Poverty and Distributional Impact of Economic Policies and External Shocks: Three Case Studies from Latin America Combining Macro and Micro Approaches

Lay, Jann Book — Digitized Version Poverty and Distributional Impact of Economic Policies and External Shocks: Three Case Studies from Latin America Combining Macro and Micro Approaches Göttinger Studien zur Entwicklungsökonomik / Göttingen Studies in Development Economics, No. 18 Provided in Cooperation with: Peter Lang International Academic Publishers Suggested Citation: Lay, Jann (2007) : Poverty and Distributional Impact of Economic Policies and External Shocks: Three Case Studies from Latin America Combining Macro and Micro Approaches, Göttinger Studien zur Entwicklungsökonomik / Göttingen Studies in Development Economics, No. 18, ISBN 978-3-631-75365-1, Peter Lang International Academic Publishers, Berlin, https://doi.org/10.3726/b13887 This Version is available at: https://hdl.handle.net/10419/182891 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Poverty and Distributional Impact of Economic Policies and External Shocks Three Case Studies from Latin America Combining Macro and Micro Approaches GÖTTINGER STUDIEN ZUR ENTWICKLUNGSÖKONOMIK / GÖTTINGEN STUDIES IN DEVELOPMENT ECONOMICS Jann Lay Economists have had much to say about the impact of economic policies on growth, but little on their distributional consequences and poverty impact. The reorientation of development policy from structural adjustment to poverty reduction as the central objective thus called for new tools to examine distributional change. This book analyzes the poverty and distributional impact of policy changes and external shocks in three case studies from Latin America: Trade liberalization in Colombia and Brazil, and the gas boom in Bolivia. It uses an innovative approach that combines computable general equilibrium and microsimulation models. The country applications illustrate that distributional consequences depend very much on the nature of the shock or policy change as well as the characteristics of the country in question. The book issues a warning against policy prescriptions being based on oversimplifying assumptions and models. Jann Lay is research associate at the Kiel Institute for the World Economy and completed a doctorate in economics at the University of Göttingen (Germany). He has worked as a consultant to different development agencies on various developing countries in Africa and Latin America. His research interests include pro-poor growth, poverty impact analysis, and the resource curse. GÖTTINGER STUDIEN ZUR ENTWICKLUNGSÖKONOMIK / GÖTTINGEN STUDIES IN DEVELOPMENT ECONOMICS Jann Lay Poverty and Distributional Impact of Economic Policies and External Shocks Poverty and Distributional Impact of Economic Policies and External Shocks Gettinger Studien zur Entwicklungsokonomik Gottingen Studies in Development Economics Herausgegeben von/Edited by Hermann Sautter und/and Stephan Klasen Bd./Vol. 18 £ PETER LANG Frankfurt am Main · Berlin · Bern · Bruxelles · New York · Oxford · Wien Jann Lay Poverty and Distributional Impact of Economic Policies and External Shocks Three Case Studies from Latin America Combining Macro and Micro Approaches PETER LANG Europaischer Verlag der Wissenschaften Open Access: The online version of this publication is published on www.peterlang.com and www.econstor.eu under the international Creative Commons License CC-BY 4.0. Learn more on how you can use and share this work: http://creativecommons. org/licenses/by/4.0. This book is available Open Access thanks to the kind support of ZBW – Leibniz-Informationszentrum Wirtschaft. ISBN 978-3-631-75365-1 (eBook) Bibliographic Information published by the Deutsche Natlonalblbllothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the internet at <http://www.d-nb.de>. Q) : f! Zugl.: Gottingen, Univ., Diss., 2006 Gratefully acknowledging the support of the lbero-Amerika-lnstitut tor Wirschaftsforschung, Gottingen. Cover illustration by Rolf Schinke D7 ISSN 1439-3395 ISBN-13: 978-3-631-56559-9 © Peter Lang GmbH Europaischer Verlag der Wissenschaften Frankfurt am Main 2007 All rights reserved. All parts of this publication are protected by copyright. Any utilisation outside the strict limits of the copyright law, without the permission of the publisher, is forbidden and liable to prosecution. This applies in particular to reproductions, translations, microfilming, and storage and processing in electronic retrieval systems. Printed in Germany 1 2 3 4 5 7 www.peterlang.de Acknowledgements I am grateful to my supervisor Jun.-Prof. Dr. Michael Grimm for his academic support and his patience. I am indebted to the Kiel Institute for the World Economy that provided an inspiring work environment as well as financial support in the past few years. I wish to thank my colleagues in Kiel, Rainer Thiele and Manfred Wiebelt, who co-authored the chapter on Bolivia, as well as Maurizio Bussolo and Dominique van der Mensbrugghe at The World Bank, who co-authored the chapter on Brazil, for the endless but productive discussions and the many hours spent together "crunching" numbers. Special thanks go to Anne-Sophie Robilliard (and again to Maurizio, Manfred, and Rainer) for introducing me to the methods used in this dissertation. Many other colleagues and friends have provided ideas, comments, and technical as well as moral support: Christiane Gebiihr, Olivier Godart, Robert Kappel, Gernot Klepper, Rolf Langhammer, Matthias Liicke, Toman Omar Mahmoud, Cornelius Patscha, Susan Steiner, Saju Thundiyil, and the team of the development economics research group at the University of Gottingen led by Prof. Stephan Klasen, Ph.D. Encouragement came from many other people as well, in particular from my family. 1. General introduction and main findings The extent to which economic growth reduces poverty has always been a central issue in development economics. Obviously, the extent depends on the distribution of the benefits of growth.' Already by the late 1950s, it became apparent that growth in "underdeveloped countries" did not trickle down to the population at large, but was instead accompanied by massive underemployment and unemployment. This "employment problem" implied that growth did not necessarily translate into poverty reduction, but rather to increasing inequalities between those who remained poor and those who were lucky enough to find employment in modern urban sectors. This was consistent with the Kuznets' (1955) hypothesis of an inverted U-shaped relationship between inequality and development, according to which inequality would tend to increase in the early stages of development. Kuznets (1955) admits that his often cited paper is "perhaps 5 percent empirical information and 95 percent speculation, some of it possibly tainted by wishful thinking." Despite these observations, it took until the 1970s and the work of Adelman and Morris (1973) and Chenery et al. (197 4) until the question of income distribution within a country explicitly entered the debate. Adelman and Morris ( 1973) even reached the conclusion that "development is accompanied by an absolute as well as relative decline in the average income of the very poor", although they were challenged by Cline (1975) and Lal (1976) that this finding was not borne out by their data. Lal (1976) harshly criticized these studies and argued that the "concern with distributional issues amongst the international agencies and American development economists marks more their acknowledgement of their neglect of what a number of Third World governments and many development economists have for a long time recognized to be a major area of concern". This assessment certainly contained some truth, but these studies had a significant influence on the research agenda. Lal (1976) went on to conclude that these studies "may perhaps do indirect damage to the prospects of the poor by not emphasizing enough that efficient growth, which raises the demand for labor is probably the single most important means available for alleviating poverty in the Third World". The latter statement illustrates the ideological nature of the discourse between those who "emphasized" growth and others who "emphasized" distribution. This is mainly owed to the fact that the debate of the 1970s still rested, to stick to Kuznets' This introduction focuses on the discourse in development economics with regard to and the empirics of the impact of economic growth and economic policies on the distribution of income, and hence on poverty. For reviews of the theories of development and distribution see Cline ( 1975) and Kanbur (2000). This focus also implies that we do not consider the reverse causal relationship from inequality to growth. See Atkinson and Bourguignon (2000), Aghion et al. ( 1999) for literature reviews and the 2006 World Development report (World Bank 2005a) for recent empirical insights. 14 INTRODUCTION wording, on very little empirical information, a lot of speculation, and possibly even more on wishful thinking. The emphasis on distributional and poverty issues however was relatively shortlived. With the arrival of the debt crisis in the early 1980s, the focus of both development policy and research shifted towards structural adjustment to current and capital account imbalances. As this went along with the arrival of conservative governments in OECD countries, the view that development and poverty reduction could best be reached through economic growth and free markets dominated. In this environment, little research effort was dedicated to resolve the issues raised by earlier empirical studies on the relationship between the distribution of income and economic development and it took until the early I 990s to put the issue back on the agenda.' In the late 1980s, concerns were raised that the costs of structural adjustment programs, which were implemented in most developing economies, were disproportionately borne by the poor (Adelman and Robinson 1989).' In the course of the I 990s, this concern was replaced by the worries, in particular voiced by non-governmental organizations, that the benefits of globalization would be concentrated on the rich in the developing world. In the policy arena, the adoption of the Millennium Development Goals in 2000, which put poverty reduction at the centre of development policies, created demand for detailed micro datasets necessary to monitor progress on the poverty reduction goals. Possibly, the major reason why income distribution was back on the research agenda is related to data availability. Investigators could rely on more data of much better quality, in particular on household survey data, thereby dramatically reducing the degree of speculation contained in earlier studies.' In recent years, a burgeoning literature has significantly improved our understanding of the relationship between growth, poverty, and inequality. Today, it is widely acknowledged that on average growth is distribution-neutral and hence reduces poverty (Ravallion and Chen 2003, Dollar and Kraay 2002). In that sense, "growth is good for the poor" (Dollar and Kraay 2002). Ravallion (2001a) however suggests that one needs to look "beyond averages", as the impact of economic growth on poverty differs substantially across countries. In addition, he notes that this impact can also vary among the poor in a given country. That it is indeed worthwhile to look "beyond averages" is confirmed by many country studies that in recent years have cast light on the very different income distribution 2 Exceptions include a major research project by The World Bank initiated in 1985, the results of which are summarized in Fields ( 1989), and some country studies e.g. on Malaysia by Anand (1983). 3 The UNICEF study "Adjustment with a human face" that examines the poverty impact of structural adjustment with a focus on children and other vulnerable groups received a lot of attention at that time. See Comia, Jolly, and Stewart ( 1987). 4 See Deaton ( 1997 and 2003) on data issues. INTRODUCTION 15 dynamics during growth both across countries and across time.' These dynamics vary case by case, which almost makes it impossible to draw any general conclusions from this literature, not even by region or confined to a specific timeperiod. The only exception is the increase in inequality throughout the late 1980s and 1990s. Inequality measured by the Gini-coefficient, increased in 6 out of 7 country cases presented in Bourguignon et al. (2005a) covering the 1980s to the mid-l 990s and in 8 out of 14 cases in the OPPG6 project covering the 1990s. The differences in country experiences of course raise the question of why is this so. It turns out that even more diversity can be found when one attempts to identify the drivers of distributional chan9 e. The studies in Bourguignon et al. (2005a) employ microsimulation methods to decompose historical changes in income distributions into "fundamental sources": ( l) Changes in the resources at a household's disposal, including human capital accumulated through education, as well as socio-demographic changes, such as changes in the area of residence, age structure, and households composition (endowment or population effects), (2) changes in market remuneration of the factors of production (price effects), and (3) changes in the occupational structure of the population, in terms of labor market participation and formal or informal sector of employment (occupational effects)." The main general lesson from the country studies is that observed changes in the distribution of income result from a number of sources, which may offset or reinforce each other. Since country experiences differ considerably, only few common patterns can be identified. One of these patterns is that price effects, i.e. changes in returns to education, were typically inequality increasing. Some 5 See Christiaensen, Demery, and Patemostro (2003) for a review of a number of Sub- Saharan cases, the country studies on pro-poor growth in the 1990s in World Bank (2005b) and Grimm, Klasen, and McKay (2006), and the collection of East Asian and Latin American experiences in Bourguignon, Ferreira, and Lustig (2005a). 6 "Operationalizing Pro-Poor Growth" was a joint research program undertaken by The World Bank and the French, German, and British donor agencies. The results can be found in World Bank (2005b) and Grimm, Klasen, and McKay (2006). 7 In the tradition of Oaxaca (1973) and Blinder (1973), poverty and distributional changes between two (or more) points in time are decomposed using two (or more) cross-sections of households. This is achieved via simulating counterfactual distributions on the basis of household income generation models that will be described in more detail below. See Bourguignon, Fournier and Gurgand (2001) and Grimm (2005) for further applications of this technique. 8 It should be noted that this empirical operationalization of distributional drivers implicitly reflects the insights from the grand long-term theories of development and distribution from Lewis ( I 954) and Kuznets ( I 955) with their focus on intersectoral movements out of a traditional subsistence into a modem enclave sector to recent models originating from endogenous growth theory where externalities, such as parent-to-child human capital or economy-wide technological externalities, drive accumulation, growth and inequality dynamics (Kanbur 2000). It also mirrors short- to-medium term distributional adjustments to external shocks and policy changes that can be explained in a simple neoclassical trade model with fixed factor supplies. 16 INTRODUCTION conditional findings include the distributional effect of increasing female labor market participation that is found to be very positive when mainly females from poorer households entered the labor market. Increasing informality generally contributes to increases in inequality and is shown to very strong impoverishing effects on the poor in Brazil. Educational expansion improves the income distribution in some cases, but it can also be inequality increasing in the presence of earnings that are highly convex in years of schooling despite improving the distribution of education. These findings illustrate the wealth of insights into the microeconomics of income distribution obtained by this type of decomposition analysis. Yet, such an approach leaves many questions open. First and foremost, these questions concern the factors that explain changes in what Bourguignon et al. (2005b) call the "fundamental" sources of distributional change. In other words, which are the factors that explain the increase in wage inequality in many countries, the increase in female labor market participation, increases/decreases in informality, the patterns of educational expansion, and socio-demographic changes? Possible explanatory factors can be grouped into phenomena related to socio-economic development in general, such as demographic changes and human capital accumulation, external shocks, and economic policies. In particular external shocks and policies related to globalization, such as increased trade and capital flows, the related pattern of technological change, and external liberalization, have recently received a lot of attention as possible reasons for the observed increase in inequality in many countries. The empirical assessment of these "fundamental causes" of distributional change is by no means trivial. It implies to analyze the poverty and distributional impact of specific external shocks, economic policies, or other relevant events, rather than analyzing the reduced-form relationship between changes in the distribution of income and economic growth. Such analyses are extremely policy-relevant, in particular when development policies are geared towards poverty reduction. The chapters in this dissertation therefore address the short to medium run poverty and distributional impact of economic policy changes and external shocks for three Latin American countries.9 More specifically, chapter 2 examines the impact of trade liberalization in Colombia in the early 1990s. Chapter 3 looks at the poverty and distributional implications of the gas boom in Bolivia. While these two chapters "ex-post" analyze past experiences, the third chapter attempts to assess "ex-ante" the possible effects of multilateral trade liberalization on poverty and the distribution of income in Brazil. The country case studies included in this dissertation hence analyse the "micro" impact of "macro" events, which raises the question of the appropriate methodology to do so. In the short to medium run, macro policies as well as external shocks affect household incomes and consumption primarily through two channels; (I) through changes in returns to factors of production, in particular to 9 The short-to-medium run corresponds to 5 to 15 years. INTRODUCTION 17 labor, and in employment, and (2) through changes in relative goods prices. The empirical challenge now lies in linking the policy or external shock to these variables in a first step, in order to assess their impact on household welfare, i.e. the distribution of income and poverty, in a second step. In some cases, where a clear link between the shock and household welfare exists, this can be achieved relatively easily. Tax or price reforms, for example, directly affect real household income. Therefore, the distributional impact of such reforms can be evaluated relatively simply on the basis of household survey data. 10 In most instances, however the analysis is complicated by the fact that there is no direct link between the shock and real household income. Another complication arises from general equilibrium effects that are ignored in the former approach. As a result, most analyses of the poverty and distributional impact of policies and external shocks have turned to Computable General Equilibrium (CGE) models. CGE models based on Social Accounting Matrices (SAM) provide a coherent analytical framework for understanding the complex mechanism through which economic policies and shocks affect household income distribution. Most CGE models applied to evaluate distributional impacts in developing countries are extended neoclassical models that incorporate important structural characteristics of these countries by assuming (a) limited substitution elasticities in various economic relationships and (b) various markets not to work properly.n They can be used for ex-ante assessment as well as for ex-post analysis in order to disentangle the effects of different shocks." CGE models for distributional analysis incorporate different representative household groups representing "classes" defined by area of residence (rural vs. urban), skill level (unskilled vs. skilled), socio-political factors (organized vs. unorganized workforce), or power and wealth (factor endowments, wealth, tenancy of land). In terms of distributional outcomes, the main defining features of 10 See Sahn and Younger (2002) for a short introduction into this approach. Often, such analyses do not account for behavioral responses, but this shortcoming could be remedied by estimating an appropriate empirical model that could be used to simulate behavioral change. 11 See Robinson (1989). Taking into account these structural characteristics was emphasized by the "structuralist" tradition of CGE modelling (see e.g. Taylor 1990). Only some "macro-structuralist" (Robinson 1989) features, such as markup-pricing and Keynesian multiplier effects did not make it into mainstream CGE modelling. The IFPRI (International Food and Policy Research Institute) standard model (Lofgren et al. 2002) in the tradition of Dervis et al. (1982) represents this current mainstream. Similar models have been applied to a number of countries. See Wiebelt ( 1996) for a detailed description of such a model for Malaysia. 12 The first big wave of studies using applied CGE models was motivated by the concern about the poverty and distributional impact of structural adjustment programs. A series of country studies undertaken by the OCED Development Centre tried to assess the impact of actual and (hypothetical) alternative structural adjustment packages on the poor (Bourguignon, de Melo, and Morrisson 1991). Sahn et al. ( 1997) contains a number of country studies from Sub-Saharan Africa. 18 INTRODUCTION these household groups in most CGE model applications are differences in their factor endowments and hence the incomes they receive. Possibly, household groups also differ in labor supply, consumption, and savings behavior. When a shock is applied to a CGE model, sectoral production changes, as do resource reallocations, factor and goods prices, as well as real income and consumption of the respective household groups. To translate the changes in real incomes of the respective household groups into poverty and distributional outcomes, one needs to specify the within-group income distributions. Two approaches have been proposed in the literature (Lofgren et al. 2002). The first, in the tradition of Adelman and Robinson (1978), is to fit ( or to estimate) parametric distributions for each household group, e.g. the log-normal distribution that fits empirical income distributions reasonably well. This implies to categorize households in different groups according to the main sources of income or to other important socioeconomic characteristics of the head of the household. The change in mean real income of the respective household groups is applied to this withingroup distribution, which is "shifted" accordingly. The within-group distributions are finally summed to give the overall income distribution. The second approach uses disaggregated household survey data, classifies households according to the CGE model household groups, and directly applies the changes in real household group income from the CGE to the survey. The calculation of poverty and inequality changes is then straightforward. This approach is also referred to as micro-accounting." The representative household group assumption implies that income distribution variations only result from changes between household groups, given that within household groups the variance is fixed." Yet, recent empirical findings on distributional change indicate that changes within the typical representative household groups of CGE models account for an important share of overall distributional change." At first sight, an obvious way out of this problem would be to increase the number of household groups, or even to incorporate all households from representative household surveys. The latter has been done e.g. by Harrison et al. (2000) in an assessment of Russia's accession to the World Trade Organization (WT0).16 They find the differences in price effects between a model with 10 representative household groups and a model with 55 000 households to be negligible. This finding is not too surprising, as the failure of CGE models to capture some of the distributional dynamics is not grounded in the failure to account for household heterogeneity in terms of factor endowments and/or consumption patterns. The problem is rooted in the fact that CGE models do not account for decisions taken at the individual level. These individual decisions, for 13 See Lay, Thiele, and Wiebelt (2006) for an application. 14 For a detailed discussion of the problems of the representative household group assumption see Bourguignon, Robilliard and Robinson (2005). 15 See again the findings in Bourguignon et al. (2005a) and similar studies cited above. 16 Cockburn (2006) is another CGE application that incorporates all households from a survey. INTRODUCTION 19 example entry into the labor market, falling into unemployment or switching between sectors or occupations, are important drivers of distributional change. In other words, the CGE model treats the factor endowments of a household or household group as fixed (although these endowments may grow at an exogenous rate) and fails to represent individual decisions that may alter these endowments dramatically. Of course, CGE models can be extended to include, e.g. unemployment and/or endogenous labor supply. Yet, in order to capture the income distribution implications, decisions would have to taken by "real" individual household members. This implies to introduce individual "fixed effects" into the model and requires the estimation of structural labor market models (Bourguignon et al. 2005b). Two approaches have been proposed to overcome these shortcomings of applied CGE models. First, individual behavior can be fully integrated into CGE models. There have been attempts to build such fully integrated models (Cogneau 2001, Cogneau and Robilliard 2001 ), but the results are mixed. In particular the formidable identification problems in estimating the structural labor market equations cast doubts on the robustness of this approach. Second, traditional CGE models have been sequentially linked to microsimulation models based on household income generation models that are estimated from household survey data. Using this approach, a pioneering study by Robilliard et al. (2002) examines the poverty and distributional effects of the Asian crisis. A comparison of the results to those obtained under the representative household group assumption reveals the superiority of using a microsimulation model. In the sequential approach, a shock is first simulated in the CGE model and then the microsimulation adjusts micro data so that values for its aggregate variables (wages and employment) are consistent with the CGE macro equilibrium. The "degree of consistency" between the macro and the micro model however differs between applications of the approach, but in a narrow sense a sequential model is confined to be inconsistent both theoretically and empirically. Theoretically, the changes, e.g. in wages and employment are driven by relative price changes, whereas the microsimulation typically only features a reduced-form representation of labor market behavior where prices do not appear as explanatory variables. Empirically, problems arise from the large differences in national accounts and household data, in particular with regard to labor value added, although some authors, e.g. Robilliard et al. (2002), manipulate survey weights to reach "empirical consistency". Yet if the analyst were able to reach complete theoretical and empirical consistency between the micro and the macro model, why not build an integrated model anyway. The strength of the sequential approach lies in the combination of a structural macro model that allows tracing the transmission channels from macro shocks to prices and quantity changes relevant to distributional outcomes and a microsimulation model that provides a detailed account of the household income generation process. The microsimulation model used by Robilliard et al. (2002) follows Bourguignon, Fournier, and Gurgand (2001) and is similar to the one used 20 INTRODUCTION in the country studies in Bourguignon et al. (2005a). In this model, household income is defined as the aggregation of earnings of individual household members, earnings from joint household activities and non-labor income, such as transfers or capital income. The econometric specification underlying the income generation process is composed of two types of equations, those describing occupational choices of the household members and earnings/profit equations. Household members typically choose between inactivity, wage employment and participation in a joint household activity, depending on individual and household characteristics. Decisions by household members are modeled sequentially, i.e. the household head's choice and related earnings enter the decision function of other household members, whereby the simultaneity of occupational choices within a household is taken into account. Of course, simulating poverty and distributional changes based on this type of household income generation models is not without shortcomings. Typically, the behavioral equations, e.g. those governing occupational choices, are estimated from cross-sectional data. It is hence assumed that the observed variation in behavior between individuals is used to simulate behavioral change of (other) individuals in time." Yet, even if panel data was available, constant parameters would have to assumed for the simulation period, which apparently becomes an increasingly problematic assumption the longer time horizon of the analysis. Two of the three country-studies in this dissertation, Bolivia and Brazil, use the sequential methodology that links a CGE and a microsimulation model -despite its imperfections. Both the CGE and the microsimulation model are adapted to the investigated shock as well as to the structural characteristics of the country in question. The studies explore the advantages of this innovative approach, but also hint at its shortcomings and suggest possible areas for improvements. The Colombia study uses a methodology that is similar in spirit. Counterfactual income distributions are generated using a household income generation model that is shocked by changes in earnings, employment, and relative goods prices. Yet, instead of using a CGE model to construct the counterfactual changes in these distributional drivers, we rely on other studies and additional descriptive analyses to identify the changes that can be related to the trade liberalization shock. The issues addressed in the three following chapters are highly relevant and hotly debated in the Latin America context, but also beyond this sub-continent. Most Latin American countries are known for their very unequal income distributions and even the middle-income countries including Brazil and Colombia therefore exhibit relatively high levels of poverty. In such an environment, the evaluation of the poverty and distributional impact of policies and external shocks is key for the design of socially sustainable (and politically feasible) economic policies. In the case of external shocks, such knowledge can be important to 17 Although this assumption seems to be very restrictive, it can be plausibly made e.g. in the context of occupational choices, which are explained mainly by individual educational attainment, age, and household composition variables. INTRODUCTION 21 cushion possible adverse distributional and poverty effects. Economists have developed adequate tools for informing policy-makers ex-ante, i.e. before decisions are taken. The Brazil study contained in this dissertation is one example of such an ex-ante policy evaluation. Of course, much can be learned from looking at past experiences, which is what the other two chapters do. The biggest "external" shock that has hit developing countries undoubtedly is globalization. As an integration process, it encompasses increased trade and capital flows as well as technology transfer. Eventually, globalization and increased interdependence in world markets are triggered by technological change and both domestic and multilateral policies, in particular trade and capital account liberalization. Hence, globalization has many facets. The chapters in this dissertation shed light on the role of some of these facets in explaining distributional change. Whether trade liberalization, which is dealt with in chapters 2 and 3, is good for the poor and whether it increases inequality has been a major concern of many observers of developing countries." The resource boom examined in the chapter on Bolivia may not be directly related to globalization, but the study's current relevance for many other developing countries stems from the fact that the globalization-related economic rise of China and India has fuelled world demand for commodities that are often exported by poor countries. The remainder of this introduction shortly summarizes the main findings and major methodological features of the country studies. The following chapter examines the impact of trade liberalization in the early 1990s on income distribution and poverty in urban Colombia. It first analyzes the effects on the urban labor market, i.e. on labor earnings and employment, and relative goods price changes. Using a microsimulation model of the type described above, the chapter then analyzes how these changes have shaped the distribution of income and how they ultimately affect poverty outcomes. Increasing informality is found to have rather small effects, although very poor groups are affected disproportionately. The increase in the unskilled-skilled wage gap however has major negative distributional implications and lowers considerably the poverty reduction potential of growth. The relative price shifts for consumer goods have an unexpectedly strong positive distributional impact. Finally, the analysis demonstrates that two factors related to a non-tradable boom, increasing female labor market participation as well as an increase in informal profits, have played an important role in cushioning the possible adverse labor market effects of trade liberalization, in particular for the very poor. The Bolivian chapter addresses the question of whether the gas boom of the 1990s has bypassed large parts of the poor population, thereby leading to increasing inequalities in an already unequal society. The chapter examines the 18 See Winters et al. (2004) for a very comprehensive review of the empirical evidence on trade and poverty. Some recent evidence with a focus on Latin America can be found in Harrison (2005). 22 INTRODUCTION transmission channels through which the large resource inflows related to the gas boom, both initial foreign investment in the sector and the subsequent export earnings, as well as large public transfer programs affect the distribution of income. These transfers may well be interpreted as a means of redistributing resource rents. The CGE model explicitly models the gas sector and takes into account its enclave character. The focus of the analysis is on the second round labor market impacts of Dutch disease type effects, in particular on shifts between formal and informal employment and changes in relative factor prices. The microsimulation model is specified accordingly. The simulation results suggest that the gas boom induces a combination of unequalizing and equalizing forces, which tend to offset each other. As net distributional change is limited, growth generated by the boom reduces poverty despite increasing informality. For the Brazilian case, the next chapter intends to evaluate the poverty effects of possible trade liberalization outcomes of the Doha round in the medium run. 19 This implies to assess the poverty impact of a Doha Round (and a Full Liberalization) counterfactual scenario against a scenario that incorporates some of the main features of medium run structural change. The chapter thus examines whether the effects of trade liberalization, in particular on poverty and the distribution of income, are still prominent in the medium run. The main povertyrelevant transmission channels incorporated in the simulation exercise are changes real factor prices and changes in the sectoral composition of the workforce, focusing on employment movements between agricultural and non-agricultural sectors. Structural change is driven by changing consumption patterns, differentials in productivity growth rates across sectors, educational upgrading of the workforce, and, finally, the trade shocks. The methodology combines again a sequentially dynamic CGE model with a microsimulation model that takes into account educational expansion on the micro level. The analysis suggests that the economic effects of the Doha round, even of an "optimistic deep" liberalization scenario, are rather limited for Brazil. Accordingly, poverty would remain largely unaffected by this trade reform, which does not appear to be biased in favor any of particularly poor groups. Yet, through a slight improvement in the urban income distribution the Doha scenario has some positive effect on poverty. In contrast, a full liberalization scenario that implies drastic domestic tariff cuts causes quite substantial welfare gains that are concentrated among some of the poorest groups of the country, in particular those in agriculture. Yet, relatively strong contractions in manufacturing sectors give reasons for concern. 19 This chapter forms part a major research program of The World Bank to assess the poverty impacts of a possible WTO agreement ("Doha" round). World market price changes caused by multilateral liberalization were calculated using a global CGE model and then passed to a number of single country cases. The results of all country case studies as well as a summary can be found in Hertel and Winters (2005). This book also contains a shorter summary version of chapter 3. CAPTURING THE TRANSMISSION CHANNELS: METHODOLOGY 29 differ for skilled and unskilled labor as well as for females and males, which implies that there are four wage labor market segments. The following set of equations describes the model. Household m has km members, which are indexed by i. logw =a +x P +e mi g(mi) mi g(mi) mi Nm= tlnd[c:<,.n +z,.;a:(mn +u;,; >Sup(o,c;(., 0 +z,,,;a;., 0 +u;J] i - 1 (I) (2) (3) (4) (5) (6) The first equation is a Mincerian wage equation, where the log wage of member i of household m depends on his/her personal characteristics. The explanatory variables include schooling years, experience, the squared terms of these two variables, and a set of regional dummies. This wage equation is estimated for each of the four labor market segments. The index function g(mi) assigns individual i in household m to a specific labor market segment. The residual term em; describes unobserved earnings determinants. The second equation represents the profit function of household m. Profits are earned if at least one member of the household is self-employed. The profit function is of a Mincer type and includes as explanatory variables the schooling of the household head, her/his experience plus the squared terms the former two variables, and regional dummies. Of course, profits also depend on the number of self-employed in household m, Nm, The residual &m captures unobserved effects. Household income is defined by the third equation. It consists of the wages and profits earned by the household members and an exogenous income Yom, This exogenous income corresponds to "other income" in the survey and may include government transfers, transfers from abroad, capital income, etc .. !Wm; is a dummy variable that equals I if member i of the household is wage-employed and 0 otherwise. Likewise, profits will only be earned if at least one family member is self-employed (N,,;>O). Household income is deflated by a household specific price index, which is defined by equation (4). The parameters denotes the expenditure shares for food- and non-food. These shares are calculated by household income quintiles. Note that the prices Pi for food and Pn/ will be I initially. The index 30 COLOMBIA function d(m) indicates to which of the five income brackets household m belongs and which food expenditure share is assigned to the household. The fifth equation explains the aforementioned dummy !Wm;. The individual will be wage-employed if the utility associated with wage-employment is higher than the utility of being self-employed or inactive. The utility of being inactive is arbitrarily set to zero, whereas the utilities of the employment options depend on a set of personal and household characteristics, =mi• These characteristics include gender, marital status, education, experience, other income, the educational attainments of other household members, and the number of children. Unobserved determinants of occupational choices are represented by the residuals. Equation (6) gives the number of self-employed. Similar to the choice in equation (5), the individual i of household m will prefer self-employment if the associated utility is higher than the utility of inactivity or wage-employment. The self-employed household members form the "household enterprise" with Nm working members. Thus, the last two equations represent the occupational choices of the household members. The occupational choice model is estimated separately for household heads, spouses, and other household members in urban and rural areas. The index function h(mi) assigns the individual to the corresponding group. The model just described gives the household income as a non-linear function of individual and household characteristics, unobserved characteristics, and the household budget shares. This function depends on three sets of parameters, which are estimated based on the 1988 survey. These parameters include ( 1) the parameters of the wage equation for each labor market segment, (2) the parameters of the profit function for "household enterprises", (3) the parameters in the utility associated with different occupational choices for heads, spouses, and other family members. The income generation model requires some comments on the assumptions behind its formulation. First of all, despite the availability of data on working time we decided to model the occupational choice as a discrete choice.28 Secondly, our model assumes that the Colombian labor market is segmented along different lines. One line of segmentation separates wage-employment from selfemployment. In a perfectly competitive labor market, the returns to labor would be equal for these two types of employment. Yet, segmentation may be justified because income from self-employment is likely to contain a rent from non-labor assets used, and its clearing mechanism may differ from that of wage employment. Information on non-labor assets is not available for Colombia, hence distinct equations need to be estimated even if the labor markets were competitive. In addition, even in those cases where information on non-labor assets is available, a segmented labor market can be justified by the fact that wage-employment may be rationed and self-employment thus "absorbs" those who do not get a job in the preferred wage work. Wage work could be preferred for generating a more steady 28 However, estimating wage equations based on hourly wages did not make a major difference in the coefficients. CAPTURING THE TRANSMISSION CHANNELS: METHODOLOGY 31 income stream or for fringe benefits related to this type of employment. Conversely, self-employment might exhibit important externalities, for example for households, in which children have to be taken care of. Self-employment of the household head may also create employment opportunities for other family members. Additional segmentation is assumed within the wage labor market. The segmentation hypothesis along the lines of skills and gender is strongly supported by the regression results. In a first step, the occupational choice model and the wage and profits equations are estimated in order to obtain an initial set of coefficients (ac;, /k, bh <5F, CH"', aHw, ci, aH') and unobserved characteristics (em;, &m, u"'m;, U 5 m;). 29 The income generation model is estimated using data from the Colombian household survey from 1988. 30 The estimated benchmark coefficients are then employed and changed in the micro-simulation. Unobserved characteristics say for the wage equation can of course only be obtained for those who are actually wageemployed. For self-employed or inactive individuals the unobserved characteristics in the wage-equations are randomly drawn from normal distributions with the respective estimated error variances. In the same way, we generate unobserved characteristics for the profit function for households without a household enterprise. As we estimate wage and profit functions using ordinary least squares (or 2 SLS), these unobserved characteristics are assumed to be normally distributed as well, again with the estimated error variances. Additionally, unobserved characteristics (or utility) need to be generated for the occupational choice model. In latent utility models, these residuals cannot be observed and are hence generated from the distribution underlying the respective model. The multinomial choice model assumes that errors follow the Gumbel ( or type I extreme value) distribution.'1 Residuals have to be drawn consistent with the observed occupational choice, i.e. the utility an observed wage earner relates to 29 The occupational choice model was estimated using a multinomial logit. The wage equations were estimated by Ordinary Least Squares. Correcting for selection bias in these equations did not lead to major changes in the results and was hence dropped. In the estimation of the profit functions, the number of self-employed was instrumented. The estimation results for the wage and profit equations as well as for the occupational choice models are reported in App. Table 2.2 and App. Table 2.3. For a more detailed discussion of the estimation methods see Bourguignon and Ferreira (2005). 30 The household survey used for estimation of the micro-simulation parameters is the Colombian Encuesta Nacional de Hogares from 1988 (EH61). After the removal of outliers, removal of individuals with top-coded earnings, and observations with missing data the survey covers 29 729 individuals living in 12 092 households in urban areas. The household weights are adjusted for the removed observations. The expenditure shares are calculated from an income and expenditure survey and matched with the EH61 based on household groups, classified according to income quintiles. For the problems of these datasets see Nunez and Jimenez ( 1997). 31 The variance of this distribution is -rr?/6. Train (2003) contains a detailed (and accessible) discussion about the distributional assumption oflogit models. 32 COLOMBIA wage-employment has to be higher than the utility associated with inactivity or self-employment. Statistically, this implies to draw these residuals conditional on the observed choice. We apply a method proposed by Bourguignon, Fournier and Gurgand ( 1998), who show how to obtain these conditional draws from a Gumbel distribution." In the simulation, changes in aggregate variables are used as target values. Individual earnings and occupational choices change to reach these targets on the aggregate level. The required individual changes are obtained by varying coefficients in the occupational choice and the earnings models. In other words, coefficients are adjusted and occupational choices and earnings change accordingly, until the results of the micro-simulation are consistent, at an aggregate level, with the given aggregates. Formally, the following constraints describe the consistency requirements. Let the right hand side variables describe the initial aggregate values, where EG corresponds to the number of wageemployed, Sc;, the number of self-employed, we;, the sum of wages paid in segment G, m·, the sum of profits paid in activity F. "G" stands for the eight labor market segments, i.e. urban male skilled and unskilled, urban female skilled and unskilled, rural male skilled and unskilled, rural female skilled and unskilled labor. Note that A indicates that the coefficients, residuals, and indicator function values result from the estimation described above. L L1 Wm; = m 1,g(mi)=G L IInd[c;(nu) + =m,ah(mi) + u::;; > Sup(o,ch(mi) + =miah(mi) + u!,; )]= Ee m i,g(mi)=G L I1ndlch(m1) + =m,ah(mi) + U~; > Sup(o,ch(mi) + =miah(mi) + u::;, )J= Sc m i,g(mi)=G L Iexp(ac; +Xm;/Jc; +em;)I~mi =We m 1,g(mi)=G (7) (8) (9) (10) 32 Another ad-hoc approach would use a random number generator for the Gumbel distribution (u=-log(-log(x)) where xis a random draw from a uniform distribution), and repeatedly draw residuals until the observed choices correspond to the simulated ones. CAPTURING THE TRANSMISSION CHANNELS: METHODOLOGY 33 The liberalization shock now produces changes in these aggregates and in the price vector. The result is a new vector of these variables, which will be identified by an asterisk (E'a, s•a, w •a, rc°). For the above constraints to hold, an appropriate vector of coefficients and prices (aa, f3a, b, o, CHw, aHw, ci, aH) is needed. For these coefficients, many solutions exist and additional constraints have to be introduced. As in Robilliard et al. (2002) our choice is to vary the constants (aa, b, cw H, c58) and leave the other coefficients unchanged. We hence assume that the changes in occupational choices and earnings are dependent on personal and household characteristics only to a limited degree. Changing the intercept in one of the wage equations implies that all individuals of the respective segment experience the same increase in log earnings. This increase does not depend on individual characteristics. The same holds for the profit function. Consistency of the microsimulation and the new aggregates hence requires the solution of the following system of equations. This is the way the microsimulation is "forced" to reproduce given changes in aggregate variables. L L!Wm; = m i,g(mi)=G L L lnd[cZ(mi) + =m;lXh(mi) + u;, > Sup(o,cZrmi) + =m/4,(mi) + U~; )]= E; m i,g(mi)=G (II) I: I: lndlcZrmi) + =m;Cll,(mi) + U~; > Sup(o,cZ7mi) + =miah(mi) + u;; )J= s; m i,g{mi)=G L I:exp(a; +Xm;Pc +em;)I~mi =w; m i,g(mi)=G L exp (be; + z mac; + i m }Ind (Nm > 0) = ,r * m (12) (13) (14) Equations (11) and (12) require the number of self-employed and wageemployed to be consistent with the target values for each of the four segments (G). This also holds for the wage equation for each of the segments and the profit function, as indicated by equations (13) and (14). Hence, the above system contains 13 restrictions. The system has four unknown constants in the wage equations, one in the profit function, and 8 in the occupational choice model." 33 Note that the constants of the occupational choice model -though estimated separately for heads, spouses, and others -are changed separately across the four labor market segments. Therefore, we have 8 unknown constants in the occupational choice model, two occupational choices in each of the four labor market segments. 34 COLOMBIA Thus we have 13 unknown constants and 13 equations. We obtain the solution by applying standard Gauss-Newton techniques. Solving the above system gives us a new set of constants (a•o, b•, c•wH, c••H), which is then used to compute occupational choices, wages, and profits. Additionally, changes in relative prices of food and non-food items are taken into account by deflating the resulting household incomes by household group specific price indices. Before we turn to the simulations we shortly would like to point to some of the shortcomings of the model. First, we just differentiate between formal and informal employment defined by being wage- or self-employed. Yet, income differences also exist e.g. between manufacturing and personal service sectors and these sectoral wage differences are ignored. In addition, self-employment activities are very heterogeneous, which is only rudimentarily reflected in the model. An interesting extension of the income generation model would hence be to differentiate between subsistence-oriented and other self-employment activities. A more general second set of drawbacks of the microsimulation is its reliance on one cross-section and the assumption of constant behavioral parameters. Choosing the constants in the occupational choice and earnings equations to vary is an additional rather restrictive assumption. It implies that the within-group wage distribution only changes through entry and exit of individuals with certain characteristics and not, for example, through changes in returns to education. This may not be relevant for unskilled labor, but certainly for skilled labor in the considered period. A related third shortcoming regards the use of an occupational choice model, which is estimated comparing occupational "states", to simulate occupational transition. For example, the simulation makes those women move into employment first, who are relatively unlikely to be inactive; these are ceteris paribus more educated women. Whether it is the more educated who, in reality, move into employment may depend a lot on how the shock impacts on labor demand. A final point should be made on the rudimentary nature of the model with regard to the expenditure side. Expenditure shares, which are only defined by income quintile, are fixed, i.e. substitution is not allowed for. Furthermore, we only consider two price indices based on baskets of food and non-food items. Household heterogeneity in terms of consumption patterns is hence quite limited. 2.4. The poverty and distributional impact of"Apertura" To set the stage for the following analysis, we examine the composition of income of urban households by the sources that we distinguish in our analysis, i.e. wage income from unskilled/skilled male/female labor as well as "informal" profits from self-employment and exogenous other income. Figure 2.1 shows the composition of household income by per capita income vintiles. As expected, informal profits constitute the main share of income for the poorer parts of the population. The poorest 5 percent of households derive almost half of their income from selfemployment. Unskilled male wage-employment accounts for most of the income earned by poorer households, whereas male as well as female skilled wage income THE POVERTY AND DISTRIBUTIONAL IMPACT OF "APERTURA" 35 is an important income source only for the richest 20 percent of the population. Unskilled female wage income is most important for the urban middle class. Figure 2.1: Composition of income by source by income vintile 5 10 Source: Author's calculations. 15 • other • profits • sk fem wage • sk m wage a us fem wage •us m wage The historical scenario: To put the effects of trade liberalization into perspective, we construct a historical labor market scenario by comparing the 1988 and the 1995 survey. In terms of employment composition, the most remarkable development in the early 1990s is the sharp increase of self-employment across all labor market segments, which for both male unskilled and skilled workers is at the expense of wage-employment (Table 2.2). Female labor market participation increases considerably, especially in self-employment activities.34 Table 2.2: Labor force composition in 1988 and 1995 1988 initial shares 1988-95 point change in shares Inactive Wage-work Se/f-empl. Inactive Wage-work Self-empt. Unskilled male 6.5 6.5 32.3 -0.3 -5.8 6.1 Skilled male 7.6 7.6 19.5 0.1 -3.0 2.9 Unskilled female 64.3 64.3 13.9 -6.8 2.1 4.7 Skilled female 48.6 48.6 9.3 -2.4 -0.5 2.9 Total urban 32.5 32.5 17.3 -4.0 0.2 3.8 Source: Author's calculations. Note: The right panel of the table displays the percentage point change with regard to the initial occupational category shares. 34 Our results are consistent with former studies, although comparability is limited due to the different segmentation choices. For an overview of labor market indicators for 1988 and 1995 see Velez et al. (2005). Ocampo et al. (2000) additionally consider the sectoral composition of employment. 36 COLOMBIA Table 2.3: Real wages and self-employment income, 1988 and 1988-95 evolution 1988-95 Initial change Wage Unskilled male 40659 2.4 Skilled male 85 331 6.5 Unskilled female 30 316 -4.8 Skilled female 57 272 6.8 Self-empl. income 44 071 13.0 Source: Author's calculations. Note: the second column shows percent changes. Table 2.3 reports (cumulative) income changes by labor market segments between 1988 and 1995. For both males and females the wage gap between unskilled and skilled labor increases considerably. This increase is more pronounced for females, as unskilled female wages even decline, which may also be caused by unskilled female entrants in the lower parts of the wage distribution. Whereas unskilled male wages grow only slightly in the period under consideration, informal profits increase substantially. These historical labor market aggregate changes should be interpreted with some caution, as they are just based on two surveys. We now pass these historical changes in employment and earnings to the microsimulation model to test whether the model is adequate to reproduce distributional and poverty change. Such an exercise will hence illustrate whether there are distributional changes the methodology cannot trace. Overall, this validation exercise suggests that the admittedly simple income generation model does reasonably well in capturing the forces of distributional change that are considered key transmission channels of trade liberalization. In Figure 2.2, we plot growth incidence curves of historical (based on the 1988 and the 1995 household survey) and of simulated per capita income changes. Overall, the microsimulation gives lower growth in per capita incomes than could be observed historically (see also Table 2.4). This is not too surprising, since human and physical capital accumulation is partly ignored in the model. For example, the simulation does not reflect the considerable educational expansion between 1988 and 1995 that is also reflected in a marked increase in the share of skilled labor (App. Table 2.1 ). The distributional effect of this expansion is negative, since earnings are highly convex in years of schooling. This convexity implies that (wage) inequality increases if educational endowments of every individual increase by a constant greater than 1. Actually, the distribution of years of schooling became more equal between 1988 and 1995, which could not offset THE POVERTY AND DISTRIBUTIONAL IMPACT OF "APERTURA" 37 Figure 2.2: Growth incidence curves, real data vs. historical simulation 1988-95 (5 th to 95 th percentile/5 ~ " ;; 0 ~"' " !; 00 OOV> ~- C 8 t .<>o .,,- '; ~ ~~ .5 s ·~ 1;0 0. 0 w ~ w w Cumulative population ranked by per capita income 1-- 88 vs. 95 survey ----- Historical simulation 100 Source: Author's calculations. the unequalizing effect of higher mean education.36 This also explains why the gap between the historical and simulated growth incidence curves widens with rising income. Another phenomenon that the simulation does not capture is rural-urban migration, which is likely to increase the number of rather young and relatively uneducated individuals in urban areas. Despite these shortcomings, the microsimulation quite well reflects the structure of distributional change, in particular the "steps" in the historical growth incidence curve around the 40 th and the 75 th percentile. Furthermore, it reproduces oscillating income changes in the lower parts of the distribution and even traces some of the spikes of the historical 35 For expositional purposes, we only plot the curves from the 5th to the 95th percentile. The entire curves can be found in App. Figure 2.1. For both the lowest and highest percentiles there are larger deviations between the simulation and real data. For the lowest percentile, the historical data shows a substantial decrease between 1988 and 1995. In addition, the microsimulation cannot capture the very large per capita increases in the highest percentiles. We can only speculate on the extremely high negative (for the poor) and positive (for the rich) growth rates found in the historical data. While conflict-induced migration of poor migrants into the cities may well be a reason for negative growth at the bottom of the distribution, the high growth rates for the rich may arise from data deficiencies. For 1995 we have to drop 80 observations due to top-coding, while for 1988 the number amounts to 122, with roughly the same sample size. Despite our efforts to adjust the household weights for dropped observations, this might explain part of the observed increase. Although this merits further investigation it is not crucial to the question examined here. 36 For details see Velez et al. (2005) and the summary chapter in Bourguignon, Ferreira, and Lustig (2005a). 38 COLOMBIA growth incidence curve. Yet, the differences between the "real" and the simulated historical scenario are also reflected in the aggregate indicators reported in Table 2.4 . The relatively pronounced differences in poverty reduction are mainly due to the failure of the microsimulation to capture the spike in the "real" growth incidence curve around the 30th percentile. Whereas the Gini-Index remains constant in the microsimulation, the survey data shows a considerable increase. In the microsimulation, the positive distributional impact of increasing informal profits and increasing female labor market participation hence compensate the negative effects of a rising wage-gap and increasing informality. In reality, educational expansion and very high growth rates of incomes in the highest ten percentiles, which the microsimulation fails to capture (or which may partly be due to data deficiencies), dominate distributional outcomes. Table 2.4: Poverty and distributional simulation results, real data vs. historical simulation 1988-95 _J Initial 1995 Hist sim Per capita income 21899.6 18.0 73 Source: Author's calculation. % point change in index Poverty Inequality PO PI P2 Gini Theil 60.4 26.S 14.9 39.9 46.7 -6.8 -4.2 -2.7 32 -3.8 -2.3 -I 4 0.0 -0.1 While the above historical scenario reflects all policy changes and shocks, the following discussion intends to evaluate employment and earnings changes attributable to trade liberalization. We review existing evidence that we complement by own descriptive survey-based evidence. First, we consider liberalization-induced changes in the employment structure, i.e. employment creation or destruction and associated increases/decreases in unemployment or employment shifts between formal and informal activities. We then discuss earnings changes. The impact of trade liberalization varied considerably across sectors. According to Ocampo (1999), light manufacturing, such as apparel and leather, was mainly hurt through the real appreciation, whereas heavy manufacturing faced strong import competition. This competitive effect was reinforced again by the appreciation. The contraction in the manufacturing sectors also translated into a reduction of wage-employment and hence to informalization." In light manufacturing, informal female employment shrank massively, which is very likely to reflect job losses for self-employed women who are sub-contracted for work at home in the leather and apparel sector. The data however suggests that these job losses have not been associated with an increase in unemployment, in line with findings by Attanasio et al. (2004). The losses of wage-employment in 37 See the changes in the sectoral composition of employment by labor market segment reported in App. Table 2.4. APPENDICES 2.6. Appendices 2.6.1. Additional figures App. Figure 2. 1: Complete growth incidence curves, real data vs. historical simulation 1988-95 "?f.o .s "° .,., °' ~ -0 ~~ 00 ~ C " to _8c-, t " §0 .5 s -~ "o ~N ~· 0 W ¼ W W 100 Cwnulative population ranked by per capita income 1--- 88 vs. 95 survey ----- Historical simulation Source: Author's calculations. App. Figure 2.2: Growth incidence curves, simulations 11 and 111 ~o~-----------------------~ .s- "' °' ~ ] 00 00 °'"' 0 w ¼ w w Cwnulative population ranked by per capita income 1--- Sim II ----- Simm! Source: Author's calculations. 100 45 46 App. Figure 2.3: Growth incidence curves, simulations IV and V 0 w ¼ ~ w Cumulative population ranked by per capita income 1--- Sim IV ----- SimV I Source: Author's calculations. 2.6.2. Additional tables 100 App. Table 2./: Composition of the labor force by gender and skill 1988 1995 Unskilled Male 43.5 36.8 Skilled Male 21.2 23.9 Unskilled Female 20.9 20.3 Skilled Female 14.4 19.0 Source: Author's calculations. COLOMBIA APPENDICES App. Table 2.2: Estimation results of wage equations and profit function Wage unsk male unskfem sk male Years of schooling 0.036 -0.092 0.045 (0.012)*** -0.105 (0.018)** Years of schooling 0.003 0.009 0.003 squared (0.001 ) ... (0.004) .. (0.001)** Experience 0.047 0.063 0.037 (0.002)* .. (0.004)*•• (0.004) ... Experience squared -0.001 -0.001 0.000 (0.000)'** (0.000)*** (0.000)• .. Education head Experience head Experience squared head Number of self-employed Constant 9.314 9.956 9.041 (0.050)* .. (0.690) ... (0.080)• .. Observations 6316 3100 2750 R-squared 0.22 0.43 0.18 Standard errors in parentheses * significant at 10%; ** significant at 5%; *** significant at I% Note: Regional dummies not reported unsk fem -0.034 -0.134 0.006 -0.005 0.045 (0.005)••· -0.001 (0.000)* .. 9.663 (0.877)• .. 2229 0.33 47 Self-empl. income 0.102 (0.004)**• 0.023 (0.005) ... 0.000 (0.000)* .. 0.959 (0.215)*** 8.656 (84.20)•• 4188 0.18 48 COLOMBIA App. Table 2.3: Estimation results of the choice models, average marginal effects heads spouses wageself- wageself- inactive employed employed inactive employed employed Male -0.271 0.149 0.123 -0.564 0.423 0.141 (0 015)* .. (0 016)••· (0.013)••· (0 038)••· (0 047)••· (0.058)••· Married -0.01 I 0.061 -0.050 -0019 0.032 -0.014 -0.008 (0011 )••· (0.010)••· -0.009 (0.011 )* (0.013)••· Years of schooling 0.008 -0.003 -0.005 -0.010 0.021 -0011 (0.003)••· -0.005 -0.005 (0.005)•• (0.006)* (0.007)••· Years of schooling squared -0.001 0.001 0.000 -0004 0.003 0.001 (0.000)* .. (0 000)••· 0.000 (0.000)••· (0.000)••· (0.000)••• Experience(age-years of -0.006 -0004 0.010 0.014 -0.023 0.009 schooline-6) (0.001 )••· (0 002)** (0.002)••· (0.002)••· (0 001 )••· (0.002)••· Experience squared 0.000 0.000 0.000 0000 0.000 0.000 (0.000)••· (0.000)* (0.000)*** (0.000)* .. (0.000)*0 (0.000)*** Individual other income 0.000 0.000 0.000 0.000 0.000 0.000 (source not soecified) (0.0001••· (0.0001••· (0 0001•• coooo1•• 0.000 0.000 No. of adult males in hh 0.017 -0.001 -0.016 0.003 -0.055 0.053 without education -0.014 -0.027 -0.024 -0.018 (0.028)* (0.025)** No. of adult females in hh 0.020 -0.055 0.035 -0.050 -0.010 0.060 without education (0.01 I)* (0.020)••· (0.019)* -0.034 (0.036)* -0.018 No. ofadult males in hh with 0.019 -0.003 -0.016 -0.042 -0.013 0.055 primary education (0 004)••· -0.009 (0 009)* (0.01 I)••• (0.007)* (0 012)••· No. of adult females in hh 0.005 0.021 -0.026 0.053 -0.002 -0.051 with orimary education -0.004 (0008)** (0.008)••· (0.0121••· (0013)••· -0009 No. of adult males with 0.015 -0.009 -0006 -0.041 -0.017 0.058 secondarv education (0.003)••· -0.007 -0006 (0.006)••· (0.009) ... (0.008)••· No. of adult females with 0.010 0.004 -0014 -0002 -0004 0.006 secondarv education (0.003} ... -0.006 (0.006)** -0.006 -0.009 -0009 No. of adult males with high 0.006 -0.010 0.003 0.019 0.007 -0.026 education -0.006 -0.01 I -0011 -0.012 -0.008 (0.010)** No. of adult females with 0.001 -0.015 0.014 -0.002 -0.024 0.026 hieh education -0005 -0.010 -0.010 (0.013)* -001 I (0016)* No. of elderly males in hh -0.047 0. 132 -0.085 0.075 -0.010 -0.066 without education (0.019)** (0.047)••· (0.042)** -0.059 -0.067 -0081 No. of elderly females in hh -0.030 0.010 0.019 0.040 -0.101 0.061 without education (0.016)* -0.033 -0.031 (0.053)* -0.050 -0 035 No. of elderly males in hh -0.026 -0.034 0.060 0.035 -0.031 -0004 with orimarv education (0.014)* -0.027 (0.026)** -0.023 -0037 -0.034 No. of elderly females in hh 0.000 -0.008 0.008 -0.049 0.041 0.008 with orimarv education -0.008 -0.020 -0.018 (0.024)* (0.028)* -0.019 No. of elderly males in hh -0.037 0.052 -0.015 0.109 -0.067 -0.042 with hieh education -0.023 -0.041 -0036 (0.043)** (0.039)* (0.023)* No. of elderly males with 0.021 -0.077 0.056 0.122 -0.170 0.048 secondarv or hieher -0.019 -0.047 -0.044 (0.034)••· (0 066)••· -0.053 No. of children aged Oto 9 0.002 -0.008 0.006 0.035 -0.004 -0.030 -0.002 (0.004)** -0.004 (0.005)••· -0003 (0.005)••· No. of children aged 10 to 18 -0.004 0.000 0.003 0.002 0.008 -0.010 -0.002 -0.004 -0.004 (0.005)** -0003 -0 005 Wage income of household 0.000 0.000 0.000 head (0.0001**• 0.000 (0.000)* .. Observations 12092 8334 Pseudo R-squared 0.1865 0.0988 APPENDICES 49 others wageself- inactive em(?loyed em(?loyed Male 0.237 -0.320 0.083 (0.012!*** (0.008)'" (0.011)"' Married -0005 0.130 -0.125 -0.012 (0.018)"' (0.020)"' Years of schooling -0013 0.012 0.001 (0.007)' (0.006)" -0.004 Years of schooling squared 0.000 -0001 0.001 0.000 (0.000)** (0000)" Experience(age-yean of 0.011 0.019 -0.030 schooling-6! (0.002)'" (0.002)*" (0 001)*" Experience squared 0.001 0.000 0000 (0.000)'" (0.000)*** (0.000)'** Individual other income 0.000 0.000 0.000 {source not sl!ecified! 0.000 (0.000)"' (0.000)'" No. of adult males in hh -0.012 -0 007 0.019 without education -0.013 -0.022 -0.019 No. of adult females in hh 0.006 0.002 -0.007 without education -0.019 -0020 -0.012 No. of adult males in hh with 0.025 -0018 -0.007 (!rima!l education (0.007)" -0.004 (0.006)"* No. of adult females in hh 0.006 0.014 -0.021 with (?rima!l education (0.005)"* (0 008)' -0.007 No. of adult males with -0.006 -0.026 0.032 seconda!J: education (0.006)**" -0.005 (0 006)'"' No. of adult females with 0.015 0.000 -0.015 seconda!J: education (0.004)"' -0005 (0.006)" No. of adult males with high 0.005 -0040 0.035 education (0.01 I)*" (0.01 I)*" -0.008 No. of adult females with -0.039 0.032 0.007 high education -0.007 (0 01 I)*" (0.011)'" No. of elderly males in hh -0075 0.035 0.040 without education -0.041 (0.013)"' -0039 No. of elderly females in hh -0.012 -0 01 I 0.023 without education -0.037 -0 023 -0.032 No. of elderly males in hh 0.032 0.016 -0.048 with l!rima!l education -0.025 (0.027)' -0.016 No. of elderly females in hh -0.016 0.001 0.015 with j!rima!l education -0.018 -0.020 -0.012 No. of elderly males in hh -0.004 -0.005 0.010 with high education -0.040 -0.027 -0.041 No. of elderly males with -0.096 0.010 0.086 seconda!J: or higher -0029 (0.042)" (0.047)' No. of children aged O to 9 0.009 0.006 -0.014 (0.003)' (0.005)'" (0.004)" No. of children aged 10 to 18 0.001 0.006 -0.007 (0.003)" -0.004 (0.004)' Wage income of household 0.000 0.000 0.000 head 0.000 (0.000)" (0.000)'" Observations 9303 Pseudo R-squared 0. 1345 Standard errors in parentheses • significant at IO¾; •• significant at 5%; ••• significant at I% 50 COLOMBIA App. Table 2.4: Percentage point changes in sectoral employment composition, urban labor force, /988-95 unskilled male skilled male unskilled female skilled female Total waee self wa2e self wae:e self wae:e self wae.e self all agr -0.6 -0.2 -0.6 1.0 -0.4 0.1 -0.1 0.0 -0.5 0.1 -0.4 min -0.1 -0.2 -0.1 0.4 0.0 0.1 -0.1 -0.7 -0.1 -0.1 -0.1 lif(ht man -0.3 0.7 -0.3 -0.6 -0.1 -4.2 -0.2 -1.9 -0.9 -0.6 -0.9 heal' man -1.0 0.7 -0.3 0.7 -2.2 0.3 1.0 1.2 -1.0 0.7 -0.7 Ulil -0.2 0.2 0.1 -0.6 -0.1 -0.1 COn.\'l I.I 9.1 I.I 4.5 0.5 0.1 0.1 0.7 -0.1 4.4 1.2 -,-~~d~-~;;,--- -5.8 2.8 0.5 ... 2.0 1.8 ···· is --3.4 0.8 0.8 -7.1 -0.8 trans 0.6 3.4 -0.3 9.1 0.0 0.1 -0.5 0.2 -0.4 2.7 0.6 ---- ---- -~ -------- ----------- comm 0.3 0.5 0.2 0.1 0.4 0.2 fin 0.8 o~_ -2.4 -1.8 -I.I 0.2 -0.9 -3.7 -0.1 0.3 -0.2 ·------ ---- --·-·-·--· -----·---- pub -2.0 -0.9 -0.2 -0.8 -0.8 -0.7 oth serv 0.6 -8.1 0.3 -6.2 4.1 2.7 -0.2 2.4 2.1 -4.2 0.3 Source: Author's calculations. App. Table 2.5: Shares of manufacturing wage-employment across labor market segments 1988 1995 point diff Unskilled Male 21.6 18.8 -2.9 Skilled Male 20.1 18.9 -1.2 Unskilled Female 23.4 20.0 -3.3 Skilled Female 14.5 14.6 0.0 Total Urban 20.7 18.3 -2.4 Source: Author's calculations. App. Table 2.6: Sectoral shares of wage-employment and skill intensity (wageemployment, 1988 and 1995 I Share in employment Skilled labor share % point 1988 1995 1988 1995 diff. %diff. agr 1.7 1.2 30.0 36.8 6.8 22.6 min 0.5 0.4 47.1 54.8 7.7 16.4 light man 17.3 16.4 28.4 35.6 7.2 25.4 heavman 10.5 9.5 34.8 47.1 12.3 35.4 util 1.2 11 53.9 63.3 9.4 17.5 canst 6.8 6.7 16.8 24.8 8.1 48.3 trade rest 20.3 21.8 39.5 50.6 Ill 28.0 trans 5.8 5.4 28.1 32.8 4.8 17.0 comm 0.6 0.9 70.7 71.1 0.4 0.6 fin 8.7 8.5 61.6 66.1 4.6 7.4 pub 5.4 4.6 61.8 78.2 16.4 26.5 othserv 21.4 23.5 48.1 54.7 6.6 13.8 Total 100.0 100.0 40.0 ·48.9 8.8 22.0 Source: Author's calculations. 3. Resource booms, inequality, and poverty: The case of gas in Bolivia 3.1. Introduction In principle, countries richly endowed with natural resources, may it be fertile soils or mineral resources, should be able to prosper and overcome poverty faster than resource-poor countries. Yet, the experience of many resource-rich countries illustrates that this is not necessarily the case. Often, resource-rich countries go through boom and bust cycles that finally leave them poorer than resource-poor countries with similar initial conditions. Through a number of different channels resource wealth may negatively affect economic development." In many instances, tradable sectors become uncompetitive and shrink excessively because export revenues are consumed quickly rather than invested. This can be particularly harmful for economic development, as these sectors, especially manufacturing, are believed to exhibit important positive externalities, whereas resource sectors are often said to have an enclave character with few spillovers to the rest of the economy. In addition to hampering economic development, wealth and income in resource-dependent economies tend to be distributed very unequally, as resource rents typically benefit a small privileged group. As a more unequal distribution of income results in a lower rate of poverty reduction for a given growth rate, it is therefore unlikely that resource-rich countries achieve pro-poor growth, i.e. a growth pattern from which the poor benefit disproportionately, without deliberate interventions in favor of the poor. There is an extensive empirical literature on the "resource curse" and the channels through which it operates. Starting with Sachs and Warner (1997), many authors have confirmed that resource-rich countries actually grow more slowly than resource-poor countries using cross-country growth regressions. Yet, the impact of resource booms (and busts) on economic development and poverty as well as the transmission channels depend on both country and resource characteristics. Therefore, quite some country case studies have been undertaken, which have sharpened our understanding of how the resource curse works. Some of these studies, for example the collection in Auty (200 I), have focused on the impact of resource wealth on long-term development, whereas other studies have rather looked at the short to medium run economic impact of a resource boom or bust. The studies in Gelb et al. (1988) examine the economic impact of the oil windfalls in the late 1970s on a number of oil-exporting developing economies. These case studies focus on relative price effects, related sectoral shifts, in particular the performance of agriculture, and the fiscal response, especially public investment programs. The case studies included in Collier and Gunning (1999a and 1999b) centre on the savings response of public and private agents when faced with trade shocks. 45 See Auty (2001) and Lay and Omar Mahmoud (2004) for surveys of the literature on the resource curse. 52 BOLIVIA The present chapter examines one particular resource shock, namely the gas boom Bolivia experienced in the late 1990s and early 2000s. Following the studies in Gelb et al. (1988), we analyze the sectoral shifts and the fiscal response to the gas shock in the short to medium run. In contrast to these studies, our focus is on the poverty and distributional effects of the shock. We consider some of our findings to be of relevance to poor resource-rich countries with similar structural characteristics. The central question is whether the gas boom really bypasses large parts of the (poor) population in Bolivia, thereby leading to increasing inequalities in an already very unequal society. We examine the transmission channels through which the large resource inflows related to the gas boom, both initial foreign investment in the sector and the subsequent export earnings, as well as large public transfer programs (that may well be interpreted as a means of redistributing resource rents) affect the distribution of income. In doing so, we focus on general equilibrium effects and the corresponding labor market impacts, in particular on shifts in formal vs. informal employment and changes in relative factor prices. These transmission channels seem to be particularly relevant, as direct employment effects of the resource boom are virtually absent. To address these issues adequately, we propose a modeling framework that captures the structural features of the Bolivian economy and allows us to trace the poverty and distributional implications of the resource-boom-induced changes on the labor market. The framework therefore consists of a multi-sectoral computable general equilibrium (CGE) model that is combined with a microsimulation model. The CGE model allows us to construct counterfactual scenarios to disentangle the effects of the gas boom from other shocks that the Bolivian economy experienced at the same time and to trace the transmission channels at work at the macroeconomic level. Changes in important factor market aggregates from the CGE model, more specifically changes in relative factor prices and in the workforce composition in terms of formal and informal sector employment, are then passed on to a microsimulation model. The microsimulation model is based on an income generation model estimated on household survey data and produces a counterfactual income distribution given the CGE model results. The remainder of the chapter is structured as follows. The first part provides an overview of the scope and scale of the gas boom that began to shape the Bolivian economy in the late 1990s and a first broad assessment of its macroeconomic impact. It also motivates the counterfactual simulations of the second part. There, we first describe our methodological framework and then present our results. The final section concludes. 3.2. The gas boom and other resource shocks We consider Bolivia a particularly interesting case for the following reasons. It is the poorest country of South America with a long history of resource-induced booms and busts, and quite some observers have argued that the country's resource wealth is the root cause of its poverty (e.g. Auty and Evia 2001). The structural THE GAS BOOM AND OTHER RESOURCE SHOCKS 53 reforms of the 1980s and 1990s have been associated with some growth, but that growth has not done the job of lifting large parts of the population out of poverty in an economy with a highly unequal distribution of income, although poverty has been reduced somewhat in the course of the 1990s. Bolivia's economy has always been and still is highly dependent on natural resources. In the 1990s, hydrocarbons and minerals have typically accounted for roughly 50 percent of exports and the hydrocarbons sector contributes significantly to public revenues. In the second half of the 1990s, huge investments have been undertaken in the gas sector to explore and exploit Bolivia's vast gas reserves. Recent reserve additions have turned Bolivia into the second largest gas reserve holder in Latin America only after Venezuela. A new gas pipeline to Brazil went into operation in 1999 through which the bulk of Bolivia's gas exports run today, and gas exports are expected to increase further in the following years. The gas shock can be split into three important components: first, the huge investment into the gas sector between 1997 and 2003, most of which was foreign direct investment; second, the gas exports through the new pipeline to Brazil; and third, the government take of the gas rents. Figure 1 illustrates the magnitude of these components as a share of GDP in the 1990s. Investments in the gas sector reached about IO percent of GDP in the peak years 1998 and 1999. Unfortunately, more recent figures beyond 2001 are not available. Figure 3.1 also shows the phasing out of gas exports to Argentina in the course of the 1990s and the gas exports to Brazil, which started in 1999. By 2003, gas exports to Brazil accounted for almost 8 percent of GDP, as they had reached the contracted volume, which under the current contract is going to be roughly constant until 2019. In addition, gas exports to Argentina are likely to rise again, as new contracts were signed in 2005. Exports will thus mainly fluctuate due to fluctuations in gas prices, which are linked to a basket of international energy prices.46 Numbers on the government take from the gas sector are difficult to obtain. Here, we draw on a study by the Energy Sector Management Assistance Programme (ESMAP) (2002) of UNDP and World Bank, which places the government take in 2001 at more than 4 percent of GDP. According to the Ley de Hidrocarburos from 1996, the government take comprises royalties, a profit tax and a tax on remittances abroad, which is why the government take cannot be easily calculated from publicly available budgetary publications. 46 This of course only holds if both Brazil and Bolivia comply with the signed contract. 54 BOLIVIA Figure 3. 1: Major gas-related resource flows in percent of GDP, 1990-2003 --Gas exports -- Mneral and rretal exports - - • • lnvestrrent in gas exploration and explo~ation -Upstream governrrent take from the gas sector Source: Authors' calculations. Note: Data on GDP, exports, and investment in gas exploration and exploitation from INE. Upstream government take from the gas sector is from a study of the joint UNDP/World Bank Energy Sector Management Assistance Programme (ESMAP) (2002). The government take includes royalties, patents, and taxes on profits. Values for 2003 are preliminary. The gas boom was arguably the biggest but not the only significant external shock. The Bolivian economy was hit by an adverse terms of trade shock in the late 1990s, which comprised falling prices for exports of the four major metals zinc, gold, tin and silver. These shocks together with a further expansion of soybean production in the Bolivian lowlands had a major impact on the composition of Bolivian exports in the second half of the 1990s and the early 2000s (Figure 3.2). Hydrocarbon exports rose steeply from less than 10 percent to more than 30 percent of total exports over the period 1999-2004, as did soy and soy derivatives exports from 5 percent in 1990 to more than 20 percent in 2004. By contrast, exports of minerals and metals experienced quite some decline from well above 40 percent in the early 1990s to about 20 percent in 2004. On balance, the already high degree of export concentration increased even further. By 2004, the three product groups accounted for 80 percent of Bolivia's exports. THE CGE MODEL Table 3.2: Sectoral Deviations.from average wages Sectors Skilled labor Unskilled labor Formal Informal Formal Informal Food Pro 2.85 1.27 3.26 2.57 OthLiMan 0.65 0.38 1.31 0.94 Const 0.65 0.29 0.97 0.54 Trade 0.42 0.27 0.50 0.49 Trans 1.67 1.03 3.72 2.30 Hot Rest 0.59 0.33 115 0.70 Source: Authors' calculations. Note:A value above (below) one indicates higher-than-average (lower-than-average) wages. Computed from the SAM. 61 The factor income generated in the production process is distributed to four different household groups -poor and rich urban households as well as poor and rich rural households -according to fixed coefficients derived from the SAM. Moreover, households receive transfers in fixed proportions to government expenditures and remittances from abroad that are fixed in foreign currency. They use their gross income to pay taxes, to save and to consume. The allocation of consumption expenditures on different goods is modeled employing a Linear Expenditure System (LES). The sectoral allocation of government and the non-gas component of investment demand are both governed by fixed coefficients, whereas investment demand for exploration and pipeline construction is exogenously given. Despite the exogeneity of gas-related investment, the model is savings-driven, i.e., non-gas investment adjusts so as to bring about the necessary ex-post identity of savings and investment. The balance of payments equilibrium is determined by the equality of exogenous foreign savings to the value of the current account. With fixed world prices, the real exchange rate serves as the equilibrating variable. In the macro closure for the government, the budget deficit is allowed to adjust in order to achieve a predetermined level of real government expenditures at a fixed household income tax schedule. A simple recursive-dynamic framework allows us to implement the gas boom as a sequence of shocks (see below) and to trace over time the changes in economic structure caused by the boom. There are four elements driving model dynamics -exogenous labor growth, investment-driven capital accumulation, exogenous growth of natural resources in the gas sector, and exogenous productivity growth. Several other exogenous variables require updating to arrive at a realistic baseline scenario. Government expenditures and transfers, for example, are assumed to grow at the same rate as GDP. Finally, the CGE model is linked to the microsimulation model. We assume that the link is sequential, i.e., the CGE model is solved first and certain target values are passed to the microsimulation that is "forced" to reproduce the aggregate 62 BOLIVIA changes in these targets. In line with our focus on skilled vs. unskilled labor and formal vs. informal employment, the following link variables are used: ( 1) the share of unskilled workers in the formal sector, (2) share of skilled workers in the formal sector, (3) mean wages for skilled workers, (4) mean wages for unskilled workers, and (5) mean informal profits." While the first four link variables have a straightforward interpretation, the fifth is based on the concept of mixed income received by self-employed workers. Accordingly, informal profits are calculated as the sum of skilled and unskilled labor income as well as informal capital income. Before we discuss the specification of the microsimulation model, it should be stressed that the CGE and the microsimulation model should not be seen as a consistent macro-micro modelling framework. Rather, the idea of combining these two types of simulation models in a sequential fashion is to get "the best of the two modeling worlds". The CGE model is a useful tool to examine the transmission channels of the resource boom. It provides a consistent modeling framework based on an empirical representation of the Bolivian economy and yields a numerical counterfactual approximation of the shocks' labor market impacts. Yet, in order to assess the ultimate poverty and distributional consequences of the shocks, the microsimulation is a much more suitable tool. It takes into account household heterogeneity in terms of factor endowments at a much more detailed level and models occupational choices and corresponding earnings changes at the individual level. Sequentially combining these models typically implies the imposition of a number of ad-hoc assumptions that may not be satisfying from a theoretical perspective. 3.3.2. The microsimulation model The microsimulation model is based on an income generation model that is estimated on household survey data." The following estimations are based on all individuals employed outside traditional agriculture, as our focus is on changes in 51 Although formal profits account for an important share in value added, they are not passed to the microsimulation for two reasons. First, most formal profits are retained and invested. Second, capital income is likely to be measured very poorly in household surveys. As formal profits increase considerably during the gas boom, we may systematically ignore an inequality-increasing factor. 52 We use the Encuesta Continua de Hogares from 2001. The survey comprises 24 996 individuals from 5 797 households. Due to a number of missing income observations, incomes were imputed based on income equations estimated separately by labor market segments (smallholder, worker in traditional agriculture, agricultural employer, worker in modem agriculture, informal self-employed/employer, informal worker, formal employer, and formal worker). This imputation is also necessary for using income data for poverty and distributional analysis. THE MICROSIMULATION MODEL 63 formal vs. informal employment. In contrast to the CGE model, the microsimulation hence assumes that smallholders remain in their occupation." The two basic components of this income generation model are a model of occupational choices that represents the "choice" between formal and informal employment" as well as earnings functions that correspond to the respective sector of employment. If individuals happen to be in ( or switch to) the formal sector they are assumed to earn a wage, whereas individuals in the informal sector are assumed to be (or become) part of a household enterprise and contribute to the profits earned by this enterprise. Table 3.3 provides an overview of the equations of the income generation model, the econometric models, the sub-samples and lists of the explanatory variables. We limit the discussion of the specification to innovative and interesting features. As estimation results correspond to expectations, we do not further comment them here, but report the detailed results in App. Table 3.1 and App. Table 3.2. The choice between informal and formal activities is estimated separately for household heads, spouses, and other household members using a logit model. The equations of the choice model are interrelated through the head's wage (and hence her choice) entering the occupational choice model of spouses and other household members. We hence assume a sequential choice with the household head deciding first. In addition to the formal-informal segmentation, we assume segmentation according to skill levels, differentiating between unskilled and skilled labor (defined in terms of years of schooling). We therefore estimate separate wage equations for skilled and the unskilled labor employed in the formal sector using OLS. The set of explanatory variables (reported in Table 3.3) is standard. The individuals in the informal sector are assumed to pool resources and work effort in a household enterprise, for which we estimate a profit function. The numbers of informal unskilled as well as informal skilled individuals enter as separate explanatory variables. As the number of household members working in the enterprise is very likely to depend on the prospective profits to be earned in informal activities, we are likely to have an endogeneity problem here. We therefore instrument the number of unskilled as well as skilled household enterprise members. As instruments we use the total number of (informal and formal) household members (unskilled or skilled, respectively, that we additionally 53 The income generation model is based on a model first proposed by Alatas and Bourguignon (2005) to decompose inequality and poverty changes between two household surveys. See Bourguignon, Ferreira, and Lustig (2005a) for a selection of country studies using this type of decomposition technique. It has also been used in a macro-micro simulation framework by Robilliard et al. (2002) and Bussolo and Lay (2005). 54 Employment is assumed to be informal if the individual is self-employed/nonremunerated household member and/or works in an enterprise with less than 5 employees. 64 BOLIVIA differentiate by gender and age) interacted with the share of formal employment of the province of residence. Table 3.3: Overview of the income generation model Explained variable and estimated Model Sub- Explanatory variables eauation samvle 1 Being in formal or informal sector Logit Heads Education (squared), P(formal = 11 X') = g(c' + X'a' experience, female, share of formal workers in province of residence 2 Being in formal or informal sector Logit Spouses Education (squared), P(formal = 11 X') = g(c' + X'a') experience, female, indigenous, geographical dummies, number of children under I 0 ( interacted with female), share of formal workers in province of residence, log wage of household head (if head formal) 3 Being in formal or informal sector Logit Others Education (squared), P(formal = 1jX 0) = g(c0 + x0 a0) experience, female, widow, number of children under 10 ( interacted with female), share of formal workers in province of residence, log wage of household head (if head formal) 4 Unskilled wage Linear Unskilled Education (squared), In wus = cus + xus pus + uwus formal experience (squared), sector female, geographical 5 Skilled wage Linear Skilled dummies lnw'' =c" +X''P'' +uw" formal sector 6 Profits of household enterprise Linear Informal Average education of lnp=c'+X'P'+up sector members (squared), enterprises Average experience of (max. one members (squared), number per of female members, household) geographical dummies, number of members These estimated relationships form the basis of the microsimulation. The microsimulation is shocked using changes in the five link variables that are passed on from the CGE analysis. To gain an understanding of how the microsimulation works it is useful to think of the set of equations summarized in Table 3.3 as a system of equations. Let employment n be the sum of n"' and n'' , the number of unskilled and skilled labor, respectively. THE MICROSIMULATION MODEL 65 fs"' denotes the formal share of employment among the unskilled. The number of formal unskilled nr is given by the sum of heads, spouses, and other household members that derive a higher ''utility" from being employed in the formal than from being employed in the informal sector where ( ind~ ,(c' + X'a' +uhl > uhO) is an indicator function assuming 1 if the condition in brackets is fulfilled and 0 otherwise. uhO, uhl, usO, usl, uoO, uol are residuals for heads, spouses, and others, respectively, that cannot be observed in latent variable models and are hence drawn consistent with the observed choice." nf" fs~=- n~ "f.indwJc' + X'a' + uhl > uhO) + "f.ind •. ,(c' + X'a' + usl >usl) ~ h + "f.indw ,(c" + X'a' + uol > uoO) = nf~ We have the same set of equations for skilled labor (2). if ,k fs'' =!'__ n'' "f.ind,, ,(c' +X'a' +uhl > uhO)+ "f.ind,, ,(c' +X'a' +usl > usl) M_h sic s + "f.ind,to(c' + X'a' +uol > uoO) = nf'' sk _o (1) (2) The mean unskilled wage is the wage of all unskilled workers in the formal sector divided by the number of unskilled formal workers. "f.indw ,( )w:_, + "f.indw_,(.)ws:, + "f.indw J )w:, mean(ww) =-w~•------~-----~----- nfw (3) where WW = exp(cw + r fr+ uww) is the individual unskilled wage with uww' the observed residual. The same equation holds for the skilled "f.ind.,.( )w;;_, + "f.ind,, ,()ws:,_, + "f.ind,, J .. )w;;, mean(w'') = "-' ''-' " ' nf'' (4) where w'' = exp( c'' + X '' fl" + uw'') . Whereas wages are summed over individuals, total profits from informal activities are summed over households (hh) 55 Residuals are drawn conditional on the observed choice, as suggested by Bourguignon, Fournier, and Gurgand (1998). We randomly draw two residuals from a Gumbel distribution (type I extreme value, variance n2/6), one for each alternative. This is equivalent to drawing one logistically distributed residual, as the difference of two Gumbel-distributed random variables is logistically distributed (with variance n2 /3). See Train (2003) for details. 66 LPhh mean(p) = -~'~' -- n -nf"' -nf'' BOLIVIA (5) where the profit phh in household hh will only be greater than O if at least one member is employed in the informal sector. p,,,, = exp(cp + X' fr+ upX-l)[(ind,( )-!)+(ind,()-!)+~ (indJ )-!)] (6) Remember also that the number of (skilled and unskilled) household enterprise members enters the profit function. (7) The above equations describe the initial distribution of labor income (with the exception of traditional agriculture). As mentioned above, the microsimulation is "forced" to reproduce the changes in the aggregates given by the CGE model. This is achieved by varying the constants in the above system of equations such that the household income generation model just reproduces the target values. Increasing e.g. the constant c" leads some heads to switch from informal to formal activities. When individuals switch from informal to formal activities, they are assigned a simulated wage residual, as we do not have an observed unexplained wage uw for her. The same holds if an individual becomes the first household member active in the informal sector (not if she joins an existing enterprise).56 The above income generation model represents only part of the income households receive. It focuses on income generation in non-agricultural sectors and translates changes in labor incomes and the formal-informal composition of employment into poverty and distributional changes. Other income sources include income from traditional agricultural activities (including home-consumed production) and all kinds of transfer incomes from remittances to public transfers. Agricultural incomes of smallholders from the household survey will be scaled up using a weighted real factor price index for traditional agriculture (real factor 56 To adjust to the targets, we need to vary the constants in the occupational choice equations differently for unskilled and skilled labour, respectively. This implies that we fix the (absolute) differences between the occupational choice model constants for heads, spouses, and others for both unskilled and skilled labour. If one thinks of this problem in terms of solving the system of equations to reach the new target values, this means that the occupational constants (i.e. for heads, spouses, and others) for each skilled segment are augmented by the same amount. We now have 5 variables (2 "vectors" of constants of the choice models and 3 wage/profit constants) and 5 equations. The system is solved using a Newton-Raphson algorithm. THE MICROSIMULATION MODEL 67 prices for land and unskilled labor are weighted with the share in value added of the sector). Changes in public transfers and changes in other income sources will not be taken into account in our counterfactual microsimulation of the gas boom. Public transfers will also grow or decrease in accordance with the CGE model. It should however be borne in mind that these income sources and changes therein do not affect individual behavior (at least in our model). 3.4. Results 3.4.1. Stylized simulations for link variables While in the gas shock simulations presented below all link variables change simultaneously, the importance of specific labor market link variables for poverty and inequality can be seen more clearly if they are considered in isolation. Consequently, we performed stylized simulations for each link variable, the results of which are reported in Table 3.4. Since the mechanics of the income generation model can be better understood when looking at the urban population only, we restrict the results to this group. It turns out that, at constant factor prices, a lower formal employment share (by 5 percentage points) leads to a significant rise in urban poverty. The effect is markedly stronger for skilled workers. Yet, urban mean per capita income declines by 2.93 percent when informality increases among unskilled labor, much more strongly than the 1. 76 percent decline recorded for skilled labor. This seemingly paradox result can be explained by looking at the distributional shifts. Whereas increasing informality is equalizing for unskilled labor, as indicated by the decrease in the Theil-Index, it is inequality increasing for skilled labor. This rise in inequality reinforces the negative poverty effect of declining incomes for skilled workers. A more detailed analysis of the distributional impact that looks at the entire distribution rather than aggregate indicators reveals striking results (see App. Figure 3.2 and App. Figure 3.3): for both unskilled and skilled labor, the very poor are affected most by increasing informality. These results can be rationalized by looking at who moves into informality as well as the size of the income loss for movers relative to both their initial income and the income losses incurred by other individuals. The size of the income loss depends on individual characteristics (as the returns to these characteristics differ between formal and informal activities) and on whether an individual joins an already existing household enterprise or establishes a new one. From the estimation that underlies the microsimulation we know that less educated younger (and hence poorer) individuals tend to move into informality first. As regards the size of the income losses, the estimation results for wages and profit functions indicate that the income loss of moving into informality is higher for more educated individuals, at least in absolute terms, when they move into an existing household enterprise. Since the constant in the wage function for skilled labor is lower than the constant in the informal profit function, it may also happen that establishing an informal enterprise increases earnings for a skilled individual conditional of course on other individual characteristics. For an 68 BOLIVIA unskilled individual, by contrast, moving into informality will always imply an income loss, which explains the overall decrease of per capita incomes. It is the combination of these effects that explains the poverty and distributional 57 outcomes. That overall income losses for unskilled labor are higher than for skilled labor can hence be explained by two factors. First, unskilled labor always loses when moving into informality. In addition, they are more likely to move into an existing enterprise instead of establishing a new one. For both the unskilled and skilled workforce, the move of less educated (and hence poor) individuals explains the strong negative income growth in the lower parts of the income distribution, as the incurred losses are relatively large compared to initial income. A similar reasoning explains why unskilled workers from middle-income classes do not suffer significant income losses. Fewer formal workers lose their jobs and if they do their relative income loss is not too high. Yet, higher educational endowments are associated with higher income losses, which is why the growth incidence curve is downward sloping for higher incomes. This effect seems to overcompensate the general effect that the relative importance of income losses decreases with higher incomes and explains the overall "positive" distributional impact among unskilled workers. The negative distributional shift when informality increases among skilled workers is not only due to the strong losses of the poor. The very rich do not seem to be affected at all or even experience slightly positive income gains. This is mainly due to the very low incidence of informal work among high-skilled workers. In addition, these individuals are likely to set up a new household enterprise instead of moving into an existing one. Table 3.4: Marginal effects of changes in link variables on urban inequality and poverty (point differences) Scenario PO Pl Theil Urban Initial 50.8 23.5 63.3 5 % point decline informal share unskilled 0.7 0.5 -1.2 5% point decline informal share skilled 1.7 0.9 1.2 IO % increase in unskilled wages -0.9 -0.7 -I.I IO % increase in skilled wages -0.7 -0.4 2.4 10 % increase in informal profits -1.6 -I.I -1.3 Source: Authors' calculations. The poverty and distributional effects of increases in different types of labor incomes are in line with expectations. An increase in unskilled wages decreases poverty and improves the distribution of income. Maybe somewhat less obvious is the relatively strong reduction in the headcount index associated with an increase 57 Furthermore, differences in the variance of the simulated residual also play a role for the distributional impact, but this effect should not be too large. THE RES UL TS 69 in skilled wages. Yet, the impact on Pl is much less pronounced, indicating that the "skilled poor" are close to the poverty line, and the overall income distribution worsens quite substantially compared to the distributional improvements that can be reached e.g. by an increase in unskilled wages. Increases in informal profits tum out to reduce poverty most effectively. 3.4.2. Gas shock simulations We now tum to a counterfactual analysis of the gas shock. We use the CGE model in combination with the microsimulation model to evaluate the distributional and poverty impacts of a gas boom over the period 1997-2005 by simulating two scenarios. The first scenario combines a positive temporary demand (investment) shock with a delayed positive supply shock. The size of the shocks roughly corresponds to what Bolivia actually experienced (see 3.2). The demand shock consists of a doubling of real investment demand in exploration and pipeline construction over the period 1998-2001, which is financed by foreign capital inflows. After 2001, foreign direct investment in the oil and gas sector is assumed to fall back to the pre-shock level. The upfront investment is assumed to induce a positive supply shock from 2004 onwards, which is modeled by quadrupling oil and gas reserves, i.e. the specific factor used in oil and gas production. The second scenario combines the first scenario with a progressive reallocation of parts of public revenues to households, as observed in the household surveys. More specifically, it is assumed that real government transfers to poor rural, poor urban, and all rich households increase by 100%, 50% and 10% (and again by 50%, 25%, and 5%), respectively, from year 2000 (from year 2004) onwards. This scenario takes into account the mainly transfer-induced rise in public deficits that occurred before the government received higher revenues from oil and gas extraction. Benchmark: In the benchmark simulation, real GDP growth is exogenously fixed and assumed to increase by four percent annually over the period 1997-2005, while the growth rate of labor productivity is calibrated for each year so as to keep growth constant over time. This implies steadily increasing labor productivity growth over the whole simulation period. Yet, the assumption that productivity growth is only labor-augmenting is not appropriate for Bolivia. We therefore assumed a balanced growth path along which capital per worker, measured in efficiency units, remains constant over time, and calibrated the growth rate of capital productivity, which keeps the capital-labor ratio constant. At given labor growth rates and given capital accumulation rates, this implies increasing capital productivity at a decreasing rate over the total simulation period.58 The most striking distributional result of the benchmark simulation is the lack of significant progress with regard to poverty reduction despite relatively high 58 In the policy simulations discussed below, these calibrated productivity parameters are kept constant, and the growth rate of GDP and the capital-labour ratio are allowed to vary endogenously. 70 BOLIVIA growth rates of 4 percent. This is largely the result of rural-urban migration, which increases the supply of urban unskilled labor, thereby depressing wages and increasing the urban wage differential between skilled and unskilled workers (App. Table 3.3). As a result, urban inequality rises while poverty stays constant (Table 3.6). At the same time, migration reduces the supply of unskilled labor in rural areas and thereby increases wages. Together with increasing land rental rates, higher wages for rural unskilled workers raise the mixed income earned in agriculture (agricultural profits in App. Table 3.3) and slightly reduce poverty and income inequality in rural areas (Table 3.6). At the national level, both the incidence of poverty (PO) and the poverty gap (Pl) hardly change over the period under consideration. Gas Shock: Given that our CGE model does not allow for multiplier effects, it is not surprising that the foreign direct investment in gas exploration and the construction of pipelines is shown to have only a minor impact on aggregate economic activity. In the short run, the demand shock causes a steep rise in the price level, which even lowers real GDP, while over the medium run production capacity is slightly higher than in the base run. The model can much more reliably capture the structural effects of the demand shock, which tum out to be sizeable. The investment boom induces a massive expansion of exploration, which clearly dominates the much less pronounced expansion of formal (pipeline) construction. As a result, the factors attracted into gas-related activities are predominantly formal capital and to a lesser extent skilled labor, exerting upward pressure on formal profits and skilled wages. Two secondary effects refine this picture. First, booming gas-related investment crowds out other investment goods, which are predominantly produced by heavy manufacturing as well as the formal and informal construction sectors. Since these sectors make intensive use of unskilled labor, the respective wage is driven down. The contraction of informal construction lowers the informalization of the economy. Second, the rise in the price level, which at a fixed exchange rate implies a real appreciation of about six percent in the medium run, leads to a contraction of trade-oriented sectors such as heavy manufacturing, mining, and modem agriculture, whereas sectors with low trade shares and non-traded sectors tend to expand. Among the gaining sectors with low trade-orientation, informal activities figure prominently. They realize higher returns on capital and demand additional labor, in particular unskilled workers, thus raising the informal share of unskilled labor. As shown in Table 3.5, the investment boom is on balance associated with higher (lower) wages for skilled (unskilled) labor, higher profits in the informal urban sectors, and an increasing informalization of urban production activities, as indicated by lower formal unskilled employment shares compared to the base run. The net result of the in formalization and the rise in the wage gap on the one hand, and increasing informal profits on the other, is that aggregate indicators of urban poverty and inequality hardly change (Table 3.6). In addition, lower wages for unskilled workers reduce incentives to migrate from traditional agriculture. APPENDICES App. Figure 3.2: Growth incidence curve, 5 point decline informal employment for unskilled labor --- Gro wth in c id ence cu rve o Mean of growth rates 0 .5 0 10 20 30 Source: Authors' calculations. 40 ll 50 Perc entil es 60 Gro wth ra te in m ean 70 80 90 1 00 77 78 BOLIVIA App. Figure 3.3: Growth incidence curve, 5 point decline informal employment for skilled labor --- Growl h i ncidence cu rve o Mean of grow th rates 0 -5 0 10 20 30 Source: Authors' calculations. 40 50 Percentiles 60 Growth ra te in mean 70 80 90 JOO APPENDICES App. Figure 3.4: Growth incidence curve, JO percent increase in transfers to all households --- Growth incidence curve Growth rate in mean o Mean of growth rates . . . ' 0 ··----- - ···· . . . .. -5 0 10 20 30 40 Source: Authors' calculations. 50 Percentiles 60 70 80 90 100 79 80 BOLIVIA 3.6.2. Additional tables App. Table 3.1: Estimation results for the /ogit choice models, choices are dichotomous variables with formal= 1, informal= 1 dep var Explanatory var Head's choice Spouse's choice Other's choice Education -0.19 -0.323 -0.219 (5.30)** (5.19)** (3.17)** Education squared 0.018 0.026 0.023 (9.30)** (7.85)** (6.06)** Experience -0.012 -0.011 0.005 (3.81)** -1.36 -0.57 Female dummy -1.089 -0.857 -0.756 (11.16)** (2.70)** (3.44)** Formal employment 4.624 3.24 4.008 share in province ( 10.38)** (3.70)** (4.51)** Indigenous dummy -0.629 (3.07)** Beni dummy 0.486 -1.75 Pando dummy 0.734 (2.03)* No of children under 0.448 0.107 10 (2.53)* -1.49 Interaction female*no -0.391 -0.221 of children (2.06)* (2.07)* Head's formal sector 0.193 0.131 wage (8.05)** (5.51)** Widow dummy 0.579 -1.07 Constant -1.633 -1.686 -2.929 (5.45)** (2. 75)** (5.03)** Observations 3385 1140 1102 Pseudo R2 0.1716 0.2693 0.1943 Robust z statistics in parentheses * significant at 5%; * * significant at I% Source: Authors' calculations. APPENDICES 81 App. Table 3.2: Estimation results for the wage and profit equations dep var Formal unskilled Explanatory var wages Formal skilled wages Informal profits Education 0.084 0.152 (9.16)** (13.53)** Experience 0.051 0.057 (11.20)** (7.96)** Experience -0.001 -0.001 squared (9.54)** (4.72)** Female dummy -0.627 -0.373 (11.37)** (8.71)** Potosi dummy -0.182 -0. 17 -0.452 (2. 18)* -1.78 (3.34)** Tarija dummy 0.307 0.15 0.137 (3.62)** -1.59 -1.34 Santa Cruz 0.193 0.246 0.065 (2.77)** (3.63)** -I.OJ Beni dummy 0.374 0.175 0. 171 (4 . 81)** (2.30)* -1.73 Pando dummy 0 0.258 0.792 0 (2.49)* (4.30)** Avg. education 0.059 (8.64)** Avg. experience 0.038 (8.44)** Avg. experience -0.001 squared (8.20)** Number of -0.602 females (13.87)** Number of 1.332 enterprise (20.07)** Constant 5.092 4.258 4.947 (45.57)** (25.40)** (48.74)** Observations 1357 1407 1905 R-squared 0.26 0.27 0. 31 Robust t statistics in parentheses * significant at 5%; ** significant at I% Source: Authors' calculations. 82 BOLIVIA App. Table 3.3: Business as Usual (BaU) scenario Variable II 1997 2001 2005 GDP growth rate 4.00 4.00 Price level 1.00 0.97 0.94 Growth rate of capital productivity 1.45 1.15 Growth rate of labor productivity 1.56 1.61 Labor demand Skilled labor 1.00 I. 11 1.23 Rural unskilled labor 1.00 1.06 1.13 Urban unskilled labor 1.00 1.16 1.34 Migration_ 1.00 0.87 0.78 C:apital demand F onnal capital 1.00 I.II 1.25 lnfonn11l Cllj)Ital 1.00 1.12 1.27 - - --···------- · Link variables Agricultural profits 1.00 1.06 1.13 Informal profits 1.00 1.00 1.00 Unskilled wage 1.00 0.97 0.97 Skilled wage 1.00 1.03 1.07 Formal share unskilled 0.38 0.38 0.39 Formal share skilled 0.72 0.71 0.71 Source: Authors' calculations. 4. Structural change and poverty reduction in Brazil: The impact of the Doha Round 4.1. Introduction Trade liberalization, in particular the liberalization of trade in agricultural products, is considered by many observers as one of the key components of a strategy to reduce poverty worldwide. In their review of the relationship between trade liberalization and poverty, Winters, McCulloch and McKay (2004), conclude that trade liberalization "may be one of the most cost-effective anti-poverty policies available to governments" although it may not be the most powerful. Yet, the complex relationship between trade, growth, distribution and poverty does not allow for a simple conclusion with regard to the impact of trade liberalization on poverty, as the poverty outcomes of trade liberalization may vary substantially from case to case. Despite the complexity of this relationship the evidence so far provides enough insights to predict at least the largest impacts (Winters, McCulloch, and McKay 2004). One of these insights is the importance of the impact of trade liberalization on wages and employment, which depends on the structure and the functioning of the labor market. For the Brazilian case, this chapter intends to evaluate the poverty effects of trade liberalization in the medium run. In doing so, we focus on the labor market, as we consider this transmission channel to be of overriding importance in this time horizon. This implies to assess the poverty impact of a Doha Round (and a Full Liberalization) 61 counterfactual scenario against a scenario that incorporates some of the main features of medium run structural change. We will thus examine whether the effects of trade liberalization, in particular on poverty and the distribution of income, are still prominent in the medium run. Recent research has demonstrated that growth can differ tremendously in its power to reduce poverty both across countries and over time.62 In particular, in high-inequality countries such as Brazil, an apparently slight worsening of the income distribution can imply that growth has very little impact on poverty. It is hence not only important by how much trade liberalization raises incomes, but how it affects the pattern of income growth. The driving forces of this pattern work through the labor market. Among them are changes in relative factor prices, but also changes in endowments play an important role in the medium run, as for 61 The global trade liberalization scenarios will not be discussed in this chapter. The trade shocks that the Brazilian economy faces were developed by a team examining the economic effects of likely Doha negotiation outcomes using a (GT AP-based) global trade model. We will comment only on the changes in tariffs and international prices, as they affect the Brazilian economy. Detailed information on the scenarios and the global trade model used to investigate their economic effects are available from the authors on request, as the project output has not been published yet. 62 See Ravallion (2001a), Ravallion and Datt (1999), World Bank (2005b), Grimm, Klasen, and McKay (2006) or Kappel, Lay and Steiner (2005). 84 BRAZIL example the workforce advances its skills. In addition, sectoral employment change can contribute significantly to poverty reduction. Such change in the structure of employment can have very large effects on poverty, as it may enable people to escape poverty traps. There is quite some evidence on the existence of such poverty traps that can arise if occupational or technology choices are discrete and there exist fixed or sunk costs to choosing a higher return occupation or technology (Barrett 2004). Moving out of agriculture where poverty rates are often much higher than in other sectors is one example for such choices. This latter issue is of particular interest in the Brazilian context, as there has been a massive reduction in agricultural employment in recent years, which we consider to be likely to continue. This reduction in agricultural employment may have contributed to poverty reduction, as poverty among agricultural households is considerably higher than among non-agricultural households. Trade liberalization that one would expect to favor agriculture in Brazil may thus work against the "natural" forces of structural change with an adverse impact on poverty reduction. However, trade liberalization may also relieve some of the pressure put on nonagricultural incomes by the reduction in agricultural employment and have some direct poverty reducing effect through raising agricultural incomes. This last example illustrates the necessity of quantifying each of these transmission channels to evaluate the overall poverty and distributional impact of trade reform. The features incorporated in our simulation exercise are therefore changes in different sources of factor income, changes in the sectoral composition of the workforce, and educational upgrading of the workforce. These changes are driven by changing consumption patterns, purely exogenous factors (at least in our model), such as differentials in productivity growth rates across sectors and differential growth rates for different types of labor, and, finally, the trade shocks. The methodology used here combines a dynamic computable general equilibrium model with a microsimulation model for Brazil. For a time horizon of 15 years, a business as usual scenario and two trade counterfactuals are developed in the COE model and crucial aggregate results on relative factor prices and resource movements from agricultural to non-agricultural sectors are linked to a microsimulation. This macro-micro model enables us to analyze the long-term poverty and distributional impact of different growth patterns. The chapter is structured as follows. We first provide some background information on the Brazilian case and motivate our approach. Then, we describe the macro and micro modules of the model. The results of our simulations are reported and commented in the following section. The last section summarizes and concludes. 4.2. Background and motivation The main objective of this chapter is to assess whether trade reform favors the Brazilian poor. It is therefore important to know who the poor are, where they live, and especially how they earn their living. In addition, it should prove helpful to identify economic trends that have been particularly important for the poor. BACKGROUND AND MOTIVATION 85 Brazil's per capita income has virtually stagnated for the past 25 years and the very unequal distribution of income has remained more or less unchanged. Accordingly, poverty in Brazil has been roughly constant over the past 25 years (Bourguignon, Ferreira, and Lustig 2005b; Verner 2004). In light of the substantial structural changes that have occurred in this period, especially increasing urbanization, a massive decline in agricultural employment, increasing unemployment, an important educational expansion and demographic changes, this appears "paradoxical", as Bourguignon et al. (2005b) put it. Yet, microimulation exercises by Ferreira and Paes de Barros (2005) show that each of the features of structural change affects poverty and inequality, but they tend to cancel out each other.6' Poverty in Brazil varies considerably between regions, rural and urban areas, and city sizes with poverty being particularly high in rural areas, small and medium towns and the metropolitan peripheries of the North and the Northeast (Ferreira, Lanjouw, and Neri 2001). In 1996, the North and the Northeast accounted for 55 percent of the poor and for 34 percent of the Brazilian population. At the national level, about 20 percent of the population lived in rural areas contributing 35 percent to total poverty.64 The high poverty rates in rural areas, particularly in the Northeast, are related to this region's predominance of employment in agriculture. The Northeast has the highest share of agriculture in employment with 34 percent in 2001 compared to only 11.5 percent in the Southeast.65 According to Ferreira, Lanjouw, and Neri (2001), 20 percent of all households had a household head employed in agriculture and these households contributed 34 percent to overall poverty in 1996. Yet, not only do poverty levels differ across regions, rural and urban areas, and activities, but also do the changes in poverty. Vemer's (2004) PNAD-based66 figures suggest that the poverty headcount in the Northeast declined from almost 60 percent in 1990 to 42.3 percent in 2001, whereas poverty in Brazil's most populous state Sao Paulo rose slightly from 8.6 to 9.4 percent during the same period. For urban areas, Ferreira and Paes de Barros (2005) show that extreme poverty increased between 1976 and 1996. According to Paes de Barros (2004) however, the poverty incidence in rural areas in general and among households engaged in agricultural activities, in particular, declined from levels of about 60 percent to around 50 percent between 1992 and 2001. One important factor for understanding these developments are the structural changes in Brazilian agriculture in the 1980s and 1990s. These changes have 63 Note that their analysis compares the income distribution of 1976 with the 1996 distribution. For detailed results see Ferreira and Paes de Barros (2005). 64 Poverty is measured by the headcount ratio. The poverty figures in this paragraph are taken from Ferreira, Lanjouw, and Neri (2001). 65 The figures on agricultural employment are own calculations based on the PNAD 1997 and the PNAD 2001. 66 The PNAD (Pesquisa Nacional por Amostra de Domicilios) is a regularly conducted representative household survey. 86 BRAZIL certainly had a profound impact on rural livelihoods and poverty in Brazil, but they may also have affected urban areas for example by putting pressure on urban labor markets through increased migration. With the exception of Paes de Barros (2004), research efforts in this direction however have focused on agricultural performance rather than on how this performance affects people's livelihoods. In their assessment of the impact of sector-specific as well as economy-wide reforms on Brazilian agriculture, Helfand and Rezende (2004) conclude that agriculture became one of the most dynamic sectors in the Brazilian economy. Between 1980 and 1998 real GDP grew by about 40 percent and real agricultural output by about 70 percent. In many sub-sectors, yields increased significantly and more harvested area was dedicated to exportables, in particular soybeans and sugarcane. Agriculture benefited from a conducive macroeconomic environment and trade reforms that led to less industrial protection and the elimination of taxes and quantitative restrictions on agricultural exports. In addition, specific agricultural reforms -in particular a reform of agricultural credit and price support policies; an agrarian reform program, including a land reform; and, finally, the deregulation of domestic markets for agricultural goods -were important drivers of the observed agricultural performance.61 The increase in agricultural productivity however was accompanied by a massive lay-off of hired labor and by important changes in the size distribution of farms. According to the agricultural census from 1996, the number of small farms declined dramatically and agricultural employment shrank by 23 percent between 1986 and 1996, although these figures should be taken with caution (Helfand and Rezende 2004). Non-agricultural activities appear to have compensated for the loss in agricultural employment in rural areas, but unemployment rates in urban areas with a previously important share of agricultural labor have risen in that period (Dias and Amaral 2002). Our analysis based on the 1997 and 2001 household surveys (PNAD) suggests that this decline in agricultural employment has continued after 1996. In 200 I, agriculture accounted for 20.6 percent of employment in Brazil down from 24.2 percent in 1997. Unemployment in rural areas has stayed constant at about 2.5 percent during this period, whereas urban unemployment has risen from 9.44 to 10.6 percent, an increase that may be related to the decline in agricultural employment. 68 Less agricultural employment opportunities may also be one of the reasons for further urbanization in Brazil although it is difficult to establish this link empirically, as we explain in more detail later. The rural population declined quite dramatically from 24.41 percent in 1991 to 21.64 percent in 1996 (IBGE 1997) and 16 percent in 2001 (PNAD 2001, authors' own calculations). The trends in rural poverty 67 See Helfand and Rezende (2004) and Dias and Amaral (2002) for details. 68 Data from employment histories in the PNAD reveal that in both 1997 and 2001 about 6 percent of those who became unemployed in the last year were employed in agricultural sectors before. Taking into account that approx. 20 percent of the workforce are employed in agriculture, this figure is rather low and may be taken as a sign that the rise in urban unemployment is not causally linked to the decline in agricultural employment. THE MACRO MODEL "_,,I_ Ld ( w J LJ fl ,,/ ieg l+T;,1 AWAGEg,l = '° d LJLiJ ,eg 93 (3) The variable A WAGE is the average wage in the respective segments and is given by equation (3). The average wage is calculated based on the net-of-tax wage rate, the rate which matters to the worker deciding to migrate or not. Labor market equilibrium conditions are based on two separate labor markets rather than the integrated market of the skilled workers. Equation (4) determines the equilibrium wage rate by segment-i.e. agriculture and non-agriculture. It sets the aggregate segment labor supply equal to the demand for labor in the same segment, i.e. it determines the variable iv' which is now indexed by both segment index as well as labor type. The model allows for inter-sectoral wage differentials, but these are exogenous in the standard model. Equation (5) evaluates the relative wages with respect to the segment-specific equilibrium wage. L' I=" Ld1 g, ~ ,, (4) ieg (5) The remaining loose end is the definition of labor supply and this is given by equations (6) and (7). It is assumed that labor supply net of migration is given in any given period. In the dynamic scenario, labor supply in each segment grows at the same exogenous rate, l and migration is subtracted from this amount in the agricultural segment, equation (6), and is added to labor supply in the nonagricultural segment, equation (7). Equation (8) determines the total economywide labor supply for each labor type. (6) (7) L;o,,l = L~ri,/ + L~agri,/ (8) Model Closures. The equilibrium condition on the balance of payments is combined with other closure conditions so that the model can be solved for each period. Firstly consider the government budget. Its surplus is fixed and the 94 BRAZIL household income tax schedule shifts in order to achieve the predetermined net government position. Secondly, investment must equal savings, which originate from households, corporations, government and rest of the world. Aggregate investment is set equal to aggregate savings, while aggregate government expenditures are exogenously fixed. Growth equations. Sectoral shifts among agriculture and non-agriculture and human capital upgrading are two of the main features that have characterized recent growth processes in Brazil, and indeed in most developing nations. To capture these features in a transparent and simple dynamic framework, productivity growth calibration is different for the agriculture and non-agriculture sectors. Equation (9) defines the growth rate of GDP at market price and equation (10) is a formula expressing a balanced growth, where capital to labor ratio in efficiency units is constant. RGDPMP =(I+ gY) RGDPMP_ 1 (9) ( 10) Equation ( 11) and (12) determine the growth rates of labor and capital productivity for the non-agricultural sectors (subscript nag). The growth rates have two comrments, a uniform factor applied in all sectors to all types of labor and capital, y and y", and a sector- and factor-specific factor, x' and x*. In defining a baseline, the growth rate of GDP is exogenous, as well as the capital to labor ratio. In this case, equation (9) is used to calibrate they' parameter and equation (10) calculates the common growth rate for capital productivity, y". In policy simulations, y1 and y" are given, and equation (9) defines the growth rate of GDP, whereas equation (10) estimates the capital output ratio. )_I -(1+yl + -vi )JI nag,/ - 11, nag,/ nag,/,-1 ( 11) ( 12) Productivity growth in agriculture is treated differently. As already mentioned, in the last decade, Brazilian agriculture recorded high productivity growth, and we impose exogenous growth rate for productivity in agriculture uniformly across all factors, as shown in the following equations. Equation (13) represents the increase in labor productivity in agricultural sectors not subject to the uniform productivity shift factor y1• Equations (14) through ( 16) update productivity of capital, land and THE MACRO MODEL 95 the sector specific factor, respectively. With agricultural productivity assumed to be uniform across all factors of production, the growth parameters ;I,;/, ;I, ;i will be the same for all agricultural sectors. (13) (14) A~g,lt = (l + Z;g,11 )A~g,lt,-1 (15) (16) Additional support for a sector specific treatment of productivity where agriculture shows total factor productivity (TFP) growth rates higher than those for manufacturing comes from a recent panel study on sectoral productivity growth in OECD and developing countries.77 In this study, depending on the estimation method, the average growth rate for agricultural TFP in middle-income developing countries ranges from 1.78 to 2.91 (in% per year). Other elements of simple dynamics include exogenous growth of labor supply, with skilled labor growing faster than unskilled labor, and investment driven capital accumulation. 78 Equation (17) determines labor supply growth for the skilled workers (unskilled labor supplies are determined in equations (6) and (7)). It simply applies an exogenous assumption about the growth of labor supply, g'', to the labor supply shift parameter. Equation ( 18) updates population. Equations (19) and (20) are similar growth equations for land and the sector-specific resource, respectively. afs =(I+ gf') af'._1 Pop = (1 + gPop) Pop _ 1 Land= (I+ g') land_1 (17) (18) (19) (20) Capital accumulation is based on the level of investment of the previous period less depreciation. Equation (21) represents the motion equation for capital growth, where 8 is the rate of depreciation and KAP is the capital stock. 77 See Martin and Mitra (1999). 78 Note that public investment, in this version of the model, has no impact on production technology. 96 BRAZIL KAP= (1- o)KAP_1 + XFz1p,-I (21) Other exogenous variables may require updating for the baseline. One obvious one is government expenditure. This is typically assumed to grow at the same rate as GDP: XFc;ov = (1 + gY )XFGov,-1 (22) Other variables that have been updated include the various transfer variables, foreign savings, exogenous world prices (i.e. the terms of trade), and fiscal policies. 4.3.2. The micro model The micro model is linked to the macro model through changes in the following set of endogenous (in the CGE model) variables: (a) changes in agricultural and non-agricultural labor income ofunskilled labor (2 variables); (b) changes in labor income of skilled labor (1 variable); (c) changes in the sectoral (agriculture vs. non-agriculture) composition of the unskilled workforce (1 variable). In addition, we take into account that unskilled and skilled labor supplies grow at different rates. In the microsimulation, we do not produce a series of cross-sections through time, but only simulate one cross-section that reflects the cumulative changes in the aforementioned exogenous and endogenous variables between 2001 and 2015. In accordance with the structure of the CGE model, the micro model treats skilled and unskilled labor differently and, in particular, simulates the decision to move from agriculture into non-agricultural sectors only for unskilled workers. In the following, we first illustrate the equations to be estimated to serve as a basis for the microsimulation and we report and comment on some estimation results.79 Then, we outline the microsimulation module that combines (static) reweighting methods to reflect the changes in the composition of the labor force, i.e. the unskilled/skilled labor ratio, and (dynamic) behavioral elements to simulate the sectoral movements and wage changes given by the CGE. We conclude this section by pointing towards possible shortcomings of the data and the methods applied. 79 The estimations are based on the 2001 PNAD. The sample includes 378 701 individuals in 112 558 households. The number of employed individuals used for the estimation of the income generation model is 166 646. Due to a relatively large number of missing income observations, labor incomes are imputed using simple OLS income regressions for different groups of employed individuals (urban/rural, agriculture/non-agriculture, worker/self-employed/employee). See Ferreira et al. (2001) for details on the PNAD dataset and the problems of using the PNAD income measure for poverty and distributional analysis. THE MICRO MODEL 97 First, we estimate sectoral mover-stayer models for unskilled heads and nonheads separately. For both heads and non-heads, we observe whether an individual has moved from agriculture into a non-agricultural sector. Our sample hence consists of those individuals who are still in agriculture and those who have moved out of agriculture within the last year. In contrast to many other household surveys, the PNAD provides information on employment histories, which allows us to identify the movers out of agriculture and, very important for our undertaking, the characteristics of the movers at the time of moving. For all the movers we thus know, for example, which type ofland right they had if they were self-employed before they moved out of agriculture. To our knowledge, this information has not been explored to date. The estimated model hence combines the idea of the mover-stayer model from the migration literature'° with the approach to modeling occupational dynamics typically applied in income generation microsimulations." In the latter approach, bi- or multinomial choice models are estimated on the entire population. In our case, this would imply comparing the characteristics of those in agriculture with those in non-agricultural sectors. Instead, the mover-stayer model compares the characteristics of only the movers with those of the stayers. This appears to be more appropriate in the current setting, as our goal is to simulate the transition from agriculture to nonagriculture. Let move be a dichotomous variable that assumes a value of I if the individual has moved out of agriculture in the last year, and O if the individual has stayed in agriculture. As indicated by equation (23), an individual will move (move= I) if the utility ( U) associated to this choice is higher than the utility of staying in agriculture."' Otherwise, the individual will stay in agriculture (24). move= I if U (move=])> 0 move = 0 otherwise. (23) (24) As indicated by equations (25) and (26), the utility of moving depends on a set of explanatory variables X and a random error term£. We hence assume a linear relationship between utility and the explanatory variables. The subscripts msh and 80 See for example Nakosteen and Zimmer ( 1980). 81 See for example Robilliard, Bourguignon, and Robinson (2002) and Bussolo and Lay (2005). The combination of retrospective information and choice models has been applied in microsimulation models e.g. to model fertility or schooling decisions. See Grimm (2005) for an application on Cote d'Ivoire. 82 The utility of staying in agriculture is assumed to be 0. This is typical identifying assumption of the logit. 98 BRAZIL msnh refer to heads and non-heads, respectively. They also serve to remind us that the parameters and variable vectors are from the mover-stayer (ms) model." U(move = 1)msnh = am,nh + X msnh/Jm,nh + 8 msnh (25) (26) where X msh includes an educational dummy for more than lO years of schooling, age, a dummy that refers to own-consumption worker, an employment category to our knowledge unique to the PNAD that describes workers who are not selfemployed, do not receive any monetary income, and work "for their own consumption"84, two dummies that refer to the type of land right held by the selfemployed in agriculture, one referring to a situation, in which the landowner agrees with the self-employed occupying the land and another to the self-employed owning the land, and a regional dummy for the northern region. X m,nh for the nonheads is a vector of similar variables, but some notable differences that will be discussed later in the regression results section. It consists of three educational dummies, experience (age-schooling-6), experience squared, a female dummy, a dummy for blacks, and the same employment category and land right dummies as before plus a dummy for non-remunerated family members or workers. Note that the reference group for the employment-related dummies are the wage-employed. In addition, the explanatory variables for the non-heads include a dummy for the household head being employed in a non-agricultural sector and another dummy for the head being a mover out of agriculture. As we cannot observe the latent utility U, the parameters of the mover-stayer model will be estimated by maximum likelihood logit techniques, i.e. we estimate the models described by equations (27) and (28), where F denotes the cumulative density function of the logistic distribution. (27) (28) In addition to the sectoral choice model, Mincer wage/profit equations for unskilled labor in agriculture (the subscript uagr) and non-agriculture (unagr), and for skilled labor (s) are estimated: 83 We do not use a subscript for indicating individual observations in the exposition of the micromodel for illustrative purposes. 84 See Notas Metodol6gicas PNAD 2001. THE MICRO MODEL In w uag, = a uag, + X uag, fJ uag, + uw uag, In wunagr = aunagr + x,magrP1magr + uwunagr Inw, =a,+X,P,+uw, 99 (29) (30) (31) where the explanatory variables in all three equations include years of education, experience, the corresponding squared terms, a female dummy and racial dummies. In addition, we include regional dummies that in equations (30) and (31) also differentiate between rural and urban areas. In the equation for the unskilled in agriculture (29), we introduce a dummy for being self-employed and the number of non-remunerated family members in order to capture their labor input. The wage/profit equations (29) to (31) are estimated using Ordinary Least Squares (OLS). Some words on the choice of this specification are in order, as estimating the two wage equations for unskilled labor using OLS may appear problematic to the reader who is familiar with the concept of selectivity bias and is aware of the available econometric methods to deal with it. The reason for estimating two wage equations is based on the assumption that the wage-setting process in agriculture is different from the one in non-agricultural sectors; and the results of our regressions confirm this assumption. When estimating two separate equations we therefore might have to account for selectivity bias. Selectivity bias refers to a bias in the coefficients of the wage equations which arises as the coefficients do not merely reflect the returns to education, seniority, or the influence of the included dummies, but also the returns of being employed in ( or selected into) the respective sector. For example, having a high level of education affects the sectoral choice, i.e. the earnings indirectly, as well as the earnings directly. Applying OLS to the estimation of two separate (sectoral) wage equations would result in coefficients that reflect both the indirect (selection) and the direct effects. Selection can also be interpreted in terms of having a kind of "comparative advantage" in the chosen sector, which is not explicitly accounted for but represented by the biased OLS coefficients. Econometricians often describe the concept of selectivity by noting that the selection into the respective sectors is nonrandom. It has become very common to correct for selectivity bias using the socalled Heckman correction or one of its many variants. Many authors have warned against the indiscriminate use of the selectivity correction methods." In line with this general skepticism, for reasons that have to do with the purpose of estimating the above equation, and due to practical estimation problems, we believe that correcting for selectivity bias is not necessary or may even lead to wrong results in the present context. The purpose of 85 See Johnston, Di Nardo ( 1997, pp. 449-450) for a short overview of the major problems involved and the citations there. 100 BRAZIL estimating the wage equations is to impute earnings for those who move between sectors. If we estimate the wage equation using OLS we implicitly assume that the returns to education and other characteristics of the individual include the indirect returns due to selection. We believe that this can be reasonably assumed, as an individual, for whom a wage is to be imputed, is actually selected into the corresponding sector."6 In addition, estimating a selection model rendered inconsistent results. The only feasible estimation strategy would then have been to reduce the number of explanatory variables in the wage equation to include only one educational dummy for tertiary education with the remaining educational dummies to be only included in the selection equation."' This of course would have been highly unsatisfactory in terms of explaining the variation in earnings. These results are mainly owed to a combination of the following two factors. First, lower levels of education are highly significant in selecting an individual into agriculture. Second, there is little variation in these variables for those in non-agricultural sectors. In light of these theoretical and practical arguments we used OLS rather than Heckman correction procedures. With few exceptions the explanatory variables of the mover-stayer models as well as in the wage/profit equations are significant at the 5 percent level." The detailed regression results are reported in the App. Table 4.1 and App. Table 4.3, but here we just want to comment on some results that we find remarkable and consider of particular importance with regard to the simulation exercise. The mover-stayer models appear to have some predictive power for the decision to move out of agriculture, as indicated for example the Pseudo R 2 of 0.07 for the heads' and 0.15 for the non-heads' mover-stayer model. It should be noted that measures of fit for logit models can only provide a rough indication of whether a model is adequate (Long and Freese 2001). In the mover-stayer model for heads, one educational dummy for IO or more years of schooling, turned out to have a significant positive influence on moving out of agriculture. Yet, we find a number of factors that negatively affect the choice of moving, among which age is the most important one. As we would expect, older individuals are less likely to move out of agriculture. The effect of a discrete change in age of some years is particularly strong for younger individuals. Working only for own-consumption also has quite a strong negative effect on the propensity to move out of agriculture. Many of these own-consumption workers are employed in the livestock sector and possibly even own livestock. In addition, if household heads own land or if they have an agreement with the landowner to occupy the land, they are more likely to stay in agriculture. Owning land or other agricultural production factors, such as livestock, hence acts as important barrier to 86 This assumption implies that there are no differences between individuals in terms of sectoral comparative advantages. If we estimated the wage equation correcting for selectivity, these differences would be reflected in the individual inverse Mills ratios. 87 These or similar problems of applications of the Heckman procedure are often noted in applied work. A case in point is Spatz (2004b). 88 Standard errors are adjusted for clustering. THE MICRO MODEL 101 intersectoral movements. Finally, household heads from the north are more likely to move out of agriculture, an interesting finding one might not necessarily expect, as the north is a region with a low share in agricultural employment. As described above, the list of explanatory variables for non-heads is longer. The strongest determinant of moving out of agriculture is the dummy indicating whether the household head is employed in a non-agricultural sector. We can think of either a self-employed head being able to offer employment to other household members in a non-agricultural household enterprise or networks of a wageemployed head that facilitate finding non-agricultural employment for relatives. In addition, the choice of the household head to leave agriculture strongly influences the choice of the non-heads. Educational dummies for having finished primary and secondary education have a significant positive effect on the probability of moving out of agriculture. This effect is strongest for having finished primary education and declines somewhat for higher educational levels. The effect of a change in experience is much stronger than the effect of the corresponding change in the squared term. As in the case of the heads, the overall marginal effect of experience (including the squared term) declines with increasing experience. The subset of coefficients for educational dummies, experience, and squared experience can be interpreted as reflecting the earnings opportunities of an individual in nonagricultural employment. In other words, these five explanatory variables can be thought of as a reduced form representation of the wage differential between agricultural and non-agricultural activities. Accordingly, the coefficients for education can be seen to reflect decreasing returns to education for the movers and the results for experience appear to catch both the effect of age being a barrier to move out of agriculture as well as the typical seniority effect in earnings, i.e. increasing but marginally decreasing returns to experience. In a similar way, the significant and negative coefficients for racial dummies can be interpreted either or both as a direct barrier to non-agricultural employment and/or an indirect effect that works through more racial discrimination in non-agricultural employment than in agriculture. Non-remunerated workers are less likely to move out of agriculture. This finding point towards the importance of externalities associated to this type of employment, as estimation results indicate that the income gains due to an additional household member engaged in the household farm are rather moderate (see App. Table 4.3). The coefficients of being an own-consumption worker or owning land are of the same sign as in the case of the heads and the changes in predicted probabilities due to a change in the dummy of equal magnitude. In sum, the results for the heads show the barriers to moving out of agriculture, whereas the results for non-heads also illustrate the possible gains of such a move. Household heads appear to respond to wage differentials between agricultural and non-agricultural sectors to a lesser degree. They hence tend to be ''trapped" in agricultural activities, possibly due to factor market imperfections. Their decision to stay or move however is of great importance for the decision of other household members. 102 BRAZIL The microsimulation involves three steps. First, households are reweighted in order to reflect the change in the skilled/unskilled labor ratio that results from different growth rates of these two types of labor over time. In a second step, unskilled labor moves out of agriculture until the new share of unskilled labor in agriculture given by the COE is reproduced. Third, wages/profits are adjusted according to the COE results taking into account the changes in the skill composition of the workforce as well as the sectoral movements of unskilled labor from agriculture into non-agricultural sectors. The reweighting procedure basically increases the weight of skilled individuals and decreases the weight of unskilled individuals to reach a new given ratio of unskilled to skilled workers following an efficient information processing rule. 89 Let weight denote the old weight (normalized to 1), and nweight the new weight of individual i. As Robilliard and Robinson (2003), we estimate the new weights by minimizing the Kullback-Leibler cross-entropy measure of the distance between the new and the old weights "' [ nweight ) Min L,nweight; - In . ; ; weight, (32) subject to the following constraints L nweight; · U; tu' _ tu -(1 + gJ -=L=-nw-e-ig_h_t;-. s-, - Is · (1 + g,) Tl° (33) Inweight, =l (34) with u (s), a dummy variable for unskilled (skilled) individuals i, tu (Is), the initial unskilled (skilled) labor force, and g" (gs), the cumulative growth rate (between 2001 and 2015) of the unskilled (skilled) labor force. tu" (Is') hence denotes the target value for the number of unskilled (skilled) labor. Equation (33) hence states that the new weights have to reflect the new skill composition (the ratio of unskilled to skilled workers) of the workforce. Equation (34) is the adding-up normalization constraint. Note that this procedure gives new individual weights for just the employed population. Yet, for our purposes we need household weights for entire population. 89 For details on maximum entropy econometrics see Golan, Judge and Miller (1996). Robilliard and Robinson (2003) apply these methods to reweight household survey weights. They also provide a GAMS code for solving this type of problems. BRAZIL IN THE NEXT DECADE 109 intensively, whereas manufacturing labor intensities are in-between agriculture and services. Table 4.4: BaU's output and trade sectoral growth rates, and employment intensities Annual average growth rates Employment percentages Labor demand by sector by skill Output Imports Exports Skilled lfnsk- Skilled Unsk. Skilled -Unsk - --·---- - ----·------- .. -·-----~--- -- 0 ---·------- --~---- Cerea/Grains 3.2 2.5 2.3 0.3 0.1 5 2 98 Oi/Seeds 3.1 2.2 2.4 0.1 -0.1 0 I 6 94 RawSugar 3.2 0.2 0.1 0 I 4 96 OtherCrops 2.9 1.3 2.5 0.0 -0.1 I 12 3 97 Livestock 3.2 1.5 0.3 0.1 2 4 IO 90 RawAnima/Products 3.3 2.5 1.6 0.4 0.3 0 3 I 99 Oi/Minera/s 3.3 3.0 2.9 1.5 1.7 0 0 15 85 LightManufacturing 3.3 0.8 3.7 1.0 1.2 I 2 16 84 Agri!ndustriesExp 3.2 0.5 3.4 1.0 1.2 2 3 16 84 WoodProductsl'aper 3.3 0.9 3.5 1.0 1.2 2 2 15 85 ChemicalsOi/Pr 3.3 1.8 2.9 I. I 1.3 2 I 30 70 Meta/Minera/Products 3.5 1.8 3.3 1.2 1.4 2 2 17 83 MachineryEquipment 3.6 1.9 3.5 1.4 1.6 3 2 28 72 ----------- ----· ~ - ---~--------- ---------- -OtherServices -3~0-- 2.6 1.7 2.1 i.3 58 30 33 67 Construction 3.2 2.3 2.5 2 8 6 94 TradeCommunication 3.1 2.4 1.8 2.2 2.4 15 18 17 83 Pub/icServices 3.1 2.7 1.7 2.2 2.4 9 4 41 59 Agri 3.0 1.9 2.4 0.0 4 27 6 94 Non-Agri 3.2 2.0 3.1 2.2 96 73 26 74 Economywide 3.2 2.0 3.1 2.0 100 100 24 76 Source: Authors' calculations. Note: The mapping of this table sectors and GTAP sectors is shown in App. Table 4.6 and App. Table 4.7. 4.4.2. Distributional and poverty results for the BaU A moderate decrease in poverty between 2001 and 2015 results from microsimulating the identified key structural trends on the Brazilian household data. Considering the full sample of all households, the headcount poverty ratio (PO) declines by about 6 percentage points (see Table 4.5). The reduction of the average normalized poverty gap (PI) and the poverty severity index (P2) indicates that those who remain poor move closer to the poverty line."' Inequality changes very 94 A short note on the interpretation of the reported poverty measures: The income-gap ratio, i.e. average income shortfall (of the poor) divided by the poverty line, can be calculated as Pl/PO. This ratio is 0.4 for all households in our case, i.e. the perfectly targeted cash transfer needed to lift every poor person out of poverty is 40 percent of 110 BRAZIL little, as indicated by the 0.1 decrease in Gini coefficient ( or as in the Theil or other inequality indices, not reported). These average indices indicate that some progress in reducing aggregate poverty and inequality is achieved in a Business as Usual scenario, but these aggregate measures may conceal relevant distributional changes at a more disaggregated level. In fact, reaching stronger poverty reduction may require specific pro-poor policies which often rely, for their successful implementation, on more detailed information about disaggregated distributional effects. A first obvious way to gather more detailed information is to analyze the poverty and inequality impacts separately for the agricultural and non-agricultural households. Table 4.5: Poverty and inequality in the BaU scenario, by sectors All households Non-agricultural Agricultural hou.fehold.f 2001 level 2001-15 2001 level 2001-15 2001 level 2001-15 chanl!e chanl!e chanl!e PC income 314.9 1.5 351.9 1.2 148.3 2.3 Gini 58.6 -0.1 57.1 0.6 56.6 -0.7 PO 23.6 -5.6 18.6 -3.1 46.2 -13.8 Pl 9.6 -3.0 7.1 -1.6 21.0 -8.0 P2 5.3 -1.8 3.7 -0.9 12.3 -5.2 Pupu/atwn% 100 81.8 3.3 18.2 -3.3 Cuntr. tu PO 64.4 8.8 35.6 -8.8 Source: Authors' calculations. Note: PC income is per capita income in 2001 R$ and the change is given as annual growth rate. All levels are in percent and changes in percentage points. A household is classified as "agricultural" when its head and/or at least two of its members are employed in agriculture. In 2001, according to this classification, agricultural households accounted for 18.2 percent of the Brazilian population, poverty incidence among them almost reached 50 percent, and their contribution to total poverty was about 36 percent (see Table 4.5). Between 2001 and 2015, the share of agricultural households in the population shrinks by 3.3 percentage points following the decline in agricultural employment of more than 5 percentage points. Poverty among agricultural households falls by more than 13 percentage points, whereas poverty among non-agricultural households decreases by only 3.1 percent. Accordingly, the contribution of agricultural households to the headcount falls by almost 9 percentage points. the poverty line times the number of the poor. Thus, 0.4 times the percentage point change in PO (here 2.4) provides a percentage point change benchmark for evaluating the change in Pl, as this would be the change in Pl that we would observe had the average income of the poor stayed constant while the headcount declined. BRAZIL IN THE NEXT DECADE 111 A more detailed analysis also shows that the lack of progress in aggregate inequality is due to the agricultural and non-agricultural groups' individual inequality indicators moving in opposite directions. Among non-agricultural households, inequality rises because skilled labor income, a major source of income for these households, grows faster than that of unskilled labor. Conversely, inequality among agricultural households falls, mainly because richer agricultural households earn a higher share of their income from non-agricultural unskilled labor and, in some cases, from skilled labor. Figure 4. I: Growth incidence curves, BaU, all, agricultural, and non-agricultural households 0 20 40 60 80 Cumulative population ranked by per capita income I== ~~n-agr ----- Agr I Source: Authors' calculations. 100 Another way of analyzing detailed distributional effects is to consider growth incidence curves. These curves plot per capita income growth at income percentiles (Ravallion and Chen 2003) and are shown in Figure 4.1 for all households as well as for the agricultural and non-agricultural groups." Reflecting the increase in unskilled agricultural wages from the CGE model's results, per capita income growth is much higher for agricultural households. In addition, the agricultural growth incidence curve illustrates a strong pro-poor distributional shift, which reflects both the increase in agricultural labor incomes and the gains resulting from moving out of agriculture. These agricultural households specific distribution shifts also explain the pro-poor changes in the national distribution, 95 The household category, i.e. agricultural or non-agricultural household, is the category the household belonged to in the base year 2001. 112 BRAZIL since only minor distributional changes are registered in the non-agricultural distribution. However, richer non-agricultural households experience somewhat higher gains than poorer households. Non-agricultural poor household incomes increase by a meager 1 to 1.5 percent annually. These more detailed analyses of the long term evolution of the Brazilian income distribution highlight the different roles played by changes in inequality and shifts in the growth rates of the average incomes. The following two relevant questions then arise: if the current (2001) distribution of income were to remain unchanged, how would additional growth help in reducing poverty? And, what is the role of the differential in the sectoral growth rates for agriculture and non-agriculture in reducing poverty? Answering these questions requires performing two additional microsimulations as follows. The first simulation generates a counterfactual distribution under the assumption that all incomes out of all sources grow by 1.5 percent annually. This implies shifting the entire income distribution "to the right" leaving its shape unchanged. Individuals do not change employment sectors and hence households retain their initial non-agricultural or agricultural classification. Results are presented in Table 4.6 and changes are given as percentage share of the BaU change (column I). In addition, we simulated a second counterfactual distribution for agricultural and non-agricultural households separately with per capita incomes of the respective household types growing with the BaU rates, i.e. by 1.3 percent annually for non-agricultural and 2.4 percent annually for the agricultural households (column II). Table 4.6: Poverty and inequality in a distributionally neutral scenario _J All households N!!.ll_~a~r!cultural household.f _ Agricultural_households %of %of %of %of %of %of 2001 BaU 2001 BaU 2001 BaU level BaU change level BaU change level BaU change change I II change I II change I II PC income 314.9 100.0 100.0 351.9 117.7 100.0 148.3 65.7 100.0 PO 23.6 91.7 102.4 18.6 139.8 133.3 45.9 56.5 90.5 PI 9.6 90.9 97.7 7.1 132.5 119.7 20.8 61.9 93.2 /'2 5.3 86.8 97.9 3.7 125.6 114.3 12.l 62.6 93.4 Source: Authors' calculations. The comparison of the counterfactual simulations of the "completely" distributionally neutral ( column I) and the "separately" neutral ( column II) scenarios shows that the growth bias in favor of agricultural households is poverty reducing. Yet, the difference between the BaU and the completely neutral scenario does not seem too pronounced. This is due to the fact that poverty among nonagricultural households is reduced much more than in the BaU, where the income distribution among these households worsens. This "slight" worsening of the BRAZIL IN THE NEXT DECADE 113 income distribution hence hampers quite strongly the potential of growth to reduce poverty among non-agricultural households. In addition, the differences between the two neutral scenarios for non-agricultural households illustrate that a 0.2 percentage point difference in annual growth rates for 14 years can make a difference in terms of poverty reduction. The last two columns of Table 4.6 show the importance of growth for reducing poverty among agricultural households as well. A 0.9 point percentage point difference in annual income growth rates for 14 years implies a reduction of about 5 percentage points less in the headcount over this time period. In contrast to what we see for non-agricultural households, the impact of the pro-poor distributional shift for agricultural households observed in the BaU is relatively small. In other words, had the income distribution among agricultural households not improved, growth would have reduced poverty by only little less. The poverty reductions recorded in the BaU scenario are due the change in skill endowments, the increase in real factor prices, and inter-sectoral movements. A main advantage of micro-simulation techniques is their ability to decompose the total effect in different partial effects that can be attributed to single causes. A slight complication arises because different causes interact. The interaction arises because factor incomes increase at different rates in agricultural and nonagricultural sectors. By simulating counterfactual distributions, where only one cause at the time, or a combination of two of them, are included, it is possible to decompose the total effect into individual or joint (interactive) contributions. Figure 4.2: Decomposition of poverty changes, BaU, all households D Skil l Upgrading • Sect. Change Source: Authors' calculations. a Factor Price Change J •Interaction Note: The figure displays the contribution of the respective component to the total change in PO and P2, respectively, in percent. The contributions add to 100. Contributions refer to reductions in the respective poverty indices. 114 BRAZIL Figure 4.2 displays the results of this decomposition. Factor price changes account for the largest share of total poverty reduction. The change in the skill composition of the workforce does not contribute much to poverty reduction, whereas the sectoral shifts are quite important, in particular for the poorer among the poor, as the higher contribution of the sectoral change component with regard to P2 indicates. This implies that households with members moving out of the agricultural sector escape poverty. We consider this issue in more detail later. The interaction component, which actually is a sum of distinct interaction components, hampers poverty reduction. Counterfactual simulations show that the interaction between sectoral movements and income changes is the most important one. It is negative since people move out of agriculture where their incomes would have increased much more than in non-agricultural sectors. In sum, the distributional and poverty analysis suggests that the BaU scenario leads to relatively little poverty reduction. Agricultural households fare quite well and the poverty incidence and intensity among them is reduced quite substantially. Decomposition analyses show that sectoral change contributes quite significantly to poverty reduction, although income growth is the most important source of poverty reduction. Micro-accounting exercises underline the importance of growth for poverty reduction, but we also illustrate that slight increases in inequality can considerably reduce the poverty reduction potential of growth in the context of a high-inequality country, such as Brazil. 4.4.3. Macro results for the full liberalization and the Doha trade policy shocks The trade shocks simulated in the dynamic CGE model consist of changes in Brazilian tariff protection against imports from the rest of the world and of exogenous changes of international prices of traded goods and export quantities demanded by foreigners.96 The shocks are assumed to take place progressively through a gradual phasing-in starting in 2005 and lasting 6 years. Table 4.7 displays these shocks as percentage changes of the final year (2015) between the BaU and the trade reform scenarios. As part of the shock and to leave the government fiscal balance unchanged, tariff revenue losses are compensated by a lump sum transfer implemented as an increase in the direct taxes paid by households. This lump sum additional tax is the least distortionary instrument that can be readily used in our model, however, in practice, the Brazilian government may chose other forms of compensatory taxes which may alter relative prices and have significant income distribution effects. 96 It should be noted that to mimic the global model results for increased demand for Brazilian exports and changes in international prices, we introduce a downward sloping export demand function as shown in equation (I) above. During a shock, for obvious reasons, we cannot target both prices and quantities and the shock is implemented by modifying both the international price index WPEindex (the price shock) and the intercept dt (the quantity shock). Our Brazil (single-country) model will then endogenously determine the quantity supplied. BRAZIL IN THE NEXT DECADE 115 Table 4. 7: Trade shock -Tariff reductions and international prices changes Own tariff reductions Change in import prices Change in export prices Full Deep Weak Full Deep Weak Full Deep Weak lib Doha Doha lib Doha Doha lib Doha Doha Cereo/Grains -100 8 2.0 2.1 16 5.9 6.0 Oi/Seeds -100 6 2.5 2.5 14 4.8 4.9 RawSugar 2 0.9 1.0 14 5.3 5.4 OtherCrops -100 0 0 2 0.9 0.9 13 4.7 4.8 Livestock -100 2 1.0 I.I 25 9.7 9.8 RawAnima/Producls -100 2 0.4 0.4 18 6.6 6.7 Oi/Minerals -100 0 0 0.1 0.1 2 I.I 1.3 LightManufacturing -100 -6 0 I I.I 1.2 9 3.8 4.0 Agri!ndustriesExp -100 -4 -1 0 0.6 0.6 7 3.0 3.2 WoodProductsPaper -100 -6 -2 0 0.0 0.0 4 1.8 2.0 Chemica/sOi/Pr -100 -1 I -3 -1 -0.1 0.0 3 1.4 1.7 Meta/Minera/Products -100 -6 -1 0 0.0 00 3 1.6 1.7 MachineryEquioment -100 -7 -2 0 0.0 00 2 1.5 1.7 OtherServ,ces 0 0.0 0.0 5 2.0 2.2 Construction 0 0.0 0.0 4 1.7 1.9 TradeCommunication 0 -0.1 -0.1 5 1.9 2.1 Pub/icServices 0 -0.1 -0.1 5 2.1 2.3 Agri -100 0 0 5 1.5 1.5 14 4.8 4.9 Non-Agri -100 -7 -2 0 0.0 0.1 4 1.9 2.1 Economywide -100 -7 -2 0 0.1 0.1 5 2.2 2.4 Source: Authors' calculations. The full liberalization scenario has the largest impacts: tariffs are completely eliminated and Brazil enjoys strong terms of trade gains; the other two shocks, representing two possible versions of the Doha negotiation outcomes, generate almost no own liberalization and fairly muted global prices effects. In order to fully appreciate their final effects, these shocks need to be mapped to the economic structure of Brazil. Table 4.8 presents this structure and helps in this regard. For instance, in the full liberalization scenario, export oriented sectors -those displaying high shares of export to domestic output -such as Oilseeds, Other Crops and the industrial sectors transforming agricultural products (AgrilndustriesExp which buys most of its inputs from agriculture) record considerable increases of their export prices. Conversely, import competing sectors, such as Chemicals and Oil derived products and capital goods, do not face high increases in their international prices. These combined export and import price movements result in fairly strong terms of trade gains, inducing significant reallocation of resources towards export oriented sectors. Additional push for this reallocation comes from Brazil's own liberalization which entails a reduction of 116 BRAZIL the anti-export bias implicit in the higher protection rates for manufacturing of the initial tariff structure. Table 4.8: Initial (year 2001) structure of the Bra=ilian economy I Tariff Sectoral Imports/ Sectoral Sectoral Exports/ DomDemof Dom rates imports output exports comp output Cereo/Grains 7 I 15 I 0 I Oi/Seeds 6 0 8 0 4 29 RawSugar 0 0 0 0 0 0 OtherCrops 9 2 3 4 8 7 Livestock 3 0 I I 0 0 Raw Animal Products 8 0 I I 0 I Oil Minerals 4 7 33 I 7 25 Light Manufacturing 17 4 5 5 3 2 AgrilndustriesExp 18 3 3 7 19 II WoodProductsPaper 9 2 5 3 7 IO ChemicalsOi/Pr 9 15 10 9 8 3 Metal Mineral Products 12 5 6 5 13 11 MachineryEquipment 19 37 27 8 20 II OtherServices 0 II 3 23 5 I Construction 0 0 0 8 0 0 TradeCommunication 0 IO 5 13 5 2 Pub/icServices 0 2 1 II I 0 Agri 8 4 4 7 12 6 Non-Agri II 96 6 93 88 4 Economywide II 100 6 100 100 4 Source: Authors' calculations. These effects are detailed in Table 4.9. The complete elimination of tariffs in the full liberalization case explains the large increase of imports (measured in volume) which, in the final year of this scenario, is 2 I% above the value in the same year of the BaU. Increases in imports of agricultural goods are much weaker: an aggregate 6% increase versus the 21 % surge of the non-agriculture bundle. The combination of lower initial tariffs and stronger international price increases for agriculture, with respect to non-agriculture, explain the difference in import response of these two aggregate sectors. Given their very limited scope of tariff reduction, the Doha scenarios imply much more contained changes of imports. With high elasticity of substitution in demand (currently set at 4), cheaper imports have the potential to displace domestic production, especially for those goods whose demand is fulfilled by a large share of foreign supply. For Brazil, this is the case for the Chemicals, and Capital goods sectors. In the full liberalization scenario, domestic production experiences significant market share losses in these BRAZIL IN THE NEXT DECADE 117 sectors; however this is not happening in the Doha cases. The competition from cheaper imports is also reflected -again only for the full liberalization case - in the decline of prices of domestic output. Table 4. 9: Brazil' structural adjustment, percent changes in the final year between BaU and trade shocks Demand side Import volumes Domestic demand of Price of domestic dom nroducts out,.ut in dom mkts Full Deep Weak Full Deep Weak Full Deep Weak lib Doha Doha lib Doha Doha lib Doha Doha Cerea/Grains -6 -3 -3 4 1 1 -2 1 1 Qi/Seeds -18 -8 -7 5 1 1 -6 0 0 RawSugar 0 0 0 -2 1 1 QtherCrops 23 1 2 1 0 0 -1 1 1 Livestock -4 1 1 3 1 1 -2 1 1 RawAnima/Products 22 4 5 2 1 1 -2 1 1 Qi/Minerals -6 0 1 1 -1 -1 -5 0 1 LightManufacturing 48 0 -3 0 1 I -5 0 0 AgrilndustriesExp 59 2 1 0 0 0 -4 0 1 WoodProductsPaper 23 4 4 -1 0 0 -4 0 1 ChemicalsQi/Pr 18 5 3 -2 -1 0 -4 0 1 Meta!Minera/Product 24 3 2 -4 -1 -1 -5 0 1 MachineryEg_ui[J'!'_e_nt_ 42 5 3 _-12 ____ -l__ -1 -6 0 1 ------- ----· ·--------- QtherServices 1 0 0 -4 0 1 Construction -14 2 3 0 0 0 -3 0 1 TradeCommunication -12 2 3 0 0 0 -3 0 1 PublicServices -13 2 3 0 0 0 -3 0 1 --·---------------- --·- --- --- -- ------ ------ ----·-----···-- -- --·· Agri 6 -2 -1 2 1 1 -2 1 1 Non-Agri 21 3 3 -1 0 0 -4 0 1 Economywide 21 3 3 -1 0 0 -4 0 1 118 BRAZIL Table 4.9 continued _J Supply side Exoort volumes Domestic outout Price of domestic Full Deep Weak Full Deep Weak Full Deep Weak lib Doha Doha lib Doha Doha lib Doha Doha Cerea/Grains 68 14 13 5 I I -2 I I Di/Seeds 60 9 8 20 3 3 -3 I I RawSugar 0 0 0 -2 I I DtherCrops 6 -3 -3 I 0 0 -1 I I Livestock 3 I I -2 I I RawAnima/Products 5 0 -1 2 I I -2 I I - -------- --··-····- -. ----··-· - -----· ... -·· Di/Minerals 26 2 I 7 0 0 -4 0 I LightManufacturing 159 62 61 5 3 3 -4 0 I Agri!ndustriesExp 30 4 4 3 I I -4 I I WoodProductsPaper II -1 -1 0 0 0 -4 0 I ChemicalsDi/Pr 9 -1 -1 -2 -1 0 -4 0 I Meta/Minera/Product 15 0 -I -2 -1 -1 -4 0 I MachinerJ.Equipment 11 -1 -2 -10 -1 -I -5 0 I DtherServices I 0 0 -4 0 I Construction 8 -I -1 0 0 0 -3 0 I TradeCommunication 6 -1 -2 0 0 0 -3 0 I PublicServices 7 -1 -2 0 0 0 -3 0 I Agri 22 I 0 3 I I -2 I I Non-Agri 21 3 2 0 0 0 -4 0 I Economywide 21 3 2 0 0 0 -4 0 I Source: Authors' calculations. These demand/imports side effects are linked to the supply response to which we now tum. For producers of exportable goods, the reduction of prices in local markets (~Pd) combined with unchanged or rising export prices creates incentives to increase the share of sales destined to foreign markets. This export response (shown in the columns "Export Volumes") varies across sectors and it is linked to the pattern of Brazil's comparative advantage and to the increase in international prices. Brazil's comparative advantage can be ascertained by considering the export orientation (Exports / Dom Output) column in Table 4.8, which highlights three sectors in particular: Oilseeds, Other Crops, and the Agricultural transformation industry. These sectors -which also enjoy large jumps in their international price -experience export surges. Due to the generally positive export price shocks, other sectors join in an overall expansion of supply to foreign markets. Rising export sales more than offset, or at least compensate, reductions of domestic sales and lead to changes observed in the columns labeled "Domestic BRAZIL IN THE NEXT DECADE 125 agricultural households. 101 The agricultural expansion following trade liberalization has only a minor effect on agricultural employment, by far not enough to offset the reduction in agricultural employment from the BaU. Accordingly, the change of the share of agricultural households due to trade liberalization is only minor, in particular for the Doha scenario. Yet, when translated in actual migrating individuals, this small share change means that almost four hundred thousand individuals -who would have become members of non-agricultural households in the BaU - in the full liberalization scenario remain in agricultural households. Despite the fact that these "potential mover households" are on average poorer than the typical "stayer household", as we illustrate below, poverty among agricultural households decreases compared to the BaU. The poor stayers hence gain under both trade scenarios although this gain is almost negligible for the Doha scenario. Table 4.15: Poverty impact of trade, agri stayers __J- , .. ,, .... PO Pl P2 Population % 44.1 20.0 11.7 Source: Authors' calculations. _ Households remaining in agri BaU 2001-15 Doha % of chan es BaU chan e -11.7 101.7 -7.0 102.4 -4.6 102.3 14.9 100.4 Ful/%of BaUchan e 109.5 108.5 108.2 101.5 As could be indirectly inferred from the analysis of the stayers, the group of the movers should experience the largest welfare gains. Indeed as illustrated in Table 4.16, in the BaU agricultural households who become non-agricultural households record a 22.4 percentage points reduction in their headcount index, down from a considerably high, especially in comparison to the stayers group, initial level of 56.6 percent. This outcome could not be derived straightforwardly from the estimation of the migration choice. In fact, the estimations showed that potential migrants were found to be poorer, in particular landless, heads, but also better educated, hence less poor, non-heads. The explicit quantitative measurements allowed by microsimulation were needed to highlight the poverty reducing role of changes in the sectoral composition of employment. IOI The initial poverty levels among those who stay in agriculture under the trade scenarios are almost identical to the initial levels among the BaU stayers, so we decided not to report them. The same holds for the movers, for whom we report results later. 126 BRAZIL The observed poverty reduction under the trade scenarios is of a moderate additional increase. This is due to the income increases trade reforms induce in the non-agriculture sectors, but also because the fewer households that still move out of agriculture under the trade scenarios are actually poorer. Table 4.16: Poverty impact of trade, sectoral movers Agri households who have become non-agri 1001 levels BaU 2001-15 Doha% of Full% of chan es BaU chan e BaU chan e PO 56.6 -22.4 105.1 108.2 Pl 26.0 -14.0 102.0 105.4 P2 15.2 -9.4 101.7 105.1 Population % 3.1 98.0 92.5 Source: Authors' calculations. One final category needs to be examined: the non-agricultural stayers. Representing 80 percent of the population, this is a large group; however, given the negligible migration out of the non-agricultural sector observed in the data, this group is explicitly excluded from the migration choice. For these households, the full liberalization brings about an additional reduction in the headcount of 0.4 percentage points, and through its favorable impact on non-agricultural unskilled wages the Doha scenario, too, makes a small but noticeable difference. Table 4. 17: Poverty impact of trade, non-agri stayers _J PO Pl P2 Population % Non-agri households before and after 2001 levels BaU 1001-15 Doha % of Full% of chan es BaU chan e BaU chan e 18.6 7.1 3.7 82.4 -3.8 -1.8 -1.0 104.0 1033 103.2 110.7 109.8 109.5 Source: Authors' calculations. Up to this point, the disaggregated analysis of the poverty impacts has been based on sectoral affiliation and thus it has been possible to link it directly to the sectoral results generated by the CGE model. However, additional policy relevant criteria can be used to identify other groups of households and to evaluate their specific trade induced poverty effects. In particular, we conduct impact analyses for rural and urban areas, by regions, by land ownership, by educational level and by occupation. Obviously, all these criteria are somehow correlated to basic categories (and variables) included in the CGE model, for example the educational BRAZIL IN TilE NEXT DECADE 127 level is linked to the skilled/unskilled factor types, or the region to the prevalence of agricultural employment. It should be noted that the criteria are not included in the COE analysis. Yet, by using the full household survey information we can generate impact profiles where household are grouped according to poverty and distributionally relevant correlates. Table 4. I 8: Poverty and inequality impact of trade, urban and rural Urban Rural 2001 Bau Doha% Full% oj 2001 Bau Doha% Full% of levels 2001-15 of BaU Bau levels 2001-15 of Bau BaU chanl!es chanl!e cham>e chanPes chanPe chanPe PO 19.6 -4.0 103.8 112.2 44.4 -12.1 103.1 108.2 Pl 7.6 -2.0 103.2 110.1 20.1 -7.1 102.5 108.2 P2 4.0 -12 103.0 109.5 11.7 -4.7 102.3 107.7 Population % 83.7 1.3 99.5 93.2 16.3 -1.3 99.5 93.2 Contr. to PO 69.4 3.9 96.6 35.5 30.6 -3.7 102.3 119.8 Source: Authors' calculations. Table 4.18 shows the poverty impact of trade by urban or rural residence. Interestingly, the share of urban households in 2001 (83.7 percent) is even higher, although not much, than the share of non-agricultural households (81.8 percent). Quite some households live in urban areas, very likely in urban peripheries, and earn their living primarily from agricultural wage-employment. Actually, only 66 percent of the agricultural households live in rural areas, while 5 percent of the non-agricultural households live in rural areas. The micro-simulations that generate the results of Table 4.18 also take into account rural-urban migration by assuming that households migrate to urban areas if all employed household members leave agriculture. In the BaU, this causes the rural population to decline by l .3 percentage points. The urban population accounts for almost 70 percent of the Brazilian poor in 2001 and this share rises in the BaU by 3.9 percentage points. Urban poverty declines under both trade scenarios with the decline being stronger under the full liberalization scenario. The Doha scenario hardly affects rural poverty, but full liberalization decreases the rural headcount by an additional percentage point. Some simple calculations can give some more meaning to these figures: The 0.5 percentage point difference in PO in the full liberalization scenario means that approx. 135 000 people are lifted out of poverty. 102 The l percentage point difference implied by the full liberalization scenario reduces the number of poor people in rural areas by approx. 115 000. Considering the very small increase in non-agricultural unskilled wages this may be somewhat surprising, but it is the urban concentration that drives this result. Some more growth in urban areas lifts more people out of poverty than very high agricultural growth. 102 To put these absolute numbers into perspective it should be noted that the Brazilian population in 2001 was approximately 165 million, of which 39 million were poor (27 million in urban and 12 million in rural areas). 128 BRAZIL Table 4. /9: Poverty impact of trade, by region ~ 2001 initial levels 2015 Population % contr. to BaU 2001- Doha %of Full% of % PO PO 15PO BaUPO BaUPO chanxe chanxe chanxe North 5.7 34.0 8.2 -7.9 101.0 107.7 Northeast 28.5 45.4 54.8 -9.3 102.6 109.5 Southeast 43.5 12.4 22.8 -3.7 101.9 107.9 South 15.2 14.7 9.5 -4.6 101.4 1078 Center-West 7.1 16.0 4.8 -5.0 119.4 121.3 Source: Authors' calculations. Due to the regional differences both in factor endowments and specialization patterns, we might expect poverty reduction patterns to differ substantially between the regions for the BaU as well as for the trade shocks. The reduction in the headcount for the BaU confirms this expectation, as poverty declines more strongly in the Northeast, South, and Center-West, the regions with the highest shares in agricultural employment. The Doha round has negligible effects across all regions although the figures in Table 4.19 suggest a different story for the Center-West. Yet, a look at the changes of Pl and P2 (not reported) demonstrates that this strong effect is due to many households being just below the poverty line in this region.1 "' The Northeast, the region with the highest incidence of poverty where more than 50 percent of the Brazilian poor reside, benefits most from the Doha liberalization and about 50 000 individuals are lifted out of poverty in this region. In the same region, full liberalization helps about 175 000 individuals to escape poverty. The poor in the North, another region with worryingly high poverty rates, gain relatively little from trade liberalization, whereas poverty in the South as well as in the Center-West decreases quite substantially due to the importance of agricultural income for the poor in these regions. Table 4.20 shows the poverty changes for landowners and agricultural households who do not own land separately. The landowning households account for approximately 40 percent of the population in agricultural households. The differences between the two groups of agricultural households are quite striking. The poverty incidence among landowning households is much lower; the difference is more than IO percentage points. Poverty decreases quite substantially for both groups in the BaU. Note that we only consider households who stay in agriculture. Poor households who do not own land benefit little from the Doha round, but full liberalization brings about an additional decrease of more than I percentage point in the headcount (affecting almost 100 000 individuals). The 103 This is a case that illustrates why we usually report not only the headcount index, as this indicator can be quite misleading in some instances. BRAZIL IN THE NEXT DECADE 129 Table 4.20: Poverty impact of trade, agricultural stayers by owning land Landowner households No land owning households BaU Doha Fu//% BaU Doha Fu//% 2001 2001-15 %of of BaU 2001 2001-15 %of of BaU levels BaU levels BaU changes change change changes change change PO 37.1 -10.5 101.7 108.0 48.5 -12.5 101.9 110.6 Pl 16.6 -6.2 102.7 109.1 22.2 -7.4 102.3 108.4 P2 9.5 -3.9 102.7 109.4 13.1 -5.0 102.2 107.8 Population % 38.4 61.6 Source: Authors' calculations. reason why they benefit more from both trade scenarios is that they are more specialized in agricultural income. This is not necessarily what one would expect, but it may well be that owning land provides the resources to set up a small nonagricultural business. Noteworthy is the finding that under both trade scenarios poverty gap as well as severity index decrease stronger than the headcount in terms of the BaU change. This again has to do with the specialization of households. The poorer landowning households derive a higher share of their income from agriculture whereas the richer households (at least among the poor) earn a higher share of income from non-agricultural activities. Table 4.21: Poverty impact of trade, by educational levels 2001 initial levels 2015 BaU2001- Doha% Full% of Popula- PO % contr. Popula- 15PO of BaU BaUPO Hh. average tion % to PO lion% change PO change schoolinl! chantle <=3 16.2 52.3 35.9 16.0 -10.7 103.5 111.5 <=5 16.9 36.0 25.7 16.6 -9.1 102.9 109.9 <=8 20.9 21.1 18.7 20.7 -5.6 105.0 111.5 <=10 12.9 11.3 6.2 12.9 -2.9 104.6 111.3 >=11 33.1 9.6 13.5 33.7 -2.0 101.7 104.6 Source: Authors' calculations. As Table 4.21 indicates, the educational level of the households is an important determinant of poverty. The headcount among households with 3 or less average years of schooling of the employed household members is well above 50 percent. This group accounts for 16.2 percent of the population but for more than a third of the poor. The poverty incidence among households with 4 or 5 years of schooling is 15 percentage points lower. For both groups with ten or more years of average schooling, the headcount is about IO percent. The Doha round does not appear to be particularly helpful for those with little educational endowment. Yet, the full liberalization scenario again leads to a substantial additional reduction in poverty. 130 BRAZIL Finally, we analyze the poverty impact of trade reform by occupational groups. In Table 4.22, we differentiate between wage-employed, self-employed and households with members engaged in both types of employment in agricultural and non-agricultural activities, respectively.104 In addition, there are households with no employed household member. One out of five poor people in Brazil comes from a self-employed agricultural household, and the agricultural wage-employed households are almost equally poor. In non-agricultural households, the difference between self-employed and wage-employed households in terms of poverty is not too pronounced. Due to their high share in the population, non-agricultural wageemployed households account for more than a third of the Brazilian poor. Poverty rates are significantly lower for households who derive their income from both wage-and self-employment. Under the Doha scenario, all non-agricultural household groups gain, whereas there are only minor gains for agricultural households. As noted above, full liberalization however helps both agricultural and non-agricultural households. Interestingly, poverty in non-agricultural activities declines more among the self-employed, whereas in agricultural activities the decline is stronger for the wage-employed. As in the case of households who do not own land, the agricultural wage-employed households derive more income from agricultural unskilled labor than agricultural self-employed households. For non-agricultural households, it is the greater importance of skilled income for the wage-employed that makes poverty decline less strongly for this group. Table 4.22: Poverty impact of trade, by occupation _J 200 l initial levels Bau 2001-15 Doha 2001-15 Full 2001-15 Popula- % Popula- PO Popula- %of Popula- %of PO contr. tion % tion % BaUPO tion % BaUPO tion % to PO change change change change change change Agri wage-empl 5.6 46.2 11.0 -1.5 -14.3 -1.5 101.5 -1.4 111.2 Agri self-empl. 10.2 48.0 20.7 -1.3 -13. 7 -1.3 101.4 -1.2 106.7 Agri both 2.4 38.8 3.9 -0.5 -14. I -0.5 101.3 -0.5 105.5 Notempl. 7.8 27.4 9.1 0.0 -4.3 0.0 100.0 0.0 100.0 Non-agri wage 48.5 17.2 35.3 1.6 -3.1 1.6 107.2 1.5 117.4 Non-agri self 15.9 22.7 15.3 I.I -2.9 I.I 107.9 I.I 121.0 Non-agri both 9.6 11.5 4.7 0.6 -2.3 0.6 108.0 0.5 123.4 Source: Authors' calculations. To sum up, the poverty changes under the deep Doha scenario are rather moderate and disappointing. With one exception, our analyses do not detect a particularly favorable effect on any of the poor and vulnerable groups that we have identified. This one exception is that the Doha scenario very slightly appears to favor the Northeast. Overall, income growth under the Doha scenario favors non- 104 If the number of self-employed household members is greater than the number of wage-employed members, the household is considered self-employed, and vice versa. Are the numbers equal, the households falls under the "both" category. CONCLUSIONS 131 agricultural activities and, accordingly, urban areas. Since the population is concentrated in urban areas, some growth can already reduce poverty considerably, in particular if accompanied by a pro-poor distributional shift, as in the Doha scenario. Our analyses show that anti-poor changes in the distribution can easily dwarf the poverty reducing potential of growth. The income growth pattern under full liberalization tends to favor poor groups. Poor agricultural and less educated households benefit considerably more from full liberalization than from the Doha liberalizations. 4.5. Conclusions Our analysis suggests that the economic effects of the Doha round, even of an "optimistic deep" liberalization scenario, are rather limited for Brazil. Accordingly, poverty would remain largely unaffected by this trade reform, which does not appear to be biased in favor any of particularly poor groups. Yet, through a slight improvement in the urban income distribution the Doha scenario has some positive effect on poverty. In contrast, a full liberalization scenario implies quite substantial welfare gains that are concentrated among some of the poorest groups of the country, in particular those in agriculture. Consequently, the rural poor and certain comparably poor regions in Brazil benefit more than proportionately. This result is driven by an export boom in agriculture and agricultural processing industries, growing labor demand and associated higher wages. Following full liberalization, a smaller number of workers remain in agriculture compared to the BaU. Given that inter sectoral migration may substantially improve the income situation of a household, one may expect full liberalization to weaken poverty reduction. This expectation is supported by the observation that moving households are on average poorer than those remaining in agriculture. However, this is not the case, as the gain in agricultural incomes overcompensates the reduced benefits from lower migration flows. The beneficial impact of the full liberalization is not limited to rural areas and agricultural activities. The urban poor benefit through higher incomes for unskilled labor also in non-agricultural sectors, which induces a pro-poor shift in the urban income distribution. In addition, the urban poor benefit indirectly from the gains in agriculture, as the pressure on non-agricultural unskilled is relieved somewhat. Trade reform, and in particular domestic trade reforms, may particularly help the Brazilian poor farmers, but only broad-based high growth will eradicate urban poverty. Whether trade can do the job of significantly raising the incomes of the urban poor is questionable. In this regard an important limitation of our analysis is that we do not assume any dynamic gains from trade liberalization. Our results might hence be taken as a lower bound of the welfare effects, as there is strong evidence of a beneficial impact of trade liberalization on productivity (Winters, McCulloch, and McKay 2004). We also acknowledge that our representation of urban labor 132 BRAZIL markets may be too simple to evaluate the precise effects on some particularly poor groups in urban areas, e.g. in informal activities. The trade policy implications for Brazil are clear-cut. As Brazil does not lose from the possible Doha scenarios and even slightly gains, there is no reason to oppose such an outcome of the negotiations. Furthermore, our analysis suggests that Brazil and especially the Brazilian poor can substantially gain from own liberalization. An obvious complementary policy to a trade policy that favors the agricultural sector is to enable poor households to participate in agricultural growth. In the Brazilian context, an important means to do so is to provide access to land. APPENDICES 4.6. Appendices 4.6.1. Additional tables App. Table 4.1: Estimation results, mover-stayer mode/for heads Number of obs Wald chi2(6) Prob> chi2 Pseudo R2 Log pseudo-likelihood= -1495.442 14365 197.120 0.000 0.068 Robust Std. 133 Coef Err. z P>lzl [95% Con[ Interval] dedu 0.419 0.278 1.510 age -0.044 0.005 -9.490 ddworkown -1.123 0.319 -3.520 dcess -0.825 0.327 -2.530 dprop -0. 736 0.172 -4.280 dgregiol 0.593 0.202 2.930 cons -1. 708 0.178 -9.580 - Changes in Predicted Probabilities for Moving out of Agriculture 0->l dedu 0.008 age ddworkown -0.012 dcess -0.009 dprop -0.010 dgregiol 0.012 Source: Authors' calculations. Note: -+sd/2 Mars.£.lft 0.007 -0.01 I -0.001 -0.018 -0.013 -0.012 0.009 0.132 0.000 0.000 0.012 0000 0.003 0.000 dedu: ddworkown: dcess: dprop: dgregiol: educational dummy for IO or more years of schooling dummy for own-consumption workers holding ceded land holding own land north. -0.126 0.963 -0.053 -0.035 -1.749 -0.497 -1.466 -0.185 -1.074 -0.399 0.196 0.990 -2.058 -1.359 134 BRAZIL App. Table 4.2: Estimation results, mover-stayer model/or non-heads Number of obs 16737 Wald chi2(6) 293.58 Prob> chi2 0.000 Pseudo R2 0.152 Log pseudo-likelihood= -1282.4805 Robust Std. Coef Err. z P>[z[ dprim3 0.782 0.229 3.410 0.001 dsecl 0.633 0.298 2.120 0.034 dsec2 0.597 0.324 1.840 0.066 exp 0.101 0.023 4.340 0.000 exp2 -0.002 0.000 -4.270 0.000 gend -0.759 0.188 -4.030 0.000 preta -0.500 0.336 -1.490 0.137 ddnonrem -0.849 0.171 -4.960 0.000 ddworkown -1.731 0.307 -5.640 0.000 dprop -1.839 0.462 -3.980 0.000 headnagr 1.122 0.167 6.710 0.000 headmover 2.468 0.374 6.610 0.000 cons -4.369 0.298 -14.640 0.000 - Changes in Predicted Probabilities for Moving out of Agriculture 0->/ -+sd/2 dprim3 0.010 dsecl 0.007 dsec2 0.007 exp 0.017 exp2 -0.024 gend -0.007 preta -0004 ddnonrem -0.007 ddworkown -0.011 dprop -0.008 headnagr 0.015 headmover 0.084 Source: Authors' calculations. Note: dprim3: dsecl: dsec2: exp: exp2: gend: preta: 9 years of schooling 11 or 11 years of schooling 12 years of schooling age minus schooling experience squared female black Mars_E(ft 0.007 0.006 0.005 0.001 0.000 -0.007 -0 004 -0.007 -0.015 -0.016 0.010 0.021 ddnonrem: ddworkown: dprop: headnagr: headmover: {95% Con[ Interval} 0.332 1.232 0.049 1.218 -0.039 1.232 0.055 0.146 -0.003 -0.001 -1.127 -0.390 -1.159 0.159 -1.185 -0.513 -2.332 -1.130 -2.745 -0.934 0.794 1.450 1.736 3.200 -4.954 -3.783 non-remunerated household member own-consumption worker holding own land household head in nonagricultural sector head has moved out of agriculture. 5. Conclusions, policy relevance, and future research The main message of the country studies included in this dissertation is that the poverty and distributional impact of external shocks and economic policies depends very much on the exact nature of the shock as well as the structural characteristics of the country in question. Hence, there are no policy blueprints. This may sound trivial, but often enough have policy prescription been based on oversimplifying assumptions without taking into account country-specificities. One such case is the belief that the poor in developing countries in general would benefit from trade liberalization through increased demand for unskilled labor.'°' The Colombian and the Brazilian case studies very well illustrate that the impact of trade liberalization on poverty and the distribution of income depends on the structure of protection in place and how it is modified, i.e. the nature of shock, as well as a number of country characteristics, in particular the functioning of the labor market and the sectoral and skill composition of the workforce. In sum, the studies demonstrate that country-specific empirical research can provide policymakers with insightful analyses to take better-informed decisions. Of course, three country-studies cannot provide the empirical basis, on which to judge whether globalization, or some of its many facets, are good or bad for the poor. The analyses may however suffice for the general tentative conclusion that globalization, at least in the Latin American context, is neither good nor bad; rather it entails threats and opportunities. In addition, what appears to be a threat at first sight may actually be seen as an opportunity, given the right policy. Trade liberalization in Colombia, for example, has increased inequality through a rising wage gap between the skilled and the unskilled. If educational policies however allow labor supply to adjust to this increase, productivity gains would be distributed more equally. The case for better policies to make a difference is even stronger in natural resource abundant countries, in particular during boom phases, as illustrated by the Bolivian study. In the latter, current public expenditure policies were even found to aggravate negative side-effects of the resource boom. A worrying tendency, to which there is no obvious policy response, is the increase in informal employment that is identified in both the Bolivian and the Colombian case. If the informal sector involves negative externalities, for example in terms of human capital accumulation, this increase can harm long-term development prospects. A fairer world trading system is been by many as an important component of a comprehensive development strategy, as reflected by the adopting the attainment of an "open trading and financial system that is rule-based, predictable and nondiscriminatory" committed to "good governance, development and poverty reduction as one of the Millennium Development targets. Even non-governmental I 05 The structural adjustment programs of the 1980s and early 1990s are another case of fairly similar policy packages applied to a number of countries, whose problems only appeared to be the same at first sight. For a critical view on these programs see Collier and Gunning (1999). 142 CONCLUSIONS organizations seem to put great hopes in particular into the effects of cutting-down agricultural subsidies in rich countries. For Brazil however, the preceding chapter suggests that the gains for the poor from a new development-and-poverty-focused round of multilateral trade negotiations would be rather limited. This finding is not limited to the Brazilian case. The other country studies included in Hertel and Winters (2005) also point to moderate effects of a "Doha Development Agenda". In some countries, poverty even increases slightly. Among the reasons why liberalizing agricultural trade does not help the poor as much as expected is that, in some countries, higher world market prices are not transmitted to poor farmers, while urban households suffer from the price increase. 106 Such findings again demonstrate that the poverty and distributional impact of economic policies depends on a whole range of country-specific factors and that assessing these effects requires very disaggregated analyses. In terms of methodology, the chapters demonstrate that the "two-step" or sequential approach provides an appropriate framework to link policies or external shocks to poverty and distributional outcomes. This approach first analyzes the impact of shocks on "distributional drivers", such as changes in prices and factor remunerations, as well as employment shifts between different types of activities or sectors. Using a CGE model in this first stage, allows for a detailed analysis of the transmission channels of the shocks at the "macro" level. In the second step, the final distributional outcome is assessed using a microsimulation model that takes into account the complexities of the income generation process through modelling individual decisions. The sequential approach brings together two strands of literature, applied CGE models, on the one hand, and poverty and distributional analyses, on the other, which were largely separated from each other. While CGE analyses tend to suffer from being too stylized and not being well informed by micro data, poverty and distributional analyses are often merely descriptive and lack an assessment of the causes of distributional change and the related transmission channels. The sequential approach attempts to get the best out of these two "modeling worlds". Its main advantage is that while it remains tractable both at the macro and the micro level, it allows for sufficiently detailed and disaggregated analyses. The microsimulation models based on household income generation models provide a powerful tool to assess the final distributional impact of changes in "distributional drivers", as illustrated by the validation exercise in the Colombian chapter. Modeling decisions at the individual level implies that household heterogeneity is not only represented in terms of factor endowments and consumption patterns. The welfare implications of discrete changes in individual behavior, such as labor market entry or sectoral movements, can thus be taken into account. The impact of individual transitions out of agriculture in the Brazil study demonstrates the possible magnitude of these discrete individual changes on 106 This is shown by Nicita (2005) for Mexico as well as Arndt et al. (2005) for Mozambique. CONCLUSIONS 143 household welfare. Here lies the major advantage of such microsimulation models vis-a-vis traditional CGE analyses based on representative household groups. The shortcomings of the income generation models are discussed at length in each of the chapters and only one major problem should be reconsidered here. It concerns the simulation of occupational transition based on state comparisons. The income generation models for Colombia and Bolivia implicitly assume that the estimated propensity to have a certain occupational status, i.e. to be inactive, or to be employed in the informal/formal or agricultural/non-agricultural sector, is closely related to the propensity to change occupational status. Whether this assumption holds true is an empirical question, which could (and possibly should) be addressed using panel datasets. The Brazilian model relies on employment histories and therefore avoids this problem, but the type of information used reflects to a certain extent short-term behavior. The Brazilian and the Bolivian chapter use CGE models to trace the transmission channels and quantify the magnitude of the effects of the respective shock. Although widely applied, these models have been criticized for a number of reasons. Analytically, most CGE models rely on the neoclassical framework, although the influence of the "structuralist" school (Taylor 1990) has led to the incorporation of a number of structural characteristics and rigidities in most developing country applications. Which structural characteristics to consider and how to precisely model rigidities, e.g. on factor markets, hence differs between country applications and the research question at hand. The models in this dissertation make an attempt to capture in a realistic way some country characteristics that are key for understanding the transmission of the respective shock. This includes for example the modeling of the gas sector and the related investment flows in the Bolivian model as well as the labor market segmentation in the Brazilian model. Clearly, the two models have their shortcomings. Assuming, for example, neoclassical price setting in the case of traditional agriculture in the Bolivian model, is at best a very rough approximation of reality. In fact, modeling of the rural sector is unsatisfactory in most applied CGE models. Disaggregated input-output data for agriculture are typically not available and agricultural surveys suffer from a lot of problems related to measurement, seasonality, and temporary shocks. In addition to data gaps, the insights from agricultural household models regarding non-separability of production and consumption decisions in rural households (Singh et al. 1986) have not yet entered standard models.107 More research effort also needs to be dedicated to modeling the informal urban sector. Its heterogeneity in terms of technology, import penetration, and export orientation needs to be addressed. This implies to incorporate the knowledge on the linkages between formal and informal activities into applied 107 See Lofgren and Robinson ( 1999), who integrate a rural household model into a standard CGE model, for an exception. 144 CONCLUSIONS models. '0' However, even with all these improvements, eventually the results of a CGE model will be driven by the assumptions made."" Econometricians challenge the empirical relevance of applied CGE models on grounds of the calibration technique based on very restricted functional forms, typically (nested) CES functions. McKitrick (1998) shows the choice of the functional form to make a considerable difference in the results. Yet, in the developing country context, data to estimate these functions is typically not available and the calibration approach overcomes these data restrictions. Furthermore, it is well known that model results are very sensitive to the assumed trade and production elasticities. Harrison et al. (1993) therefore suggest to perform systematic sensitivity analyses and to provide confidence intervals for the results. The CGE analyses in this dissertation, as most CGE model applications, perhaps do too little sensitivity analysis and rely too much on parameters "typically assumed in the literature". Yet, an assessment of the validity of CGE model results also depends on the purpose of the model. If the analysis is expected to provide a precise numerical estimate of the effects of a specific policy change, the above criticisms have to be taken very seriously. In contrast, if CGE models are seen as a rather stylized, yet empirically underpinned, analytical tool to better understand the transmission channels of a shock through counterfactual analysis and approximate their relative importance, the critique is less relevant. In this dissertation, CGE models are considered such a tool. This is not to say that the numbers resulting from CGE models are without meaning. They should be taken as the results of a model, given a specific set of assumptions. Claiming that CGE modeling experiments would yield "real world forecasts" appears to be exaggerated.110 Instead of using a CGE model, the Colombian study relies on secondary sources and additional descriptive analyses to construct the counterfactual scenarios of "distributional drivers" that can be linked to trade reform. 111 Linking ex-post econometric studies on the impact of policies or shocks on "distributional 108 See Grimm and Giinther (2006) for a recent study on formal-informal linkages in Burkina Faso and a short literature review. I 09 See de Maio et el. ( 1999) and the reply by Sahn et al. ( 1999) for an exemplary discussion on specific aspects of CGE models applied to developing countries. These aspects include the macroeconomic and labor market closures as well as the assumption on price setting mechanisms. De Maio et al. ( 1999) challenge the results of a study by Sahn et al. ( 1997) on the poverty impacts of structural adjustment in Sub-Saharan Africa as reflecting only the assumptions made in the CGE models, and not reality. 110 Admittedly, the CGE applications in this dissertation, in particular the Brazilian chapter, sometimes tend to treat the numbers as being "forecasts". l l l A similar approach is followed by Nicita (2005) who uses his results on price transmission of changes in world market prices in Mexico to simulate the possible effects of trade liberalization on household welfare based on a simple income generation model. CONCLUSIONS 145 drivers", such wages or prices, to household income generation models seems to be a promising approach for future research. Quite some studies have examined e.g. the imr,act of trade liberalization on the distribution of wages and employment. 12 The micro analyses in this dissertation have demonstrated that looking at the impact on wages and employment alone does not say much about final distributional outcomes. Increasing female labor market participation provides an example for the complex relationships between employment and wages, on the one hand, and the changes in the distribution of per capita incomes, on the other. Labor market entry of females from poorer households typically worsens the distribution of wages, but may improve welfare of those poorer households considerably. Of course, identifying the effects of a specific shock expost is not trivial and requires data with the variation across sectors, regions, and/or time that allows for doing so. The approach of the Colombian chapter basically combines two "reducedform" models; a reduced-form model that links trade and labor market outcomes, and a second one that links labor market outcomes and household income. Therefore the exact pathways through which the shocks affect distributional outcomes remain unclear and counterfactual experiments cannot be conducted in such a framework. This would be possible in a general equilibrium model that incorporates heterogeneous individuals. As argued in the introduction, building such an applied model based on a "full-blown, micro-based general equilibrium theory of income distribution and income inequality" 111 does not seem to be feasible for developing countries. Eventually, the appropriate methodology will depend on the shock or policy to be examined and the data available. Data availability and quality is of central importance to measuring progress towards the Millennium Development goals and to the type of analysis conducted in this dissertation. Applying macro and micro simulation models implies working with different types of data sources including national accounts and primary surveys. The experiences gathered during this research hints at large systematic discrepancies between these different sources. As building Social Accounting Matrices for distributional analysis requires the use of household survey data, applied CGE modelers have also noted the considerable inconsistencies between the two data sources (Round 2003, Robilliard and Robinson 2003). The remainder of this conclusion therefore argues in favor of taking data issues much more seriously and putting them at the heart of the research agenda in development economics. Deaton (2005) illustrates the scope of the problem. Consumption estimated from surveys is typically lower than consumption from the national accounts by approximately 20 percent, with regional differences. Survey income is on average less than 60 percent of GDP. More worrying than these static comparisons are the differences in growth rates. According to the surveys, average annual real per 112 See Arbache et al. (2004) and Winters et al. (2004) for reviews. 113 Quote from James Heckman on the back cover ofBourguignon et al. (2005). 146 CONCLUSIONS capita consumption growth in the 1990s has been 2.3 percent if the simple average is computed and 1.9 percent if log growth rates are regressed on a time trend. National accounts give growth rates of 3.8 and 4.5 percent, respectively (Deaton 2005). Even if the distribution remains unchanged, consumption growth as measured by the surveys would be too sluggish to make a dent in poverty reduction, while growth measured by national accounts would reduce poverty substantially. Deaton (2005) provides a discussion of the reasons behind the discrepancies and points to differences that result from differences in definitions, e.g. regarding items to be included into consumption, and differences in meeting those definitions, e.g. in measuring production."' National accounts are known to capture production for own consumption, which constitutes an important share of production in poor countries, only to a limited extent. In general, national accounts, in contrast to surveys, are more likely to capture larger transactions than smaller ones (Deaton 2005). As these small transactions are those reflecting the living standards of the poor, Deaton (2005) concludes that poverty can only be measured using household surveys. However, understanding the relationship between growth, inequality and poverty will require a reconciliation of macro and micro data. In light of these findings, the discipline dedicates astonishingly little effort to data issues. The main problem of basing empirical research on these obviously flawed datasets lies in the fact that the reasons for these differences are likely by to systematic. Yet, as biases are systematic, they can and should be investigated. The described data inconsistencies are not a technical question of interest to insiders, but a key problem when progress towards the Millennium Development Goals needs to be measured: Where is the point of setting quantitative development targets ifwe lack the means to measure them? 114 See also Robilliard and Robinson (2003) and Ravallion (2001b). 6. References Adelman, I. and C.T. Morris ( 1973). Economic Growth and Social Equity in Developing Countries. Stanford: Stanford University Press. Adelman I. and S. Robinson (1978). Income Distribution Policy in Developing Countries: A Case Study of Korea. New York: Oxford University Press. Adelman, I. and S. Robinson (1989). Income Distribution and Development. In: Chenery, H. and T.N. Srinivasan (eds.). Handbook of Development Economics Volume II. Amsterdam: North Holland, pp. 949-1004. Aghion, P., E. Caroli, and C. Garcia-Penalosa. (1999). Inequality and Economic Growth: The Perspective of the New Growth Theories. Journal of Economic Literature, 37 (4), pp. 1615-1660. Alatas, V. and F. Bourguignon (2005). The Evolution of Income Distribution during Indonesia's fast Growth, 1980-96. In: Bourguignon, F., F.H.G. Ferreira, and N. Lustig (eds.). The Microeconomics of Income Distribution Dynamics. Washington, D.C.: The World Bank and Oxford University Press. Anand, S. (1983). Inequality and Poverty in Malaysia. Oxford: Oxford University Press. Arndt, C. (2005). The Doha Round and Mozambique. In: Hertel, T. and L. A. Winters (2005). Poverty and the WTO. Impacts of the Doha Development Agenda. Washington D.C.: Palgrave and The World Bank. Arbache, J.S., A. Dickerson, and F. Green (2004). Trade Liberalisation and Wages in Developing Countries. Economic Journal, 114 (493), pp. F73-96. Armington, P.S. (1969). A Theory of Demand for Products Distinguished by Place of Production. IMF Staff Papers, 16, pp. 159-178. Atkinson, A. and F. Bourguignon (2000). Income Distribution and Economics. In Atkinson, A. and F. Bourguignon, (eds.). Handbook of Income Distribution, Vol. 1, Amsterdam: North-Holland, pp. 1-58. Attanasio, 0., P. Goldberg, and N. Pavcnik (2004). Trade Reforms and Wage Inequality in Colombia. Journal of Development Economics, 74 (2), pp. 331-366. Auty, R.M. (ed.) (2001). Resource Abundance and Economic Development. Oxford: Oxford University Press. Auty, R.M. and J.L. Evia (2001). A Growth Collapse with Point Resources: Bolivia. In: Auty, R.M. (ed.). Resource Abundance and Economic Development. Oxford: Oxford University Press. Barrett, C.B. (2004). Rural Poverty Dynamics: Development Policy Implications. Paper prepared for invited presentation at the 25 th International Conference of Agricultural Economists, Durban, South Africa. Blinder, A.S. (1973). Wage Discrimination: Reduced Form and Structural Estimates. Journal of Human Resources 8 (4), pp. 436-455. 148 REFERENCES Bourguignon, F., F.H.G. Ferreira, and P.G. Leite (2002). Ex-ante Evaluation of Conditional Cash Transfer Programs: The Case of Bolsa Escola. World Bank Policy Research Working Paper No. 2916. Bourguignon, F. and F.H.G. Ferreira (2005). Decomposing Changes in the Distribution of Household Incomes: Methodological Aspects. In: Bourguignon, F., F.H.G. Ferreira, N. Lustig (eds.). The Microeconomics of Income Distribution Dynamics. Washington, D.C.: The World Bank and Oxford University Press. Bourguignon F., M. Fournier, and M. Gurgand (1998). Wage and Labor Force Participation Behavior in Taiwan, 1979-1994. Mimeo. Bourguignon F., M. Fournier, and M. Gurgand (2001). Fast Development with a Stable Income Distribution:Taiwan, 1979-1994. Review of Income and Wealth, 47 (2), pp. 139-163. Bourguignon, F., A.-S. Robilliard, and S. Robinson (2005). Representative vs. Real Households in the Macroeconomic Modeling oflnequality. In: Kehoe, T.J., T.N. Srinivasan, and J. Whalley (eds.). Frontiers in Applied General Equilibrium Modeling. Cambridge: Cambridge University Press. Bourguignon, F., F.H.G. Ferreira, and N. Lustig (eds.) (2005a). The Microeconomics of Income Distribution Dynamics. Washington, D.C.: The World Bank and Oxford University Press. Bourguignon, F., F.H.G. Ferreira, and N. Lustig (2005b). A Synthesis of the Results. In: F. Bourguignon, F.H.G. Ferreira, and N. Lustig (eds.) (2005). The Microeconomics of Income Distribution Dynamics. Washington, D.C.: The World Bank and Oxford University Press. Bourguignon, F., J. de Melo, J. and C. Morrisson (1991) Poverty and Income Distribution During Adjustment. World Development, 19 (11), pp. 1485-1508. Bussolo, M., and J. Lay (2005). Globalisation and Poverty Changes in Colombia. In: Bussolo, M. and J. Round (eds.). Globalisation and Poverty -Channels and Policy Responses. London: Routledge. Chenery, H., M. Ahluwalia, C. Bell, J. Duloy, and R. Jolly (1974). Redistribution with Growth. Oxford: Oxford University Press. Christiaensen, L., L. Demery, and S. Patemostro (2003). Macro and Micro Perspectives of Growth and Poverty in Africa. World Bank Economic Review 17 (3): 317-347. Cline, W.R. (1975). Distribution and Development. A Survey of the Literature. Journal of Development Economics, I (4), pp. 359-400. Cockburn, J. (2006). Trade liberalisation and poverty in Nepal : a computable general equilibrium micro-simulation analysis In: Bussolo, M. and J.R. Round (eds.). Globalisation and Poverty: Channels and Policy Responses. London: Routledge., pp. 171-194. Cogneau, D. (2001). Formation du revenue, segmentation et discrimination sur le marche du travail d'une ville en developpement: Antananarivo fin de siecle. DIAL DT/2001/18. REFERENCES 149 Cogneau, D. and A.S. Robilliard (200 I). Growth, Distribution and Poverty in Madagascar: Learning from a Microsimulation Model in a General Equilibrium Framework. TMD Discussion Paper 61. Washington, D.C.: IFPRI. Collier, P. and W. Gunning (1999). The IMF'S Role in Structural Adjustment. The Economic Journal, I 09 ( 459), pp. 634-651. Collier P. and J.W. Gunning (1999a). Trade Shocks in Developing Countries. Vol. 1: Africa. Oxford: Oxford University Press. Collier P. and J.W. Gunning (1999b). Trade Shocks in Developing Countries. Vol. 2: Asia and Latin America. Oxford: Oxford University Press. Comia, G., R. Jolly, F. Stewart (1987). Adjustment with a Human Face. Oxford: Oxford University Press. De Maio, L., F. Stewart, and R. van der Hoeven (1999). Computable general equilibrium models, adjustment, and the poor in Africa. World Development, 27 (3), pp. 453- 470 Deaton, A. ( 1997). The Analysis of Household Surveys. A Microeconometric Approach to Development Policy. Washington D.C.: The World Bank and John Hopkins University Press. Deaton, A. (2004). Measuring Poverty in a Growing World (or Measuring Growth in a Poor World). Review of Economics and Statistics, 87 (1), pp. l-19. Dervis, K., J. de Melo, and S. Robinson (l 982). General Equilibrium Models for Development Policy. Cambridge: Cambridge University Press. Dias, G.L.S., C.M. Amaral (2002). Structural Change in Brazilian Agriculture, 1980-98. In: Baumann, R. (2002). Brazil in the 1990s -An Economy in Transition. Houndsmill: Palgrave. Dollar, D. and A. Kraay (2002). Growth is Good for the Poor. Journal of Economic Growth, 7, pp. 195-225. Edwards, S. (1998). Openness, Productivity, and Growth. What do we really know? Economic Journal, 108 (447), pp. 383-398. Elbers, C., J.O. Lanjouw, P. Lanjouw, and P.G. Leite (2004). Poverty and Inequality in Brazil: New Estimates from Combined PPV-PNAD Data. The Word Bank. Mimeo. Escobar, F. and 0. Nina (2004). Pension Reform in Bolivia: A Review of Approach and Experience. Study No. GJ-E6. La Paz: Grupo Integral. ESMAP (2002). Estudio Comparativo sobre la Distribuci6n de la Renta Petrolera £studio de Casos: Bolivia, Colombia, Ecuador y Peru. Mimeo. Falvey, R. and C.D. Kim (1992). Timing and Sequencing Issues in Trade Liberalization. Economic Journal, 102 (413), pp. 908-924. Feenstra, C. and G.H. Hanson (1995). Foreign Direct Investment and Relative Wages: Evidence from Mexico's Maquiladoras. NBER Working Paper No. W5122. Ferreira, F.H.G. and J. Litchfield (1999). Calm After the Storms: Income Distribution and Welfare in Chile, 1987-1994. World Bank Economic Review, l3 (3), pp. 509-538. 150 REFERENCES Ferreira, H.G.F. and R. Paes de Barros (2005). The Slippery Slope: Explaining the Increase in Extreme Poverty in Urban Brazil, 1976-96. In: F. Bourguignon, F.H.G. Ferreira, and N. Lustig (eds.). The Microeconomics of Income Distribution Dynamics. Wasington, D.C.: The World Bank and Oxford University Press. Ferreira, H.G., P. Lanjouw, and M. Neri (2001). A Robust Poverty Profile for Brazil Using Multiple Data Sources. Paper Presented at the LACEA 2000 and ANPEC 200 I meetings. Fields, G. ( 1989). Changes in Poverty and Inequality in Developing Countries. World Bank Research Observer, 4 (2), pp. 167-184. Fiess, N.M. and D. Verner (2003). Migration and Human Capital in Brazil During the I 990s. World Bank Policy Research Working Paper No. 3093. Gelb, A.H. (1988). Oil Windfalls: Blessing or Curse? New York: Oxford University Press. Golan, A., G. Judge, and D. Miller (1996). Maximum Entropy Econometrics. Robust Estimation with Limited Data. Chichester: John Wiley and Sons. Goldberg, P. and N. Pavcnik (2003). The Response of the Informal Sector to Trade Liberalization. Journal of Development Economics, 72 (2), pp. 463-496. Goldberg, P. and N. Pavcnik (2004a). The Effects of the Colombian Trade Liberalization on Urban Poverty. Paper prepared for the NBER Globalization and Poverty Conference, Cape Cod. Goldberg, P. and N. Pavcnik (2004b). Trade, Inequality, and Poverty: What Do We Know? Evidence from Recent Trade Liberalization Episodes in Developing Countries. NBER Working Paper No. I 0593. Goldberg, P. and N. Pavcnik (2005). Trade, Wages, and the Political Economy of Trade Protection: Evidence from the Colombian Trade Reforms. Journal of International Economics, 66 (I), pp. 75-105. Grimm, M. (2005), Educational policies and poverty reduction in Cote d'Ivoire. Journal of Policy Modeling, 27, pp. 231-247. Grimm, M. and I. Giinther (2006), Inter- and intra-household linkages between the informal and formal sector. A case study for urban Burkina Faso. In: Guha- Khasnobis, B. and R. Kanbur (eds.). Informal Labor Markets and Development, London: Palgrave Macmillan. Grimm, M., S. Klasen, and A. McKay (2006). Determinants of Pro-Poor Growth: Analytical Issues and Findings from Country Cases. London: Palgrave Macmillan, forthcoming. Harrison, A. (2005). Globalization and Poverty: An NBER Study. University of California at Berkeley and NBER (Draft). Harrison, G.W., R. Jones, L. J. Kimbell, R. Wigle (1993). How robust is applied general equilibrium analysis? Journal of Policy Modeling, 15 (I), pp. 99-115 Harrison, G., Rutherford, T., and Tarr, D. (2000) Trade liberalization, poverty and efficient equity. Journal of Development Economics (2003) 71, pp. 97-128.