 24 Table A.2 - Trade Openness and Poverty Gap in Africa-GMM System excluding Outliers Poverty gap (log) (1) (2) (3) (4) Trade openness (log) 0.894* 1.842*** 5.896* 3.729* (1.93) (3.59) (1.83) (1.94) GDP per capita (log) -1.003*** -0.743*** -1.198*** -0.672** (-2.82) (-3.42) (-3.95) (-2.10) Inflation (log) 0.299** 0.249** 0.147 0.295** (2.29) (2.07) (1.31) (2.23) Education (log) -1.041** -0.348 5.399* -0.712* (-2.49) (-1.12) (1.71) (-1.94) Law and order 0.295 0.210 -0.092 4.623** (1.28) (1.08) (-0.62) (2.06) Private Credit/GDP (log) -0.579** 1.968* -0.113 -0.561** (-2.00) (1.84) (-1.17) (-2.06) Private Credit/GDP *Trade openness -0.654** (-2.57) Education* Trade openness -1.538* (-1.87) Law and order *Trade openness -1.122* (-1.96) Constant 9.963*** 2.226 -9.267 -4.501 (2.70) (1.05) (-0.80) (-0.51) Observations 61 61 61 61 Number of countries 28 28 28 28 Sargan/Hansen test 0.62 0.30 0.93 0.75 AR(2) 0.41 0.78 0.81 0.88 Notes: data are averaged over five years. Absolute value of z statistics in parentheses; * significant at 10%; ** significant at 5%; ***significant at 1%. 25 Appendix 6 - Classification of Countries with regard to the Thresholds (1981-2010) Financial development Education level Quality of institutions Under the threshold Above the threshold Under the threshold Above the threshold Under the threshold Above the threshold Angola Algeria Angola Algeria Algeria Botswana Botswana Cote d’Ivoire*** Burkina Faso Botswana Angola Burkina Faso Burkina Faso Egypt, Arab Rep. Cote d’Ivoire Cameroon Cameroon Egypt, Arab Rep. Cameroon Kenya Ethiopia Congo, Dem. Rep. Congo, Dem. Rep. Ethiopia Congo, Dem. Rep Morocco Guinea Congo, Rep. Congo, Rep. Gambia, The Congo, Rep. Senegal Guinea-Bissau Egypt, Arab Rep. Cote d’Ivoire Morocco Ethiopia South Africa Kenya Gabon Gabon Tanzania Gabon Togo Liberia Gambia, The Ghana Tunisia Gambia Tunisia*** Madagascar Ghana Guinea Ghana Mali Malawi Guinea Bissau Guinea Mozambique Morocco Kenya Guinea-Bissau Niger South Africa Liberia Liberia Nigeria Tanzania Madagascar Madagascar Senegal Togo Malawi Malawi Uganda Tunisia Mali Mali Mozambique Mozambique Niger Niger Nigeria Nigeria Senegal Tanzania South Africa Uganda Togo Uganda Notes: Countries in bold are under the calculated threshold for each of the three characteristics.