“Operationalising Pro- Poor Growth” A joint initiative of AFD, BMZ (GTZ, KfW Development Bank), DFID, and the World Bank A Country Case Study on Bolivia Stephan Klasen, Melanie Grosse, Rainer Thiele, Jann Lay, Julius Spatz, Manfred Wiebelt October 2004 This paper belongs to a series of 14 country case studies spanning Africa, Asia, Latin America and Eastern Europe. The series is part of the ‘Operationalising Pro-Poor Growth (OPPG)’ work programme, a joint initiative of AFD, BMZ (GTZ, KfW Development Bank), DFID and the World Bank. The OPPG work programme aims to provide better advice to governments on policies that facilitate the participation of poor people in the growth process. Other outputs of the OPPG initiative include a joint synthesis report, a note on methodological approaches to analysing the distributional impact of growth, cross-country econometric work, literature reviews, and six synthesis papers on: macroeconomics and structural policies, institutions, labour markets, agriculture and rural development, pro-poor spending, and gender. The country case studies and synthesis papers will be disseminated in 2005. The entire set of country case studies can be found on the websites of the participating organisations: BMZ www.bmz.de , DFID www.dfid.gov.uk , GTZ www.gtz.de , KfW Development Bank www.kfw- entwicklungsbank.de/EN/Fachinformationen and the World Bank www.worldbank.org . For further information, please contact: AFD: Jacky Amprou [email protected]BMZ: Birgit Pickel [email protected]DFID: Manu Manthri [email protected]and Christian Rogg [email protected]GTZ: Hartmut Janus [email protected]KfW Development Bank: Annette Langhammer [email protected]World Bank: Louise Cord [email protected]and Ignacio Fiestas [email protected]
Transcript
“Operationalising Pro- Poor Growth”
A joint initiative of AFD, BMZ (GTZ, KfW Development Bank), DFID, and the World Bank
A Country Case Study on Bolivia
Stephan Klasen, Melanie Grosse, Rainer Thiele, Jann Lay, Julius Spatz, Manfred Wiebelt
October 2004
This paper belongs to a series of 14 country case studies spanning Africa, Asia, Latin America and Eastern Europe. The series is part of the ‘Operationalising Pro-Poor Growth (OPPG)’ work programme, a joint initiative of AFD, BMZ (GTZ, KfW Development Bank), DFID and the World Bank. The OPPG work programme aims to provide better advice to governments on policies that facilitate the participation of poor people in the growth process. Other outputs of the OPPG initiative include a joint synthesis report, a note on methodological approaches to analysing the distributional impact of growth, cross-country econometric work, literature reviews, and six synthesis papers on: macroeconomics and structural policies, institutions, labour markets, agriculture and rural development, pro-poor spending, and gender. The country case studies and synthesis papers will be disseminated in 2005. The entire set of country case studies can be found on the websites of the participating organisations: BMZ www.bmz.de, DFID www.dfid.gov.uk, GTZ www.gtz.de, KfW Development Bank www.kfw-entwicklungsbank.de/EN/Fachinformationen and the World Bank www.worldbank.org. For further information, please contact: AFD: Jacky Amprou [email protected] BMZ: Birgit Pickel [email protected] DFID: Manu Manthri [email protected] and Christian Rogg [email protected] GTZ: Hartmut Janus [email protected] KfW Development Bank: Annette Langhammer [email protected] World Bank: Louise Cord [email protected] and Ignacio Fiestas [email protected]
Manfred Wiebelt Kiel Institute for World Economics
Operationalizing Pro-Poor Growth
Country Case Study: Bolivia
Final Report, September 28, 2004
Table of Contents
Executive Summary i
Chapter 1: Historical Context 1
Chapter 2: Analysis of Growth and Its Distributional and Poverty Impact 9
Chapter 3: Factors Affecting the Participation of the Poor in Growth 25
Chapter 4: Possible Trade-Offs between Growth and Poverty Reduction 44
Chapter 5: Recommendations for Policy-Making 46
References 50
Acknowledgements
We would like to thank Juan-Carlos Aguilar and Stefan Zeeb for generous support during two visits to Bolivia as well as for providing valuable inputs, comments, and documentation. We also want to thank Annette Langhammer and Louise Cord for valuable comments throughout the drafting of this document. In addition, we like to thank Berk Ozler, Omar Arias, Fernando Landa, Wilson Jimenez, Sara Calvo, participants and discussants at workshops at the World Bank, in Frankfurt, and in La Paz for valuable comments and discussion. Funding from the German Federal Ministry for Economic Cooperation and Development via the KfW Entwicklungsbank (KfW Development Bank) is gratefully acknowledged.
i
Executive Summary
Introduction 1. This case study examines to what extent Bolivia has been able to achieve pro-poor growth, what the mechanisms of achieving (or failing to achieve) pro-poor growth have been, and what options are available to ensure higher rates of pro-poor growth. The analysis focuses on the period from 1989 to 2003, which spans a time of relatively high growth in the 1990s, and low growth with social and political turmoil in the past few years. In contrast, there have been notable and sustained improvements in social indicators which continued to improve despite the economic slowdown.
2. Bolivia, a landlocked country with poorly developed infrastructure and a very uneven population distribution, has had a legacy of high economic and social inequality with a strong ethnic dimension. The political system has always been dominated by an urban-based elite and has only recently opened to serious indigenous representation. After a disastrous bout of hyperinflation, Bolivia embarked on a path of structural reforms in the late 1980s, which brought stability and fairly high growth throughout most of the 1990s. Growth decelerated since as a result of external shocks, which reversed some of the gains made in the previous decade. A large share of the population is dependent on subsistence agriculture and informal activities (some illegal including the production of coca leaves), with a small modern agricultural sector, a small formal sector, and a capital-intensive natural resource sector, which generates a large share of export earnings.
Poverty Trends, Profiles, and Pro-Poor Growth 3. As there are no national poverty data before 1997, we have created a new time series of poverty data from 1989 to 2003 by linking information from urban household surveys with nationally representative Demographic and Health Surveys. The new time series, which is robust to different sensitivity analyses, indicates large differentials in poverty between urban and rural areas. In addition, poverty rates in urban areas responded rapidly to economic opportunities (and the recent slowdown), while poverty in rural areas followed its own dynamic. The extent of poverty reduction in rural areas was moderate, did not affect the headcount ratio much and is partly sensitive to the assumption made in the data matching exercise. Using the Ravallion-Chen measure of pro-poor growth, we find that growth was pro-poor but relatively low throughout the 1990s, but became sharply anti-poor in urban areas since then. In rural areas, growth was slower, but generally more pro-poor. Due to the recent slowdown, pro-poor growth over the entire 1989-2002 period was too slow to lead to significant poverty reduction. A decomposition of poverty reduction shows that about 2/3 of poverty reduction was due to income growth with the remaining share being allocated to a redistribution component which, however, also includes the effect of favorable price shifts for the goods consumed by the poor.
4. A poverty profile shows considerable regional inequality, with the central highland and valley provinces being affected by much higher poverty, compared to the outlying valley and lowland provinces. The most important correlates of poverty are, apart from the urban/rural divide, ethnic background and education. There is comparatively little gender bias in education (but serious gender gaps persist elsewhere in the formal economy and in the home).
5. We link the record of pro-poor growth to the sectoral composition of growth and find that urban incomes were closely tied to macroeconomic developments, while rural incomes were more dependent on weather conditions and the coca economy. Consistent with the poverty profile, we also find that Bolivia is a highly segmented society with relatively sharp segmentations along a formal-informal divide, a rural-urban divide, and an ethnic divide. The formal-informal divide is related, among other things, to tight labor market regulation in the
ii
urban formal market, poor credit access for informal producers and other barriers to formalization, relatively little opportunities for migrant workers to gain entry into the formal economy, and the small inherent size of the formal sector. The urban-rural and the ethnic divide are closely related and are partly a legacy of strong discrimination against the indigenous population, little success in modernizing highland agriculture, and little success in generating an income base in rural areas of the highlands and central valleys beyond the coca economy.
Initial Conditions, Policies, and Pro-poor Growth 6. Initial conditions were unfavorable for linking the poor to the growth process. Among them are an uneven population distribution, high initial inequalities (of land, other assets, human capital, and incomes), and comparative advantages in highly capital-intensive agricultural and resource extraction activities. Moreover, poor governance and the divisive and strife-torn political economy of Bolivia have made stable economic policy-making difficult.
7. Bolivia’s macro policies were narrowly focused on stability, liberalization, and growth with little direct concern for distributional issues. Such a policy stance was feasible as long as the policy environment produced stable growth and some poverty reduction. In the current slowdown, which is largely caused by events beyond Bolivia’s control but amplified by its liberalized economy, the legitimacy of this economic model has been seriously questioned.
8. The tax system is not progressive and the expenditure system generally reaches the poor but is not particularly well targeted. Despite this, a rapid expansion of social sector spending beginning in the mid-1990s, aided by funds freed from the HIPC II debt reduction initiative, has contributed to rapid improvements in health and education indicators (from a relatively low level). Unfortunately, the sustainability of this expansion is highly doubtful given the economic slowdown, the associated decline in tax revenues, and the emergence of huge budget deficits. Buying support for economic reforms through an expansion in social sector spending does not seem to be feasible anymore.
9. Using a dynamic CGE model we then assess the impact of shocks and policies on pro-poor growth, both to account for the developments of the past and to investigate policy options for the future. In an optimistic baseline scenario, Bolivia could achieve a sustainable 4.7% rate of growth per year with moderate poverty reduction, but a widening urban-rural gap. External shocks such as terms-of-trade shocks, El Niño, and declining capital inflows all served to lower economic growth in the latter half of the 1990s and contributed to rising poverty. Given Bolivia’s high degree of dollarization and its dependence on foreign capital, exchange rate and monetary policies can do little to cushion the blow from external shocks.
10. As far as forward-looking policies are concerned, expansion of natural gas exports will boost growth and reduce urban poverty somewhat, but will lead to rising inequality and rising rural poverty. Labor market and tax reforms have the potential to increase growth and urban poverty reduction, with relatively little impact on rural poverty. The combination of gas exports and labor market and tax reforms would yield the highest outcome in terms of economic growth. If they were combined with transfer programs targeted at the rural poor, they would also lead to significant poverty reduction there. Other targeted interventions in favor of the poor such as improvements in credit access, agricultural technologies, and rural infrastructure have only a small impact on poverty reduction in the medium term, although the impacts are likely to be larger over a longer time horizon.
Institutions and Pro-Poor Growth 11. Bolivia’s institutional environment is difficult and has recently deteriorated considerably given the political uncertainty and social instability. Bolivia scores particularly low on
iii
political stability and government effectiveness which is largely due to high perceived levels of corruption and low judicial reliability. Lack of transparency and voice in the public sector appears to be the main factor responsible for the high levels of corruption. Well-intentioned decentralization aimed to bring the government closer to the people and involve the poor have not (yet?) had the desired outcome due to difficulties in implementation, the loss of fiscal control, and the inability to manage the high expectations of the population. Bolivia’s PRSP process, once hailed as a model and enshrined in a permanent National Dialogue Law, is now largely seen as a failure. The goals were too ambitious, there was a serious disconnect between the consultation and the write-up of the strategy, it was too focused on determining how to allocate HIPC resources, there was no thorough discussion of economic policy-making, there was too little emphasis on strengthening the productive capacities of the poor, and by now nobody seems to own this document. As a result, revisions of the PRSP and the associated National Dialogue have stalled. It thus appears that the pay-off to the ambitious decentralization and PRPS processes has been quite low in Bolivia and might have contributed to some of the polarized political debates that currently undermine Bolivia’s political and social stability.
Trade Offs between Growth and Poverty Reduction 12. Using the CGE model, we investigate trade-offs and win-win situations for growth and poverty reduction. Among the win-win scenarios would be a reform of urban labor markets and a tax reform, although the urban poor would benefit more than their rural counterparts. But both policies might face stiff opposition from interest groups and thus are not easily implemented.
13. The expansion of the natural gas sector appears to cause a trade-off between growth and rural poverty reduction. It raises the growth rate but leads to sharply increasing inequality so that nationwide poverty would fall only moderately, while rural poverty would actually go up. Only if the receipts of gas were channeled as transfer or investment programs into rural areas, could this trade-off be mitigated. The largest effect for pro-poor growth could be achieved if the gas exports, tax and labor market reform were combined with transfer programs that are better targeted to the rural poor than currently.
14. More fundamentally, the model-based assessments suggest that incremental reforms will have a limited impact on putting Bolivia on a sustainable pro-poor growth trajectory. In particular, it highlights the fundamental constraint imposed by the very low domestic savings rate, which limits growth, increases vulnerability to external events, and limits opportunities for pro-poor policy-making. In addition, the high dualism of the economy is sharply reducing the poverty impact of growth. It thus appears to be necessary to confront some of the deep-seated inequalities in opportunities, resources, and power.
Recommendations for Policy-Making
15. We find that there is a range of incremental policies that could lift growth and poverty reduction in urban areas, where, in the absence of shocks, poverty reduction is expected to continue in coming years. Among them are policies to develop the gas sector, deregulation of urban labor markets, and income tax reform. The options to reduce rural poverty are much more limited. Our model-based estimates suggest that transfer programs (such as a demand-side transfer program linked to human capital investments) might be the best option, although a combination of investments in rural infrastructure, micro-credit, and agricultural productivity might also be of some help. A combination of such transfer and investment programs with gas exports, tax and labor market reforms might be a politically and economically feasible option.
iv
16. In addition, there are clear opportunities for improvements in policy-making at the macro and fiscal level. At the macro level, it is critical to develop policies that raise the domestic savings rate. They could include institutional reforms to widen the coverage of savings, greater public savings (e.g. from the proceeds of gas exports) and, at the international level, further debt relief. In addition, it is necessary to implement, to the extent feasible, policies to reduce dollarization of the economy in order to increase the room to maneuver for an active monetary and exchange rate policy that could support growth and poverty reduction.
17. Similarly, there are opportunities to increase the progressivity of the tax system and improve the poverty impact of public spending. In addition, policies to strengthen the productive capacities of the poor (such as the Cadenas Productivas Initiative and other pro-active policies) should receive the same attention as the expansion of social sector spending has.
18. Apart from these incremental reforms, it appears urgently necessary to confront some of the deep-seated inequalities in assets, opportunities, resources, and power in Bolivia. Among the policies to consider are revisiting the stalled land reform program, policies to transfer proceeds from natural gas directly to the poor, and policies to increase the voice of Bolivia’s marginalized indigenous communities.
1
Chapter 1: Historical Context
1. Bolivia is a large land-locked country with low population density (8 people per km2), difficult terrain, and consequently poorly developed transport and communications infrastructure (see Table 1). It is characterized by great economic and social inequalities with deep historical roots. Apart from a Spanish-speaking population (consisting of people of Spanish and mixed descent) that has dominated political and social affairs since independence in the early 19th century, Bolivia also has a very large indigenous population that comprises Aymara-speaking people in the highlands, Quechua-speaking people in the valleys, and smaller ethnic groups in the lowlands and the rainforest. Consequently, Bolivia is one of the most ethnically diverse countries in Latin America. Its index of ethnic fractionalization in 1998 stood at 0.74, compared to an average for Latin America and the Caribbean of 0.42 (Alesina et al. 2003).1
2. Until the revolutionary government of Victor Paz Estenzoro installed in 1952, most indigenous people lived in serf-like arrangements in rural areas. The agrarian reform in 1953 freed the peasants in the highlands and gave them access to land. Since then, subdivisions of land and population pressure have created smaller and smaller land-holdings (minifundismo) and landlessness has recently become a problem. In other parts of the country, particularly the lowlands, large estates dedicated to commercial farming predominate. As a result, the Gini coefficient for land inequality stood at 0.768 in 1989, indicating overall high land concentration similar to other Latin American countries (Deininger and Squire 1998). The other main source of incomes in the highlands, tin and silver mining, became progressively less lucrative and was sharply curtailed in the 1980s. Also here, the indigenous people had been used as forced labor for many centuries and as free miners since the 1950s, who organized themselves in unions. The mines became the breeding ground for considerable labor unrest throughout much of the 1970s and 1980s.
3. In contrast, the previously largely unpopulated lowlands surrounding Santa Cruz have become the focus of settlement and growth in recent decades, fuelled by a large-scale farming sector as well as the discovery of natural resources (oil and gas).
4. Starting in the 1970s, Bolivia became a major exporter of coca leaves, the input to cocaine, which became Bolivia’s most lucrative cash crop. The coca growing regions (Chapare and Yungas) became the focus of much in-migration (temporary and permanent) from other rural areas, generating considerable remittances. At the same time, under pressure from the United States, Bolivian governments promised coca eradication and pursued it with varying degrees of intensity. In the late 1990s and early 2000s, coca eradication was pursued much more vigorously, leading to a decline in production of some 80% (World Bank 2004b). The ebb and flow of these eradication efforts have played a significant role in the income sources of poor rural households.
1 The index measures the likelihood that two randomly drawn people from the population belong to different
ethnic groups.
Table 1: Bolivia in a Comparative Latin American Perspective, 2001 Bolivia Argentina Brazil Chile Ecuador Guatemala Paraguay Peru
Economic Indicators GNI per capita (PPP $) 2240.00 10980.00 7070.00 8840.00 2960.00 4380.00 5180.00 4470.00 Average GDP Growth 1994–2001 (%) 3.46 1.48 2.86 5.14 1.64 3.85 1.76 4.30 Average Population Growth 1994–2001 (%) 2.30 1.26 1.30 1.34 1.95 2.63 2.55 1.69 Population density (people per km2) 7.85 13.70 20.39 20.57 46.52 107.75 14.18 20.58 Average Inflation 1999–2001 2.79 -1.06 6.25 3.58 62.00 6.16 7.67 3.07 Average GDP Shares 1999–2001 of Agriculture 15.27 4.84 7.99 8.57 10.98 22.83 20.85 8.58 Industry 28.92 27.36 29.88 34.53 36.86 19.82 26.08 29.90 Services 55.81 67.80 62.13 56.90 52.16 57.35 53.07 61.51 Exports 17.66 10.70 11.58 30.44 36.89 19.30 22.42 15.53 Current Account Deficit -4.97 -3.01 -4.52 -1.24 3.01 -5.72 -2.92 -2.61 Budget Deficit -4.14 -2.82 n.a. -0.49 n.a. n.a. -2.71 -1.98 Gross Domestic Savings 7.80 15.73 19.74 23.07 24.85 7.72 10.24 18.11 Aid 7.23 0.04 0.05 0.08 1.04 1.36 0.97 0.82 External Debt 65.03 51.15 43.72 51.21 95.52 22.49 41.03 53.73 Human Development and Infrastructure Life Expectancy at birth (years) 63.06 74.08 68.31 75.79 70.04 65.23 70.58 69.57 Immunization, DPT (% of children under 12 months) 81.00 82.00 97.00 97.00 90.00 82.00 66.00 85.00 Hospital beds (per 1,000 people) 1.67 3.29 3.11 2.67 1.55 0.98 1.34 1.47 Total Years of Schooling (15+) 2000 5.58 8.83 4.88 7.55 6.41 3.49 6.18 7.58 Adult Illiteracy (%) 14.00 3.09 12.70 4.10 8.16 30.79 6.50 9.80 Female Illiteracy (%) 20.06 3.09 12.75 4.26 9.75 38.21 7.55 14.27 Roads, paved (% of total roads) 6.50 29.40 5.50 19.40 18.90 34.50 9.50 12.80 Roads to surface area (%) 4.90 7.75 20.18 10.55 15.23 12.97 7.25 5.67 Roads to total population (per ‘000) 6.46 5.89 10.14 5.25 3.42 1.27 5.73 2.85 Telephone mainlines (per 1,000 people) 62.21 223.83 217.84 232.51 103.71 64.68 51.24 77.50 Poverty and Inequality Data Year 1997 2001 2001 2000 1998 2000 1999 2000 PPP $1 Poverty Incidence 29.40 3.33 8.17 0.97 17.67 15.95 14.86 9.14 PPP $2 Poverty Incidence 51.69 14.31 22.43 9.58 40.77 37.36 30.29 37.71 Gini Coefficient 0.585 0.522 0.585 0.571 0.522 0.483 0.568 0.498 Source: http://www.worldbank.org/research/povmonitor/regional/Latin_America_and_the_Caribbean.htm; Barro and Lee (2000); World Bank (2003a).
3
Table 2: Basic Economic and Human Development Indicators for Bolivia
1985-1989 1989-1994 1994-1999 1999-2002 Economic Indicators Real GDP growth 1.62 4.08 3.93 2.18
Export growth (goods and services) 15.56 4.08 1.54 0.02 Export growth (merchandise) 5.04 5.89 -0.89 0.09 Export growth (mineral and hydrocarbon) -0.81 -2.49 -2.81 0.18 Ave. share of mineral and hydrocarbon exports to GDP 13.68 10.17 7.57 7.65 Ave. share of agricultural exports to GDP 2.14 3.87 5.16 5.28 Current Account Deficit -5.28 -3.53 -6.05 -4.38 Budget Balance -0.38 -1.92 -2.33 -5.06 Inflation 2414.35 13.41 7.43 3.10 Savings Rate (domestic) 10.91 9.05 10.53 7.52 Investment Rate 14.42 15.15 18.70 15.09 Human Development Indicators Population Growth 2.18 2.41 2.33 2.16 Child Mortality 146 122 97 80 Life Expectancy 56.19 58.81 61.03 62.56 Primary Enrollment (male) 100.81 103.29 111.34 116.66 Primary Enrollment (female) 89.80 94.71 106.26 115.07 Secondary Enrollment (male) 42.16 41.69 60.28 81.34 Secondary Enrollment (female) 35.92 35.74 54.54 77.87 Note: Data on GDP growth and current account is taken from UDAPE (various issues) and INE (various issues). Data on exports is taken from UDAPE (various Issues) and WDI (2003). All other data are taken from WDI (World Bank 2003a), covering the years up to 2001.
Source: WDI 2003; UDAPE (various issues); INE (various issues).
5. Politically, Bolivia oscillated between military dictatorships and civilian rule between the 1950s and the early 1980s when the latest military government was replaced with a democratic one, and democracy has persisted ever since. Bolivia’s politics were dominated by three main political parties (MNR, MIR, and ADN) and a few smaller ones and all governments since 1982 have been coalition governments, where the coalitions only lasted for one term and then were replaced by another coalition among the three major parties (or coalitions involving smaller ones; all possible permutations of coalitions among the three major parties existed in the past 20 years); this was aided by the constitutional provision that a president can only serve one term in office. All three parties represented the Spanish-speaking population with little representation from the indigenous populations. As a result of these arrangements, horse-trading and patronage became central elements in Bolivia’s political system, both to ensure the support of indigenous populations in elections and to generate coalition governments between groups with substantially different ideological agendas (Kaufman et al. 2003). This led to an increasing alienation and frustration of the population with the political process and led to the rise of powerful extra-parliamentary opposition forces, such as the coca growers’ union and other civil society groups, which were in hostile opposition to the government.
6. The latest election in 2002 brought major breakthroughs for new parties aligned with indigenous groups, which for the first time have a major representation in parliament. In particular, a party allied to coca growers (MAS) was able to gain major representation in parliament. Apart from representing coca growers, they have also taken on a range of populist positions on macro and trade issues. In this new environment, politics as usual continued and a coalition between MNR and MIR brought Gonzalo Sanchez de Lozada back into power (he had been president before between 1993
4
and 1997). Some of the proposed reforms and measures of the government, in particular a poorly communicated tax reform in early 2003 and a proposal to sell liquefied natural gas via Chile to the USA, led to such opposition (within and outside of parliament) and civil unrest that the government was forced out of power in October 2003 and the vice-president, Carlos Mesa, took over as the constitutional successor to form an independent government. Despite enjoying some popular support (based on his background in media and his strong stance against corruption), he has little support in parliament and it is unclear whether he will be able to bring back stability to the country. A constitutional assembly has been called for 2005 tasked reassessing the entire political and economic model that has been followed in the past, with great uncertainties about what outcome this might generate.
7. Regarding economic policies, Bolivia had pursued a state-led import-substitution regime until the 1980s, which was largely financed through the export of raw materials (tin and silver). The first democratic government under Siles-Zuazo (1982-85) faced a very difficult internal (drought, social unrest) and external environment (debt crisis, global recession and collapse in tin prices in 1985) and was unable to stabilize the country but instead allowed a hyperinflation to develop which led to a collapse of the government in 1985. Victor Paz Estenssoro took over and first undertook a strict stabilization plan, which ended hyperinflation and brought back internal and external stability (for details see Sachs and Larrain 1998).
8. In addition, the Paz Estenzoro government designed and began implementation of a Nueva Politica Economica, which included a wide range of structural reforms, which were supported thereafter by structural adjustment programs of the World Bank and the IMF. These reforms , which in the early 1990s shifted to second generation structural reforms, were continued by most of the successive governments so that Bolivia stands out as a country having undertaken more structural reforms in line with the so-called ‘Washington Consensus’ than most developing countries (Rodrik 2003; Lora 2001). They included:
⎯ Product market deregulation (freeing of prices, regulation of natural monopolies)
⎯ Capital market deregulation (freeing of interest rates, reduction in reserve requirements, liberalization of the external capital market)
⎯ Fiscal reforms involving the simplification of the tax structure where a value-added tax and an income tax (both at 13% where individuals could deduce value-added tax payments from the income tax bill) became the central revenue source and tax collection increased significantly as a share of GDP. On the expenditure side, there was a considerable expansion of expenditure in the social sectors (health and education), while expenditures on state-owned companies were sharply reduced through the privatization program.
⎯ Trade liberalization (simplification and sharp reduction of import tariffs, elimination of non-tariff barriers, efforts to promote non-traditional exports)
⎯ Liberalization of the FDI regime (regulatory framework, investor protection, equal treatment of domestic and foreign investors)
⎯ Restructuring, closing, and ‘capitalization’ of the large state-owned companies. The latter refers to a scheme where public companies sold a 50% stake to strategic investors (where the proceeds remained with the companies to finance a pre-specified investment program). The proceeds from the remaining shares are being used to finance an annual old age pension (the Bonosol) for all citizens over the age of 65. This way, electricity, railway, telecommunications, mining, the national airline, and the national hydrocarbon company were transferred to (mostly foreign) strategic investors who took management control of these companies.
9. The one area where there were only few reforms was the labor market. Here, only government intervention in wage setting was reduced and there was some reduction in wages and benefits for public sector employees. The Labor Law of 1942 is still largely in force with quite high
5
costs of dismissal, few options for temporary work, substantial requirements to meet occupational health and safety standards, a prohibition of employment agencies, and other regulations which were aimed primarily at the mining sector but have since become a stumbling bloc for a smoother operation of the formal labor market.
10. In addition, the first government of Sanchez de Lozada (1993-97) undertook an ambitious decentralization program in the 1994 Popular Participation Law and the 1995 Decentralisation Law, which transferred a considerable amount of resources (and responsibilities) to Bolivia’s 314 municipalities. In addition, the municipalities were also awarded all additional resources that were freed up as a result of the HIPC II initiative which were the focus of attention in Bolivia’s first PRSP, concluded in 2000.
11. In several dimensions, Bolivia’s structural reforms produced positive outcomes. Macroeconomic stability was achieved and maintained throughout the period with low inflation, low fiscal deficits, and a relatively stable exchange rate. The fiscal reforms, combined with the reform of the state sector, ensured that the fiscal situation improved dramatically over the 1990s. Exports, including non-traditional exports, improved, and there were significant improvements in human development indicators, particularly education (see Tables 1 and 2). While Bolivia remains a lot poorer than all of its neighbors, has higher poverty rates and lower life expectancy, it compares favorably in education indicators with some richer Latin American countries such as Guatemala or even Brazil (see Table 1).2 Economic growth also improved and Bolivia grew at around 4% per year from 1990-1998, but only about 1.5% in per capita terms. This relatively positive performance was aided by a favorable external environment, with high growth of Bolivia’s main trading partners, the expansion of natural resource exports, and a surge in foreign direct investment that accompanied the capitalization process. The combination of strong memories of the 1985 hyperinflation, an open capital account, and high political and economic uncertainty of a small open economy led to high and increasing dollarization in the economy, which permeates the financial system and significantly limits the options for an active monetary and exchange rate policy. There were few attempts to combat dollarization, which is extremely high to this day.
12. Exports, while improving throughout the 1990s, remained largely focused on primary products with the mix shifting from a heavy reliance on minerals to a much greater importance of hydrobarbons and agricultural cash crops produced by commercial agriculture (i.e. soybeans, sugar, and wood). The lack of diversification and the failure to develop manufactured exports appears to be due to a combination of geographical factors (land-locked country, poor infrastructure, high transport costs), economic risks and volatility (i.e. exchange rate risks and volatility vis-à-vis trading partners), Dutch disease problems associated with the primary exports, and institutional constraints (weak protection of property rights, high corruption, contraband economy, high regulatory burden for start-ups, high informality of the economy, e.g. Kaufman et al. 2001; World Bank 2004b). A continuing concern is also the very low domestic savings rate (see Tables 1, 2 and below), making Bolivia heavily dependent on capital inflows to finance investment.
13. Since 1998, economic growth has decelerated to an average of only about 1.5% per year and has become negative in per capita terms. The main causes for this slowdown are a series of external economic shocks that have affected the economy, including particularly the strong devaluations and recessions in Brazil and Argentina in 1999 and 2002, respectively, while the Boliviano appreciated significantly alongside the US$. This led to a sharply overvalued currency and the (independent) monetary authorities did little to combat this due to the risks of devaluations in a dollarized economy, but instead stuck to their policy of allowing only very small devaluations against the
2 One should note that the findings on poverty and inequality are quite sensitive to the choice of the survey, and to
whether income or expenditure is being used as the indicator. When one uses expenditures and the 1999 MECOVI survey, the Gini stands at only 0.45 and the poverty headcount of below $1 a day falls to 14.4%. We report the income-based figures in Table 1 as the data from the other countries are also based on incomes.
6
dollar (some 8% in 2001, falling to 4% in 2002). Instead, the economy slowed down considerably, credit contracted sharply as the financial sector experienced build-up of non-performing loans; as a result of the recession and costly amendments to a pension reform, budget deficits have soared to unsustainable levels, adding economic uncertainty to the already existing explosive political and social situation (World Bank 2004a). The financing of the large budget deficit through domestic and international borrowing has placed Bolivia in an increasingly vulnerable situation where rising shares of government spending must be allocated to debt service payments, thereby partially wiping out some of the gains realized by the HIPC debt relief (World Bank, 2004a). As the dollar has fallen recently against the currencies of Bolivia’s main trading partners and raw material prices have increased, the external environment has improved somewhat and growth is projected to at 3.8% and 4.5% for 2004 and 2005, respectively. .
14. Regarding poverty and inequality trends, one first has to note that nationally representative household surveys with income and expenditure information are only available from 1997 onwards.3 Before, there are income surveys for departmental capitals (plus El Alto) going back to 1989, and some spotty survey information from non-urban areas (see Annex 1). Thus rural areas (comprising about 40% of the population in 1994, with the share falling over time) and towns (comprising 12% of the population in 1994 with the share rising over time) were excluded from these surveys. In addition, there are three national censuses (1976, 1992, and 2001) and three nationally representative Demographic and Health Surveys (DHS in 1989, 1994, and 1998) none of which contain income information.4 As a result there have been considerable disagreements about the actual trends in poverty in Bolivia as shown in Tables 1 and 2 in Annex 1 which compiles all poverty estimates we could find. Nevertheless, most of the studies agree on the following three stylized facts: First, in the late 1990s, poverty is much higher in rural than urban areas; second, there was some decline in poverty in capital cities since 1989 with an upturn in poverty again after 1997; third, non-income measures of poverty have declined stronger than income measures throughout the 1990s, particularly in urban areas.
15. For the purposes of this study, it was critical to generate nationally representative poverty data going as far back as 1989. In order to achieve this, we employed two alternative methodologies to generate national poverty data and poverty profiles for the time prior to 1997. The first uses information from the DHS to generate an asset index as a proxy for income following proposals from Sahn and Stiefel (2003) and Pritchett and Filmer (2001). Due to limitations in the data, we can do this only for 1994 and 1998.5 The second combines information from the urban household surveys with the DHS to generate income and poverty information for the entire country from 1989 to 2002. The precise methodology and all of the statistical and econometric issues are discussed in Annex 1.
16. The most important results regarding poverty and inequality, based on the second methodology, are summarized in Table 3 below. We present our main estimates but also include (in brackets) the results of a sensitivity analysis of one of our key assumptions underlying the simulation which might lead to an overestimate in the decline of poverty in rural areas.6 Moreover,
3 The 1997 survey is also not comparable to later surveys so that a consistent national time series only emerges in
1999. 4 There are further restrictions on the DHS. The 1989 DHS only includes households with women of reproductive
age (15-49), while the later ones include a representative sample. The 2003 DHS is due to be out in June. We will be able to report on some summary information from the survey which was made available to us below.
5 We are also not convinced that this approach will be appropriate for inter-temporal comparisons of welfare and poverty as changes in tastes and relative prices might systematically distort such an inter-temporal assessment. See Annex 1 for a further discussion. We nevertheless used this method primarily as a robustness check on our other approach.
6 In particular, we assume that the difference in returns to assets and endowments between rural, urban, and capital cities did not change between 1989 and 1999. In our sensitivity analyses we replace the fixed difference assumption
7
one should note that the use of consumption (including auto-consumption) as the welfare measure in rural areas and income as the welfare measure in capital cities (the nine departmental capitals and the city of El Alto) and towns (all other cities and towns), as is standard practice in Bolivia (e.g. INE-UDAPE, 2002), will lead to lower levels of inequality compared to using incomes in rural areas which are reported to be considerably smaller. Using incomes for rural areas as well would raise the Gini in 2002 to about 0.598. But as incomes in rural areas are implausibly low (about 25% lower than consumption with many households reported extremely low incomes--including incomes from own-consumed goods--that are impossible to survive on), we believe that it is preferable to stick to the mixed definition.7 Lastly, we should point out that the poverty lines used here are based a regionally differentiated basket of goods that allows sufficient caloric consumption which has been updated using local price data on these goods. The extreme poverty line is derived by just allowing for enough caloric consumption while the moderate poverty line also makes allowance for non-food items (see annex 1 for further discussion). As will be shown below (and in annex 1), the updating of the poverty line is not in line with the developments of overall prices as the prices of the poor have risen less than the overall CPI.
17. With these caveats in mind, the following observations are noteworthy: First, using our methodology, we are able to reproduce actual poverty trends in capital cities (where we have actual data for comparison) fairly well, particularly for the poverty gap measure, which is quite reassuring. We tend to slightly underpredict the headcount ratio (poverty rate) most of the time but also here, the most important trends (in capital cities where we can make a comparison) are accurately reflected.8 Second, consistent with other studies, there is a steep gradient in poverty levels between capital cities, towns, and rural areas, with poverty being much higher in the latter. As far as the poverty rate is concerned, this differential between capital cities and rural areas gets larger over time (from about 25 percentage points in 1989 to nearly 29 percentage points in 2002). This is not true, however, when we consider the poverty gap, for which the differential gap has somewhat narrowed. This suggests that the very poor have been able to make some gains in the 1990s while rural dwellers close to the poverty line did not benefit as much. Third, there is a clear poverty trend in capital cities, which closely mirrors macroeconomic conditions. Thus poverty (using the headcount or the poverty gap measure) declines considerably between 1989 and 1999 and then increases again between 1999 and 2002. In towns and rural areas, in contrast, the dynamics of poverty are not as closely aligned to macroeconomic developments. In particular, there is no poverty reduction at all in rural areas between 1989 and 1994, then considerable poverty reduction between 1994 and 1999, and stagnation (headcount) or slight further reductions (poverty gap) between 1999 and 2002. Note also that the pace of poverty reduction in rural areas is smaller in our sensitivity analysis but does not change the general picture (see figures in brackets).
18. Using the first approach (see Annex 1 for tables and discussion) to generate poverty data largely confirms the findings above for the time period 1994 to 1998, but with some slightly different nuances. While the asset index which we use as a proxy for incomes increases overall and in all three regions, which is consistent with the findings above, the increase in the asset index is largest in towns, followed by capital cities, and smallest in rural areas (see Annex 1), suggesting that rural poverty reduction measured this way has been somewhat smaller than urban poverty reduction.
with the assumption that the difference in the impact of assets moved in accordance with the overall growth rates or rural areas, towns, and capital cities which show that rural incomes increased more slowly than incomes elsewhere.
7 At the same time, we acknowledge that using consumption in one area and income in another may also lead to biases that are hard to quantify. It is not possible to use expenditure throughout as expenditure data are not available prior to 1999.
8 In Annex 1 (Table 5), we show that most of the differences in our prediction are due to our specification of the error term in the underlying regression where we assume a normal distribution. We will experiment with other distributional assumption to address this issue.
8
19. Regarding inequality, the trends follow closely the poverty discussion, but with some additional features. In particular, the sharp increase in inequality in capital cities between 1999 and 2002 is noteworthy. Measures that are more sensitive to the bottom of the distribution, such as the Atkinson measure with e=2, show even more dramatic deteriorations (see Annex 1) suggesting that the urban poor have fared particularly badly in the last few years. In other areas, inequality seems to have fallen and thereby somewhat offsetting the dramatic worsening of inequality in capital cities. Overall, the Gini in 2002 is similar to 1989. It thus appears that the fate of the urban population, including the urban poor, has been closely linked to macro developments and has recently led to a significant deterioration in poverty and inequality. In contrast, the much poorer rural poor have been more detached from improvements and deteriorations in the overall economic environment and their poverty trends have followed another logic.
Table 3: Poverty and Inequality Trends using Moderate Poverty Line*
Capital Cities** 67.2 64.8 59.5 57.4 51.1 48.1 55.1
Towns 81.1 (80.7)+
75.1 (74.3)
69.1 64.2 67.7
Rural 89.7 (87.8)
89.6 (87.8)
83.4 79.1 83.8
Total 76.9 (76.0)
72.4 (71.6)
65.2 60.3 67.2
Poverty Gap
Capital Cities** 32.9 32.9 25.7 25.3 21.0 21.3 24.4
Towns 51.3 (50.7)
44.7 (44.0)
34.7 33.6 32.9
Rural 58.3 (55.2)
60.9 (58.2)
47.7 43.1 44.9
Total 45.5 (44.1)
41.9 (40.7)
32.5 30.1 32.9
Gini Coefficient
Capital Cities** 0.505 0.497 0.481 0.455 0.480 0.488 0.540
Towns 0.547 0.537 0.455 0.500 0.452
Rural 0.475 0.497 0.423 0.443 0.421
Total 0.555 0.555 0.525 0.531 0.551
*The moderate poverty line is, in line with standard practice in Bolivia, applied to income in urban areas, and consumption in rural areas (as income data are considered not to be reliable there and consumption data are not available for the urban household surveys prior to 1997). While the extreme poverty line in Bolivia is only based on ensuring adequate nutrition, the moderate poverty line also makes allowance for some non-food expenditures. The moderate poverty line stood at about US$40 per capita and month, the extreme poverty line at about US$20. For details on the poverty lines and the results for the extreme poverty line, refer to annex 1.
**Capital cities refer to the 9 departmental capitals and El Alto (the city adjacent to La Paz).
+ The figures in brackets refer to sensitivity analyses which no longer assume that the impact of endowments on growth did not change between urban and rural areas between 1989 and 1998 but that it changed in proportion with the differential in aggregate growth performance in the three areas. See Annex 1 for details and full results.
20. One should point out that Bolivia’s record in non-income dimensions of poverty is considerably more favorable than its record in income poverty reduction. As shown in Table 2, Bolivia has
9
achieved impressive improvements in the reduction of child mortality and the expansion of primary and secondary education. More recent data suggest that the decline in infant and child mortality as well as the expansion of reproductive services and immunization coverage has continued at a rapid pace, including in rural areas (INE, 2004), while education data suggest that the poorest quintile have (in contrast to richer groups) suffered from slight declines in enrolment and attendance rates (World Bank 2004b). The index of unsatisfied basic needs which combines information on housing, sanitation, education, and health care, also shows strong improvements between 1992 and 2001; but the improvements are much smaller in rural areas where in 2001 91% of the population continues to suffer from unsatisfied basic needs (see Annex 1 and World Bank, 2004b). The apparent disconnect between rapidly improving social indicators and only moderate improvements in income poverty are one of the conundrums of Bolivia’s economy (see below).
Chapter 2: Analysis of Growth and Its Distributional and Poverty Impact
21. Growth Decomposition and Pro-Poor Growth. Two ways to provide further insights about the links between poverty, inequality, and growth trends is to do a decomposition of the observed poverty reduction and provide estimates of the rates of pro poor growth (Datt and Ravallion 1992; Ravallion and Chen, 2003). The decomposition of the observed poverty reduction into a growth and an inequality contribution (and an interaction term which cancels if one the average of a ‘forward’ and ‘backward’ decomposition) is using the methods proposed by Ravallion and Datt (1992). As discussed in detail in the Grimm and Günther (2004), the distribution component in this decomposition also implicitly includes the impact of changes in the real value of the poverty line (i.e. how prices paid by the poor have moved relative to the overall price level). As shown in Table 4 of Annex 1, the prices paid by the poor (in particular food prices) have risen somewhat less than the overall price level (particularly in recent years) so that the purchasing power of the poor has increased by more than suggested by the change in their real incomes. This is implicitly captured in the decomposition as a distributional shift favoring the poor.
1989–1999 1999–2002 1989–2002 Total Bolivia Change in poverty -0.118 0.020 -0.099 Growth component -0.080 0.018 -0.064 Redistribution component -0.038 0.002 -0.035 Departmental Capitals Change in poverty -0.163 0.040 -0.123 Growth component -0.105 0.025 -0.080 Redistribution component -0.057 0.015 -0.043 Other Urban Areas Change in poverty -0.117 -0.015 -0.132 Growth component -0.067 0.017 -0.074 Redistribution component -0.050 -0.032 -0.058 Rural Areas Change in poverty -0.068 0.005 -0.064 Growth component -0.041 -0.005 -0.039 Redistribution component -0.028 0.010 -0.025 Notes: Calculated using the Datt-Ravaillion (1992) method of growth-inequaltiy decomposition.
Source: Own calculations. For the extreme poverty line, see Table 12 in Annex 1.
10
22. The result of the decomposition analysis (Table 4) for the entire period show that about two-thirds of the 10 percentage point decline in poverty for total Bolivia is attributable to growth, and about one-third to a distributional shift favoring the poor. 9 As the income distribution hardly shifted between the two periods (see Table 3)10, most of this distributional shift is actually due to the poverty line effect which increased the real purchasing power of the poor. Considering sub-periods and different parts of the country shows a more differentiated picture. In the period 1989-99 both the growth and redistribution (and/or poverty line) effect served to reduce poverty in all parts of the country. In the latter three years, the picture has changed drastically. Now poverty has increased in capital cities nationally, and particularly in capital cities where 60% is due to falling incomes and 40% due to adverse distributional shifts.
23. When one splits out this poverty line effect from the distributional component (results not shown), we find that ‘pure’ redistribution helped to lower poverty in all of Bolivia between 1989 and 1999 as well as capital cities and towns, while the redistribution component was essentially zero in rural areas. Between 1999 and 2002, the redistribution component served to increase poverty in all regions and Bolivia as a whole. For the overall period (1989-2002), this ‘pure’ redistribution effect had a slightly poverty-increasing effect for Bolivia as a whole so that the poverty decline that happened occurred mostly due to growth and a favorable development of the prices paid by the poor. This adverse distributional effect is entirely driven by an adverse distributional shift in capital cities which dominates a favorable distributional shift in towns and rural areas.
24. A second way to examine the linkages between growth, inequality, and poverty is the Ravallion-Chen measure of Pro-poor Growth which takes the average of growth rates of the quantiles of the population that were poor in the initial period (see Ravallion and Chen 2003).11 The growth incidence curves underlying this analysis are shown below for the entire period (1989-2002); for sub-periods they are available in Annex 1. For the entire country and the entire period, they are above 0 for all groups, and moderately downward sloping from the 10th to the 90th percentile suggesting that, on the whole, the poor gained proportionately more from growth than the rich. This is not true below the 10th percentile and above the 90th percentile suggesting that the extremely poor were not benefiting as much and that the very rich were benefiting more from growth.12 Matters are different when one considers the different parts of the country. In departmental capitals (and El Alto), growth over the period was anti-poor with the poor gaining less than the rich (particularly due to the influence of the period after 1999), while it was strongly pro-poor in towns, and moderately pro poor in rural areas.
25. The annual rate of pro poor growth, shown in Table 5, summarizes the information provided in the growth incidence curves.13 We also show the results of our sensitivity analysis for towns and
9 This changes very slightly in our sensitivity analysis which is available on request. 10 Whether income distribution in Bolivia worsened between 1989 and 2002 is sensitive to the choice of inequality
indicators which give different weights to different parts of the distribution. But all show that whatever distributional shifts occurred were small.
11 There are various criticisms of this approach of measuring pro-poor growth some of which can be found in Klasen (2004).
12 One should note that measurement error might have a considerable influence at the two tails of the distribution so that these results should be treated with some caution.
13 We should point out that Jimenez and Landa (2004) from UDAPE have, for the World Bank poverty assessment (World Bank 2004b), also been calculating rates of pro poor growth using the Ravallion and Chen method whose results, on the surface are quite different from ours. Their growth incidence curves for 1999-2002 point to sharply rising inequality in rural areas and somewhat rising inequality in urban areas (combining capital cities and towns); the calculated annual rates of pro poor growth are -6% per year. Where we use the same information (per capita incomes for capital cities between 1999 and 2002), our findings are virtually identical. The most important reasons for the discrepancy appear to be that they use income as the welfare indicator in rural areas while we use
11
rural areas (and by implication, all Bolivia) in brackets. The most important findings are the following. Overall, there was Pro Poor Growth between 1.9 and 2.2% per year between 1989 and 2002, which was mostly due to high pro poor growth in towns and some pro poor growth in rural areas, while pro poor growth in capital cities was negligible. As before, it is useful to consider sub-periods. Between 1989 and 1999, there was a considerable amount of pro-poor growth in total Bolivia, in capital cities, towns, and rural areas, regardless of the poverty line. Also, the rate of pro-poor growth exceeded the growth rate in the mean, suggesting that growth was accompanied by falling inequality. The particularly high growth rate in total Bolivia (2.23%) is due to growth in the three areas plus a shift in the composition of the population from the poorer rural areas to the richer urban areas. Between 1999 and 2002, we find that there was a strongly anti-poor contraction in capital cities, wiping out most of the gains the urban poor have made in the ten previous years. In other urban areas, the contraction was not particularly anti-poor so that the poor had roughly stagnant incomes. In rural areas, incomes continued to rise, although very slowly, and growth continued to be somewhat higher for the poor than for the non-poor. Given that the rural poor predominate among the poor, overall growth was only slightly anti-poor between 1999 and 2002, and this finding is sensitive to the choice of the poverty line. In the sensitivity analysis, growth and pro poor growth is somewhat smaller in total Bolivia and more significantly so in rural areas which hardly experienced any growth mean income growth between 1989 and 2002; but the rates of pro- poor growth remain between 1.2 and 1.4% suggesting that the poor were able to make some gains over the period. Figure 1 — Growth Incidence Curve for Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100 Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
consumption, in line with the usual practice in Bolivia. Using the income indicator for rural areas shows massive declines in per capita income which are implausible in two ways. First, they imply income levels in rural areas that are unlikely to assure basic survival and second the growth rates, -20% per year for the poorest quintile over three years, is not consistent with all the known information about economic developments between 1999 and 2002 (where per capita incomes declined slightly, but not by these magnitudes). For the period prior to 1999 (1993-1999), they calculate only very moderate pro poor growth rates in capital cities, in contrast to our higher figures; this discrepancy is probably largely due to the different time periods considered. Beginning in 1993 omits high (and pro-poor) growth from 1989 to 1993. Thus we find those figures to be roughly consistent with ours (which they should given that we both use incomes and use a similar income definition).
12
Figure 2 — Growth Incidence Curve for the Departmental Capitals of Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
Figure 3 — Growth Incidence Curve for Other Urban Areas of Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest %Growth Rates in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
13
Figure 4 — Growth Incidence Curve for Rural Areas of Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
26. With the exception of the strongly anti-poor growth in capital cities in recent years, it appears that growth has been quite pro-poor throughout most of the last 15 years, and particularly so in towns and (moderately so) in rural areas. One may wonder how this squares with the results in Table 3 which showed only slowly falling poverty rates in rural areas in the 1990s. But these results are entirely consistent with each other when one notes that the depth of poverty in rural areas is so large that even considerable pro-poor growth does not lift many of the poor above the poverty line (but does reduce the poverty gap as indeed happened, particularly between 1994 and 1999). Thus the problem of Bolivia’s poverty is not so much that growth in the 1990s has been biased against the poor, but that overall growth has not been very high throughout the period and that the initial inequality was so large that the poor remained poor despite some improvements in incomes. It would probably have taken another decade of such growth to make serious inroads into poverty, particularly in rural areas. Unfortunately, that did not happen. With the type of growth experienced since 1999, rural poverty will not change much and urban poverty is on a sharply increasing trend.
Source: Own calculations. Growth rates use the actually observed levels of income/expenditure where available (in capital cities throughout and elsewhere from 1999 onwards). Figures in brackets are based on sensitivity analysis as discussed in text (footnote 6) and in Annex 1.
27. (Sectoral) Sources and Proximate Determinants of Growth. Before discussing the determinants of pro-poor growth, it is important to first discuss the sources of overall growth in Bolivia in the past 15 years. Table 6 gives an overview over the sectoral composition of GDP and its growth. Regarding the sectoral composition of GDP in 2002, agriculture makes up about 14%, about half of which is subsistence agriculture where many of the rural poor live. About 10% of GDP is generated by mines, oil, and gas and only about 16% by manufacturing. Most of this manufacturing consists of food processing and the processing of raw materials (wood, oil, and minerals), with hardly any light or heavy industry present in the country. The remainder of GDP consists of services of various kinds, which includes mostly services that involve the rural and urban poor (such as trade and transport services). Employment shares differ radically from this sectoral composition of GDP (see Table 7). Agriculture employs 60% of the workforce, sales employs another 10% of the workforce, while manufacturing, oil and gas, and high-value services employ only a small fraction of the workforce. Thus Bolivia is a highly dualistic economy with a
15
large employment in low value agriculture and the small-scale service sector and very small employment in manufacturing.
28. Overall GDP growth between 1989 and 1999 was driven largely by sharp growth in commercial agriculture, oil and gas production (and associated construction and production in the electricity, gas and water sector), some small-scale food processing industries, and some services. In contrast, subsistence agriculture, mining, hotels and restaurants, and public administration grew less than proportionately. Between 1999 and 2002, virtually all sectors grew slower, with the exception of oil and gas, which expanded production due to enhanced exports to Brazil. The figures for coca production show a continuous and sharp decline between 1989 and 2002. This decline in reported coca production is very likely overstating the actual decline. While eradication efforts were intensified throughout the 1990s, the enforcement varied considerably. It was particularly strong under the Banzer regime (1997-2002), but it is still likely that clandestine production is much larger than reported here (and it is also likely that coca production increased considerably recently as enforcement has flagged).
29. One should also note that the oil, gas and mineral sectors only account for about 10% of Bolivia’s GDP and less than 1% of its employment, but more than 40% of Bolivia’s exports, so that the importance of these sectors for Bolivia’s external position is much larger than its GDP share. Thus we find that Bolivia has a highly dualistic economy, with the most dynamic sectors being the oil and gas sector, industrial agriculture (concentrated in the lowlands) and some high-value service sectors. The remainder of the economy showed a much more moderate evolution.
30. TFP Analysis. Another way to examine the proximate sources of growth is to examine the influence of input factors (labor, capital, human capital) and the residual component, total factor productivity (TFP). This can be done using a growth accounting framework based on the Solow growth model and is such an analysis was done by Loayza et al. (2002). Depending on whether human capital and the input factors are adjusted for capital utilization, the results show that the contribution of capital to GDP growth was negative on average in the 1981-1990 period (-0.26 to -0.31% per year) indicating very low investment rates. Similarly, TFP growth was negative indicating worsening efficiency. In the 1991-2000 period, things turned around with capital contributing about 0.45% to annual growth and TFP contributing about 1.23-1.66% per year depending on the assumptions. Labor throughout both periods contributed about 1.4-1.7% per year and its contribution was very stable. While the findings for the crisis-ridden 1980s are to be expected, the remarkable finding for the 1990s is the very low capital contribution to growth, suggesting very low investment rates that are barely able to make up for depreciation. This, in turn, is related to Bolivia’s very low domestic savings rate (Table 1) which, even with generous aid and capital inflows, leads to only a moderate investment rate and thus quite low growth attributable to capital deepening (see Table 2 and below).
16
Table 6: Sectoral Composition of GDP and its Growth, 1989–2002
Production in 1990 Bs ('000) Annual Growth 1989 1999 2002 1989-99 1999-2002
A. PRIVATE SECTOR 11876 18054 19209 4.3 2.1 1. AGRICULTURE 2267 3071 3296 3.1 2.4 - Non-industrial Agricultural Products 1062 1358 1437 2.5 1.9 - Industrial Agricultural Products 212 558 605 10.2 2.7 - Coca 193 74 39 -9.1 -18.9 - Cattle and other Livestock 669 896 1005 3.0 3.9 - Forestry, Hunting and Fishing 130 185 211 3.6 4.5 2. MINING AND QUARRYING 1470 2017 2191 3.2 2.8 - Crude Oil and Natural Gas 644 978 1189 4.3 6.7 - Metal and Non-Metal Minerals 826 1039 1002 2.3 -1.2 3. MANUFACTURING 2430 3633 3849 4.1 1.9 - Food, Drinks and Tobacco 1109 1745 1975 4.6 4.2 - Other Industries 1321 1889 1874 3.6 -0.3 4. ELECTRICITY, GAS, AND WATER 235 452 475 6.7 1.7 5. CONSTRUCTION 462 819 819 5.9 0.0 6. TRADE AND COMMERCE 1270 1820 1937 3.7 2.1 7. LOGISTICS & COMMUNICATIONS 1365 2331 2562 5.5 3.2 8. FINANCIAL AND BUSINESS SERVICES 1528 3161 3103 7.5 -0.6 - Financial Services 242 974 914 14.9 -2.1 - Business Services 382 1113 1040 11.3 -2.2 - Real Estate 904 1075 1149 1.7 2.3 9. PERSONAL SERVICES (INCKL. DOMESTIC SERVICES) 667 973 1073 3.8 3.3 10. RESTAURANTS Y HOTELS 507 688 734 3.1 2.2 11. IMPUTED BANKING SERVICES -234 -911 -830 14.5 -3.1 B. PUBLIC SECTOR 1570 1991 2140 2.4 2.4 TOTAL A AT FACTOR COSTS 13537 20045 21350 4.0 2.1 INDIRECT TAXES 1222 1764 1916 3.7 2.8 TOTAL AT MARKET PRICES 14759 21809 23266 4.0 2.2 Source: UDAPE (various isues).
Table 7: Employment Shares, 1999
A. PRIVATE SECTOR 2528708 95.1% 1. AGRICULTURE 1598358 60.1% 2. MINING AND QUARRYING 44051 1.7% 3. MANUFACTURING 249167 9.4% 4. ELECTRICITY, GAS, AND WATER 3986 0.1% 5. CONSTRUCTION 116845 4.4% 6. TRADE AND COMMERCE 253974 9.5% 7. LOGISTICS & COMMUNICATIONS 66776 2.5% 8. FINANCIAL AND BUSINESS SERVICES 12802 0.5% 9. PERSONAL SERVICES (INCKL. DOMESTIC SERVICES) 107401 4.0% 10. RESTAURANTS AND HOTELS 75348 2.8%
B. PUBLIC SECTOR 131464 4.9%
TOTAL 2660172 100.0%
Source: MECOVI survey. 31. Poverty Profile. These analyses have so far provided quite an aggregative picture of developments in poverty as well as of GDP growth. We now turn to a detailed poverty profile to
17
give a better sense of who and where the poor are, and what they mainly live off. In Tables 8 and 9, we present our results for the poverty gap, which also captures the depth of poverty.14 Apart from the already noted rural-urban divide, there are very large regional variations in the poverty gap in the different departments. In particular, poverty gaps are very high in the two highland and valley departments of Chuquisaca and Potosi, while they are much lower in the lowland departments of Santa Cruz, Beni, Pando, and the valley department of Tarija. The former two provinces are particularly dependent on subsistence agriculture, while the latter three are the home to large-scale farming, as well as most oil and gas production. The three provinces La Paz, Oruro, and Cochabamba take on an intermediary position.
Table 8: Regional Disaggregation of the Poverty Gap
Source: Own calculations. Standard errors are in brackets (only applicable to the simulated poverty rates). For 1999 and 2002, we use the actual poverty rates.
32. Regarding household characteristics of the poor, large households, those with many dependents, and those with a young head are significantly poorer, although the latter influence is quite small. This suggests an important influence of fertility on pro-poor growth, where fertility decline could make a significant contribution to the decline of poverty and inequality (see Box 1). Particularly striking are the very large differences in poverty by language and education. The poverty gap of those speaking an indigenous language is nearly twice as large when the moderate poverty line is applied, and three times as large when the extreme poverty line is used. Similarly, there is hardly any poverty among those with more than completed secondary education, while there are very high poverty rates among those with less than 5 years of schooling. Given the differences between employment shares and sectoral contributions to GDP (as shown in Tables 5 and 6), it is not 14 The poverty gap index (or P1 from the FGT family of indices) divides the percentage average shortfall of the poor
from the poverty line with the total population. Other results can be found in Annex 1. We should note that the surveys were not designed to be representative at the level of departments so that the results presented here should be treated with some caution (particularly in the case of the smaller departments such as Beni and Pando).
18
surprising to find considerable differences in poverty rates by the sectoral employment of the household head. In particular, those working in agriculture have a much larger poverty gap than those working in any other profession. Unemployed heads also have very large poverty rates while the poverty rate among white-collar workers is predictably low. It is also interesting to note that the gender of the household head does not appear to have a big impact on poverty. If anything, female-headed households are less poor than male-headed households, a finding common to many Latin American countries (see Marcoux 1998). Similarly, education gaps by gender, an important cause of poverty, have largely disappeared. But females continue to be disadvantaged in other ways, particularly in the labor market but also in the household (see Box 2).
33. There are no dramatic trends in terms of changes of the characteristics of the poor over time.15 But a few changes are noteworthy. In particular, the poverty gap in Chuquisaca appears to have declined the least so that it surpassed Potosi as the poorest province by 2002. In contrast, in Tarija, Beni and Pando, poverty reduction appears to have been particularly rapid. As far as household characteristics are concerned, small households appear to have reduced poverty more rapidly than large households, particularly in relative terms. While the absolute reduction for the poorly educated and those speaking indigenous languages were larger than for others, in relative terms it was smaller so that the relative gap between them and the rest has widened. Similarly, the relative gap between farming households and while-collar households has widened considerably in the past 15 years even if the poverty gap was reduced considerably in farming households.
34. Using the asset index confirms most of the results shown above, but in a somewhat more accentuated fashion (see Annex 1). The difference between rural areas, towns, and capital cities in the asset index is larger than in the simulated incomes leading to starker differences in poverty rates between the three areas. The poverty profile confirms that larger households16 and those with less educated and younger household heads are poorer, and find even stronger differences in poverty rates by language and education of the household head and spouse. Also here, female-headed households are less poor than male-headed ones. Thus we find large and significant differences in poverty rates among different groups.
35. Accounting for Inequality Change. It is useful to further examine the causes of the observed changes in inequality over the past 10 years. Here we draw on findings from Gasparini et al (2003), which decompose changes in inequality in (equivalized) household labor income in capital cities between 1993 and 1997 and urban and rural areas from 1997 to 2002. Between 1993 and 1997, they find a slight increase in household labor income in capital cities which is mostly driven by a rising employment gap between the highly educated and the less educated, a slight shift in educational inequality, and a significant increase of in the returns to unobservable characteristics, while returns to education were equalizing. Between 1997 and 2002, inequality in household labor incomes increased considerably in capital cities and here all factors (returns to education, inequality in employment, inequality in education, and inequality in returns to unobservables) all contributed to this rise in inequality.17 The importance of the rising inequality in unobservables points to increasing disparities in the returns to characteristics such as educational quality, labor market connections, and unmeasured skills. While reducing educational and employment inequality would serve to reduce inequality and thus help with poverty reduction, the high returns to inequality in unobservables points to deeper segmentations of the Bolivian economy, to which we turn now. 15 To a limited extent, this is true by construction as we use correlates of incomes to simulate incomes which include
the characteristics listed in the table. But since we allow these correlates to vary over time, we would be able to discern if there have been significant changes in the determinants of poverty.
16 The effect of household size on poverty is found to be smaller using the asset index than with the simulated per capita incomes. This is to be expected given that large households are likely to possess more assets and thus appear less poor in an asset-based index than in an income-based one. See also the discussion below.
17 In rural areas they find declines in inequality between 1997 and 2002. As this is based on reported labor income, it only captures a small portion of the rural economy.
19
Table 9: Disaggregation of the Poverty Gap by Household Characteristics (Total Bolivia)
Moderate Poverty Gap Extreme Poverty Gap
1989 1994 1999 2002 1989 1994 1999 2002
Total 45.45 41.89 32.53 32.94 27.53 25.21 15.73 15.32 (0.35) (0.25) (0.34) (0.22) By Hh Size
(1.00) (0.59) (0.59) (0.33) Sectoral Employment of Head
White Collar 24.06 (0.72)
15.60 (0.62) 13.17 9.68
11.16 (0.58)
5.79 (0.37) 3.68 2.62
Blue Collar 43.84 (0.75)
39.06 (0.56) 30.31 32.63
24.62 (0.67)
19.59 (0.49) 11.28 13.40
Agriculture 65.51 (0.59)
68.22 (0.38) 52.21 48.35
46.15 (0.72)
50.73 (0.47) 31.60 26.52
Sales & Services 37.27 (1.06)
30.26 (0.69) 22.80 19.45
19.11 (0.89)
13.19 (0.50) 8.72 6.07
Not Employed 48.52 (2.33)
43.16 (1.63) 32.45 29.57
29.80 (2.32)
24.05 (1.49) 19.15 13.67
No Partner 41.42 (0.89)
36.80 (0.61) n.a. n.a.
23.55 (0.79)
20.95 (0.50) n.a. n.a.
Source: Own calculations. Standard errors are in brackets (only applicable to the simulated poverty rates). For 1999 and 2002, we use the actual poverty rates. In the DHS, the employment of the head is not listed directly and can only indirectly be inferred by the employment of the partner of the women who was the respondent. In some cases, this partner of the respondent might not be the head of household so that there might be some inaccuracies here. For 1999 and 2002, the head’s employment is listed in the data and we do not need to rely on the partner’s employment status (and thus the option of ‘no respondent’ no longer exists). The change in poverty between 1994 and 1999 by head’s employment should therefore be interpreted with some caution.
20
Box 1: Population Growth, Household Size, Poverty, and Pro-poor Growth Bolivia has a surprisingly high population growth rate. The intercensal annual growth rate of the population went up from 2.1% between 1976 and 1992 to over 2.7% between 1992 and 2001 (INE 2003a). Based on revised census counts, officials at INE argue that the correct intercensal growth rates would be 2.4% for both periods, which still indicate very high population growth by South American standards (see Table 1). Part of the high population growth is due to continued high fertility. The 2001 census estimates the TFR to be at 4.4, while the DHS 2003 reports it to be at 3.8 (INE, 2003b; INE, 2004). The impact of this TFR (and the much higher levels of TFR in the past) generates a considerable demographic momentum through the impact of large increases in the number of women of reproductive age. The second source of high population growth has been a sharp fall in mortality levels in the past 15 years and as such is a welcome development (see Table 2).
There has been a considerable fertility decline in the past 20 years, which is now clearly visible in the age structure of the population where the absolute number of 0-4 year olds has recently begun to decline. If these trends continue, Bolivia will soon be about to enter the phase which has been referred to as a ‘demographic gift’ by Bloom and Williamson (1998), where the share of the working age population will be particularly large (and dependency rates correspondingly low), enabling the country to save more, to invest more in the quality of children, and, if employment opportunities are there for the large working age population, to boost growth of per capita incomes.
The ‘demographic gift’ is likely to make growth more pro-poor as it is particularly the poor who are now in the process of further reducing their household size and thus benefiting from reduced dependency rates (see also Klasen 2003; Klasen and Woltermann, 2004; Eastwood and Lipton 2001). This can be seen when considering two factors. First, as shown in Table 8, poverty rates are highly correlated with household size. This also holds if we calculate poverty rates based on adult equivalents (rather than based on per capita incomes), which assumes that children need fewer resources and that households benefit from considerable economies of scale (see Annex 2). More importantly, it appears that the poverty risk of household size has sharply increased over time. Based on adult equivalent incomes, the poverty rate of households with more than 6 members was nearly 20 percentage points higher in 2002 than of those with less than 4 members, up from a difference of less than 10 percentage points in 1989. The differential has similarly widened when using per capita incomes.
Second, the poor have much larger families and thus disproportionately suffer the costs of large families. Using the unsatisfied basic needs index and applying it to the 2002 Census, ‘marginal’ households have a TFR of 6.9, compared to a rate of 2.1 for those with satisfied basic needs. If fertility decline reaches the poor, it is likely to have a major impact on poverty reduction as it did elsewhere in recent years (e.g. in East Asia and in countries such as Brazil).
Policies that would further such a development would be a combination of further improvements in education and health access for the poor, combined with the availability of low-cost family planning for the poor which still appears to be a problem among uneducated women in some rural areas (see INE 2004).
21
Box 2: Gender and Pro-poor Growth
Compared to other Latin American countries, Bolivia had considerable gender inequality in a variety of indicators of well-being, human capital, access to resources, and income earning opportunities. For example, as late as 1976, there was a 24 percentage point gap in literacy rates among adults (INE, 2003). On many fronts, there has been considerable progress in closing these gaps. The gap in education has closed the fastest. In 2001, the gap in literacy rates has narrowed to 12 percentage points (with the remaining gap being largely due to past discrimination) and gaps in enrolments or progression are now limited to a few pockets in more remote municipalities (Anderson and Molina, 2004). As shown in the international literature on the subject, the closing of the gender gap in education could have important positive effects for growth and human development (e.g. World Bank, 2001; Klasen, 2002). This is due to the direct impact of the removal of an artificial distortion that limits the potential of women to contribute to economic development and through the indirect impact female education has on fertility, mortality, and education of children. The impact of female education on fertility is well-documented in Bolivia. Females with more than 12 years of education have only 1.9 children, compared to 6.7 for (the by now very few) females with no education (INE 2003b). Thus the closing of the gender gap will further accelerate the on-going fertility decline with the potential benefits described in Box 1.
As shown in the poverty profile below, it is also noteworthy that female-headed households are generally less poor than male-headed households, which is a common finding in Latin America but much rarer elsewhere (e.g. Marcoux, 1998). But one should caution that female-headed households represent a very heterogeneous group of households (e.g. single female elderly, single professional women, divorced women with or without children, women of migrant workers with or without children, etc) so that it may well be that sub-groups are particularly vulnerable to poverty (an issue that deserves further examination). Less positive is the record on female opportunities in employment. Here we find that females have much fewer employment opportunities, making up only about a third of formal sector employees, while constituting about 50% of informal and self-employment. In all three sectors they then suffer from considerably wage gaps with gender having one of the largest effects on wages (Tannuri-Pianto et al. 2004); these are also considerably larger than in other Latin American regions (World Bank 2004b). Also, female migrants (who are over-represented among rural to urban migrants) were, prior to 2002, had to accept lower wages than the (already depressed) female wages of non-migrant counterparts in urban areas (Pianto et al. 2004). As shown in Klasen and Lamanna (2004), such discrimination in employment has also been found to reduce economic growth due to the distortion such discrimination brings about. Moreover, it is likely to have an adverse effect on poverty reduction as female earnings increase their bargaining power within households which has been found to increase investments in education, health, and nutrition of children (e.g. Thomas, 1997). Lastly, a particularly worrying development is the high incidence of domestic violence in Bolivian households, particularly against women. About 10% of women report being beaten regularly, nearly half occasionally, and some 15% of women report occasional incidences of forced sexual relations (INE, 2004). These problems are not only well-being issues for the women concerned, but also clearly affect their and their children’s ability to contribute to, and profit from economic development opportunities. In sum, there has been notable progress in closing gender gaps, particularly in education, but gender gaps in employment and pay as well as an unusually high incidence of domestic violence continue to generate formidable obstacles for women to contribute to pro-poor growth.
22
36. Segmentation of the Bolivian Economy and its Impact on Growth and Poverty. Based on the findings above, particularly the large gaps by region, education, and economic sector, it is clear that Bolivia’s economy suffers from considerable segmentation, with the poor being largely separated from the income-generating and growth processes that tend to favor urban areas as well as resource-based sectors and modernized agriculture. In this section we want to discuss various forms of segmentation of the economy in some more detail.
(i) Urban-rural and formal-informal divide 37. The urban-rural divide in Bolivia is particularly strong as was shown by the poverty rates, the depth of poverty, and the poverty profile. The particular mechanisms leading to this large divide relate to initial conditions, the dynamics of internal migration, the educational system, and the urban labor market.
38. The initial conditions relate largely to Bolivia’s population distribution. Bolivia’s poor are heavily concentrated in the rural areas of the altiplano (highlands) and the valles, following Bolivia’s historical settlement pattern which focused on these areas. As these rural areas face difficult ecological and climatic conditions for agricultural production, and suffer from the proliferation of tiny plots, it is not surprising that poverty rates are higher there. In addition, Bolivia’s poor have been relatively slow to settle in the areas of dynamic economic development in the lowlands, partly for climate and health reasons as well as the lack of support networks in these areas. Thus much migration of the poor has involved moving to urban areas in the altiplano and valles as well as rural-rural migration within these areas (with particular emphasis on migration related to the coca economy) (CODEPO 2002, Pianto et al. 2004). In 1997, 46% of recent rural migrants went to other rural areas, presumably due to agricultural employment (esp. coca production) as well as family reasons. By 2002, this share had dropped to 37% (with metropolitan areas taking in a larger share of migrants), probably in line with the sharp decline in coca production (Pianto et al. 2004). Probably as a result of the economic crisis which is concentrated in the capital cities, return migration from them to rural areas (as well as to towns) was between 1997 and 2002 very large, making up about 43% of total migration flows between the three regions. It is important to note that female heads of households are over-represented among the rural-urban migrants who apparently see better economic opportunities in large cities; at the same time, migrants do not come from the poorest areas of the country. As far as the economic success of migrants are concerned, they are mostly able to earn as well as their urban non-migrant counterparts, which then raises the question why there is not more migration to equate the large earnings differentials between the regions (World Bank 2004b).18
39. In sum, migration currently does not appear to be a reliable mechanism for ensuring quick convergence of regional disparities and thus is currently unlikely to contribute much towards poverty reduction; this might be an issue for the attention of policy-makers concerned about the poverty impact of these regional disparities (see below).19 Another point of note is that rural-urban migrants retain a connection to rural areas to which they can return to, suggesting that the segmentation between the regions is not so relevant for this group. Lastly, not only has economic performance of smaller towns has outperformed rural areas and departmental capitals (also in terms of pro poor growth), but they are also the beneficiaries of considerable in-migration. If the past performance is any guide, encouraging and supporting migration from rural areas to these towns could make a significant contribution to poverty reduction.
18 It is important to note that females were not able to benefit economically as much from migration in the 1990s
although things appear to have improved (Pianto et al. 2004). 19 This is also borne out by regional growth figures which show that the departments with the lowest poverty rates grew
the fastest. It is true, that these departments (i.e. Tarija, Santa Cruz, Pando) were the targets of considerable in-migration from poorer departments but this did not enough to reduce disparities (see World Bank 2004b).
23
40. The large educational divide between urban and rural areas amplified this distinction. In 1976, the average years of schooling of the rural adult population stood at a dismal 1.8 years, compared to 6.1 years in urban areas (INE 2003a). While investments in rural education have led to better outcomes, the differences remain substantial. In 2001, average years of schooling were 9.2 in urban and 4.2 in rural areas (INE 2003a). In addition, analyses of the selectivity of migrants clearly show that young, well-educated people speaking Spanish as their main language have a much higher likelihood to migrate, thus contributing to a drain of skills from rural areas (Pianto et al. 2004). Taken together, these differentials continue to seriously compromise the economic opportunities of the poor population.
41. Third, tight regulation of the formal labor market especially with respect to dismissal protection and high costs of formality restrict the access of rural workers and the urban poor, who predominate in the informal and self-employed sectors, to income-earning jobs and keep employment in the formal labor market below levels that would otherwise be possible. This is exacerbated by particularly high institutional and regulatory barriers to formalization in Bolivia, which sharply reduces the incentive for firms to formalize (Kaufmann et al. 2001). As a result of these two problems, the share of informal employment in total employment is among the highest in Latin America (World Bank 2004b). In the 1990s, this was less of a concern since, as a result of other macroeconomic reforms, the demand for labor in the formal sector grew nonetheless and the participation rates in urban areas increased despite considerable rural-urban migration (Jimenez, Pereira, and Hernany 2001; Spatz 2004). However, when the Bolivian economy was hit by external shocks, the tight regulation became a more serious problem and reduced the employment opportunities for recent arrivals and urban informals, as evidenced by the drop of the formal sector share in urban employment to only 50% of employment in 2001 (from 55% in 1997) (Spatz 2004).20
42. While one should not see the urban formal sector as completely closed to rural-urban migrants and the urban poor (Pianto et al. 2004; Tannuri-Pianto et al. 2004), the conditions of entry into the formal sector are distinctly less favorable for these groups. In particular, participation equations in the formal sector suggest that formal sector employment is particularly difficult to achieve for women, for people with a non-Spanish mother tongue, and for the poorly educated (Tannuri-Pianto et al. 2004). In addition, selectivity-corrected earnings regressions show that earnings in the formal sector are much lower for these same groups suggesting that their ability to enter formal sector employment is restricted and happens under worse conditions, both factors that militate against urban formal employment as being an important tool for poverty reduction.
43. A last barrier associated with the formal-informal and rural-urban divide is the very restricted credit access for self-employed and informal producers. Despite the fact that some of Bolivia’s microfinance institutions have been hailed as models to ensure sustainable credit access, data show that the expansion of the portfolio of micro-credit institutions in recent years was associated with a contraction in the portfolio of banks, and that it only covers about 10% of the population operating in 68 of Bolivia’s 314 municipalities. In rural areas, credit is virtually unobtainable for anyone except very large producers, and also in urban areas it is highly restricted. These problems are exacerbated by little movement to restructure state wholesale finance institutions and years of inconclusive debate about the possibilities for bringing in informal institutions into the regulatory system.
20 Using different data and a slightly different definition, the formal sector share in urban employment dropped to only
32% of employment in 2002 (from 44% in 1997) (Tannuri-Pianto et al. 2004).
24
(ii) The ethnic divide 44. The urban-rural and formal-informal divide is posing particular problems for the non-Spanish populations of Bolivia, who predominate in rural and urban informal sectors. As shown by Andersen, Mercado, and Muriel (2003), there is very large inequality in educational attainments between indigenous and non-indigenous populations. Analyses of earnings regressions show that lower education levels and lower quality of education account for most of the earnings differences in urban and rural areas between Spanish-speaking and indigenous populations. It is important to note here that Bolivia has, compared to 13 other Latin American countries included in a comparative study, the lowest educational output in public education in terms of language test scores of fourth graders, while private educational institutions, which largely serve the Spanish-speaking urban populations, exhibit much higher scores (Mercado 2003). The earnings differentials are further exacerbated by occupational crowding of indigenous people in sales, agricultural, domestic service, and other low earnings occupations in the informal and self-employment sectors (Andersen, Mercado and Muriel 2003; Tannuri-Pianto et al. 2004). It is particularly interesting to note that this occupational crowding extends with equal force to public sector employment in education, health, administration and the like (Anderson, Mercado, and Muriel 2003). This adds to a perception of powerlessness among indigenous groups and their consequent mistrust of the government, which until recently had hardly any indigenous representation. It also can explain the finding that subjective poverty rates among particularly Quechua-speaking populations are even higher than objective poverty rates, presumably due to a felt sense of discrimination and powerlessness (World Bank 2004b). This sense is also supported by the finding of particularly low social mobility in Bolivia, compared to other Latin American countries, which transmits poverty intergenerationally (Andersen 2003).
(iii) Highlands agriculture versus the resource-based economy 45. These divides would be less important if it had been possible to ensure that productivity in highlands agriculture, the mainstay of incomes for many of the poor, had improved in past decades. But here, success has proved elusive for the majority of producers. Qualitative work by Tuchschneider (2001) shows that about 90% of highland producers find that their yields have deteriorated throughout the 1990s due to worsening climatic conditions (higher temperatures and less rain), lack of irrigation, deterioration of soil quality due to overexploitation, lack of land, population pressure and lack of modern inputs. Most development projects to support highland agriculture (which, interestingly, were concentrated on areas that were better integrated avoiding the most remote parts of rural areas) have been deemed unsuccessful by the respondents. Among the reasons cited are that they were often not focused on the central problem of declining agricultural productivity, did not take a holistic approach that addressed the technical, institutional, and economic aspects of the problem, and had little local participation, input, and support. Those few who claim to have benefited particularly value that the projects were focused on achieving substantial changes in the production and crop systems, improved access to irrigation and improved seed varieties and inputs. Given the importance of highland agriculture for employment and incomes, the failure to improve productivity there is critical. It is therefore not surprising that there was little improvement in rural incomes and poverty which is consistent with the international experience that stresses the importance of smallholder agricultural productivity for pro-poor growth (e.g. Eastwood and Lipton 2001; Timmer 1997, Klasen 2004).
46. In contrast, the most dynamic sectors of the economy have been capital-intensive, export-oriented lowlands agriculture, and the resource-based economy involving oil and gas, which are also highly capital-intensive with little linkages to the poor. In this context it is of particular importance to stress that the decline of tin mining since the mid-1980s (when the tin price collapsed and the government largely abandoned tin mining) took away one form of income from which the poor in some of Bolivia’s poorest provinces (particularly in Potosi) had benefited directly and indirectly.
25
47. Linking Growth to Poverty Reduction. From the analysis above it is not surprising to find high levels of inequality and poverty in Bolivia and the relatively poor record in poverty reduction in the past 15 years. If anything, it is somewhat surprising that during the 1990s the poor were able to improve their incomes somewhat (from a very low level), including in rural areas, which is true even in our sensitivity analysis (although the rate of pro poor growth is very low). It is likely that this is driven partly by temporarily favorable weather conditions in agriculture, a recovery from particularly poor conditions in the late 1980s, some spill-overs of the growth in the urban formal economy to the rural sector (through migration and remittance linkages), and the considerable importance of the coca economy which is likely to have had a significant direct and indirect impact on incomes and expenditures in some of the poorest rural areas (such as Chuquisaca, Potosi, and Cochabamba). Chapter 3: Factors Affecting the Participation of the Poor in Growth
48. In this chapter we are discussing the impact of initial conditions and policies on the participation of the poor in the growth process. We approach this question using two different methods. First, we discuss the role of initial conditions as well as macro and public spending policies on pro-poor growth in the past 15 years. We then turn to a formal assessment of the role of policies using a dynamic CGE model. While the particular policy simulations used are trying to assess the potential for pro-poor policy reforms as we look into the future, we also use these simulations to explain the record in the past.
a) Role of Initial Conditions 49. Bolivia’s initial conditions are generally unfavorable for achieving pro-poor growth. Bolivia suffers from being a large land-locked country with a poorly developed infrastructure that increases the remoteness of many of the poor from markets and potential income-earning opportunities. Remoteness is also likely to account for the high poverty rates and the few changes in Bolivia’s most remote and poorly developed highland and valley provinces (particularly Potosi and Chuquisaca). For those poor the otherwise positive impact of a liberalizing and integrating economy has been found to be much smaller (World Bank, 2004b), a finding similar to Christiaensen, Demery and Paternostro for a sample of African countries (Christiaensen, Demery and Paternostro 2002). Thus whatever growth has taken place has been concentrated in areas with lower poverty, while the poorest provinces have experienced below average growth and thus lower poverty reduction. The high initial income, land, and ethnic inequality is making poverty-reducing growth more difficult to begin with (World Bank 2000, Klasen, 2004) and the high ethnic fractionalization is making policy-making for pro-poor growth particularly difficult (e.g. Alesina et al. 2003). In fact, Alesina et al. (2003) cite Bolivia as a particular example of how high ethnic fractionalization contributed to poor policy choices and poor quality of public goods in the 1980s. While governance clearly improved after the crises and reforms of the mid-1980s, governance indicators remain weak compared to Latin American averages in past years (Kaufmann et al. 2003, see also below). A particular governance challenge that was inherited from decades of military rule and coalition governments was the rise of patronage relationships and informal procedures and practices in the public sector, which in turn supported a growing informalization of the private economy which severely restricts its growth potential (Kaufman et al. 2001).
50. Lastly, Bolivia is hurt by an economic structure where (with the exception of coca) its most lucrative exportable products, particularly oil, gas, some mining products, soya and other agricultural exportables, are highly capital-intensive and provide few linkages to the poor. This is partly due to nature and geography, but also linked to the inability to develop a dynamic smallholder agricultural sector with a great potential for import substitution and exports. As a result (as will be shown below in some of the simulations), the potential impact of these export products for poverty reduction is small and can even be negative if the Dutch Disease effects dominate the otherwise potential beneficial effects of greater exports for poverty reduction.
26
Box 3: Land Distribution and Reform and Pro-poor Growth As part of the revolutionary government of Victor Paz Estenzorro in the early 1950s, Bolivia enacted comprehensive land reform that included a freeing of peasants from labor duties and gave them access to some land. Despite some short-term success, the reform did not solve the problems of poverty in the highlands. Instead, subdivisions and increasing land pressure led to increasingly smaller plots with little modern inputs. The process largely stopped in the 1970s and due to the emergence of landed estates in the lowlands, land inequality in Bolivia is as high as elsewhere in Latin America. A new push for agrarian reform came with a new land reform law in 1996. The focus here is on titling although the expectation was that this would lead to restoration of illegally seized land, redistribution of public lands, and a renewed push for land reform. Unfortunately, despite relatively good progress in titling (which is nevertheless somewhat behind schedule by now, see Anderson and Evia, 2003), little redistribution has happened as a result of this process, leading to considerably disappointment with land reform process and calls for redistributive land reforms. Plan Tierra tried to address this by redistributing some public lands in specific projects, but this was widely seen as too little, too late, and not broad enough. Landless movements have sprung up and demands for further land reforms are considerable. At the same time, it is clear that there is not much land available for redistribution in the highlands and valles (apart from some public lands) and most land for redistribution would be in the lowlands. At the same time, there might be some scope for land reform by using land taxes to bring underutilized private to the market, a subsidy and support program for emerging farmers, and further redistribution of public lands. Land reform is among the most pressing social and political issues currently under discussion and it is clear that political and social stability will depend on successful resolution of these issues. Source: Urioste (2003), World Bank (2003b).
b) Macro Policies and Pro-poor Spending 51. Bolivia’s macro and public expenditure approach to poverty reduction can be seen as closely following the World Bank’s 1990 model for poverty reduction, which focused on developing a growth-oriented strategy and accompany it with investment in human resources and safety nets for the poor (World Bank 1990). Consequently, the growth agenda was largely seen as separate from a poverty-reduction agenda in the belief that high growth would ensure sufficient poverty reduction. This should be aided by investment in human resources that could then allow for a participation of the poor in the growth process. For those who were left behind, safety nets (such as the Bolivian Social Fund and public works programs) should try to address this problem. While this approach worked as long as growth was satisfactory and there were increasing resources to be used for expanding social sector spending, it failed to sufficiently promote the productive activities of the poor and in addressing the large equity problems permeating Bolivia’s society.
52. Macro policies and poverty. As already discussed in Chapter 1, the macro policy agenda was mainly one of stabilization and liberalization. One particular focus of macro policy was to ensure low and stable inflation and this has been achieved for the past 20 years (see Table 2). Given that inflation tends to hurt the poor disproportionately, this macro policy is likely to have supported poverty reduction. In addition, the external capital account was liberalized and a friendly foreign investment regime was established. While this allowed a significant increase in foreign direct investments throughout the 1990s (Schweickert et al. 2002, World Bank 2004a) the liberalization of the external capital account contributed to a further increase in dollarization of the economy which severely limited the room to maneuver as far as exchange rate policies are concerned. There were no efforts to combat dollarization, which remains high to this day. While the policy of allowing
27
dollarization per se is neither pro- nor anti-poor, it increases the vulnerability of the economy to exchange rate shocks (such as the appreciation of the US$, or the sharp devaluation of the Brazilian and Argentinian currencies as it occurred in the 1998-2001 period) which can hit the poor more as they are unable to shield themselves against such shocks. Also, it was not possible to use exchange rate policies to affect the distribution of income through the use of a real depreciation that would favor poor export and import-competing producers at the expense of wealthier import-consuming parts of the population (see Klasen 2004). In addition, strong trade liberalization, while improving the allocative efficiency of Bolivia’s economy, further undermined the ability of Bolivia to manage its external environment to support poor producers, particularly in an environment of sharply fluctuating currency movements with its trading partners.
53. Fiscal and public expenditure policies. In the areas of fiscal and public expenditure policies, the pro-growth agenda initially dominated policy-making and fiscal policy aimed at low budget deficits, which was achieved through tax reforms, prudent expenditure policies, and divestiture from loss-making state-owned enterprises. Tax reforms largely focused on broadening the tax base through a value added tax and a transactions tax which together make up some 60% of tax revenues in 2002. A hydrocarbons tax is the only other significant tax generating about 18% of revenues (Servicio de Impuestos Internos 2003). As a result, there is no progressivity in the tax system, which could be achieved via an income tax on those employed in the formal sector, a serious tax on large land-holdings (also to facilitate the sale of underused land by large landowners) or other real estate, or significant surcharges for particular items mainly consumed by the non-poor.21
54. As long as growth was relatively high in the 1990s, and tax revenues continued to rise, the government was able to maintain relatively low budget deficits while at the same time ensuring rising expenditures for priority social sectors, such as health and education. Public expenditures as a share of GDP rose sharply in the 1990s, and also public capital expenditures were high by international standards (World Bank 2004a); excluding social security, Bolivia now devotes the second-highest share of its GDP to public social expenditures in all of Latin America (World Bank 2004c).
55. This was also supported by generous aid flows and complemented by funds made available by HIPC I and II, with the latter being channeled entirely to the municipalities to fund priority investments, mostly in the social sectors and recently also in infrastructure.22 Public expenditures in health, education, and infrastructure do reach the poor, although to varying degrees (World Bank 2004a, c). In particular, while the poor get the same absolute amount of resources as the rich of public health expenditures, they get slightly more than that in public primary education (the poorest 40% capture about 50% of public expenditures on education, presumably due to their larger families and greater use of the public system). Public funds for secondary education are proportional or slightly pro-rich depending on the study (World Bank 2004a, c), largely due to an income gradient in enrolment rates, and public spending is sharply pro-rich at the tertiary level. As universities take up 32% of total education spending in 2003, the total effect of education spending is proportional, i.e. it reaches the poor about as much as the rich. Payments from the public pension system are strongly pro-rich with the non-poor (who make up 40% of the population in the study) collecting 83% of benefits. As a result of this strong bias, total current public social spending was pro-rich, i.e. with 54% of the benefits going to the non-poor and only 46% to the poor.
21 Such taxes could, however, face compliance problems in an economy where contraband is widely available and tax
evasion is considerable (Delgadillo 2000). In 2003, a few taxes were changed to increase tax revenues but they did not seriously affect the progressivity of the tax system.
22 The World Bank Public Expenditure Report notes that from 1996 to 2001, 75-80% of municipal investment was in the social sectors, while in 2002 infrastructure received 34% with social sectors falling to 58% (World Bank 2004a). Investments in productive activities were always low and only amounted to 3-6% in the period.
28
56. Investments in infrastructure at the municipal level are also strongly pro-rich, with the richest quintile capturing about twice the absolute amount of subsidies as the poorest fifth in 2000 (World Bank 2004a). This bias has been considerably reduced in the last several years, partly as a result of the National Dialogue Law, which required spending from HIPCII resources to be targeted to poorer communities (World Bank 2004c).
57. Thus public spending presents a mixed picture as far as its poverty and equity impact is concerned. While it does reach the poor to some extent, it could do so much more than is currently the case. Better targeting could be achieved through demanding greater co-payments for health and education for the non-poor, as well as boosting access to health services and enrollments in secondary and higher education for the poor (through dedicated programs including subsidies). Moreover, as the pension system overwhelmingly benefits the non-poor, limiting the obligations would automatically reduce the anti-poor bias of the overall social expenditure system.
58. The ability to combine the maintenance of fiscal discipline with rising social expenditures collapsed in the late 1990s leading to ever-rising and now unsustainable budget deficits (reaching 9% of GDP in 2002). Three factors largely account for this deterioration (World Bank 2004a). First, tax revenues plummeted in line with the economic slowdown that began in 1999, while expenditures continued to rise. Second, a mismanaged pension reform led to much higher than anticipated costs. It is now costing 5% of GDP while providing benefits to only 2% of the population, most of them non-poor formal sector retirees in urban areas. Third, due to Bolivia’s decentralization program and the dedication of HIPC funds to the municipalities, there is little central control or even information over expenditures, thereby weakening the ability of the central government to maintain fiscal discipline.
59. As will be discussed in more detail below, the bargain that maintained economic growth and social stability in the 1990s and involved the continuation of an economic model fashioned on the Washington Consensus, while ensuring social stability through significant transfers to the social sectors, thus became unstuck. It appears that in a situation where inequality, social and ethnic tension are high, such a bargain proves unsustainable and extremely fragile. It has contributed to great opposition to the government, calls for populist reforms, and now has thrown the debate about the appropriate economic model wide open in the form of the constitutional assembly that will now address these issues. A continuation of the current bargain seems not possible, nor is it fiscally feasible, and a new approach to economic policy-making is required.
c) Assessment of the Impact of Shocks and Policies on Pro-poor Growth 60. In this section, we make use of the dynamic CGE model described in Annex 2 to assess the impact of shocks and policies on Pro-poor growth. While the general approach in this section is to simulate forward-looking policies and focuses on the impact of shocks and policies on the ability to reduce poverty and inequality, we will also comment on the extent to which they are able to explain past performance in growth, inequality, and poverty reduction.
61. In its Poverty Reduction Strategy Paper (PRSP), which was completed in May 2001, the Bolivian government formulated ambitious social goals to be achieved over the period 2001–2015 República de Bolivia 2001). Among the improvements the PRSP envisages are the following targets with respect to income poverty:
a reduction of the nationwide poverty incidence from 63 % to 41 %;
a reduction of the urban poverty incidence from 47 % to 32 %;
a reduction of the rural poverty incidence from 82 % to 52 %.
61. Success in reaching these and other social targets will to a large extent depend on Bolivia’s ability to achieve higher growth. The PRSP calls for average growth in excess of 5 % over the
29
period under consideration, compared with an average growth rate of about 4 % in the 1990s. It acknowledges that faster growth will require additional structural reforms – in particular a more flexible labor market and a more efficient tax system – which enable the country to boost private investment, and that only measures specifically tailored to poverty groups, such as investments in rural infrastructure, can make growth more pro-poor than in the past.
62. So far, the expectations raised in the PRSP have not materialized. During the protracted economic slowdown of the last 5 years, both per capita income and the incidence of poverty have stagnated at best. In a recent revision of the PRSP, the Bolivian economy is projected to grow at an average rate of 4.8% between 2006 and 2015 (UDAPE 2003a), which is somewhat below the original projections. Moreover, poverty elasticities with respect to overall growth have been revised downwards from -0.77 to -0.60 and from -0.54 to -0.26 for urban and rural areas, respectively. Given these estimates, which are extremely low in international perspective, and the target growth of 4.8%, the headcount index is now only expected to fall to 54 % nationwide and to 45 % and 75 % in urban and rural areas, respectively, until 2015.
63. Against the background of these fairly disparate projections, Bolivia’s prospects for achieving pro-poor growth will be evaluated using the CGE model. In particular, it will be examined how external shocks and policy reforms in different areas, ranging from macroeconomic stabilization to poverty-focused interventions, might affect the trajectory of the Bolivian economy and the evolution of poverty.
64. While assessing model results, two central characteristics of the model have to be kept in mind. First, economic growth is determined by changes in the endowments of the primary factors capital and labor as well as the efficiency with which these factors are used. As far as efficiency of using factors is concerned, the model assumes an exogenously given rate of TFP growth of 2% per year.23 Thus all changes in the simulations will depend on changes in the primary factors capital and labor. The major driving forces of labor dynamics are population change, migration, the rate of labor productivity growth, and the change in human capital. Of these, the model only takes account of population changes, which are kept constant over simulations, and migration. The driving forces for capital accumulation are domestic savings and foreign capital inflows as well as relative returns on financial (domestic and foreign) and physical assets. Since net capital inflows are exogenously given in most simulations24, differences in growth rates across simulations are the result of changes in total domestic savings. Second, the model assumes a full employment situation for all types of labor and capital categories in each period of the simulation horizon. Hence, unlike other models for Bolivia that analyze short-run issues (e.g. Jemio 2001; Jemio, Wiebelt 2003; Thiele, Wiebelt 2004), this model neglects Keynesian multiplier effects that might result from changes in current and investment expenditures. With the exception of some taxes and limited intersectoral mobility of certain factor categories, there is essentially no difference between the technologically feasible production possibility set and the resulting transformation set reflecting market behavior and institutional characteristics of the economy. All markets are cleared and overall output is almost fixed within individual periods. Taken together, these two model characteristics imply that the model cannot be viewed as a short-run projections model and is not intended for that purpose. It is better suited to explain medium- to long-term trends and structural responses to changes in external conditions and development policy. But the problems facing pro-poor growth are inherently longer- 23 Two comments are in order. First, while the assumption of TFP growth of 2% is relatively high (slightly higher than
was found for the 1990s), for an optimistic baseline scenario it appears realistic. Second, one may speculate that some of the policies proposed will have an impact on TFP growth. This is possible but there is very sparse literature on this subject, both in an international as well as in a Bolivian context. One might therefore speculate that the effects of some policies might be larger because of knock-on effects on TFP growth.
24 Exceptions are the simulations of declining capital inflows, where capital inflows are changed exogenously, and of a devaluation, which changes the domestic currency value of capital inflows.
30
run problems, and the model thus naturally reflects an emphasis on the long run rather than on the short run, on growth rather than stabilization, and on trends rather than cyclical variation.25
(i) An Optimistic Baseline Scenario 65. A scenario that describes how the Bolivian economy might evolve in the absence of shocks and policy changes serves as a benchmark against which all alternative developments will be evaluated. In this scenario, the economy exhibits smooth economic growth of about 4.7 percent on average over a ten-year period (Table 10)26, where economic growth is driven by capital accumulation, (exogenous) growth of the labor force, and (exogenous) technical progress (2% TFP growth). This not only describes an optimistic forward-looking scenario, but is also a good description of the record of Bolivia in the 1990s. The growth process is associated with roughly constant domestic savings and investment ratios, which implies that the large savings gap is not closed over time. The continuing savings gap corresponds to a persistent current account deficit, and both are reflected in a fairly stable real exchange rate.
66. In line with past experience, the structural change projected in the baseline scenario is rather moderate. The shares of the broad aggregates “Agriculture”, “Industry” and “Services” in total value added barely change over time. More pronounced shifts of resources are taking place within these three sectors. Within agriculture, for example, the more productive export-oriented segment gains at the expense of the traditional, subsistence-like segment. The same pattern prevails in the services sector, where higher productivity growth and a higher income elasticity of demand raise the provision of formal relative to informal services.
67. From a distributional point of view, the baseline scenario suggests that without further policy reforms and without external shocks the rise in urban inequality observed over the 1990s will continue, and that the rural-urban gap in income levels will widen. In addition, inequality within rural areas will also increase. In both urban and rural areas inequality is already at very high levels, which is why aggregate growth in Bolivia barely translates into poverty reduction. As the following figures indicate, this holds in particular for rural areas. In the course of the simulated 10-year period, the national headcount merely declines from 63.6 percent to 55.3 percent. The moderate reduction results from a decrease in the urban headcount from 49.7 percent to 38.9 percent, and a reduction in rural poverty of only 4 percentage points from 86.9 percent to 82.8 percent. Even under this optimistic scenario, Bolivia would just manage to reach the revised national poverty reduction target. According to our model, however, poverty reduction in rural areas falls short of the reduction predicted in the revised PRSP, while urban poverty declines faster.
(ii) Accounting for External Shocks 68. The assumption made in the baseline scenario that no external shocks occur during the simulation period is highly unrealistic in the Bolivian context. Most predictably, the agricultural sector is recurrently hit by the El Niño phenomenon. Furthermore, a resource-based economy such as Bolivia’s has to cope with volatile terms of trade for primary goods. And as the Brazilian crisis forcefully demonstrated, Bolivia’s small open economy can hardly avoid being affected by instabilities in neighboring countries. In particular, the experience of the Brazilian crisis suggests that the high level of foreign capital inflows realized in the mid-1990s should not be taken for granted. Doubts about the sustainability of these inflows are reinforced by the fact that most foreign
25 As such, it is also not possible to use the model to track the actual development over the 1990s where short-run
cyclical variations play a considerable role. Instead, we can comment on the impact of the various shocks and policies on past growth and poverty reduction.
26 For more detailed information, see Annex 4, which also includes other poverty measures and inequality measures. It should be pointed out that the poverty gap often shows larger results than the poverty headcount, as has been the case in the past.
31
direct investment was related to the capitalization process, which was more or less completed at the end of the 1990s.
69. El Niño. Among the external shocks threatening Bolivia, El Niño carries the highest cost in terms of short-run agricultural output losses. An El Niño shock of average size may lower GDP growth by about one percentage point in the year of its occurrence. Since this is only partly compensated by higher growth in subsequent periods, and since El Niño tends to occur every three years, the losses add up to significantly lower average growth rates (Table 10). In addition, the fall in agricultural exports during El Niño years leads to a deterioration of the current account balance and a real depreciation of the Boliviano. The real depreciation, in turn, gives rise to a reallocation of resources from inward-oriented sectors such as utilities, construction, and informal services to more outward-oriented sectors such as intermediate goods and mining.
70. The direct distributional consequence of El Niño is that smallholders and agricultural workers suffer income losses. The same is true for employers, who obtain a significant share of their capital income from investments undertaken in modern agriculture. The negative impact on these three household groups is somewhat dampened by a slight increase in domestic agricultural prices as a result of supply shortages.27 The price increases are not strong enough, however, to show up in lower real incomes for urban food consumers. In urban areas, the only major effect of El Niño on household incomes runs via the real devaluation, which makes the providers of non-traded informal services worse off. By contrast, the overall income position of non-agricultural workers and employees is hardly affected as their gains in tradable sectors tend to offset their losses in non-tradable sectors. The decline of urban informal income results in a quite considerable increase in the urban poverty incidence. In some periods, the urban headcount increases by more than 1.5 percentage points compared to the baseline scenario. The rise in urban poverty is comparable to that in rural areas where the headcount is on average about one percentage point higher than in the base run. Inequality rises somewhat within urban areas, and quite considerably in rural areas, which is mainly due to the fact that the losses of the employers in modern agriculture are less pronounced than those for rural workers and smallholders.
71. Terms-of-Trade Shock. Terms of Trade shocks, such as the 10-percent decrease in world market prices for agricultural and mineral products considered here, differ from supply shocks, such as El Niño, in one main respect: they do not impair production capacities and thus do not lead to major output losses as long as the economy operates at or near full employment.28 Rather, the direct effect of the Terms of Trade shock is to reduce relative output prices and the incomes earned in the affected sectors, which induces a shift of production towards non-affected sectors. Further economy-wide repercussions are caused by the fall in agricultural prices, which lowers intermediate input costs for processed agricultural products and thereby boosts output in the consumer goods sector, and – analogous to the case of El Niño – by the real devaluation that follows the widening of the current account deficit.
72. The negative income effect of the Terms of Trade shock is most strongly felt by agricultural workers as the export-oriented modern agricultural sector, where they earn their living, experiences a marked decline. For smallholders, who mainly serve the domestic market, the drop in real income levels is much less severe. For the urban household groups, various factors work in opposite directions, and they appear to cancel out. Urban informals, for instance, benefit from lower food prices, while the real devaluation and an increase in rural-urban migration exert downward pressure
27 It is assumed that domestic agricultural prices are to a large extent determined by world market conditions so that
there is only limited scope for independent domestic price movements. 28 The assumption of full employment, which does not rule out the existence of underemployment, seems to be
justified at least in non-recession years where open unemployment tends to be very low. But even shocks which mainly affect agriculture are unlikely to drive up open unemployment as the rural population cannot fall back on a social safety net.
32
on their real income. Therefore, urban poverty remains virtually constant, whereas poverty in rural areas increases slightly. In rural areas, the income distribution worsens to some extent because of the income losses experienced by the workers in modern agriculture.
73. Declining Capital Inflows. If foreign direct investment (FDI) falls by almost a third, as has been the case in Bolivia in the year 2000, this causes only about half the immediate output losses of El Niño, but the impact turns out to be much more persistent.29 Even after ten years, growth has not fully recovered. The fall in FDI lowers the domestic investment ratio by about 2.5 percentage points, which narrows the savings gap because domestic savings hardly change. Correspondingly, the current account deficit improves, with rising exports and falling imports. The losses incurred at the macro level are spread over all sectors using formal capital, and over those producing capital goods. Since traditional agriculture and informal services lack access to formal capital, smallholders and urban informals are the only household groups, which do not immediately suffer from the shock. A real devaluation of the Boliviano, however, indirectly hurts the urban informals, while it benefits modern agriculture to such an extent that the sector’s loss of FDI is overcompensated.
74. Temporarily, smallholders and agricultural workers gain slightly from the shock, but lose somewhat in the medium run as a result of lower economic growth. This drives the evolution of the rural poverty incidence, which decreases slightly in the shock period, but increases by up to 0.7 percentage points afterwards. Over the entire simulation period, the rural income distribution improves quite considerably. By contrast, all urban households are negatively affected by the decrease in FDI. During adjustment to the shock, the urban headcount increases by almost two percentage points compared to the baseline scenario. The urban income distribution worsens somewhat, which is mainly due to the very pronounced decrease in urban informal income.
75. A similar drop in portfolio investment, which was the main consequence of the Brazilian crisis, strongly reinforces the negative short-run impact of the fall in FDI. During the initial year, both shocks combined drive the growth rate of real GDP down by about 1.5 percentage points. Since lower portfolio investment reduces growth only temporarily, the medium-run impact is dominated by the reduction of FDI. Average growth over the whole simulation period is about 0.3 percentage points lower than in the baseline scenario. The medium-run poverty outcome for the combined shocks is very similar to the FDI shock alone. In the short run, however, the impact on urban poverty is much stronger, with an increase of the urban headcount by 3 percentage points during the first period.
76. All in all, a realistic baseline scenario for Bolivia’s medium-run development prospects would have to acknowledge that under the current policy framework average growth rates are unlikely to lie markedly above 4 percent. Compared to the optimistic scenario, this implies worse prospects for poverty reduction.
77. Apart from helping to arrive at a realistic assessment of Bolivia’s medium-run development prospects, the simulations discussed above may also contribute to explain the past evolution of the economy.30 Over the 1990s, the foreign investment boom triggered by the capitalization process has in all likelihood been a key factor behind the comparably high growth rates until 1998. And according to our results, the poverty impact of foreign-investment-driven growth tends to be biased in favor of urban households, but does not completely bypass the rural economy. Compared to this effect, agricultural shocks have only played a minor role between 1990 and 1998. The only major exception is a severe El Niño, which in combination with declining prices for some primary products led to negative per capita income growth in 1992 and thus may have worsened the overall 29 Since a dramatic fall in FDI can be expected to lead to temporary open unemployment on a significant scale, the true
short-run losses probably exceed those reported here. 30 With a grain of salt, the simulation results may be used for the explanation of past developments because the
baseline scenario considered here is basically an extrapolation of the situation prevailing in the 1990s before the recent recession.
33
poverty outcome for rural areas in the early 1990s (as found above). A favorable external environment for agricultural production may at least partly explain the result obtained in Chapter 2 that the rural poor fared exceptionally well over the period 1994 to 1998. The recession that started in 1999 has in all likelihood hit urban households harder than rural households because the sharp decline in foreign capital inflows in the year 2000 has clearly dominated the agricultural shocks (El Niño, deterioration of terms of trade), which set in a year earlier.
(iii) Macro policies 78. One of Bolivia’s biggest achievements since the beginning of reforms in 1985 has been the containment of inflation by means of prudent monetary, fiscal and exchange rate policies.31 It might be argued that now, with an internal equilibrium that is firmly established, the exchange rate could be used to improve the external competitiveness of the Bolivian economy and affect its income distribution, given that the Boliviano has always been quite strong (Schweickert et al. 2003). The macroeconomic policy instruments are also still needed to bring about the real devaluations required in the face of negative external shocks.
79. Nominal Devaluation. A higher yearly devaluation of the Boliviano within the crawling peg regime causes an almost complete exchange rate pass-through to domestic prices, which will rise by nearly the same amount as the Boliviano is devalued as Bolivia’s ability to respond to higher import prices with competitive import-substituting domestic production is quite limited.32 The resulting real devaluation is therefore too small to provide the incentives for a significant reallocation of resources, and the minor real adjustment that occurs has no discernible effect on aggregate economic performance. Real effects, however, originate from the financial side of the economy, which is strongly and directly affected by the devaluation. This is because in the highly dollarized Bolivian economy the value of most assets and liabilities is indexed to the Dollar exchange rate. As a consequence, the net wealth position of net creditors in the financial system improves, while that of net debtors worsens. Since the economy as a whole – in particular the government – is a strong net foreign debtor, the overall wealth effect of the devaluation is negative. The deterioration of the domestic wealth position leads to a drop in aggregate real investment and a fall in the growth rate, which accelerates over time due to a compound interest effect. Among the household groups, the two richest, employers and employees are the only major actors on financial markets. Both are net creditors and thus benefit from higher net wealth and interest income, i.e. the revaluation of assets caused by the devaluation tends to reinforce existing wealth disparities. All other household groups are adversely affected by the negative growth effect. Unskilled workers and urban informals are most severely hurt because many of them are employed in the construction sector where production growth is lower because of lower real investment demand. As a consequence, urban inequality increases and urban poverty rises somewhat more than rural poverty. Thus such a policy would be counter-productive and poverty-increasing in the current environment of near complete pass-through to domestic prices, high dollarization, and large foreign debt. Conversely, tackling any of these three constraints successfully could change the assessment of such a policy considerably.
31 Since inflation is unlikely to be among the major policy issues in the foreseeable future, measures aimed at
containing inflation will not be analyzed here. 32 During recessions such as the current one, when capacities are underutilized, the pass-through will of course be
lower, but the question here is whether the exchange rate can contribute to improve Bolivia’s competitiveness in the medium run.
34
Table 10: The Impact of Shocks and Policies on Growth and Poverty
Average Growth
(%)
Nationala Headcount
Urbana Headcount
Rurala Headcount
Baseline Scenario 4.7 55.3 38.9 82.8
Terms-of-Trade Shock 4.7 55.6 39.1 83.3 El Niño 4.4 56.3 39.7 84.1 Declining Capital Inflows 4.4 56.3 40.3 83.3 Nominal Devaluation 4.5 56.7 40.4 83.9 Real Devaluation (restrictive monetary policy)
Gas Projects (constant government consumption) plus Labor Market Reform plus Tax Reform
5.8 52.0 33.5 83.0
Improved Access to Credit for Smallholders
4.7 55.2 38.8 82.8
Investment in Rural Infrastructure (high productivity effect)
4.8 55.0 38.6 82.4
Investment in Rural Infrastructure (low productivity effect)
4.8 55.1 38.7 82.5
Industrial Policy (modern agriculture) 4.7 55.7 39.6 82.7 Industrial Policy (consumer goods) 4.6 54.9 38.3 82.8 Transfer Program (lower government consumption)
4.7 53.8 37.9 80.5
Transfer Program (lower public investment)
4.5 54.6 38.9 81.1
Gas Projects plus Transfer Program 5.1 53.5 37.1 81.0 aRatio at the end of the 10-year simulation period. Please note that the initial poverty headcounts are: 63.6% national, 49.7% urban, and 86.9% rural.
Source: Own calculations based on the CGE model.
80. Real Devaluation. A real devaluation can be achieved if the Central Bank conducts a restrictive monetary policy. By constraining the opportunities of private banks to supply credit, such a policy temporarily lowers aggregate real investment demand and thereby exerts downward pressure on the domestic price level. The drop in real investment, in turn, causes a temporary economic slowdown. Specifically, it has a contractionary impact on capital-intensive sectors and on the sectors providing capital goods. This effect dominates the restructuring of production resulting from the real devaluation, which brings about an improvement of the current account. Household incomes are only moderately affected by these adjustments. While the investment slowdown makes non-agricultural workers, in particular construction workers, slightly less well off, the real depreciation entails minor losses for urban informals and minor gains for rural households. After the short-run adjustments, the economy soon shifts back to the old growth path, and household incomes evolve largely as in the base run. As construction output rebounds rather strongly, non-agricultural workers and urban informals even realize small medium-run gains so that urban poverty
35
declines somewhat towards the end of the simulation period. All in all, the negative impact on poor households often attributed to real devaluations is unlikely to occur in Bolivia.
(iv) Structural reforms 81. By Latin American standards, Bolivia has also made remarkable progress in the area of structural reforms (see, e.g., Lora 2001). The main exception is labor market reform, where Bolivia lags behind most other Latin American countries. Among the labor market distortions that still prevail, the segmentation of the urban labor market into formal and informal parts stands out. The tax system is another area where further reforms may be warranted. In particular, the question arises of whether the income tax, which hitherto has been of only marginal importance, should become a major source of government revenues.
82. Labor Market Reform. If the government makes it easier for urban informals to be employed as unskilled workers in the formal labor market, e.g. by lowering the costs of dismissal or by granting more options for temporary work, the obvious direct effect is that average real wages go down for unskilled workers and up for urban informals. Better earning opportunities in the urban informal sector, in turn, induce rural-urban migration on a significant scale, which moderately increases the incomes of those who stay in traditional agriculture. At the macro level, the efficiency gains achieved by reducing labor market segmentation – the wage differential between informal labor and unskilled labor is roughly halved – translate into average economic growth rates which are more than 0.3 percentage points higher than in the base run.
83. On balance, these developments cause negligible distributional shifts in urban areas because higher incomes for informals are offset by lower incomes for unskilled workers. Nonetheless, urban poverty decreases because of higher growth, but it takes some periods for the positive growth effect to materialize. The rural income distribution changes somewhat in favor of poorer groups due to the gains experienced by smallholders. This change and a slight increase in rural growth do not show up in the poverty headcount, but the rural poverty gap falls moderately (see Annex 2).
84. Tax Reform. A rise in income taxes for all household groups except smallholders and urban informals directly forces the two richest household groups, employers and employees, to consume and save less. For worker households the impact on disposable income is not strong enough to alter consumption and savings significantly.33 The main indirect effect runs via a tax-induced fall in aggregate private consumption, which lowers the prices received by smallholders and urban informals und thus worsens their real income position. The growth effect of the tax increase depends on how the receipts are allocated between consumption and investment. Under the assumption that the government broadly retains the original structure of expenditures, it is likely to be moderately contractionary in the medium run as the rise in public investment does not suffice to fully offset the fall in private investment. This in turn would have a negative impact on the factor incomes of all household groups.
85. If higher income taxes are combined with lower indirect taxes so as to arrive at a revenue-neutral tax reform, the economy-wide outcome is different.34 Lower indirect tax rates cause an expansion of capital-intensive industries (oil and gas, mining, intermediate goods), where the indirect tax burden is highest, and thus boost investment and growth. Overall, given the current tax structure, a revenue-neutral tax reform can be expected to improve Bolivia’s growth performance. As for household incomes, the decrease in indirect taxes raises private consumption expenditures, thereby offsetting the negative demand effect that higher income taxes have on smallholders and
33 The impact of the tax increase on aggregate poverty and income distribution cannot be calculated because the
household survey on which the distributional measures are based does not contain information on income tax payments.
34 To arrive at a revenue-neutral tax reform, tax rates in those sectors that are mainly subject to indirect taxes have to be lowered by roughly 20 percent.
36
urban informals. The main beneficiaries of the reform are non-agricultural workers, many of whom work in the mining and the intermediate goods sector, as well as in construction, which benefits from higher investment demand. The expansion of the construction sector additionally favors urban informals so that on balance their incomes are also significantly higher than in the base run. The gains of these two groups reduce the urban headcount ratio by up to 2 percentage points, and even rural poverty falls a little due to the growth effect.
(v) Gas Projects 86. Perhaps more than any macroeconomic and structural policy reform, the development of the natural gas sector promises to change the medium-run growth path of the Bolivian economy. Two large export-oriented hydrocarbon projects with Brazil and Argentina are already being implemented and another project involving the export of liquefied natural gas to North America has entered the planning stage but is currently on hold (IMF 2004). Taken together, these projects could roughly double the share of oil and gas in total domestic production from 5 to 10 percent within a decade, oil and gas could finally account for as much as 50 percent of total exports. Since the sector is an “enclave” in the sense that it uses negligible domestic inputs and generates little employment, its main link to the economy is through the fiscal accounts via increased revenues from taxes, and through its effects on the balance of payments – the current account improves and the exchange rate appreciates in real terms.
87. The natural gas boom translates into markedly higher economic growth. In 2008 and 2009, when the liquefied natural gas project is assumed to reach full capacity, the growth rate is likely to approach 6 percent.35 These average gains are likely to be somewhat underestimated because the upfront investment necessary to construct and develop large gas projects is not taken into account.36 The size of the growth effect will also depend on how the government uses its additional revenues. If the receipts are channeled into consumption, the average gains over the simulation period will only be about two thirds as large as if consumption growth is left constant and the resources are instead used to prop up public savings. Choosing the latter option would increase the overall domestic savings ratio by up to 3 percentage points compared to the base run, a remarkable improvement which macroeconomic and structural reforms are unlikely to achieve.
88. The real appreciation of the Boliviano, which in the peak years of the resource boom might reach 8 to 9 percent, leads to a contraction of export-oriented sectors such as modern agriculture, mining, and consumer goods, and an expansion of nontradables, in particular construction. This is the well-known Dutch Disease effect of resource booms, which, however, turns out to be rather moderate except for the two peak years. By keeping consumption growth constant, the government can slightly dampen the Dutch Disease effect. As a further economy-wide repercussion, lower consumer goods production reduces intermediate demand for agricultural raw materials so that modern agricultural activities contract even more, while smallholders suffer from declining prices as they can hardly adjust supply. A restructuring of final demand away from private consumption reinforces the pressure on smallholders’ prices and also hurts urban informals. Together with the fact that rural-urban migration rises considerably, this explains that urban informals are slightly worse off as a result of the gas projects even though they benefit from the real appreciation and the expansion of the construction sector. Overall, rural areas, i.e. smallholders as well as agricultural workers, suffer significant income losses, in particular in the two peak years. In urban areas, both unskilled and skilled workers gain, with the gain of skilled workers, who are for the most part employed in the public sector, being much more pronounced if government consumption expands.
35 The growth results obtained here come quite close to the projections reported in IMF (2004). 36 While most of the inputs, in particular capital goods, will need to be imported, some domestic activities such as
construction and business services might benefit during the early phases of the gas projects. The pipeline construction to Brazil is estimated to have contributed some 1.5% of GDP in 1996-97.
37
89. These changes in relative factor prices induce major distributional and poverty changes. From a national perspective, inequality increases substantially, which is due to both rising inequality between and within urban and rural areas. In the scenario with higher government consumption, the national Gini coefficient increases by about one percentage point. The results regarding the evolution of poverty during the gas boom are disappointing. Despite considerably higher growth rates, the decrease in nation-wide poverty is only moderate compared to the baseline scenario. More remarkably, rural poverty even increases substantially, with a rural headcount that falls by up to one percentage point. The rural poverty gap ratio, which during the second half of the simulation period is about 2.5 points higher than in the baseline scenario, illustrates that many of those who were already poor incur income losses.
90. A somewhat more favorable outcome could be expected if the government refrained from raising consumption expenditures. In this case, the headcount would be significantly lower in urban areas, but rural households would hardly benefit and thus would remain markedly worse off than without the gas projects. In addition, the rise in inequality would be somewhat less severe due to the dampened Dutch Disease effect, with an increase in the Gini coefficient of about 0.5 percentage points.
91. A fairly large medium-term boost for the Bolivian economy might become possible if the gas projects were combined with the structural reforms discussed above. Such a policy package could raise average economic growth by more than 1 percentage point and lower the national headcount by more than 3 percentage points. The gains would, however, exclusively accrue to urban households. They would benefit from a substantial drop in the poverty rate by more than 5 percentage points compared to the base run, while rural poverty would even rise a little. Clearly, such a strategy maximizes growth, but not pro-poor growth.
(vi) Targeted interventions in favor of the poor 92. The low poverty elasticities with respect to growth point to the problem that many of Bolivia’s poor are not well integrated into the economy and are additionally too poor to be lifted above the poverty line as a result of moderate economic growth. Among the policies which might help increase the productivity of the poor, particularly in rural areas where poverty is most persistent, improved access to credit for smallholders and investment in public goods such as rural infrastructure and agricultural research figure prominently (Thiele 2003). More direct ways of raising incomes of poor households could involve the subsidization of activities where many of the poor are employed (Klasen 2004), or the implementation of traditional transfer programs.
93. Improved Access to Credit for Smallholders. Efforts to improve credit availability for smallholders, e.g. by making land tenure more secure, are likely to raise investment in traditional agriculture significantly, albeit from a very low base. The impact tends to decelerate over time, but even after 10 years real investment could still exceed the base-run level by almost 50 percent. However, since the contribution of capital to sectoral value added is very small (and is assumed to have no impact on TFP growth), the investment boom only raises output by about 1 percent in the short run and by about 3 percent in the medium run. This supply response is too moderate to induce major adjustments in the rest of the economy. Aggregate investment rises slightly and average economic growth is less than 0.05 percentage points higher than in the base run. Smallholders’ real income position improves somewhat as a result of the output expansion, but this does not show up in the rural headcount, and even the poverty gap falls only marginally. Hence, the loosening of smallholders’ credit restrictions must be regarded as largely ineffective in the medium term, at least without further complementary measures. In the long term, such a policy is likely to be more beneficial as it allows the build-up of a capital stock and thus progressively raise the contribution of capital to value added.
94. Investment in Rural Infrastructure. Specific investments in public goods – e.g. the development of more productive crop varieties or the construction of rural roads – might constitute
38
one such complementary measure. However, even if public investments are tailored to smallholders’ needs, its impact is constrained by the difficult natural conditions prevailing in the Bolivian highlands.37 Here we consider two different scenarios: a fairly optimistic one where we assume that smallholder’s average output is raised by about 12 percent compared to the base run, and a more pessimistic one where the rise in output is only half that size. In both cases, the expansion of smallholder agriculture comes partly at the expense of modern agriculture so that smallholders realize income gains, whereas agricultural workers experience a less pronounced decline in wages. Although there is a countervailing force in the form of a small price decrease, smallholders benefit to such an extent that fewer of them migrate to urban areas. Together with a small real appreciation, this slightly improves the income position of urban informals.
95. Despite considerably higher income gains in rural areas, reductions in the urban and rural headcount are roughly equal. The difference between the two regions manifests itself in a significantly higher fall of the rural poverty gap (see Annex 4), which again reveals that many smallholders are far below the poverty line. In addition, inequality within rural areas decreases slightly. As for a comparison between the optimistic and the pessimistic scenario, all the mechanisms described above are more pronounced with stronger productivity effects of public investment. While this does hardly show up in economic growth rates and the poverty incidence, it is clearly reflected in the rural poverty gap, which goes down by more than 1 percentage point in the optimistic scenario as compared to 0.5 percentage points in the pessimistic scenario.
96. Industrial Policy. While the measures just mentioned aim at augmenting the asset base of poor households, a pro-poor industrial policy instead aims at raising the returns on existing assets, in particular on unskilled labor. One option in this area would be to support the development of modern agriculture. If the government, for instance, granted a 20 percent export subsidy, the sector would become markedly more important, particularly in terms of its share in total exports, which might increase from 15 to 25 percent. The expansion of modern agriculture is fuelled by rural migrants who are attracted by steeply increasing agricultural wages, and by a reallocation of capital. It is thus associated with lower output growth in traditional agriculture and in the capital-intensive sectors. The policy-induced structural change leads to an improvement of the current account, a small real appreciation of the Boliviano, and minor efficiency losses for the economy. With respect to household incomes, the out-migration of smallholders to some extent benefits those who stay in traditional agriculture. The migration effect on smallholders’ incomes remains limited because the workforce required in modern agriculture is very small compared to the number of smallholders. Lower production growth in capital-intensive sectors implies lower real incomes for workers and urban informals, which translates into moderately higher urban poverty over the whole simulation period. The rural headcount falls by almost one percentage point in the first period, but the deviation from the base run gradually disappears as the gains of agricultural workers become smaller over time.
97. An alternative option would be to support agriculture-based industrialization rather than primary agricultural activities. Subsidizing the consumer goods industry would entail stronger economy-wide adjustments than subsidizing modern agriculture and, as a consequence, efficiency losses would be higher. Most importantly, the Boliviano would appreciate considerably, and intermediate demand for agricultural raw materials would go up. The backward linkage to agriculture boosts the production of both agricultural sectors, but for modern agriculture this effect is overcompensated by the loss of competitiveness caused by the real appreciation. This implies that smallholders receive higher real incomes, whereas agricultural workers incur minor losses. The real appreciation improves the income situation of urban informals. Non-agricultural workers benefit from the expansion of the consumer goods industry, but as this expansion largely occurs at the expense of
37 Since little is known about the likely productivity effects of public investment in Bolivia’s highland agriculture, the
results presented here should be regarded as very tentative.
39
other sectors where they mainly work, their overall income position is not improved. Nonetheless, urban poverty declines substantially due to the gains realized by urban informals. The reduction in the rural headcount again tends to disappear over time, in this case because smallholders mainly benefit in the short run.
98. Transfer Programs. Transfer payments constitute the most direct means of enhancing the real income position of the poor. Here we assume that the government expands existing transfer programs so that gross incomes of the poor household groups are raised every year by roughly five percent compared to the base run. Thus these programs assume that it is possible to target such an enhancement of transfers to the poor, but that among the poor, the transfers are distributed in line with the receipts of transfers in 1999, i.e. no further improvement in targeting is assumed. As shown above, transfer and social expenditures are not particularly well-targeted, that is the effects could be more beneficial to the poor than shown here through better targeting (see also World Bank 2004a). Whether the impact of such programs goes beyond the direct beneficiaries largely depends on how the government finances the outlays. If it substitutes transfer payments for consumption expenditures, economy-wide repercussions are negligible and average growth is not affected. The only significant change is the fall in consumption expenditures itself, which leads to somewhat lower real incomes for public employees. As a consequence of the transfers that mainly benefit smallholders and urban informals, both urban and rural poverty falls markedly. The evolution of inequality appears to be less favorable. While the nation-wide Gini coefficient falls somewhat as the income change is stronger in rural than in urban areas, urban inequality remains constant and rural inequality even widens. These surprising regional results can be explained by the fact that the transfers tend to reach richer rather than poorer segments of the smallholders and urban informals.
99. Financing transfer programs through cuts in investment spending has a much stronger impact on the economy as it lowers aggregate investment and saving ratios by over one percentage point and thereby leads to reduced economic growth. The investment slowdown is most strongly felt in the construction sector, which implies that factor incomes of workers and urban informals decline. For urban informals, the decline is cushioned by a restructuring of final demand towards private consumption. The shift in final demand equally raises smallholders’ factor incomes so that they enjoy direct and indirect benefits. Overall, the secondary effects via the fall in investment fully offset the transfer-induced urban poverty reduction, whereas rural poverty alleviation remains sizeable.
100. As transfer programs have the biggest impact on reducing rural poverty, the combination with a natural gas project might yield a favorable scenario. We show such a combined scenario in Table 10 and it indeed is able to deliver higher growth and lower poverty in urban and rural areas. If such a policy package was also combined with labor market and tax reform as well as improvements in the targeting of transfers among the poor, the poverty impact could be substantial and conceivably allow the government to reach its PRSP targets also in rural areas.
101. To summarize the findings from the policy simulations, a few points are worth noting. Regarding an explanation of the impact of shocks and policies on pro-poor growth in the past, the following conclusions are warranted. First, one can nicely see how the evolution of poverty and growth in urban areas has varied with foreign capital inflows. Rural development has been more dependent on climatic conditions, and the lack of private and public capital. The failure to implement labor market reforms appears to have held back growth and urban poverty reduction. A deregulation of the urban labor market would also have had a positive if limited impact on rural incomes by providing an incentive for additional rural-urban migration.38
38 There are indications that the model underestimates the response of migration to changes in wage differentials.
Additional empirical research is needed to see whether the modeling of migration assumes the right amount of adjustment compared to actual migration patterns in Bolivia.
40
102. Looking forward, one can draw conclusions about the policy options for pro-poor growth as well as the constraints of Bolivia’s economy that the model analysis has served to highlight. Turning to the former issue, the main conclusion to be drawn from the model analysis is that, currently, the opportunities for achieving pro-poor growth are much better in urban than in rural areas. Given the available policy choices, Bolivia could clearly exceed the targets for urban poverty reduction set in the revised PRSP. Rural poverty reduction, by contrast, risks falling short of the targets due to a combination of recurrent external shocks and limited policy options. In particular, the implementation of gas projects is likely to bypass rural areas and significantly increase inequality. Improvements in access to credit and rural infrastructure have a positive but fairly small effect on poverty (particularly on the poverty gap). Only a coordinated policy package involving the gas project, labor market and tax reforms, and targeted transfer programs and interventions would allow Bolivia to achieve higher pro-poor growth and allow significant poverty reduction, also in rural areas.
103. Turning to the latter issue, the analysis has clearly shown up significant structural weaknesses of Bolivia’s economy that also need to be addressed in order to accelerate pro-poor growth. A critical weakness is Bolivia’s low domestic savings rate, which is contributes to Bolivia’s reliance on foreign capital inflows, the high degree of dollarisation, its high foreign debt, its vulnerability to external shocks, and its inability to manage its external trade and monetary environment to support pro-poor growth. A related second weakness is Bolivia’s high dependence on natural resources which have few linkages to the poor, but can have significant anti-poor effects. Third, Bolivia’ economy exhibits such a great degree of dualism that well-managed policies to generate higher growth do not reach the poor in rural areas or have little impact on their poverty. Lastly, Bolivia’s high initial inequality militates against success in poverty reduction, particularly in rural areas. While the policy packages discussed above can help with pro-poor growth, only success in tackling these four deep-seated issues will enable Bolivia to enter a sustainable path towards higher growth and poverty reduction.
d) Institutions, Political Economy, and Pro-poor Growth 104. This section will discuss selected institutional aspects of policy-making in Bolivia as they relate to pro-poor growth. We will focus particularly on an assessment of institutional weaknesses, then discuss the impact of decentralization on policy-making, and lastly consider the PRSP and National Dialogue process and its impact on policy-making for pro-poor growth.
(i) Governance weaknesses and their link to poverty and inequality 105. Based on the most comprehensive source of governance indicators (Kaufman et al. 2003), Table 11 shows the evolution of composite indicators of governance in the Bolivian case. The indicators are scaled so that they have a mean of 0 and a standard deviation of 1 for all countries included in the sample. The first group summarizes indicators of the political process, civil and political rights. The political instability indicator measures the likelihood of government overthrow or destabilization, the third cluster summarizes measures of public service provision, competence and quality of the bureaucracy and the last one summarizes the incidence of market-unfriendly policies. It is notable that in all four governance measures Bolivia exhibits a downward trend in the last few years, ending up close to or below the average for all countries.39 This is particularly the case for government effectiveness and regulatory quality. As most of these measures are based on survey-based evidence, it appears that the trust in the policy-making apparatus and the state bureaucracy has weakened already well before the recent protests over tax reform and gas exports.
106. More detailed investigations of the role of institutions in Bolivia by Kaufman et al (2002) yield further important insights. In particular, business surveys reveal that corruption and the lack
39 Somewhat ironically, the deterioration is least apparent in the political stability measure, given that the government
was indeed forced out of office by popular protests in 2003.
41
of the rule of law stand out as particular constraints in Bolivia, while issues of macro management and financing are not seen as a problem at all. Corruption, bribery, and a weak judiciary are also named as significant barriers preventing sales growth. This suggests that the structural reforms of the 1990s were having positive effects in their areas of focus but did not tackle the more deep-seated problems of corruption and a weak judiciary. As far as causes of corruption and low government effectiveness is concerned the results suggest that lack of transparency and voice of citizens appear to be the most important reasons, according to surveys of public officials. Also here, much work remains to be done.
107. These institutional weaknesses not only retard economic growth by preventing effective policy-making (Rodrik, 2003), but they also are implicated in sustaining high inequality. A study by Chong and Gradstein (2004) show that indicators of institutional weakness are strongly associated with inequality in Latin America. The causality is bi-directional with the causal link from inequality to institutions being somewhat stronger than the reverse one. The discussion suggests that also here there are two sets of policy issues emerging. On the one hand, there is clear need to tackle the institutional weaknesses in Bolivia, particularly corruption, lack of a reliable judiciary, and lack of transparency and voice in the public sector. On the other hand, it is clear that such reforms will be more difficult in an environment of high inequality that sustains these institutional problems, which provides a further case to address these inequalities.
Table 11: Trends in Governance Indicators over Time
1996 1998 2000 2002
Voice and Accountability 0.10 0.35 0.23 0.01
Political Stability -0.28 0.00 -0.42 -0.20
Government Effectiveness -0.49 -0.09 -0.35 -0.53
Regulatory Quality 0.66 0.90 0.65 -0.11
Source: Kaufmann et al (2003).
(ii) Decentralization and pro-poor growth 108. Bolivia embarked in 1994–1996 on an ambitious decentralization program, which transferred a large share of resources and associated responsibilities to Bolivia’s municipalities. Municipalities were assigned responsibilities for investment expenditures in the social sectors and infrastructure. In addition to funds from the central government, they were granted, as an outcome of the National Dialogue Law, the entire amount of HIPC II debt relief for investments at the local level (targeted to communities with higher rates of unsatisfied basic needs) and they have spent most of these funds, as reported above, on social sector investments and most recently also on infrastructure. In addition, an intermediate (centrally appointed) layer of government (prefectures) was introduced and also given considerable spending and implementation authorities.
109. In view of the governance problems cited above, such decentralization could be of help to address the problems of lack of transparency and voice as decisions and implementation are brought closer to the people affected. This was also supported by popular oversight mechanisms. At the same time, it is unclear whether such an ambitious decentralization program can tackle the problems of low government effectiveness and corruption. Clearly, it can be only one avenue of governance reforms and other items (including judicial reform and improvements in voice and transparency throughout the public sector) remain of importance.
110. In addition, decentralization in its current form appears to be plagued by a number of problems. First, the roles of municipalities versus prefectures are not clarified and lead to cumbersome coordination and oversight problems. Secondly, the institutional capacity of smaller municipalities is too weak to undertake many of tasks they are being charged with. Third, revenue
42
generation at the municipal level remained low, with the municipalities depending on average for 80% of their resources on the central government (World Bank 2004a). Moreover, they have no control over the public sector wage bill for the services under their control (e.g. health and education). This has effectively softened the budget constraint and led in some municipalities to the build-up of considerable debt. The planning and oversight systems are not working well due to weak institutional capacity and lead to very slow and cumbersome implementation of investment projects.
111. It appears that the process of decentralization has proceeded too quickly and overtaxed the capacity of the developing municipal authorities. This has led to capacity constraints, delays in spending, complaints of poor and intransparent governance at the municipal level, and left the central government with little ability to direct public policy towards priority sectors (World Bank 2004a). While many local investment projects which, if successfully executed, were found to have had a significant positive impact on health outcomes and water access (Newman et al. 2002), the cumbersome procedures, large overheads, and poor capacity have led to very slow implementation. The capacity constraints might also explain the focus on social sector spending and infrastructure, at the expense of efforts to promote productive sectors. Programs to promote productive sectors (such as credit programs, cluster initiatives, subsidy programs) tend to be much harder to implement than the expansion of health and education programs and the construction of physical infrastructure. This surely contributed to the considerable improvement in social indicators over the past few years, but did little to reduce income poverty. This focus on social sectors at the expense of promoting productive sectors has become one of the main criticisms that have been voiced in the current round of the National Dialogue centered on the revision of the PRSP.
112. As shown above, it has also, at least initially, led to an anti-poor spending pattern on infrastructure. As far as the outcome of the decentralization process on poverty is concerned, there is little reliable data to date. Viana Sarabia (2004) undertakes an econometric analysis (ordered logit) of the impact of three indicators of decentralization on a non-income measure of poverty, the so-called NBI (Unmet basic needs). She finds that municipalities with lower levels of unmet basic needs have a larger resource envelope per capita to spend, confirming that decentralization did not equate per capita spending (or target them to the poorest areas). In addition, the most important determinant of the degree of unmet basic needs is the share of own resources spent by a municipality. Given that this share is particularly large in richer municipalities, the analysis supports the notion that a decentralization process that largely depends on central government transfers for the poorest communities will not have a significant impact on poverty outcomes and might in fact exacerbate existing inequalities. More recently, the situation appears to have improved somewhat through the better targeting of HIPCII resources to communities and through a greater focus on productive sectors (which mainly means infrastructure). In fact, some poor communities have received much more funds and are having difficulties in spending it productively (World Bank 2004a).
113. While decentralization might be a way to improve governance and can also enhance pro-poor spending, it can generate new problems of capacity constraints, poor fiscal control and oversight, and failure to tackle income poverty issues. If capacity constraints or other institutional weaknesses are correlated with income poverty of a municipality, it can even exacerbate existing inequalities unless ways are found to assist poorer and smaller municipalities. Alternatively, it is important to keep options open for pro-poor interventions that are directed and supported from the central government.
(iii) PRSP and national dialogue 114. Bolivia was among the first to complete an Interim PRSP and a full PRSP in 2000 and thus to enjoy HIPC debt relief. The PRSP was the culmination of a systematic National Dialogue Process, which was written into a law in 2001 as a permanent institution to take the PRSP process forward. At the time, Bolivia’s PRSP has been widely lauded for both its content and process.
43
115. But soon after completion of the process, serious disappointment with the process and the outcome emerged, and the PRSP began to be considered a ‘dead’ document by most stakeholders (ISS 2003a, 2003b). Apart from the well-known problems with PRSPs elsewhere (overambitious targets, ‘laundry lists’ with little prioritization, unclear relation to the government macro and fiscal strategy, too focused on trying to please donors in order to get HIPC funds), there appear to have been particular problems associated with the National Dialogue and the creation of the PRSP. Among the problems encountered in the process, according to an analysis undertaken by the Institute of Social Studies, was that the Interim PRSP was never fully discussed, and that there was a significant disconnect between the National Dialogue, which was open, transparent, and had significant non-governmental participation, and the writing of the PSRP that was then relegated to a group of consultants who drafted a strategy that was only partly based on the outcomes of the dialogue but more influenced by inputs from the donor community and the desire to please the international community to get HIPC funds (ISS 2003a, 2003b). This sharply reduced the popular ownership of the final strategy, which in addition became largely obsolete when the macroeconomic conditions departed sharply from the rosy projections and plans included in the PRSP. As such, the PRSP had little impact beyond the decision in the National Dialogue Law to transfer all HIPC resources to the municipalities, which strengthened social sector investments at those levels.
116. Regarding the content of the strategy, there appears to have been considerable unease over the almost exclusive focus on social sector spending as the route out of poverty, the neglect to discuss macro issues and consider alternatives to the current economic model guiding Bolivia’s economy, the overoptimistic macroeconomic projections, the unsustainable associated expenditure plans, and the neglect to focus on strengthening the productive sectors as a means to achieve sustainable poverty reduction.
117. The revised draft PRSP tabled by UDAPE in late 2003 already included some changes in focus and put the development of micro, small and medium-sized enterprises at the heart of the poverty reduction strategy. This development was supposed to be promoted through a combination of a land policy focusing on titling and increased security of tenure, national productive clusters promoting the supply chains for 14 products through joint public-private initiatives (cadenas productivas, see Box 4), and through efforts to promote local economic development. While this focus on productive sectors addressed one central complaint about the original PRSP, there continued to be disappointment that the revised strategy did not consider alternative economic models of development, did not include more radical land redistribution programs, and was not far-reaching enough in focusing on poverty reduction through strengthening productive sectors (including public support for productive sectors that go far beyond the cadenas productivas approach). As a result, this draft is by now also seen as insufficient as the debate has now moved to the constitutional assembly where the questions of the use of natural resources, the economic model, and the land distribution are likely to figure prominently.
118. These (necessarily cursory) discussions of institutional issues in policy-making appear to suggest the following conclusions for pro-poor policy-making. First, in a situation of poor and deteriorating governance, it is critical to address public sector governance issues as a first priority. A decentralization program can be of help in some aspects, but it is particularly risky to embark on major new initiatives that further impair the management capacity of the public sector. The Bolivian government appeared to have pushed institutional changes to policy-making at a speed that eventually weakened its ability to implement effective policies. Second, decentralization need not improve governance nor does it necessarily improve the poverty focus of public expenditures. It merits consideration whether the central government should consider the build-up of a centrally managed approach to promoting equity and assisting with pro-poor growth. In addition, it appears that decentralization of the expenditure side without the necessary abilities to raise revenues can undermine the success of a decentralization effort. Third, regarding pro-poor policy-making, it is very risky and ultimately counter-productive to establish a participatory process of pro-poor policy-making and then strictly limit the agenda to certain priority actions. This is particularly risky if
44
there is significant mistrust of the government, which is grounded on long-standing inequalities and social and political exclusion. If a process is to be participatory, all aspects of economic policy-making must be openly discussed even if this can lengthen the process and pose risks for the outcome. The Bolivian case seems to suggest that the route taken here has now led to a very polarized political debate that poses even greater risks for political and economic stability than an early discussion of all aspects of economic policy-making (see also ISS 2003).
Box 4: Cadenas Productivas The idea was to promote the supply chain through coordinated public and private actions for the following 14 products: poultry, bananas, cows, Brazil nuts, leather, timber, oil products, palm heart, quinoa, textile and cotton, wheat, grapes, (wines and liquors) and tourism (Sucre-Potosi-Uyuni salt lake circuit). These were selected as they already made up a significant share of GDP, generated some 400,000 jobs, were often activities undertaken by small farmers or entrepreneurs, and took place in many parts of the country. It was intended to strengthen the supply chain by developing capacities to refine these products within the country, and by improving national and particularly international sales opportunities. The plan was to generate public-private partnerships that would develop coordinated plans to achieve these goals. Specific measures would include technical and technological assistance, support services for production, help with market access, and productive infrastructure. The preparation of this strategy would involve three steps (strategic vision, strategic plan of coordinated actions, and a conclusion of a public-private pact as part of the 2003 National Dialogue). Initial documents set aside public funds from municipal and departmental budgets as well as HIPC funds to support these investments. With the shift of the debate to the constitutional assembly, the future of this approach to support productive sectors is uncertain.
Sources: Ministerio de la Presidencia (2003); (UDAPE 2003a).
Chapter 4: Possible Trade-Offs between Growth and Poverty Reduction
119. Based on the assessments from the CGE model, we can not only assess the impact of individual policies on pro-poor growth, but also consider possible trade-offs and complementarities. When it comes to choosing a policy package for pro-poor growth from the available options, it is important to know whether particular measures promise to create win-win situations in that they help achieve growth and distributional objectives at the same time, or whether there are trade-offs involved.
120. In the area of macroeconomic policy, higher yearly devaluations of the Boliviano risk to fail on both accounts as a result of adverse balance sheet effects. A tightening of monetary policy, by contrast, may bring about the real devaluations that are regularly required to adjust to external shocks at a negligible short-run cost for the poor.
121. Among the two structural reforms considered here, a deregulation of the urban labor market carries the potential to make growth considerably more pro-poor by removing a substantial part of the existing wedge between formal and informal wages. Such a measure would, however, meet with strong resistance from formal workers, who arguably are much better organized than the diverse group of people working in the informal sector. This difficult political situation is probably the key factor behind the fact that profound labor market reforms have not yet been initiated. Also, it would not do enough to reduce rural poverty.
122. Similar pressure from powerful interest groups – in this case mainly from public employees – stands in the way of a comprehensive tax reform. Provided that this pressure can be overcome, the introduction of a revenue-neutral tax reform may improve efficiency and reduce poverty. A pure income tax increase, by contrast, is unlikely to serve these objectives. If income taxes are set at moderate rates as assumed above, they are likely to be only mildly progressive and may even raise poverty. And with substantially higher tax rates, the efficiency losses may well turn out to be intolerable.
45
123. In the development of Bolivia’s gas sector, there appears to be a trade-off between growth on the one hand and the participation of the poor – in particular the rural poor – in the growth process on the other hand. Given the prospect that nation-wide poverty might decrease only moderately as a result of the resource boom, and that rural poverty might even increase, the rationale behind the recent social unrest becomes obvious. The trade-off is, however, hard to avoid as the gas sector is highly capital intensive, generates little employment, and uses limited national inputs. To what extent growth and poverty objectives can be reconciled depends on how the government allocates the additional revenues it receives. While an increase in public savings might cushion the trade-off, more specific pro-poor measures are likely to be required in order to make the impact of the gas projects socially acceptable.
124. Given that rural poverty constitutes the most severe problem, measures targeted at augmenting the asset base of smallholders suggest themselves as possible win-win options. It has to be taken into account, however, that natural conditions in the highlands are not very favorable, and that the growth process would have to start from very low initial capital endowments. This implies that the medium-run supply response, and thus the impact on rural poverty, will probably remain limited.
125. With respect to pro-poor industrial policies, the key question is whether favorable poverty outcomes can be achieved at low efficiency losses. Our simulation results indicate that efficiency losses may be kept at moderate levels, but that neither a strategy based on export-oriented agriculture nor a strategy based on agricultural processing are likely to bring about lasting improvements for the rural poor.
126. Transfer programs targeted towards the poor can in principle alleviate poverty without compromising growth objectives, but the precondition for this to happen – a more or less complete financing of the programs out of other current expenditures – appears to be very demanding. If investment spending has to bear the lion’s share of the costs, the economic losses can become considerable.
127. In the coming years, the gas receipts may provide a way out of this trade-off by loosening the budget constraint of the Bolivian government. If gas revenues are used instead of public investment funds to finance transfers, the combination of the gas projects and the transfer programs produces a clear win-win situation, with higher growth and a marked alleviation of rural and urban poverty. The only major drawback of this policy option is that both the gas projects and the transfer programs lead to a significant increase in rural inequality, raising the rural Gini coefficient by almost three percentage points in the second half of the simulation period. This is driven by the (model) assumption that the transfer programs are targeted in the same way as existing programs. If targeting were to be improved, this rise in inequality could be significantly reduced.
128. One way to summarize the trade-off is to examine the combined scenarios in Table 10. The optimal pro-growth scenario combines a gas projects (with constant government consumption) labor market and revenue-neutral tax reform. Growth is boosted to nearly 6% per year. But the poverty impact is only moderate and entirely focused on urban areas; inequality and rural poverty are expected to increase. At the other extreme, we have a pure transfer program which, if coming at the expense of public investment reduces growth but also reduces rural and urban poverty, as well as reducing overall inequality. To achieve high rates of pro-poor growth given current constraints, the combination of both policy scenarios is likely to be best for sustainable poverty reduction in urban and rural areas, with an added focus on improving the targeting of the transfer program.
129. Apart from these model-based assessments, it is also important to consider more fundamentally the trade-offs involved between a narrow growth agenda that is largely guided by policies informed by the Washington Consensus and its associated political economy risks. It appears that given Bolivia’s unfavorable initial conditions as well as its history of high inequality and large social and ethnic tensions, a technocratic focus on liberalization and macro stability might not deliver benefits in terms of poverty and inequality reduction quickly enough to prevent serious
46
setbacks as have been experienced in Bolivia in recent years. Going further down that route and hoping that the possible high growth associated with natural gas and commercial agriculture exports will deliver enough benefits to maintain social stability are likely to prove elusive and might provoke populist backlashes with serious consequences for growth and poverty reduction. Instead it appears necessary to confront the issues of deep-seated inequalities in resources, opportunities, and power more directly rather than hoping than one can grow out of them. Some of these questions will be taken up below.
Chapter 5: Recommendations for Policy-Making
130. Bolivia is now facing a serious economic and political situation. It is in a state of serious political and social unrest, economic conditions are not favorable (although they have stabilized recently), and there are loud demands for more spending, more redistribution, and a total abandonment of the current economic policy stance. In this situation, it is not easy to come up with a policy framework that will successfully steer Bolivia to a path of significant pro-poor growth.
131. In this chapter we will start from a combination of incremental policies that all could serve to improve the overall slow record of poverty reduction in Bolivia. We will then move towards asking whether a more radical reform of economic policy-making is needed, both in terms of content as well as process. We will begin by summarizing some of the conclusions from our model-based assessments.
132. The main general conclusion to be drawn from the foregoing model-based analysis is that in Bolivia the opportunities for achieving pro-poor growth differ enormously between urban and rural areas. This has been true for the 1990s, where until the outbreak of the recent crisis urban households have benefited disproportionately from foreign investment led growth. And it is also true with respect to future prospects, which are less favorable for rural households due to several factors. To support our argument, we reorder in Table 12 the main policy experiments from Table 10 according to whether the impact is mainly on urban or rural households. First, the rural economy will inevitably have to cope with recurrent disruptions caused by external shocks. Second, difficult natural conditions in combination with very low initial capital endowments will limit the impact of efforts to increase the asset base of poor farmers. Nonetheless, investments in public goods such as rural infrastructure or agricultural research should be taken into consideration as they could at least entail some productivity improvements.40 If it turns out that significant productivity gains can be expected, measures aimed at improving smallholders’ access to credit such as increased tenure security or additional micro-credit initiatives (including tackling issues of non-traditional forms of guarantees), might also have a positive pay-off in that they help realize complementary private investment. Also, community-driven investments in irrigation, improved seed varieties, and modern inputs should remain firmly on the agenda.
133. Third, the modern, dynamic segment of the agricultural sector is too small to absorb a sizeable part of the poor rural workforce so that a pro-poor industrial policy based on modern agriculture does not appear to be promising. Finally, the development of the gas sector will largely bypass the rural economy, and will even raise rural poverty via the economy-wide repercussions it entails. As the recent protests have shown forcefully, such an uneven distribution of benefits will meet strong resistance. This does of course not imply that Bolivia should forego the gains to be expected from the gas exports, but rather that rural households should be able to share in the gains to an extent that more than compensates the expected losses to them. To achieve this, direct transfers constitute the only realistic option as only direct transfers can raise incomes significantly in the short to medium run. Such transfers should, however, be targeted very carefully. By simply expanding existing transfer programs, as assumed in the simulation reported above, the government
40 Unfortunately, evaluations of the impact of public investments in Bolivia are lacking. Hence, their productivity
effects can only be guessed, and priority areas for public investment can hardly be identified.
47
will miss most of the poorest households. As a further caveat, the financing of transfers that are large enough to accomplish a significant poverty reduction is only sustainable as long as the gas boom endures. All in all, given current constraints, it is thus likely that the prospects of fostering rural development will to a large extent rely on dynamic growth of the urban economy, which would indirectly raise rural incomes via increased rural-urban migration, higher intermediate demand for agricultural raw materials, and higher consumption demand for food. In this context, the dynamic development, in terms of population growth, economic growth, and poverty reduction, of small and medium-sized towns in Bolivia could provide one avenue for providing employment and income-earning opportunities for the rural poor. Supporting such growth centers through infrastructure investments and support for decentralized industrial activities might be one option to consider.
Table 12: Shocks, Policies and Pro-Poor Growth in Rural and Urban Areas
Shock/Policy Average Growth (%)a
Rural Headcounta
Urban Headcounta
Main Effect on Rural Households El Niño -0.3 1.3 0.8 Terms-of-Trade Shock 0.0 0.5 0.2 Investment in Rural Infrastructure 0.1 -0.4 -0.3 Improved Access to Credit for Smallholders
0.0 0.0 -0.1
Industrial Policy (modern agriculture) 0.0 -0.1 0.7 Transfer Program 0.0 -2.3 -1.0
aPercentage points deviation from base run. Source: Based on Table 10.
134. Several options can be pursued to raise urban growth and to alleviate urban poverty. Despite limited linkages to the rest of the economy, the development of the gas sector will benefit urban areas. The positive effect will be the stronger the more the gas projects boost domestic investment. Beside the funds earmarked for pro-poor spending, a substantial part of the gas revenues should thus be used to prop up public savings. The difficult task for the government then is to withstand pressures and keep public consumption under control.
135. In addition, the two big remaining structural reforms, a deregulation of the urban labor market and an income tax reform, would both have a significantly positive impact on growth and poverty and should thus be initiated. The main problem with these reforms is that the potential losers – non-agricultural workers and employees, respectively – can effectively lobby against them. Perhaps it will become somewhat easier to overcome their resistance if the structural reforms are carried out in combination with the gas projects as all urban household groups stand to benefit from the latter. If this were further combined with transfer programs aimed at the rural poor, it might be possible to generate a politically and economically feasible policy package for pro-poor growth.
136. Beyond the model-based simulations, there are further policy options to consider. As mentioned above, there is great need to switch to a tax regime that is more progressive than the current system and could also go beyond the revenue neutral system that is proposed in the model. On the expenditure side, there is considerable scope for increasing the poverty focus of public
48
expenditures. In health and education sectors, this can be achieved by requiring larger co-payments from the better-off and ensuring better access and utilization of higher education and higher order health facilities for the poor. In order to free up resources, reforms of the pension system must proceed in a way that limits government spending on this program that has little poverty impact. Expenditure reform should also include a greater poverty focus of decentralized public expenditures. This can either be ensured through greater equity in inter-governmental transfers, with a particular emphasis on increasing the resource envelope of poorer municipalities (using formulas that are not only based on population), or through additional central government programs in these areas.
137. In line with the demands made from many sides, it is important to overcome the disconnect between social sector performance and economic performance. While the social sector improvements are impressive and need to be sustained as important achievements in their own right, their impact on income poverty will largely materialize in the longer term and will require complementary measures to support the productive sectors. Consequently, public expenditures should not only focus on social sectors (important as they are for reaching the MDGs), but include rising expenditures for measures to strengthen production among the poor. The planned efforts to strengthen the cadenas productivas appear to be a step in the right direction. In addition, municipalities should be encouraged to experiment with various ways of strengthening their productive sectors through infrastructure, credit, and subsidy programs. This is particularly important as neither the national nor the international experience suggest clear best-practice models. Based on evaluations of successes and failures, better-managed programs should then be mainstreamed across the country.
138. Success on such ventures will greatly depend on the ability to overcome institutional weaknesses within the public sector. In particular, control of corruption and low government effectiveness should be the primary focus of public sector reform efforts. This will require greater transparency and voice in all tiers of government. Decentralization can support such a process but only if the current defects of the process are fixed. They include unclear roles for municipalities versus prefectures, capacity constraints in the poorer and smaller municipalities, too little central oversight and fiscal control, too little ability to affect pro-poor spending and implementation at the local level.
139. But the analysis has also shown that there are a range of basic constraints that need to be addressed if Bolivia is to succeed in developing and implementing a pro-poor growth agenda. A first critical constraint is the low domestic savings rate which leads to dependence on foreign capital, leads to foreign debt, and contributes to the vulnerability of the economy to external shocks against which it cannot act due to the high degree of dollarization. Here a combination of policy measures ought to be considered. First, at the international level the question should be re-opened whether debt relief in the case of Bolivia was deep enough. After several years of low growth, the foreign debt burden is high, putting pressure on the fiscal side and sharply curtailing any room for devaluations to improve the competitiveness of Bolivia’s economy. Second, measures to raise the domestic savings should be strengthened. They should include both institutional strengthening of the financial sector, particularly the availability of reliable savings institutions also in small towns and rural areas, policies to shield savings from the risks of inflation (through index-linked products and possibly a complete indexation of the economy), and policies to promote public savings (through limiting obligations on the pension system, savings proceeds from the gas project, and further debt relief or an increasing share of grant aid). Third, measures to reduce dollarization should be pursued more vigorously to increase the ability of the monetary authorities to engage in pro-poor monetary policies. Given the by now long record of low inflation, it should be in the interest of the government to begin pushing back the dollarization of its economy. This could be done via a set of incentives such as the recently passed financial transactions tax which is only levied on $-denominated transactions, differential reserve requirements for $ versus Boliviano denominated assets, and a push to popularize inflation-indexed securities as the main form of
49
issuing government debt. Once dollarization has been reduced, a much more active management of the exchange rate would become possible to ensure international competitiveness and also address distributional issues. To maintain this flexibility, controls on capital inflows might be needed to ensure that they do not destabilize the currency and financial markets. Should natural gas exports increase to the level envisaged, management of the exchange rate through sterilization policies would be critical to limit Dutch disease effects.
140. Secondly, it appears urgently necessary to confront some of the more deep-seated inequalities in Bolivia’s economy. As part of the on-going discussions in the National Dialogue as well as in the Constitutional Assembly, a national plan for the redistribution of assets should be considered. Elements of such a plan could include greater attention to land redistribution (in addition to titling) from public lands, market and subsidy-based land reforms using land taxes to increase the land brought to the market. In addition, such a strategy could include a mechanism that would transfer part of the benefits from natural resources directly to the poor to ensure their direct access to the proceeds from these assets. The use of demand-side transfers such as those pioneered in Mexico (Progresa and Oportunidad) or Brazil (bolsa escolar) might be a good way to proceed. Third, it appears that much of Bolivia’s current social and political turmoil stems from the fact that its indigenous population was largely excluded from the political process. Measures to increase their voice and power through quotas and other mechanisms might be considered to involve them more directly in policy-making.
141. Given Bolivia’s history and its current problems, achieving sustained high rates of pro-poor growth will be very difficult unless these deep-seated inequalities are addressed.
50
References
Alesina, A., A. Devleeschauwer, W. Easterly, S. Kurlat, and R. Wacziarg (2003). Fractionalization. Journal of Economic Growth 8(2): 155–194.
Andersen, L.E., and O. Molina (2004). Análysis Estadístico y Economico sobre las Characteristicas de la Permanencia y Acceso Diferenciodo por Género en el Systema Educativo Boliviano a Nivel Municipal. Ministerio de Educacion. La Paz. Bolivia.
Anderson, L.E. (2003). Baja movilidad social en Bolivia. Revista Latinoamericana de Desarollo Economico 1: 11-37.
Anderson, L.E., and J. E. Evia. (2003). The Effectiveness of Foreign Aid in Bolivia. IISEC Working Paper 10/03.
Andersen, L.E., A.F. Mercado, and B. Muriel (2003). Discriminacion Etnica en Bolivia: en el Sistema Educativo y el Mercado de Trabajo. Working Paper 03/03. Instituto de Investigaciones Socio-Economicas. LaPaz.
Barro and Lee (2000). International Data on Educational Attainment: Updates and Implications. Center for International Development. Working Paper 42.
Bloom, D.E., and J.G. Williamson (1998). Demographic Transitions and Economic Miracles in Emerging Asia. The World Bank Economic Review 12(3): 419–455.
Chong, A. and M. Gradstein. (2004). Inequality and Institutions. Mimeographed, Inter-American Development Bank.
Christiaensen, L., L. Demery, and S. Paternostro (2002). Reforms, Economic Growth, and Poverty Reduction in Africa: Messages from the 1990s. World Bank, Washington, D.C. Processed.
COEDPO (2002). Migracion Interna. Documento de Trabajo No. 2. La Paz: UNFPA.
de Janvry, A., and E. Sadoulet (2000). Growth, Poverty, and Inequality in Latin America: A Causal Analysis, 1970-94. The Review of Income and Wealth 46(3): 267–288.
Datt, Gaurav, and Martin Ravallion (1992). Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s. Journal of Development Economics 38(2): 275-295.
Deininger, K., and L. Squire (1998). New Ways of Looking at Old Issues: Inequality and Growth. Journal of Development Economics 57(2): 259–287.
Delgadillo, M. (2000). Es bueno el sistema tributario en Bolivia? La Paz: UDAPE.
DHS (Demographic and Health Survey) (1989). Bolivia. Enquesta Nacional de Demografía y Salud. INE and MSD. Online Database.
DHS (Demographic and Health Survey) (1994). Bolivia. Enquesta Nacional de Demografía y Salud. INE and MSD. Online Database.
DHS (Demographic and Health Survey) (1998). Bolivia. Enquesta Nacional de Demografía y Salud. INE and MSD. Online Database.
51
DHS (Demographic and Health Survey) (2003). Bolivia. Enquesta Nacional de Demografía y Salud. INE and MSD. Informe Preliminar.
Eastwood R., and M. Lipton (2000). Pro-Poor Growth and Pro-Growth Poverty Reduction: Meaning, Evidence, and Policy Implications. Asian Development Review 18(2): 1–37.
Filmer, D., and L. Prichett (1998). Estimating Wealth Effects without Expenditure Data – or Tears: An Application to Educational Enrollments in States of India. Demography 38(1): 115–132.
Gasparini, L. M. Cicowiez, F. Gutierrez, and M. Marchionni. Simulating Income Distribution Change in Bolivia: a Microeconometric Approach. World Bank Poverty Assessment Background Paper. Mimeographed. Washington DC: The World Bank.
Grimm et al. (2004) Grimm, Michael, and Isabel Günther (2004). Operationalizing Pro-Poor Growth – Country Case Study Burkina Faso. Paper Presented at the Joint World Bank, KfW, and GTZ Workshop “Operationalizing Pro-Poor Growth” in Eschborn, 15-16 July 2004.
IDB. (2003). Good Jobs Wanted. Labor Markets in Latin America. Washington DC: IDB.
IMF (2004). Bolivia: Second Review Under the Stand-By Arrangement and Request for Waiver of Applicability and Modification of Performance Criteria. IMF Country Report 04/5. Washington, D.C.
INE (Instituto Nacional de Estadística de Bolivia) (2003a). Characterísticas Demográficas. La Paz.
INE (Instituto Nacional de Estadística de Bolivia) (2003b). Bolivia: Niveles, Tendencias, y diferenciales en la Fecundidad. Serie IV: Estudios tematicos. La Paz: INE.
INE (2004). Encuesta Nacional de Demografia y Salud 2003. Informe Preliminar. La Paz: INE.
INE-UDAPE (2002). Mapa de Pobreza 2001. Mimeographed. La Paz: INE, UDAPE.
INE (various issues). Información Estadística. Online Data Base. Instituto Nacional de Estadística. http://www.ine.gov.bo.
ISS (Institute of Social Studies) (2003a). Evaluacion y Monitoreo de las Estrategias de Reduccion de la Pobreza (EPR) in America Latina. La Estrategia Boliviana de Reduccion de Pobreza: ¿”La Nueva Brillante Idea”?. Informe País: Bolivia. La Haya.
ISS (Institute of Social Studies) (2003b). Evaluation and Monitoring of Poverty Reduction Strategies in Latin America. The Bolivian Poverty Reduction Strategy: ”Yet Another Brilliant Idea”? Executive Summary. Country Report Bolivia. Bolivia. La Haya.
Jemio, L.C. (2001). Debt , Crisis and Reform in Bolivia: Biting the Bullet. Basingstoke: Palgrave in association with Institute of Social Studies.
Jemio, L.C., and M. Wiebelt (2003). ¿Existe Espacio para Políticas Anti-Shocks en Bolivia? Lecciones de un Análisis basado en un Modelo de Equilibrio General Computable (with L.C. Jemio). Revista Latino Americana de Desarollo Económico 1(1), 2003, pp. 37-68. Jiménez, W., and F. Landa (2004). ¿Bolivia Tuvo un Crecimiento “Pro-Pobre” entre los Años 1993 y 2002?. UDAPE. La Paz.
Jimenez, W. And F. Landa. 2004. Bolivia tuvo crecimiento « pro-pobre » entre los anos 1993 y 2002 ? Mimeographed, La Paz : UDAPE.
52
Jimenez, W., R. Pereira, and W. Hernany (2001). Bolivia: Efectos de la Liberacion sobre el Crescimiento, Empleo, Distribución y Pobreza. In: UNDP (ed.), Liberalización, Desigualdad y Pobreza: America Latina y el Caribe en los 90 (2001): 201–254.
Kaufmann, D., A. Kraay, and M. Mastruzzi (2003). Governance Indicators for 1996-2002. Policy Research Working Paper 3106. The World Bank, Washington, D.C.
Kaufmann, D. M. Mastruzzi, and D. Zavaleta. 2002. Sustained Macroeconomic Reforms, Tepid Growth: A Governance Puzzle for Bolivia? The World Bank, Washington DC.
Klasen, S. (2003). Population Growth, (Per Capita) Economic Growth, and Poverty Reduction in Uganda: A brief Summary of Theory and Evidence. Working Paper, University of Göttingen.
Klasen, S. (2004). In Search of the Holy Grail. How to Achieve Pro-poor Growth. In: B. Tungodden and N. Stern (eds.), Towards Pro-poor Policies. Proceedings from the ABCDE Europe Conference. Washington, D.C.
Klasen, S. and F. Lamanna (2004). The Impact of Gender Inequality in Education and Employment on Economic Growth in the Middle East and North Africa. Background Paper for World Bank Policy Research Report: Women in the Public Sphere. Washington DC: The World Bank.
Loayza, N., P. Fajnzylber, and C Calderón. 2002. Economic Growth in Latin America and The Caribbean: Stylized Facts, Explanations, and Forecasts. Working Paper of the Central Bank of Chile No. 265.
Lora, E. (2001). Structural Reforms in Latin America: What Has Been Reformed and How to Measure it. Inter-American Development Bank, Research Department, Working Paper 466.
Marcoux, A. (1998). The Feminization of Poverty: Claims, Facts, and Data Needs. Population and Development Review 24(1): 131–139.
Mercador, A. (2003). Movilibidad Social: la clave para el desarrollo. Programa de Investigacion Estrategica en Bolivia.
Minsterio de la Presidencia. (2003). Pactos Productivos. Mimeographed, La Paz: UDAPE.
Muñoz, R., and A. Palma (2003). Determinants of Firm Size and Capacity Utilization in Bolivian Manufacturing Sector. The World Bank. Mimeo.
Newman, J., M. Pradhan, L. Rawlings, G. Ridder, R. Coa, and J. Evia. 2002. An Impact Evaluation of Education, Health, and Water Supply Investments by the Bolivian Social Investment Fund. World Bank Economic Review 16: 241-274.
Pianto, D., M. Tuannuri-Pianto, and O. Arias (2004): Rural-Urban Migration and Human Capital in Bolivia. World Bank Poverty Assessment. Mimeo.
Ravallion, M., and S. Chen (2003). Measuring Pro-poor Growth. Economics Letters 78(1): 93–99.
República de Bolivia (2001). Bolivia: Poverty Reduction Strategy. La Paz.
Rodrik, D. (2003). Growth Strategies. NBER Working Paper 10050.
53
Sachs, J.D., and F.B. Larraín (1998). Bolivia 1985-1992: Reforms, Results, and Challenges. In: H. Costin (ed.), Economic Reform in Latin America. Fort Worth.
Sahn, D., and D. Stifel (2003). Exploring Alternative Measures of Welfare in the Absence of Expenditure Data. Review of Income and Wealth 49(4): 463–489.
Schweickert, R., R. Thiele, and M. Wiebelt (2003). Makroökonomische Reformen und Armutsbekämpfung in Bolivien: Ebnet die HIPC-Initiative den Weg zu sozialverträglicher Anpassung? Kiel Discussion Papers 398. Institute for World Economics, Kiel.
Servicio de Impuestos Internos (2003). Recaudacion renta interna por tipo de impuestos. La Paz.
Spatz, J. (2004). The Impact of Structural Reforms on Wages and Employment: The Case of Formal versus Informal Workers in Urban Bolivia. Latin American Journal of Economic Development 1(2): 91–122.
Tannuri-Pianto, M., D.M. Pianto, and O. Arias (2003). Formal, Informal, and Self-employed earnings in Urban Bolivia: Accounting for Sample Selection in Multinomial Choice Models. The World Bank Bolivia Poverty Assessment. Washington, D.C.
Tannuri-Pianto, M., D. Pianto, and O. Arias (2004). Informal Employment in Bolivia. A lost Propostion? World Bank Poverty Assessment. Mimeo.
Thiele, R. (2003). The Social Impact of Structural Adjustment in Bolivia. Journal of International Development 15(1): 1–21.
Thiele, R., and D. Piazolo (2003). A Social Accounting Matrix for Bolivia Featuring Formal and Informal Activities. Cuadernos de Economía 40: 285–318.
Thiele, R., and M Wiebelt (2004). Growth, Poverty, and Income Distribution in Bolivia: A Regional and Sectoral Perspective (with R. Thiele). In M. Krakowski (ed.), Attacking Poverty: What makes growth pro-poor? HWWA Studies 75, Baden-Baden: Nomos.
Thomas, D. 1997. Incomes, Expenditures and Health Outcomes: Evidence on Intrahousehold Resource Allocation. In Haddad, L. J. Hoddinott, and H. Alderman (eds.). Intrahousehol d Resource Allocation in Developing Countries. Baltimore: Johns Hopkins Press.
Timmer, P. (1997). The Agricultural Transformation. In H. Chenery and T. N. Srinivasan (eds.), Handbook of Development Economics, Vol. 1. Amsterdam: North-Holland.
Tuchschneider, D. (2001). Más Allá de las Cifras. La Vision Campesiana de los Cambios Estructurales den la Productividad Rural y los Recursos Naturales. Ministerio del Desarollo Sostenible y Planificacion.
UDAPE (2003a). Estrategia Boliviana de Reducción de la Pobreza: Informe de Avance y Perspectivas. La Paz.
UDAPE (2003b). Pobreza y Desigualdad en Municipios de Bolivia: Estimacion del Gasto de Consumo Combinando el Censo 2001 y las Encuestas de Hogares. La Paz.
UDAPE (2003c). Bolivia: Evaluacion de la Economia 2002. La Paz.
UDAPE (various issues). Dossier de Estadísticas Sociales y Económicas de Bolivia. Unidad de Análisis de Políticas Sociales y Económicas. La Paz.
54
Urioste, M. (2003). La Reforma Agraria Anandonada: Valles y Altiplano. La Paz: Fundacion Tierra.
Viana Sarabia, V. (2004) Impacto de la Decentralisacion en la satisfaccion de las necessidades basicas de la poblacion Boliviana. Master’s Thesis. La Paz: Universidad Catolica.
Wiebelt, M. (2004). GEMPIA – A Dynamic Real-Financial General Equilibrium Model for Poverty Impact Analysis. Kiel Working Papers (in print). Institute for World Economics, Kiel.
World Bank (1990). World Development Report. Washington, D.C.
World Bank (2000). World Development Report. Washington, D.C.
World Bank (2001). Engendering Development. Washington DC: The World Bank.
World Bank (2003a). World Development Indicators. CD-ROM, World Bank, Washington, D.C.
World Bank (2003b). Land Policies for Growth and Poverty Reduction. Washington DC: The World Bank.
World Bank (2004a). Bolivia: Public Expenditure Management for Fiscal Sustainability and Equitable and Efficient Public Services. Report 28519-Bo.
World Bank (2004b). Bolivia Poverty Assessment: Establishing the Basis for more Pro Poor Growth. Report No. 28068-Bo.
World Bank (2004c). Social Expenditure and its Relation to Poverty and Equity in Bolivia. Poverty and Social Impact Analysis. La Paz: World Bank.
Stephan Klasen
Department of Economics University of Göttingen
Rainer Thiele Kiel Institute for World Economics
Melanie Grosse, University of Göttingen
Jann Lay, Kiel Institute Julius Spatz, Kiel Institute
Manfred Wiebelt, Kiel Institute
Operationalizing Pro Poor Growth
Country Case Study: Bolivia
ANNEXES
Annex 1 Creating National Poverty Profiles and Growth Incidence Curves With Incomplete Income or Expenditure Data 1
Annex 2 The CGE Model 75
Annex 3 Description of Policy Simulations 79
Annex 4 Simulation Results 82
September 28, 2004
1
Annex 1 - Creating National Poverty Profiles and Growth Incidence Curves With Incomplete Income or Expenditure Data
1 Motivation A review of the literature reveals that it is not possible to obtain consistent time series of basic poverty measures1 – let alone inter-temporally comparable poverty profiles – for total Bolivia drawing on the results of previous empirical studies (Tables 1 and 2). The main obstacle to this exercise is the lack of reliable pre-1997 microdata on incomes and/or consumption expenditures and, thus, reliable pre-1997 estimates of basic poverty measures outside the departmental capitals of Bolivia.2 Until 1997, the Bolivian living standard measurement surveys (LSMS), which include reasonable proxies of incomes and/or consumption expenditures, cover only the departmental capitals of Bolivia. At the same time, the three rounds of the Bolivian Census of 1976, 1992, and 2001, and the three rounds of the Bolivian Demographic and Health Surveys (DHS) of 1989, 1994 and 1998, which cover the entire country, did not collect information on incomes, consumption expenditures, or any other monetary welfare indicator. Additionally, the construction of consistent time series of basic poverty measures out of the numerous point estimates and time series fractions of previous empirical studies is prevented by the use of different analysis units (households versus household members) and different welfare indicators (income versus adjusted income versus observed consumption expenditure).
However, the compilation of poverty headcount indices in Tables 1 and 2 can at least be used to derive two tentative hypotheses on the evolution of poverty in Bolivia in the era of structural reforms. First, there is a substantial urban-rural divide. In 2001, the difference in the incidence of poverty between households in rural areas and those in departmental capitals was 51.8 percentage points according to the unsatisfied basic needs approach, 27.2 percentage points if the poor are identified with a moderate poverty line, and 37.4 percentage points if the poor are identified with an extreme poverty line. Second, this divide seems to have widened in the era of structural reforms. The incidence of poverty fell – at least until 1999 – in departmental capitals, but remained more or less constant in rural areas.
A simple exercise based on growth differentials in GDP per capita provides further evidence that rural areas may have under-performed in the era of structural reforms. Since the national accounts of Bolivia do not report separate growth rates of GDP per capita for departmental capitals, other urban areas, and rural areas, we impute them in Table 3 by multiplying for each economic sector the average annual growth rate of value added per capita over the period 1989 to 2002 (taken from the national accounts) by the employment shares of those sectors in departmental capitals, other urban areas, and rural areas, respectively (estimated from the LSMS 1999).3 We find that the imputed average annual growth rate of GDP per capita between 1989 and 2002 was 1.25 percent in departmental capitals, 1.09 percent in other urban areas, but only 0.62 percent in rural areas.4 This result is mainly driven by slow growth in agriculture and fast growth in services.
1 A brief description of the poverty and inequality measures used can be found in Box A1 in the Appendix. We define
an individual as poor if her income (urban areas) or her consumption expenditure (rural areas) is below the official poverty line. For an overview of alternative concepts of poverty, see Hemmer and Wilhelm (2000).
2 Pre-1997 microdata on incomes and/or consumption expenditures for other urban areas and rural areas of Bolivia are, if at all available, derived from low-quality household surveys. See, for instance, Burger and Pradhan’s (1998) criticism on the Encuesta Socio-Económica de Hogares 1995, which collected microdata on only 381 rural households in only four departments of Bolivia (La Paz, Oruro, Potosí, and Cochabamba).
3 For comparison, we also report the sectoral GDP share in 1999 in Table 3. 4 If we account for changes in the population shares of the three regions between 1989 and 2002, the imputed growth
rate of GDP per capita was 1.24 percent in departmental capitals, 0.86 in other urban areas, and 0.67 in rural areas.
Table 1 — Poverty Headcount Indices in the Departmental Capitals of Bolivia Source Analysis Unit Welfare Indicator 1976 1986 1989 1990 1991 1992 1993 1994 1995 1996 1997 1999 2000 2001 2002
Unsatisfied Basic Needs INE (2001) Household Members Unsat. Basic Needs 66.3 53.1 39.0
Moderate Poverty Line CEPAL (2002) Households Adj. Incomea 49.4 45.6 46.8 42.3 Pereira and Jiménez (1998) Households Adj. Incomea 53.3 49.0 51.2 49.1 45.1 Jiménez and Yañez (1997) Households Adj. Incomea 53.3 49.0 51.2 49.1 45.1 47.8 Gray-Molina et al. (1999) Households Adj. Incomea 53.3 49.0 51.2 49.1 45.1 47.8 Antelo (2000) Households Adj. Incomea 53.3 49.1 46.9 UDAPSO (1995) Households Consumption Exp. 52.9 55.2 53.3 World Bank (1996) Households Consumption Exp. 51.6 52.6 Vos et al. (1998) Household Members Income 70.8 56.9 59.3 Jiménez and Landa (2004) Household Members Income 52.0 50.7 46.4 52.0 50.5 51.0 Wodon (2000) Household Members Adj. Incomeb 70.0 73.6 73.4 75.5 64.5 Psacharopoulos et al. (1993) Household Members Adj. Incomec 51.5 54.0 Hernany et al. (2001) Household Members Adj. Incomea 59.7 52.4 World Bank (2000) Household Members Adj. Incomea 52.0 50.7 48.5c Vos et al. (1998) Household Members Consumption Exp. 60.9 60.3 World Bank (1996) Household Members Consumption Exp. 60.1 61.6
Extreme Poverty Line CEPAL (2002) Households Adj. Incomea 22.1 16.8 19.2 16.4 Pereira and Jiménez (1998) Households Adj. Incomea 26.2 21.1 24.0 22.3 17.9 Jiménez and Yañez (1997) Households Adj. Incomea 26.2 21.1 24.0 22.3 18.0 20.8 Gray-Molina et al. (1999) Households Adj. Incomea 26.2 21.1 24.0 22.3 18.0 20.8 Antelo (2000) Households Adj. Incomea 24.5 20.9 19.3 World Bank (1996) Households Consumption Exp. 21.8 22.4 UDAPSO (1995) Households Consumption Exp. 23.1 Vos et al. (1998) Household Members Income 46.0 30.0 32.2 Jiménez and Landa (2004) Household Members Income 23.7 21.3 20.7 25.7 22.3 23.9 Wodon (2000) Household Members Adj. Incomeb 41.2 52.0 42.0 46.0 32.3 Psacharopoulos et al. (1993) Household Members Adj. Incomec 22.3 23.2 Hernany et al. (2001) Household Members Adj. Incomea 29.3 23.3 World Bank (2000) Household Members Adj. Incomea 25.5 21.5 22.5d Vos et al. (1998) Household Members Consumption Exp. 27.9 28.3 World Bank (1996) Household Members Consumption Exp. 28.1 29.3
Notes: a Incomes are adjusted according to the methodology of CEPAL (1995). b The adjustment factor is equal to the ratio of consumption expenditure per capita from the national accounts to the mean income per capita from the LSMS. c The adjustment factor is equal to the ratio of GDP (taken from the national accounts) to aggregate household income (estimated from the LSMS). d Arithmetic mean of the poverty indices estimated from the LSMS of March 1999 and Nov. 1999. The welfare indicator of the LSMS of Nov. 1999 is consumption expenditure per capita.
Source: Own compilation.
Table 2 — Poverty Headcount Indices in the Rural Areas of Bolivia
Unsatisfied Basic Needs INE (2001) Household Members Unsat. Basic Needs 98.6 95.3 90.8
Moderate Poverty Line UDAPSO (1995) Households Consumption Exp. 68.8a World Bank (1996) Households Consumption Exp. 82.4a Vos et al. (1998) Household Members Income 77.1a Jiménez and Landa (2004) Household Members Income 78.0 84.0 87.0 77.7 83.4 World Bank (2000) Household Members Adj. Income b 77.3 Vos et al. (1998) Household Members Consumption Exp. 88.3a World Bank (1996) Household Members Consumption Exp. 87.7a World Bank (2000) Household Members Consumption Exp. 81.7
Extreme Poverty Line UDAPSO (1995) Households Consumption Exp. 58.6a World Bank (1996) Households Consumption Exp. 72.7a Vos et al. (1998) Household Members Income 73.3a Jiménez and Landa (2004) Household Members Income 59.0 69.9 75.0 59.7 66.8 World Bank (2000) Household Members Adj. Incomeb 58.2 Vos et al. (1998) Household Members Consumption Exp. 85.8a World Bank (1996) Household Members Consumption Exp. 79.1a World Bank (2000) Household Members Consumption Exp. 58.8
Notes: a Microdata are derived from low-quality income and expenditure surveys. b Incomes are adjusted according to the methodology of CEPAL (1995).
Source: Own compilation.
4
Table 3 — Imputation of Average Annual Growth Rates of GDP per Capita, 1989 to 2002 Value Added Growth
1989 to 2002 Employment Share
in 1999 GDP Share
in 1999 Total
Bolivia Depart-mental
Capitalsa
Other Urban Areasb
Rural Areasc
Total Bolivia
Total Bolivia
Agriculture 0.54 1.44 12.10 85.64 41.56 14.66 Mining and Quarrying 0.73 0.88 0.72 2.16 1.45 9.62 Manufacturing 1.20 17.23 22.16 2.55 11.07 17.34 Electricity, Gas, and Water 3.11 0.31 0.15 0.15 0.22 2.15 Construction 2.08 8.69 8.72 2.09 5.65 3.91 Trade and Commerce 0.91 28.48 22.29 2.76 15.87 8.68 Logistics and Communication 2.53 9.23 6.35 0.46 4.84 11.12 Financial Services 3.16 5.34 1.19 0.11 2.42 15.09 Hotels and Restaurants 0.51 6.21 6.38 0.95 3.80 3.28 Personal Servicesd 1.33 10.64 9.70 1.03 6.47 4.64 Public Administration 0.04
X
11.55 10.25 2.11 6.66 9.50
| | | | | | | |
Imputed GDP Growth 1989 to 2002
1.25 1.09 0.62 0.95e
Notes: a Comprise the cities of Sucre, La Paz (incl. El Alto), Cochabamba, Oruro, Potosí, Santa Cruz, Tarija, Trinidad, and El Alto. – b Municipalities outside the departmental capitals with more than 10,000 inhabitants. – c Municipalities with less than 10,000 inhabitants. – d Including domestic services. e The observed average annual growth rate of GDP per capita for total Bolivia is 1.17 percent.
Source: Own calculations. To explore the trends in the urban-rural divide as well as other dimensions of poverty in more
depth and detail irrespective of the above mentioned data constraints, we set up a dynamic cross-survey microsimulation methodology. In Section 2, we start by developing the methodology and describing the data used. The empirical application for the case of Bolivia in Section 3 is carried out in three steps. First, we generate an inter-temporally comparable microdata set of simulated incomes for total Bolivia (i.e., departmental capitals, other urban areas, and rural areas) between 1989 and 2002, and check the consistency between observed and simulated incomes where the former are available. Second, we use the simulated incomes to estimate detailed national poverty profiles by place of residence and by household characteristics to track the evolution of poverty for different subgroups of the population over time. Third, we evaluate the “pro-poorness” of the simulated 1989-to-2002 income changes with the help of growth incidence curves. In Section 2.3, sensitivity analyses are performed (a) to check the robustness of our results to two alternative model specifications and (b) to compare our results with those derived from the asset-index (or wealth-index) approach developed by Filmer and Pritchett (2001), and Sahn and Stifel (2000, 2003). Section 4 discusses the results.
2 Data and Approach Our methodology to create national poverty profiles and growth incidence curves with incomplete income or consumption expenditure data builds upon the static cross-survey microsimulation methodology of Hentschel et al. (2000) and Elbers et al. (2003). Their objective is to analyze the spatial dimension of poverty in detailed poverty maps of national coverage for Ecuador. Their problem is that the Ecuadorian LSMS did not collect consumption expenditures for all households but only for a nationally representative sample of two-stage randomly selected households. The two-stage sample design, first selecting clusters and then households within the selected clusters, generates a sample in which the households are not randomly distributed over space, but are geographically grouped. Their solution to this problem is to combine the LSMS data with concurrent unit-record Census data of all Ecuadorian households and impute consumption
5
expenditures for those municipalities which were not included in the LSMS sample. To this end, they estimate a consumption expenditure model in the LSMS data restricting the set of covariates to those which are also available in the Census data. Then they multiply for each household in the Census its covariates with the corresponding regression coefficient from the consumption expenditure model and add a randomly distributed error term.
We have a similar objective but face more severe data constraints. The pre-1997 LSMS of Bolivia are not even nationally representative, but cover only the departmental capitals. Additionally, the concurrent Bolivian rounds of LSMS and Census are only available for 1992 and 2001, but not for the early years of the structural reform process. To overcome these data constraints, we extend the static cross-survey microsimulation methodology of Hentschel et al. (2000) and Elbers et al. (2003) by a dynamic component and replace the Census data by DHS data. The analysis proceeds in three steps. First, we choose a base period t in which we dispose of a nationally representative LSMS as well as a nationally representative DHS, and develop an empirical model of a monetary welfare indicator y (hereafter referred to as income) using the LSMS data. Similar to above, the set of covariates X is restricted to those which are also available in the corresponding DHS. The choice of the covariates is further guided by (a) the highest possible consistency between LSMS and DHS data as well as over time, and (b) the best possible fit of the regression model. We then construct a 3 x 3 block diagonal structure of the covariates by interacting them with three regional dummies, and run the weighted standard semi-log OLS regression model
tRt
Tt
Ct
Rt
Tt
Ct
Rt
Tt
Ct
XX
X
yyy
εβββ
+⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
⋅⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
=⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
000000
. (1)
where the indices C, T and R stand for departmental capitals, other urban areas, and rural areas, respectively, β are coefficient vectors, and ε is an independent error term. We account for heteroskedasticity using the covariance matrix estimator proposed by White (1980).5
Second, we check the consistency between the observed incomes of the LSMS and the simulated incomes of the DHS in period t. To this end, we apply the coefficient estimates β̂ from regression model (1) to the DHS covariates X~ and generate simulated incomes
⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
+⎟⎟⎟⎟
⎠
⎞
⎜⎜⎜⎜
⎝
⎛
⋅⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
=⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
Rt
Tt
Ct
Rt
Tt
Ct
Rt
Tt
Ct
Rt
Tt
Ct
uuu
XX
X
yyy
βββ
ˆˆˆ
~000~000~
~~~
0
. (2)
Since the regression model explains only a fraction of the variance, we add the realization of normally distributed random variables uC, uT, and uR with mean zero and a variance equal to the variance of the error term in the respective region. This simulation procedure is repeated 200 times to create 200 nationally representative samples of simulated incomes. Letting )~̂(yP be a poverty or inequality measure based on the simulated income distribution, we can then generate the conditional distribution of )~̂(yP , in particular, its mean point estimate and its prediction error, from the 200 samples of simulated incomes. The fit of the imputation can be evaluated by comparing the poverty
5 Unfortunately, the primary sample units of the pre-1997 LSMS are not available in Bolivia so that we cannot split
the error term into a spatial and an idiosyncratic component as in Elbers et al. (2003).
6
and inequality measures estimated from observed incomes of the LSMS, )(yP , with those
estimated from simulated incomes of the DHS, )~̂(yP .
Third, we choose an earlier period t–1 in which the LSMS covers only the departmental capitals and partially re-run our regression model
Ct
Ct
Ct
Ct Xy 1111 −−−− +⋅= εβ (3)
to obtain the coefficient estimates and the variance of the error term for the departmental capitals in period t–1. We assume that the absolute differences in the regression coefficients between departmental capitals on the one hand, and other urban areas and rural areas on the other hand, remain constant between period t–1 and t,6 and arrive at the coefficient estimates for other urban areas and rural areas, respectively, in period t–1
)ˆˆ(ˆˆ11
Ct
Tt
Ct
Tt ββββ −+= −− and )ˆˆ(ˆˆ
11Ct
Rt
Ct
Rt ββββ −+= −− . (4)
In a similar vein, by assuming that the relative change in the variances of the error terms between period t–1 and t is identical for all three regions, we obtain the variances of the error terms for other urban areas and rural areas, respectively, in period t–1
)var()var()var()var( 1
1 Ct
CtT
tTt ε
εεε −
− ⋅= and .)var()var()var()var( 1
1 Ct
CtR
tRt ε
εεε −
− ⋅= (5)
Repeating the simulation exercise (2) with the coefficient estimates from equations (3) to (5) and the DHS data in period t–1, we can create 200 nationally representative samples of simulated incomes in period t–1. Again we can compare the poverty and inequality measures between the two household surveys. In contrast to above, however, this exercise is only possible for the departmental capitals where observed incomes are available. After this consistency check, we can use the simulated incomes (a) to construct inter-temporally comparable poverty profiles of national coverage for Bolivia and (b) to evaluate the “pro-poorness” of changes in the distribution of simulated incomes over time with the help of growth incidence curves.
Our set of LSMS consists of four multi-purpose household surveys conducted by the Instituto Nacional de Estadísticas de Bolivia (National Statistical Office of Bolivia): the 2nd round (Nov. 1989) and the 7th round (July to Dec. 1994) of the Encuesta Integrada de Hogares (EIH), and the 1st round (Nov. 1999) and the 4th round (Nov. 2002) of the Encuesta Continua de Hogares (ECH). The EIH cover only the departmental capitals of Bolivia, while the ECH are nationally representative. Two-stage sampling techniques were used in selecting the sample of households, and sampling was done in a way to ensure self-weighting. The purpose of the LSMS is to collect individual, household, and community level data to measure the welfare level of the sampled population and its changes over time. In addition to income and/or expenditure data, the LSMS provide information on demographics, asset ownership, education, employment, and health.
In order to be able to compare our results with earlier empirical studies, we largely use per capita incomes as our income indicator. Only when we turn to the impact of household size on poverty, do we check the sensitivity of our results by also using equivalized incomes. As welfare indicator, we use monthly consumption expenditures including own consumption (excluding annualized costs for
6 We check the robustness of our results to an alternative assumption on the evolution of the regression coefficients
between period t–1 and t in Section 4.1.
7
durable consumer goods) for rural areas, and monthly labor income (excluding fringe benefits)7 plus monthly capital income for urban areas. The choice of the mixed measurement unit, which is common for Bolivia (see, for instance, INE-UDAPE 2002), can be justified by that (a) an all- expenditure specification is not possible since the EIHs collected only income but no expenditure data, and (b) an all-income specification is not preferable since incomes only poorly reflect the long-term welfare in rural areas due to large seasonal income fluctuations, a high degree of own consumption in agricultural households (Deaton and Zaidi 2002). On the practical level, it appears that incomes in rural areas are seriously under-estimated leading to implausible poverty figures, a common finding in many developing countries. In order to account for non-declaration of incomes, we apply a statistical matching approach similar to Hernany (1999). By contrast, we do not adjust for sub-declaration (under-reporting) of incomes (e.g. by scaling up the mean income and mean consumption expenditures in the LSMS to those in the national accounts) because (a) it is a priori not clear whether national account data or LSMS data are more accurate, and (b) Bolivia does not report separate national account data for departmental capitals, other urban areas, and rural areas.8
To identify the poor, we use the two sets of poverty lines provided by the Unidad de Análisis de Políticas Sociales y Económicas (UDAPE) (Table 4). The extreme poverty lines are given by the costs of food baskets which reflect (a) the nutritional requirements of adults, and (b) the local eating habits of the middle quintile of the income distribution. The moderate poverty lines additionally include the costs of non-nutritional basic needs and are obtained by multiplying the extreme poverty lines by the inverse of local Engel coefficients. Since no rural poverty lines are available for 1989 and 1994, we extrapolate the relative difference between the rural poverty line and the weighted average urban poverty line of 1999. The poverty lines are updated using the (regionally-disaggregated) prices for the goods included in the poverty basket.9
In Table 4 we also report a poverty line that simply takes the 1989 values and inflates them to 2002 using the Consumer Price Index (2002cpi) to see whether the prices paid by the poor have developed differently to the overall price level. The results show that overall prices have risen faster in all parts of the country than the prices faced by the poor. The difference between 1989 and 2002 is on the order of 20-30%. To the extent that the poverty basket has not changed over this time period, 10 this implies that the poor have been benefiting from falling relative prices of the goods they consume most and this will have enabled some of the poor to escape poverty.
Our set of Demographic and Health Surveys (DHS) consists of the first three Bolivian rounds which were conducted in 1989, 1994, and 1998.11 Two-stage sampling techniques were used to select nationally representative samples of women aged between 15 and 49 who serve as respondents of the DHS. The main objective of the DHS is to collect information on health and fertility trends. Additionally, the questionnaire includes some questions on the educational attainment and the employment situation of the respondent and her partner, as well as on the asset ownership of the household.
7 Only if we exclude fringe benefits is the measurement unit inter-temporally comparable between 1989 and 2002.
This is because the EIHs collected, if at all, only the incidence and type of fringe benefits but not their monetary equivalent. As a consequence, our poverty estimates for 1999 and 2002 are somewhat higher than the official figures provided by INE (var. iss.).
8 For an description and evaluation of, and an analysis of the sensitivity of poverty measures to, different adjustment methods see Székely et al. (2000).
9 Further information on the construction of the Bolivian poverty lines can be found in Box A2 in the Appendix. 10 One should bear in mind that the poverty line is a Laspeyres Index using a fixed basket. As a result of changing
preferences and prices, the poor might have changed their consumption habits over time which would obviously affect the assessment of the differences between the poverty line escalation and the development of the CPI.
11 The fourth Bolivian DHS round of 2003 has not been made publicly available until finishing this study.
8
The covariates taken from the two data sources and their sample means are listed in Tables A1 and A2 in the Appendix to this Annex. They can be grouped into five categories: information on (a) demographics of the household, (b) asset ownership of the household, (c) educational attainment of adult men and women, (d) employment situation of adult men and women, and (e) health situation of children. By choosing suitable variables and dummy categories, we obtained a high degree of consistency both across surveys and over time.
Table 4 — Poverty Lines for Bolivia (in current Bolivianos)
Urban Areas Chuquisaca 138.5 241.8 335.4 335.6 395.5 73.3 127.9 169.4 169.5 209.2 La Paz (Capital City) 135.3 227.9 324.0 327.0 383.3 75.2 126.6 180.1 181.8 214.6 La Paz (El Alto) 116.6 192.6 270.4 272.6 332.9 70.7 116.7 164.1 165.5 201.8 Cochabamba 142.1 253.2 351.1 351.3 405.8 71.8 127.6 177.3 177.4 204.9 Oruro 123.0 207.1 294.7 297.4 351.1 75.2 126.6 163.9 165.3 214.6 Potosí 113.1 190.5 271.0 273.5 323.0 75.2 126.6 150.7 152.1 214.6 Tarija 144.3 257.3 356.8 351.3 412.1 71.8 127.9 178.6 177.4 204.9 Santa Cruz 141.8 237.8 354.7 343.9 404.8 72.0 120.7 180.2 174.7 205.5 Beni 141.8 237.8 354.7 343.9 404.8 72.0 120.7 180.2 174.7 205.5 Pando 141.8 237.8 354.7 343.9 404.8 72.0 120.7 180.2 174.7 205.5 Rural Areas 96.9b 164.4b 233.6 233.4 276.6 55.2b 93.4b 131.2 133.0 157.6 Pop. Weighted Average 119.5 204.8 299.3 298.1 351.2 65.9 112.3 160.6 160.3 190.5 Notes: a Since no poverty lines are available for the 2nd round (Nov. 1989) of the EIH, they are constructed as the
arithmetic mean of the poverty lines for the 1st round (March 1989) and the 3rd round (Sept. 1990) of the EIH. –b Constructed by extrapolating the relative difference between the rural poverty line and the weighted average urban poverty line of 1999. –c 1989 poverty lines inflated with the CPI.
Source: Own compilation based on unpublished data of UDAPE.
3 Empirical Results We build our methodology around the base period 1999 and then apply it to the earlier periods 1989 and 1994. Additional data constraints impede our empirical analysis in three respects. First, to create inter-temporally comparable samples of simulated incomes for Bolivia it would be ideal to use a set of covariates which is available in all three pairs of concurrent household surveys of 1989, 1994, and 1999. At the same time, however, the availability of covariates in the LSMS and the DHS changes over time due to changes in their questionnaires. In order to avoid too small a set of covariates we, thus, decided to use three different sets of covariates to (a) check the consistency between the LSMS and the DHS data in 1999, (b) to create 200 samples of simulated incomes in the DHS 1989 data, and (c) to create 200 samples of simulated incomes in the DHS 1994 data.12
Second, since no Bolivian DHS round was conducted in 1999, we have to use the DHS 1998 data for our consistency check. That is, we compare the poverty and inequality measures based on observed incomes of the LSMS 1999 with those based on simulated incomes of the DHS 1998, assuming that the distribution of the covariates remained reasonably constant in between. By contrast, for 1989 and 1994 we dispose of concurrent rounds of LSMS and DHS. Third, due to its focus on health and fertility trends, the DHS data (in 1989) only include households with at least
12 To put it more formally, we only require that the set of covariates be identical for the LSMS and the DHS in period
t–1 as well as for the LSMS in period t. To check for robustness, we also performed our subsequent empirical analysis for the smaller set of common covariates. While, as expected, the consistency check performed worse, the empirical results did not change qualitatively.
9
one woman of reproductive age (i.e., aged between 15 and 49). We, thus, have to replicate this implicit sample selection in the LSMS data.13
3.1 Consistency Check In Tables 5 and 6, we provide four sets of poverty and inequality measures: (a) their point estimates from observed incomes of all households in the LSMS, (b) their point estimates from observed incomes of households with at least one woman of reproductive age in the LSMS, (c) their mean point estimates and standard deviations from 200 samples of predicted incomes in the LSMS, and (d) their mean point estimates and standard deviations from 200 samples of simulated incomes in the DHS.14 Taking differences between successive members of this series enables us to decompose the overall difference between observed and simulated poverty and inequality measures into three components related to (a-b) the implicit sample selection in the DHS data, (b-c) the specification of the error term in the underlying regression model, and (c-d) differences in the distribution of the covariates between LSMS and DHS.
For 1989 and 1994, for which the consistency check is limited to departmental capitals, the results are very encouraging. Restricting the sample to households with at least one woman of reproductive age does not induce a serious bias in estimating poverty and inequality measures. Using a normally distributed error term (rather than drawing observed residuals) to create 200 samples of predicted incomes in the LSMS, only slightly understates the poverty headcount, renders a very close fit for the poverty gap, and only slightly overstates the squared poverty gap.15 It also only slightly understates income inequality as evidenced by lower values of the Gini coefficient and the Atkinson indices. The transition from LSMS data to DHS data does, if at all, only slightly reduce the poverty and inequality measures.
For 1999, the situation is somewhat less favorable. Only the inequality measures continue to be unbiased by sample selection, while the poverty measures seem to be upward biased. Our specification of the error term seriously underestimates the Atkinson index with 2=ε in departmental capitals. Most striking, however, are the large differences between predicted and simulated poverty indices, particularly so in rural areas. The underlying reason is most probably the lack of consistency with respect to the collection period of the two underlying household surveys. The DHS 1998 data, the covariates of which were used to create the simulated incomes, were collected during an economic boom. By contrast the observed incomes of the LSMS 1999 were collected after a sharp economic downturn when Bolivia experienced strongly negative growth in GDP per capita.
These inconsistencies notwithstanding, we are confident that the conditions for applying our dynamic cross-survey microsimulation methodology are fulfilled for the case of Bolivia. First, the simulations can accurately reproduce the observed poverty trends in departmental capitals, where we have observed incomes for comparison. The differences between observed and simulated poverty measures are small compared to their changes over time. Second, the DHS 1998 data, which are least consistent to those of the corresponding LSMS, are not used in the subsequent analysis. Only the poverty profiles and growth incidence curves for 1989 and 1994 draw on
13 For 1994 and 1998 (but not for 1989), the DHS provide an additional data module on – and responded by – male
adults. We opted against using this data module for two reasons: (a) the information was not collected for the husbands and partners of all women included in the main module so that we would have had to reduce the sample size and possibly would have introduced another sample-selection bias, and (b) our microdata set of simulated incomes would no longer be inter-temporally comparable over the whole observation period from 1989 to 2002.
14 The underlying regression results are not reported here, but are available upon request. 15 This is because the distributions of the error terms is slightly skewed to the right. The kernel density graphs of the
errors terms are not reported here, but are available upon request.
10
simulated incomes of the DHS. Those for 1999 and 2002 are based on observed incomes of the LSMS.
In Section 2, we assumed that the absolute difference in the regression coefficients between departmental capitals on the one hand, and other urban areas and rural areas on the other hand, remained constant between 1989 and 1999. If this assumption does not hold, i.e., if the coefficients in rural areas deteriorated relative to those in urban areas, the decline in poverty in rural areas shown in the subsequent analysis would be overstated. We address this potential bias in Section 4.1. Another factor that may contribute to overstating the decline in poverty – albeit in this case not limited to rural areas – is that the degree of underreporting, which is common to all income and expenditure surveys, may have fallen over time due to improvements in the questionnaire design. Taken together, we, thus, caution to treat the reduction in poverty as an upper bound, and particularly so in rural areas.
3.2 Poverty Profiles After having completed this consistency check, we can proceed to construct inter-temporally comparable poverty profiles of national coverage for Bolivia to get an understanding of where and who the poor are. Where possible – in departmental capitals throughout the entire observation period and in the rest of the country for 1999 and 2002 – we use poverty measures estimated from observed incomes of the LSMS. The remaining gaps are filled with the mean point estimates and the standard deviations of poverty measures from 200 samples of simulated incomes in the DHS. In what follows, we focus on delineating major poverty trends of Bolivia during the era of structural reforms. The discussion of their underlying causes is deferred to Section 5.
We start our empirical analysis with a disaggregation of the poverty headcount by place of residence in Table 7.16 Between 1989 and 2002, total Bolivia experienced a significant reduction in the incidence of poverty. Moderate poverty decreased from three quarters to two thirds of the population. The reduction in extreme poverty was even more spectacular; it decreased by 17 percentage points. Yet, the picture is not all favorable. In the late 1990s, the poverty trend reversed and the incidence of moderate and extreme poverty in total Bolivia started to increase again.
As expected, rural households were more likely to be poor than those in departmental capitals and other urban areas, even after controlling for local cost-of-living differences. What is more of concern here is that rural households did not fully participate in the reduction of moderate poverty between 1989 and 1999. Departmental capitals and other urban areas could reduce the incidence of moderate poverty by 16 and 12 percentage points, respectively. In rural areas, this reduction was only 6 percentage points – despite starting from a higher level of poverty.17 By contrast, households in departmental capitals were most affected by the economic downturn in the late 1990s, accounting for almost the entire increase in the incidence of moderate and extreme poverty in total Bolivia between 1999 and 2002. Taken together, the poverty trends suggest that rural areas were quite detached from improvements and deteriorations in the overall economic environment.
16 For the corresponding tables for the poverty gap and the squared poverty gap see Tables A3 and A4 in the
Appendix. 17 That is, in relative terms, the performance of rural areas was even worse. As concerns extreme poverty, rural areas
also experienced the lowest absolute (!) reduction the poverty headcount index between 1989 and 1999.
11
Table 5 — Comparison of Poverty Indices Based on Observed and Simulated Incomes 1989 1994 1999 LSMS Data DHS
areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on predicted and simulated incomes). – a The covariates for the simulation exercise are taken from the third Bolivian DHS round, which was conducted in 1998. * denotes that the 95-percent confidence interval includes the corresponding index value in the “Sample” column.
Source: Own calculations.
12
Table 6 — Comparison of Inequality Indices Based on Observed and Simulated Incomes 1989 1994 1999
areas, and mixed income-expenditure data for total Bolivia. Standard errors of the inequality indices in brackets (only applicable to those based on predicted and simulated incomes). – a The covariates for the simulation exercise are taken from the third Bolivian DHS round, which was conducted in 1998. * denotes that the 95-percent confidence interval includes the corresponding index value in the “Sample” column.
Source: Own calculations.
13
There are also substantial differences in the incidence of poverty across the nine departments of Bolivia. The moderate poverty headcount in 1989 ranged from 62 percent in Santa Cruz to 92 percent in Potosí. The corresponding figures for the extreme poverty headcount were 31 percent and 79 percent, respectively. The departmental distribution of the poverty headcount index was also very stable in Bolivia. While Santa Cruz, which is a major host of commercial agriculture and food-processing industry, had the lowest incidence of poverty throughout the entire observation period, it was highest in Potosí, followed by Chuquisaca, which are particularly dependent on subsistence agriculture.
To gain insights into other dimensions of poverty, we proceed with a disaggregation of the poverty headcount index by household characteristics for total Bolivia as well as for its departmental capitals, other urban areas, and rural areas in Tables 8 to 11.18 By far the most important determinant of poverty and its change over time is education. Households where the average education of adult members was primary schooling or less (i.e., <= 5 years) rarely escaped poverty, even in departmental capitals. Secondary schooling (i.e., 6 to 12 years) and tertiary schooling (i.e., 13 years and above) substantially reduced the likelihood of poverty. Their contribution to reducing the incidence of poverty (relative to the next lower schooling category) was highest in rural areas and lowest in other urban areas. Over time, the distribution of the poverty headcount indices across schooling groups changed significantly. While the incidence of poverty fell in all three schooling groups, the returns to secondary schooling declined somewhat while the returns to tertiary schooling increased substantially.
Table 7 — Spatial Disaggregation of the Poverty Headcount in Bolivia, 1989 to 2002
Notes: Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
18 For the corresponding tables for the poverty gap and the squared poverty gap see Tables A5 to A12 in the Appendix.
14
Table 8 — Disaggregation of the Poverty Headcount in Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 76.88 72.37 65.21 67.22 56.24 50.43 38.35 39.24 (0.50) (0.45) (0.61) (0.45) By Hh Size
(0.53) (0.82) (0.75) (0.92) Notes: Poverty indices are calculated using mixed income-expenditure data. Standard errors of the poverty indices in brackets
(only applicable to those based on simulated data). – a Women aged between 15 and 49 and their husbands and partners. – b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. – c Women aged between 15 and 49.
Source: Own calculations.
15 Table 9 — Disaggregation of the Poverty Headcount in the Departmental Capitals of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 67.21 59.49 51.05 55.13 39.38 28.78 24.22 27.03 By Hh Size
Notes: Poverty indices are calculated using income data. a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations.
16
Table 10 — Disaggregation of the Poverty Headcount in Other Urban Areas of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 81.05 75.13 69.09 67.70 62.84 53.31 34.31 36.65 (1.32) (1.16) (1.44) (1.22) By Hh Size
(1.22) (1.69) (1.84) (1.91) Notes: Poverty indices are calculated using income data. Standard errors of the poverty indices in brackets (only applicable to those
based on simulated data). – a Women aged between 15 and 49 and their husbands and partners. – b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. – c Women aged between 15 and 49.
Source: Own calculations.
17
Table 11 — Disaggregation of the Poverty Headcount in Rural Areas of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 89.66 89.55 83.37 83.83 74.59 76.05 59.98 57.24 (0.59) (0.47) (0.92) (0.62) By Hh Size
(0.59) (1.02) (0.99) (1.30) Notes: Poverty indices are calculated using expenditure data. Standard errors of the poverty indices in brackets (only applicable to
those based on simulated data). – a Women aged between 15 and 49 and their husbands and partners. – b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. – c Women aged between 15 and 49.
Source: Own calculations.
18
We find that a large number of children is also an important factor in shaping the distribution of poverty, namely in two respects. First, large households were on average poorer than small households and the relationship between poverty and household size strengthened over time, above all in rural areas where large households did not participate at all in the overall reduction of poverty.19 Second, households where the share of members in working age was below 50 percent were more likely to be poor than other households. The relationship between the age composition of households and poverty is strongest in the departmental cities and weakest in rural areas, but its strength increased over time in all three regions.
To analyze the impact of employment on poverty, we first look at the profession of the principal wage earner.20 Given the large differences between the sectoral employment shares and the sectoral GDP shares (as shown in Table 3), it is not surprising to find a steep gradient in the poverty incidence across professions. White-collar workers were by far least likely to be poor in 1989, followed by workers in sales & services.21 At the other end of the spectrum were agricultural and blue-collar workers. Like above, we find that the differences in the poverty incidence across professions increased over time. The absolute (!) poverty headcount index of the relatively rich white collar workers and workers in sales & services fell more than twice as much as the poverty headcount index of the relatively poor agricultural and blue collar workers. The figures on unemployed principal wage earners are less straightforward to interpret. In departmental capitals, households with unemployed principal wage earners took an intermediate position in the poverty ranking throughout the entire observation period. This may be due to two opposing factors. On the one hand, total household income is reduced if the principal wage earner does not have gainful employment. On the other hand, in the absence of unemployment and other social benefits, only rich households can afford long unemployment spells (or long schooling periods) of the principal wage earner. By contrast, in other urban areas and rural areas, they started from being most likely to be poor in 1989, and ended up in an intermediate position in the poverty ranking in 2002. However, we caution not to read too much into this finding. The group of households with unemployed principal wage earners is so small (especially outside departmental capitals) that this finding is very sensitive and its information content is very low. Second, we turn to female labor market participation. Households where no adult woman had gainful employment were more likely to be poor than other households in departmental capitals and other urban areas, but less likely to be poor than other households in rural areas (except in 1989). Female labor market participation, thus, seemed to be a successful strategy to lift households out of poverty in the former two regions. By contrast, in rural areas, poverty seems to have forced women to work.
The role of the age of the household head in shaping the distribution of poverty is small and not straightforward. In 1989, older household heads tended to be richer than younger household heads in departmental capitals and other urban areas but poorer in rural areas. Between 1989 and 2002, households with heads aged between 50 and 65 outperformed the other age groups in all three regions. As a result, the relationship between poverty and the age of the household head turned U-shaped in departmental cities and other urban areas. As expected, the incidence of poverty was smaller among households with Spanish-speaking heads. Additionally, their welfare seemed to be more volatile. They benefited more from the economic boom between 1989 and 1999, but also suffered more from the subsequent economic downturn. This finding is partly due to that Spanish-speaking household heads are over-represented in departmental capitals. The explanatory power of the gender of the household head is negligible. If at all, households headed by women were slightly 19 We check the robustness of this result using incomes per adult equivalent (rather than income per capita) as welfare
indicator in Section 3.2. 20 Unfortunately, data constraints prevent us from further disaggregating the professional categories. Our
disaggregation is most problematic in the case of “sales & services” where we have to lump together bankers with street vendors. For the exact definition of the term “principal wage earner” see the notes of Tables 8 to 11.
21 It could be argued that the poverty headcount index in the latter category is downward biased since the incomes of self-employed, who are over-represented in sales & services, may not always be measured net of costs. However, we find the same ranking in rural areas, where we use consumption expenditures rather than incomes.
19
better off, a finding common to many Latin American countries (see Marcoux 1998). But we caution that female-headed households represent a very heterogeneous group (e.g., single female elderly, single female professionals, divorced women, and women of migrant workers) so that it may well be that sub-groups of female-headed households are particularly vulnerable to poverty.
3.3 Growth Inequality Decomposition and Growth Incidence Curves Poverty profiles are suitable means to track the evolution of the incidence, intensity, and severity of poverty for different subgroups of the population over time. However, they can only poorly disentangle to what extent the observed poverty trends are due (a) to changes in mean income or (b) to changes in the relative income distribution. Two ways to provide further insights about the links between poverty, inequality, and growth trends: the first is to do a growth inequality decomposition of the observed poverty reduction (Datt and Ravallion 1992) and the second is to estimate the rates of pro-poor growth and the growth incidence curves (Ravallion and Chen 2003).
The decomposition of the observed poverty reduction into a growth and an inequality contribution (and an interaction term which cancels if one calculates the average of a ‘forward’ and ‘backward’ decomposition) is using the methods proposed by Datt and Ravallion (1992). As discussed in detail in Grimm and Günther (2004), the distribution component in this decomposition also implicitly includes the impact of changes in the real value of the poverty line (i.e., how prices paid by the poor have moved relative to the overall price level). As shown in Table 4, the prices paid by the poor (in particular food prices) have risen somewhat less than the overall price level (particularly in recent years) so that the purchasing power of the poor has increased by more than suggested by the change in their real incomes. This is implicitly captured in the decomposition as a distributional shift favoring the poor.
1989–1999 1999–2002 1989–2002 Total Bolivia Change in poverty -0.181 0.008 -0.173 Growth component -0.090 0.019 -0.075 Redistribution component -0.091 -0.011 -0.098 Departamental Capitals Change in poverty -0.157 0.027 -0.130 Growth component -0.077 0.015 -0.073 Redistribution component -0.079 0.012 -0.056 Other Urban Areas Change in poverty -0.270 0.021 -0.250 Growth component -0.136 0.038 -0.080 Redistribution component -0.135 -0.017 -0.170 Rural Areas Change in poverty -0.157 -0.027 -0.184 Growth component -0.056 -0.008 -0.071 Redistribution component -0.100 -0.020 -0.113
Notes: Calculated using the Datt-Ravaillion (1992) method of growth-inequaltiy decomposition.
Source: Own calculations.
The result of the decomposition analysis (Table 12a) for the entire period show that about two thirds of the 10 percentage points decline in poverty for total Bolivia is attributable to growth, and about one third to a distributional shift favoring the poor. As the income distribution hardly shifted between the two periods (Table 3 of the main document), most of this distributional shift is actually due to the poverty line effect which increased the real purchasing power of the poor.22 Considering sub-periods and different parts of the country shows a more differentiated picture. In the period 1989-1999 both the growth and redistribution (and/or poverty line) effect served to reduce poverty in all parts of the country. In the latter three years, the picture has changed drastically. Now poverty has increased nationally, and particularly in capital cities where 60% is due to falling incomes and 40% due to adverse distributional shifts. For the extreme poverty line (Table 12b), the growth component seems to be less important in poverty reduction, but the redistribution component becomes more important. In the period 1989-2002, of the 17 percentage points poverty reduction, more than one half is due to redistribution (and/or the poverty line effect which is even larger here) and less than one half is due to growth.
22 When we additionally split out the poverty line effect (results are not shown here, but are available upon request),
we find for the period 1989 to 1999 the “pure” redistribution to contribute to poverty reduction in departmental capitals and other urban areas and zero for rural areas in the case of the moderate poverty line. For the extreme poverty line, the “pure” redistribution also becomes positive in rural areas. From 1999 to 2002, the “pure” redistribution effect leads to a poverty increase in all regions for both poverty lines. For the whole period from 1989 to 2002, the “pure” redistribution was poverty increasing in nearly all regions, except using the extreme poverty line for other urban areas and rural areas. The “poverty line shift” redistribution is poverty decreasing in all areas for all periods and both poverty lines. As explained above, this is due to the slower increase of food prices compared to overall prices.
21
To evaluate whether the simulated income changes over time were “pro-poor” in the sense that the poor benefited more from economic growth than the rich, we apply the methodology of growth incidence curves (GIC) developed by Ravallion and Chen (2003). Comparing two periods, t-1 and t, the growth incidence curve plots the cumulative share of the population (depicted on the x-axis) against the income growth rate of the ξ-th quantile (depicted on the y-axis) when the analysis units are ranked in ascending order of their income. It is given by
1)('
)('1
)()(
:)(111
−⋅=−=−−− ξξ
ξξ
ξt
t
t
t
t
tt L
Lyy
yy
g , (6)
where )(' ξL is the slope of the Lorenz curve at the ξ-th quantile, and y is the mean income. It can be shown that the area under the GIC up to the poverty headcount index 0P gives (minus one times) the rate of change of the Watts index23 over time
∫∫ ⋅=⋅=−00
00
)()(log tt P
t
Ptt dgd
dtyd
dtdW
ξξξξ
. (7)
The desirable axiomatic properties of the Watts index motivate evaluating the “pro-poorness” of economic growth by comparing the growth rate of mean income with the mean of the income growth rates of the poor in period t–1
∫−
⋅⋅=−
01
00
1
)(1:tP
tt
t dgP
PPG ξξ (8)
which Ravallion and Chen (2003) coined the “rate of pro-poor growth”.24
The comparison of the growth rates25 is shown in Table 13. Between 1989 and 1999, economic growth in Bolivia can be classified as pro-poor. For both poverty lines and for all three regions, the rates of pro-poor growth exceeded the growth rate of mean income suggesting that economic growth was accompanied by falling inequality.26 For departmental capitals, the income distribution of 1999 even first-order dominates the income distribution of 1989 as evidenced by that the GIC lies above 0 for all ξ (Figure A2 in the Appendix). For other urban areas and rural areas, this condition is met at least for all poor. That is, abstracting from individual income mobility across quantiles, the welfare of all citizens in departmental capitals, and of all poor citizens in the rest of the country, improved during the 1990s.
Between 1999 and 2002, the economic growth performance differed substantially between the three regions. The departmental capitals experienced a strongly anti-poor contraction, which wiped out a substantial part of the gains the poor had made in the previous ten years. In other urban areas, this contraction was pro-poor so that, despite negative growth rates in mean income, the poor could more or less keep their living standard. In rural areas, incomes even continued to rise (albeit very slowly), and income growth continued to be somewhat higher for the poor than for the non-poor. Given that (a) most income is generated in urban areas, but (b) most poor live in rural areas, economic growth in total Bolivia was negative between 1999 and 2002, but only slightly anti-poor or even pro-poor depending on the choice of the poverty line.
23 See Box A3 in the Appendix for a description of the Watts index and of its axiomatic properties. 24 Alternative approaches of measuring pro-poor growth can be found in Klasen (2004) and Son (2003). 25 For the corresponding growth incidence curves see Figures A1 to A12 in the Appendix. 26 The particularly high growth rate of mean income for total Bolivia (2.23 percent) is due to a shift in the composition
of the population from poorer rural areas to richer urban areas.
22
Table 13 — Annual Average Income Growth per Capita, 1989 to 2002
1989–1999 1999–2002 1989–2002
Total Bolivia Growth Rate of Mean Income 2.23 -1.29 1.41 Mean of Income Growth Rates of
Notes: Annual average income growth rates are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia.
Source: Own calculations.
With the exception of the strongly anti-poor contraction in departmental capitals in recent years, economic growth in Bolivia has been pro-poor since 1989, and particularly so in rural areas.27 This result seems to be at odds with Table 7 which shows only slowly falling poverty rates in rural areas since 1989. However, this puzzle resolves when taking into account that the depth of poverty in rural areas is so large that even substantial pro-poor growth did not lift the poor above the poverty line.28 Hence, the prime concern is not that economic growth in the 1990s was anti-poor, but that it was so low and that the initial income inequality was so high that the poor remained poor despite some welfare improvements. It would take another decade of such economic growth to make serious inroads into poverty. Unfortunately, the future prospects are even bleaker. If the meager growth performance of the Bolivian economy since 1999 continues, rural poverty will decline even less and urban poverty will rise sharply.
27 Jimenez and Landa (2004) also provide pro-poor growth rates for total Bolivia between 1999 and 2002. They find
that the anti-poorness of the recent contraction was not restricted to departmental capitals. The main difference to our analysis is that they rely on incomes (rather than consumption expenditure) as welfare indicator in rural areas. Rural per-capita income exhibits much lower and falling levels between 1999 and 2002, while rural per–capita consumption expenditure remained roughly constant.
28 But it did reduce the poverty gap in rural areas as evidenced in Table A3 in the Appendix.
23
4 Sensitivity Analyses Before drawing conclusions from the national poverty profiles and growth incidence curves, we perform three sensitivity analyses. First, we check the robustness of our results to alternative assumptions on the dynamics of the cross-survey microsimulation methodology. Second, we analyze how our results change if welfare is measured by income per adult equivalent rather than income per capita. Third, we contrast our results with those derived from the asset-index (or wealth-index) approach developed by Filmer and Pritchett (2001), and Sahn and Stifel (2000, 2003).
4.1 Accounting for Growth Differentials in GDP per Capita between Urban and Rural Areas One of the basic assumptions of our dynamic cross-survey microsimulation methodology is that the absolute difference in the regression coefficients between departmental capitals on the one hand, and other urban areas and rural areas on the other hand, remained constant between 1989 and 1999. The widening of the urban-rural divide during that time is, thus, entirely attributed (a) to changes in the endowment of covariates in favor of urban areas, and (b) to nationwide changes in the return to covariates in favor of those covariates which are relatively abundant in urban areas. If this assumption does not hold, i.e., if additionally (c) the returns to covariates in rural areas deteriorated relative to those in urban areas, the widening of the urban-rural divide would be understated. To get an idea of the possible size of this bias we have to simulate the opposite scenario where we assume that the widening of the urban-rural divide between 1989 and 1999 is entirely due to a deterioration of the returns to covariates in rural areas relative to those in urban areas. Since it is a priori not clear which covariates are affected and to what extent, we take a rather simple approach and attribute the regional growth differentials in GDP per capita to growth differentials in the regression coefficients of the regional dummies.
This sensitivity analysis proceeds in three steps. First, using the same approach as in Table 3, we impute the 1989-to-1994 and the 1994-to-1999 cumulative growth differentials in GDP per capita between departmental capitals on the one hand, and other urban areas and rural areas on the other hand. We find that the economic growth performance was nearly identical across the three regions in the first half of observation period, but it differed substantially thereafter. Between 1989 and 1994, departmental capitals (cumulatively) grew by only 0.3 percent faster than other urban areas and also by only 0.3 percent faster than rural areas. The corresponding figures for the period from 1994 to 1999 are 2.13 percent and 9.19 percent, respectively. Second, we sterilize the growth differentials in GDP per capita by adding (a) for other urban areas and (b) for rural areas, the 1994-to-1999 growth differential in GDP per capita (relative to departmental capitals) to the 1994 regression coefficient of the corresponding regional dummy, and sum of the 1989-to-1994 and the 1994-to-1999 growth differential in GDP per capita (relative to departmental capitals) to the 1989 regression coefficient of the corresponding regional dummy. Third, we partially re-run our simulation with the adjusted coefficients to generate an adjusted spatial disaggregation of the poverty headcount in Bolivia in Table 14a.29
Comparing the results with the corresponding entries in Table 7 reveals that the bias of neglecting a possible deterioration of the returns to covariates in rural areas relative to those in urban areas is small. Sterilizing the regional growth differentials in GDP per capita reduces the incidence of moderate poverty in rural areas in 1989 by less than 2 percentage points and the incidence of extreme poverty by less than 4 percentage points. This implies that the inferior performance of rural areas in reducing the poverty headcount index is not primarily due to urban-rural growth differentials in GDP per capita. Instead, due to high initial inequality, only relatively few rural households were initially just below the poverty lines so that a given growth of GDP per capita between 1989 and 2002 lifted only relatively few rural households over the poverty lines.
Table 14b calculates the corresponding rates of pro-poor growth for the various regions. Due to lower growth in rural areas and towns, overall (mean) growth in Bolivia is now smaller between 29 For the corresponding tables for the poverty gap and the squared poverty gap, see Tables A13 and A14 in the
Appendix.
24
1989 and 1999, and the growth is also less pro-poor as the rate of growth in rural areas, whose population predominates among the poor, is now estimated to have been lower. But the qualitative results from above do not change.
Table 14 a— Adjusted Spatial Disaggregation of the Poverty Headcount in Bolivia, 1989 to 2002
(0.60) (0.49) (0.90) (0.65) Notes: Only poverty indices based on simulated data changed relative to Table 7. Poverty indices are calculated using income data
for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: own calculations.
Table 14.b — Adjusted Annual Average Income Growth per Capita, 1989 to 2002 1989–1999 1999–2002 1989–2002
Total Bolivia Growth Rate of Mean Income 2.02 -1.29 1.25 Mean of Income Growth Rates of
Notes: Annual average income growth rates are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia.
Source: Own calculations.
25
4.2 Adult Equivalent Scales To this point, welfare was measured by income per capita in departmental capitals and other urban areas, and consumption expenditure per capita in rural areas. To account for different preferences and needs of adults and children as well as for economies of scale within the household, we check the robustness of the impact of household size on poverty using income per adult equivalent and consumption expenditure per adult equivalent, respectively. The number of adult equivalents is defined as
θκκ )( 2211 chichiaduAE ⋅+⋅+= , (9)
where adu is the number of adults (age ≥ 15 years), chi1 the number of children aged between 6 and 14 years, and chi2 the number of children aged 5 years and below. 1κ and 2κ reflect the costs of children relative to the costs of an adult and, thus, correct for the age composition of the household, and θ controls the extent of economies of scale within the household (National Research Council 1995). In line with Gasparini et al. (2003), we set 75.01 =κ and 5.02 =κ . Since our objective is to check the robustness of the relationship between poverty and household size, we choose a rather low value of 75.0=θ .30 We partially re-run our simulations to generate an adjusted disaggregation of the poverty headcount by household size in Table 15.31
The comparison of the results with the corresponding entries in Tables 8 to 11 shows that, as expected, the differences in the poverty headcount index across the household size categories decline substantially. However, the use of adult equivalent scales does not qualitatively change our findings. Large households are still more likely to be poor than small households, and the relationship between poverty and household size strengthened over time.
4.3 The Asset Index Approach The asset-index approach to construct national time series of basic poverty measures goes back to Filmer and Pritchett (2001) and Sahn and Stifel (2000, 2003). To proxy welfare in the absence of income or expenditure data, they assume that the asset ownership of households closely reflects their living standard. Using DHS data, they define a set of assets32 and construct a metric asset index
( ) ( ) ( )K
KjKn
k
kjkkjj
aasaasaasAI
σσσ−
++−
++−
= KK1
111 , (10)
where ks is the “scoring factor” or the weight of the asset k, jka takes the value of 1 if household j owns asset k and 0 otherwise, ka is the mean value of jka over all households, and kσ is its standard deviation.
Following Filmer and Pritchett (2001), we use the principal component analysis (rather than the closely related factor analysis as in Sahn and Stifel (2000, 2003)) to determine the asset weights ks . The underlying idea is to find a linear combination of the variables – the principal component or the asset index – which contains most of the common information of the variables and is interpreted as a background variable contained in all of them. Hence, the asset-index approach is valid if (and only if) welfare is indeed the main determinant of asset variability among households. We apply the asset-index approach to track the evolution of poverty between period t–1 and t. Since the mean 30 For comparison, Gasparini et al. (2003) set θ=0.9. 31 For the corresponding tables for the poverty gap and the squared poverty gap see Tables A15 and A16 in the
Appendix. 32 The asset definition is rather broad and includes not only real estate and financial assets, but also consumer durables
and the household’s endowment with human capital.
26
value of the asset index is zero by construction, we do not estimate equation (10) for each period separately but over a pooled sample of the periods t–1 and t.
Table 15 — Influence of Adult Equivalent Scales on the Poverty Headcount Disaggregated by Household Size
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Bolivia Total 58.67 53.32 40.11 39.87 34.27 31.65 17.16 15.18 (0.61) (0.46) (0.59) (0.39) By Hh Size
(1.50) (1.12) (1.74) (1.27) Notes: Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural
areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
In contrast to our dynamic cross-survey microsimulation methodology, the creation of national poverty profiles on the basis of the asset index requires a common set of assets for all observation years. Unfortunately, there was a change in the DHS questionnaire design: the DHS 1994 and 1998 collected information on more and other assets than the DHS 1989.33 The set of common assets over all three Bolivian DHS rounds would have been very small so that we decided to restrict our empirical analysis to the years 1994 and 1998. The derivation of the asset index and the summary statistics of the assets included therein are shown in Table 16. We use 25 assets – 17 tangible assets 33 The DHS 1989 was conducted under the DHS1 questionnaire design, the DHS 1994 and 1998 under the DHS3
questionnaire design. The lack of consistency applies especially to consumer durables (Table A2 in the Appendix).
27
and 8 human capital variables – to capture the welfare of households.34 The eigenvalues of the principal component analysis suggest that the asset index is indeed an important determinant for the asset distribution among households. The first principal component explains 21.7 percent of total asset variability.
Since all tangible assets are dummy variables, their scoring factors have a simple interpretation. A move from “non-ownership” to “ownership” of the asset changes the asset index by kks σ/ . For example, having private telephone connection increases the asset index by 0.83 in 1994 and 0.59 in 1998.35 In the case of the human capital variables, kks σ/ gives the change in the asset index if the average education of adult household members switches from the reference state (“less than complete basic schooling or unknown”) to the respective schooling category.
Table 16 — The Derivation of the Asset Index, 1994 and 1998
Notes: a Gas, kerosene, and electricity. – b Husbands and partners of women aged between 15 and 49. – c Women
aged between 15 and 49.
Source: Own calculations.
34 To check the robustness of our empirical results, we also estimated the asset index without human capital variables.
The empirical results, which are available upon request, do not change qualitatively. 35 The reduction in the asset weight reflects the fact that private telephone connection has become more affordable and,
thus, more widespread in Bolivia (from 11 percent of all households in 1989 to 25 percent in 1998).
28
As expected, consumer durables, such as telephone, radio, television, and fridge, have high scoring factors suggesting that they are powerful welfare predictors. By contrast, the ownership of a house or of a plot of agricultural land indicates poverty, which can mainly be explained by the widespread subsistence agriculture in rural areas of Bolivia. The quality of the dwelling unit also provides information on the welfare of households. Access to public utilities, high quality cooking materials, high quality toilet facilities, high quality floor materials, and a large number of sleeping rooms all increase the asset index. The scoring factors of the human capital variables are more difficult to reconcile. We find negative returns to schooling up to lower secondary schooling (9 years of schooling), which we attribute to that (a) our reference state includes “unknown” and that (b) the returns to basic and secondary schooling are indeed very small in Bolivia.
The asset-index value of the individual household is obtained by multiplying the deviation of the household asset endowment from the mean asset endowment with the vector of normalized scoring factors according to equation (10). Aggregating the asset-index values over all households, we find an increase in the mean asset index from –0.37 in 1994 to 0.38 in 1998 suggesting a favorable development of the living standard in Bolivia, which is consistent with findings using the Unmet Needs indicator (see Table 1). Based on the estimates of the asset-index values at household level, we can carry out two consistency checks between our dynamic cross-survey microsimulation methodology and the asset-index approach of Filmer and Pritchett (2001), and Sahn and Stifel (2000, 2003). First, we rank the households according to (a) their simulated incomes and (b) their asset-index values, and calculate the Spearman rank correlation coefficient between the two welfare indicators. We find a close relationship between the simulated incomes and the asset-index values. The Spearman rank correlation coefficient is 0.834 in 1994, and 0.792 in 1998.
Second, we construct poverty profiles based on asset-index values and compare them to those in Section 3.2. To this end, we again rank the households according to their asset-index values and calibrate the thresholds (i.e., poverty lines) between extremely poor, moderately poor, and non-poor so as to ensure that the incidence of poverty at the aggregated national level (i.e., in the first line of the poverty profile) in 1994 coincides with the one of the dynamic cross-survey microsimulation methodology, which is shown in Table 7.36 We keep this threshold level of 1994 constant and apply it also to the 1998 data. The spatial poverty profile based on asset-index values is shown in Table 17.
Although the direction of change and determinants are qualitative similar to our findings using the microsimulation approach, there are some differences. The most striking difference between the asset index and the microsimulation methodology is that overall poverty reduction from 1994 to 1998 appears much stronger using the asset index. Keeping the threshold of 1994 constant yields a 5.1 percentage points higher poverty reduction using the moderate poverty line and 2.0 percentage points using the extreme poverty line compared to the results shown in Table 7. We suspect that this sharper reduction in poverty using the asset index is due to a combination of changes in preferences favoring some assets (e.g. telephones and televisions), relative price reductions of some assets (e.g. telephones), and public investments in education which have not (yet) translated into income gains. Thus the sharper poverty reduction using the asset index says more about developments in non-income dimensions of well-being than being the most reliable proxy for the income dimension.
Furthermore, taking the corresponding results of the dynamic cross-survey microsimulation methodology in Table 7 as reference point, we find that the asset-index approach strongly underpredicts poverty in departmental capitals and other urban areas, and strongly overpredicts poverty in rural areas as the asset endowments there are much lower. In doing so, the results of the asset-index approach are closer to those of the unsatisfied-basic-needs approach37 (see first entries
36 The distribution of the assets among extremely poor, moderately poor and non-poor are given in Table A17 in the
Appendix. 37 The unsatisfied-basic-needs approach is very similar to the asset-index approach. It generates a weighted average of
welfare indicators (e.g., educational attainment, housing quality, access to public utilities, and access to basic health services, in the case of Bolivia) and classifies households as poor if their weighted average indicator value is below
29
in Tables 1 and 2) than those of the dynamic cross-survey microsimulation methodology. Additionally, not only the level but also the change in the incidence of poverty is more unevenly distributed across the three regions. While, according to the dynamic cross-survey microsimulation methodology rural areas participated –albeit less than proportionately– in the overall poverty reduction, they experienced nearly no progress in reducing poverty according to the asset-index approach. These differences are partly due to that only the dynamic cross-survey microsimulation methodology accounts for differences in the local price levels (Table 4); they also show that progress in improving the asset base in rural areas have been much slower in the 1990s.
Table 17 — Spatial Disaggregation of the Poverty Headcount Based on Asset-Index Values in Bolivia, 1994 and 1998
Moderate Poverty Line Extreme Poverty Line 1994 1998 1994 1998
Total 72.37 60.13 50.44 36.40
By Type of Municipality Departmental Capitals 51.20 38.61 19.14 8.91 Other Urban Areas 71.06 57.87 36.21 22.74 Rural Areas 98.17 97.01 92.27 88.37
By contrast, Table 17 shows less variation in the incidence of poverty across departments. The 1994 moderate poverty headcount index ranged only from 66 percent in Santa Cruz and Tarija to 84 percent in Potosí. For comparison, the corresponding figures of the dynamic cross-survey microsimulation methodology were 58 percent and 88 percent, respectively. As concerns the departmental poverty ranking, we find greater consistency between the two approaches.38 Santa Cruz is the richest department and Potosí and Chuquisaca are the poorest departments. The notable exception is Oruro, which is relatively poor according to the dynamic cross-survey microsimulation methodology, and relatively rich according to the asset-index approach.
The disaggregation of the poverty headcount index based on asset-index values by household characteristics is shown in Tables 18 to 21. To facilitate the comparison with the corresponding results of the dynamic cross-survey microsimulation methodology in Tables 8 to 11, we calibrate the thresholds between extremely poor, moderately poor, and non-poor for each poverty profile anew. The poverty profile for total Bolivia is calibrated to match the poverty headcount index for total Bolivia in 1994 (keeping the threshold constant for 1998), while the poverty profiles for departmental capitals, other urban areas, and rural areas are calibrated to match the poverty
a certain threshold. In contrast to the asset-index approach, the indicator weights are set arbitrarily. For a more detailed description of the unsatisfied-basic-needs approach and its application to Bolivia, see Hernany (1999).
38 This result becomes even more obvious when we compare the departmental disaggregation of the poverty headcount by quintiles rather than only at the thresholds between extremely poor, moderately poor, and non-poor (results are not reported here, but are available upon request).
30
headcount indices in the respective region in 1994 (again keeping the threshold constant for 1998).39
Table 18 — Disaggregation of the Poverty Headcount Based on Asset-Index Values in Bolivia by Household Characteristics, 1994 and 1998
By % of Hh Members between 15 and 65 Years <= 0.5 79.39 71.45 58.06 47.08 > 0.5 63.46 47.72 40.75 24.69
By Age of Hh Head <=34 76.42 70.97 51.84 39.53 35-49 72.56 57.58 50.98 35.95 50-65 66.18 49.51 47.37 32.16 >=66 61.63 47.87 46.02 34.30
By Language of Hh Head Spanish 61.06 50.06 33.05 22.68 Indigenous 98.82 96.78 91.08 86.37
By Gender of Hh Head Male 72.96 61.15 51.48 37.62 Female 69.52 55.32 45.42 30.63
By Average Years of Schooling of Adultsa <=5 97.36 93.43 83.82 72.45 6-12 64.29 51.37 32.08 21.41 >=13 8.77 9.19 1.22 1.93
By Profession of Principal Wage Earnerb White-Collar Worker 28.06 18.38 10.61 6.83 Blue-Collar Worker 79.94 68.08 45.51 27.81 Agriculture 99.00 96.86 95.32 90.93 Sales & Services 64.54 49.38 29.67 15.84 Not Employed 52.39 44.43 26.99 19.55
By % of Adult Womenc out of Employment < 100 72.71 56.34 51.61 33.41 = 100 71.66 66.95 47.97 41.78
Notes: a Women aged between 15 and 49 and their husbands and partners. – b Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. – c Women aged between 15 and 49.
Source: Own calculations.
39 The poverty headcount indices for 1994 for the calibration exercise are taken from Table 7.
31
Table 19 — Disaggregation of the Poverty Headcount Based on Asset-Index Values in the Departmental Capitals of Bolivia by Household Characteristics, 1994 and 1998
By Age of Hh Head <=34 67.89 65.54 37.16 26.13 35-49 59.31 42.85 27.01 14.20 50-65 47.93 32.60 18.72 8.64 >=66 30.84 26.84 9.48 9.09
By % of Hh Members between 15 and 65 Years <= 0.5 67.70 58.84 36.82 25.34 > 0.5 51.62 39.66 21.07 10.78
By Language of Hh Head Spanish 56.69 46.31 25.43 15.74 Indigenous 95.58 85.29 71.96 47.60
By Gender of Hh Head Male 58.93 48.72 29.03 17.03 Female 61.99 43.68 27.72 16.49
By Average Years of Schooling of Adultsa <=5 95.03 88.62 68.52 51.23 6-12 62.49 49.78 24.29 13.03 >=13 11.38 10.44 0.42 0.53
By Profession of Principal Wage Earnerb White-Collar Worker 24.65 14.50 8.20 1.99 Blue-Collar Worker 80.18 70.81 45.66 29.98 Agriculture 86.69 62.99 70.33 35.95 Sales & Services 67.18 52.90 28.78 18.03 Not Employed 48.30 41.09 14.23 8.88
By % of Adult Womenc in Employment > 0 57.93 44.59 27.34 16.30 = 0 62.36 54.61 31.44 18.27
Notes: a Women aged between 15 and 49 and their husbands and partners. – b Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. – c Women aged between 15 and 49.
Source: Own calculations.
32
Table 20 — Disaggregation of the Poverty Headcount Based on Asset-Index Values in Other Urban Areas of Bolivia by Household Characteristics, 1994 and 1998
By Age of Hh Head <=34 77.75 73.15 53.83 45.83 35-49 75.30 58.74 56.42 34.07 50-65 72.45 55.50 48.71 35.06 >=66 64.76 45.35 42.16 25.17
By % of Hh Members between 15 and 65 Years <= 0.5 79.57 69.02 58.81 44.51 > 0.5 69.00 52.64 45.96 28.22
By Language of Hh Head Spanish 72.90 59.05 50.22 34.08 Indigenous 94.33 91.28 80.81 71.95
By Gender of Hh Head Male 75.49 61.70 53.14 36.91 Female 73.75 62.92 54.63 39.75
By Average Years of Schooling of Adultsa <=5 95.95 91.27 82.32 72.67 6-12 70.33 58.08 43.01 27.13 >=13 20.39 17.19 2.85 6.54
By Profession of Principal Wage Earnerb White-Collar Worker 48.85 26.29 26.76 10.33 Blue-Collar Worker 87.56 75.96 61.32 47.16 Agriculture 95.73 88.58 90.50 74.50 Sales & Services 67.41 61.99 42.17 31.46 Not Employed 65.76 48.80 39.75 24.63
By % of Adult Womenc in Employment > 0 73.01 58.95 50.72 34.46 = 0 78.94 66.66 58.25 42.22
Notes: a Women aged between 15 and 49 and their husbands and partners. – b Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. – c Women aged between 15 and 49.
Source: Own calculations.
33
Table 21 — Disaggregation of the Poverty Headcount Based on Asset-Index Values in Rural Areas of Bolivia by Household Characteristics, 1994 and 1998
By Age of Hh Head <=34 88.55 84.00 73.06 68.03 35-49 89.70 84.47 78.16 65.82 50-65 91.81 84.01 77.87 69.90 >=66 87.90 85.08 75.61 72.19
By % of Hh Members between 15 and 65 Years <= 0.5 90.70 86.12 78.07 70.27 > 0.5 87.54 80.51 72.68 62.17
By Language of Hh Head Spanish 78.11 75.13 59.52 57.21 Indigenous 96.23 91.19 85.81 75.47
By Gender of Hh Head Male 90.25 85.21 77.76 69.02 Female 85.58 78.32 66.69 58.71
By Average Years of Schooling of Adultsa <=5 95.95 92.21 85.89 79.51 6-12 77.83 69.23 56.91 45.40 >=13 18.00 30.54 3.50 1.57
By Profession of Principal Wage Earnerb White-Collar Worker 59.98 49.31 34.28 21.41 Blue-Collar Worker 80.57 67.29 58.56 41.57 Agriculture 96.36 93.89 88.27 80.90 Sales & Services 70.37 53.66 39.10 30.02 Not Employed 77.75 75.35 56.28 53.02
By % of Adult Womenc in Employment > 0 90.03 85.34 77.06 69.73 = 0 88.28 82.74 73.62 64.57
Notes: a Women aged between 15 and 49 and their husbands and partners. – b Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. – c Women aged between 15 and 49.
Source: Own calculations.
34
Education continues to be the most important determinant of poverty and its changes over time.40 The distribution of the headcount index across schooling groups is even more polarized according to the asset-index approach. By contrast, we find a strikingly different pattern for the changes in the distribution of the headcount indices across schooling groups between 1994 and 1998. For tertiary schooling, not only did the returns to schooling decrease over time, we also find that the incidence of poverty among household where the average education of adult members was 13 years of schooling or more rose in absolute (!) terms.41
The impact of household size on poverty is found to be smaller for asset-index values than for simulated per-capita incomes. The relationship between poverty and household size is U-shaped in departmental capitals and other urban areas. In rural areas, large households continue to be poorer than small households, but the relationship was relatively weak in 1994, and became even weaker (not stronger as in the dynamic cross-survey microsimulation methodology) in 1998. We attribute these inconsistencies, which cannot be reconciled by the use of realistically defined equivalent scales (Table 15), to that the strong reliance on tangible assets in the asset-index approach may overstate the economies of scale within the household. As concerns the age composition of the household, the asset-index approach corroborates the earlier findings. Households where the share of members in working age was below 50 percent were more likely to be poor – particularly so in departmental capitals and less so in rural areas. Additionally, we again find that this relationship strengthened over time.
With respect to the impact of employment on poverty, there is also much agreement between the dynamic cross-survey microsimulation methodology and the asset-index approach. First, the incidence of poverty and its change over time are more favorable for white-collar workers and workers in sales & services than for agricultural and blue-collar workers. Second, female labor market participation was a successful strategy to lift households out of poverty in departmental capitals and other urban areas, but not in rural areas. However, like in the case of education, we find that the distribution of the poverty headcount index across professions was more accentuated according to the asset-index approach.
The asset-index approach depicts the poverty incidence of households with old heads in a more favorable light than the dynamic cross-survey microsimulation methodology. From a static perspective, the gradient in the poverty incidence based on asset-index values of 1994 was steeper across age groups in departmental capitals and other urban areas, and the relationship between poverty and the age of the household was flat rather than increasing in rural areas. From a dynamic perspective, Tables 18 to 21 show that – depending on the poverty line – households with heads aged 34 or below only less than proportionately participated in the overall poverty reduction between 1994 and 1998. A plausible explanation for these differences between the dynamic cross-survey microsimulation methodology and the asset-index approach is that household heads may accumulate tangible assets over the life cycle so that once they are old they possess more but less valuable assets.42 With respect to the other characteristics of the household head, we find more similarities between the two approaches. Being Non-Spanish speaking substantially increases the likelihood of being poor. The gradient in the poverty incidence between Spanish and Non-Spanish speaking household heads was again even steeper according to the asset-index approach. The explanatory power of the gender of the household head continues to be negligible.
40 This finding continues to hold if we exclude the human capital variables from the estimation of the asset-index
values. 41 However, especially concerning tertiary education, the sample size in some cases becomes very small, so one should
not interpret the numbers carefully. 42 The DHS data do not contain information on the age of the tangible assets so that we cannot check the validity of
this hypothesis.
35
5 Discussion In the preceding sections, we developed a new methodology to create national poverty profiles and growth incidence curves with incomplete income or expenditure data, and applied it to the case of Bolivia between 1989 and 2002. The analysis revealed that there are four main determinants of poverty and its changes over time. First, there is evidence for a large urban-rural divide in Bolivia. Following the historical settlement patterns, Bolivia’s poor are still concentrated in the rural areas of the highlands (altiplano and valles), where they face difficult ecological and climatic conditions for agricultural production, and suffer from the proliferation of tiny plots. The migration of the poor to urban areas and the more dynamic lowlands (llanos) has been limited due to reasons of disease ecology as well as the lack of support networks.43 The large and persistent differences in education levels between urban and rural areas have perpetuated the urban-rural divide. In 1976, the average years of schooling of the adult rural population stood at only 1.8 years, compared to 6.1 years in urban areas. While recent investments in rural education have led to some improvements, the differences remain substantial. In 2001, average years of schooling were 4.2 in rural areas and 9.2 in urban areas (INE var. iss.). Additionally, there is very restricted credit access for informal enterprises, and particularly so in rural areas. Despite the fact that some of Bolivia’s microfinance institutions have been hailed as models to ensure sustainable credit access, they still operate in only 68 of Bolivia’s 329 municipalities and cover only about 10 percent of the population. In rural areas, credit is virtually unavailable for anyone except very large producers. These problems are exacerbated by little progress in restructuring Bolivia’s wholesale financial institutions, and by years of inconclusive debate about the possibilities for incorporating microfinance institutions into the national regulatory system.
However, what is more of concern here is that the urban-rural divide seems to have widened over the last 20 years. An obvious explanation for the inferior performance of rural areas in reducing the poverty incidence is that the agricultural sector suffered most from the adverse El Niño/La Niña weather phenomenon, which hit Bolivia twice during the 1990s, and from the continuing decline in world-market prices for agricultural products. Eastwood and Lipton (1999, 2004) offer three additional explanations: (a) structural reforms might have contributed to the widening of the urban-rural divide because well educated households, which are concentrated in urban areas, might be better at exploiting economic opportunities following the structural reform process, and urban economic activity, which had been most regulated before the structural reforms, had most to gain. (b) there might have been two pro-urban demographic trends: selective rural-urban migration might have left behind a core of old, poorly educated individuals in rural areas who are weak in reaping “trickle down effects” from economic growth, and faster fertility transition might have increased per-capita income of urban households. (c) there might have been a persistent urban bias in public spending – at least until the transfer of large parts of the tax revenues from the national to the municipal level which marked the start of the decentralization process in 1994.44 While these explanations played a role in widening the urban-rural divide,45 our results suggest that neither anti-rural economic growth, nor anti-poor economic growth within rural areas were the main reasons for the inferior performance of rural areas in reducing the poverty headcount index. It is rather because due to high income inequality in rural areas, only relatively few rural households were initially just
43 Only very recently have the migration patterns of Bolivia’s poor changed. In 1997, 46 percent of the rural migrants
went to other rural areas, presumably due to agricultural employment (especially coca production) and family reasons. By 2001, this share has dropped to 37 percent with departmental capitals taking in the larger share of migrants (Tannuri-Pianto et al. 2004).
44 Table 7 provides some supportive evidence for this hypothesis. Before the start of the decentralization process in 1994, there was no progress in reducing the incidence of rural poverty. Thereafter, it declined more or less in line with the incidence of urban poverty – at least in absolute terms.
45 See, for instance, Tannuri-Pianto el al. (2004) who provide evidence on that there has indeed been selective rural-urban migration in Bolivia. Young, well-educated people speaking Spanish as their mother tongue have a much higher likelihood to migrate, thus, contributing to the brain drain from rural areas.
36
below the poverty lines so that a given growth of GDP per capita between 1989 and 2002 lifted only relatively few rural households over the poverty lines.
Second, a large number of children has become an increasingly powerful poverty predictor in Bolivia as evidenced by that (a) the incidence of poverty was higher among large households and among households with few members in working age, and that (b) these relationships strengthened over time. These trends reflect the considerable fertility decline in Bolivia over the past 20 years, which is now clearly visible in the age structure of the population where the absolute number of 0-4 year olds has recently begun to decline. If the fertility decline continues, the country can expect two types of welfare improvements: (a) Bolivia is likely to enter the phase which has been referred to as “demographic gift” by Bloom and Williamson (1998), where the share of the working age population will be particularly large. Under these conditions, the country can save more, invest more in physical and human capital, and, if sufficient employment opportunities are available, spur growth of GDP per capita. (b) economic growth is likely to become more pro-poor as it is particularly the poor who are now in the process of further reducing their household size and of benefiting from lower dependency rates (Klasen 2004; Eastwood and Lipton 2000).46 Once the fertility decline has reached the poor in Bolivia, it can be a major driving force of poverty reduction, as it was elsewhere in recent years (e.g., in East Asia and Brazil).
Third, average education of adult household members strongly shaped the likelihood of being poor. There is a steep gradient in the poverty headcount index between basic, secondary, and tertiary schooling. However, the two approaches to estimate basic poverty measures with incomplete income or expenditure data applied in Sections 3 and 4 yield conflicting evidence on the changes in the returns to schooling (in terms of reducing the incidence of poverty relative to the next lower schooling category) between 1989 and 2001. According to our dynamic cross-survey microsimulation methodology the returns to secondary schooling declined somewhat while the returns to tertiary schooling increased substantially. Using the asset-index approach developed by Filmer and Pritchett (2001), and Sahn and Stifel (2000, 2003), we find that not only did the returns to tertiary schooling decrease over time, but the incidence of poverty among household where the average education of adult members was 13 years of schooling or more also rose in absolute (!) terms. These inconsistencies deserve more attention. Data constraints prevent us from exploring them in more depth and detail on a national scale. This is because the DHS data on education of male adults are deficient: (a) they provide the schooling category only for husbands and partners of women aged between 15 and 49, but not for single, divorced, or widowed men or men whose wife or partner is aged 50 or more, (b) the information on the education of male adults rely on the recall of their wives and partners, who serve as respondents of the DHS so that many men are classified to have “non or unknown schooling”, and (c) the schooling categories asked in the DHS are rather broad – it would at least be desirable to distinguish between educación intermedia (lower secondary schooling) and educación media (higher secondary schooling).
Fourth, we find that the explanatory power of the profession of the principal wage earner has considerably increased since 1989. The poverty headcount index of the relatively rich white-collar workers and workers in sales & services fell more than twice as much as the poverty headcount index of the relatively poor agricultural and blue-collar workers. This finding, however, can be criticized on several grounds. Like above, the DHS data of the mainly male principal wage earner rely on the recall of the female respondent. Additionally, the professional categories (a) represent very heterogeneous groups (e.g., “sales and services” can include anything from bankers to street vendors), and (b) are highly correlated to education level of the job holder since they reflect a mixture of academic credentials, job requirements, and industry characteristics.
46 The poor still have much larger families. Using the unsatisfied-basic-needs approach and applying it to the 2001
Census, extremely poor households in Bolivia had a total fertility rate of 6.9, compared to 2.1 for households with satisfied basic needs.
37
6 References – Annex 1
Antelo, Eduardo (2000). Políticas de Estabilización y de Reformas Estructurales en Bolivia a partir de 1985. In: Luis Carlos Jemio and Eduardo Antelo (eds.), Quince Años de Reformas Estructurales en Bolivia. CEPAL and Universidad Católica Boliviana. La Paz.
Atkinson, Anthony B. (1970). On the Measurement of Inequality. Journal of Economic Theory 2(3): 244–263.
Bloom, David E., and Jeffrey G. Williamson (1998). Demographic Transitions and Economic Miracles in Emerging Asia. The World Bank Economic Review 12(3): 419–455.
Burger, Kees, and Menno Pradhan (1998). Rural Poverty: The Survey of 1995. In: Pitou van Dijck (ed.), The Bolivian Experiment: Structural Adjustment and Poverty Alleviation. Amsterdam.
CEPAL (1995). La Medición de los Ingresos en la Perspectiva de los Estudios de Pobreza. Santiago de Chile.
CEPAL (2002). Social Panorama of Latin America 2000/2001. Santiago de Chile.
Cowell, Frank A. (1995). Measuring Inequality. London.
Cowell, Frank A. (2000). Measurement of Inequality. In: Anthony B. Atkinson and François Bourguignon (eds.), Handbook of Income Distribution. Volume 1. Amsterdam.
Datt, Gaurav, and Martin Ravallion (1992). Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s. Journal of Development Economics 38(2): 275-295.
Deaton, Angus, and Salman Zaidi (2002). Guidelines for Constructing Consumption Aggregates for Welfare Analysis. LSMS Working Paper No. 135. World Bank. Washington D.C.
Eastwood, Robert, and Michael Lipton (1999). The Impact of Changes in Human Fertility on Poverty. Journal of Development Studies 36(1): 1-30.
Eastwood, Robert, and Michael Lipton (2000). Pro-poor Growth and Pro-Growth Poverty Reduction: Meaning, Evidence, and Policy Implications. Asian Development Review 18(2): 1-37.
Eastwood, Robert, and Michael Lipton (2004). Rural and Urban Income Inequality and Poverty: Does Convergence between Sectors Offset Divergence within Them? In: Corina, Giovanni A. (ed.), Inequality, Growth, and Poverty in an Era of Liberalization and Globalization. Oxford.
Elbers, Chris, Jean O. Lanjouw, and Peter Lanjouw (2003). Micro-Level Estimation of Poverty and Inequality. Econometrica 71(1): 355–364.
Filmer, Deon, and Lant Prichett (2001). Estimating Wealth Effects without Expenditure Data – or Tears: An Application to Educational Enrollments in States of India. Demography 38(1): 115–132.
Foster, James E., Joel Greer, and Erik Thorbecke (1984). A Class of Decomposable Poverty Indices. Econometrica 52: 761–766.
38
Gasparini, Leonardo, Martín Cicowiez, Federico Gutiérrez, and Mariana Marchionni (2003). Simulating Income Distribution Changes in Bolivia: a Microeconometric Approach. World Bank Bolivia Poverty Assessment. Washington D.C.
Gray-Molina, George, Wilson Jimenez, Ernesto Pérez de Rada, and Ernesto Yañez (1999). Pobreza y Activos en Bolivia: Qué Papel Desempeña el Capital Social? El Trimestre Económico 66(3): 365–417.
Grimm, Michael, and Isabel Günther (2004). Operationalizing Pro-Poor Growth – Country Case Study Burkina Faso. Paper Presented at the Joint World Bank, KfW, and GTZ Workshop “Operationalizing Pro-Poor Growth” in Eschborn, 15-16 July 2004.
Hemmer, Hans-Rimbert and Rainer Wilhelm (2000). Fighting Poverty in Developing Countries: Principles for Economic Policy. Frankfurt am Main.
Hentschel, Jesko, Jean Olson Lanjouw, Peter Lanjouw, and Javier Poggi (2000). Combining Census and Survey Data to Trace the Spatial Dimensions of Poverty: A Case Study of Ecuador. World Bank Economic Review 14(1): 147–165.
Hernany, Werner (1999). La Pobreza en el Área Urbana de Bolivia: Período 1989–1997. Tesis para la Obtención de la Licenciatura en Economía. Universidad Católica Boliviana. La Paz.
Hernany, Werner, Wilson Jiménez, and Rodney Pereira (2001). Bolivia: Efectos de la Liberalización sobre el Crecimiento, Empleo, Distribución y Pobreza. In: Enrique Ganuza, Ricardo Paes de Barros, Lance Taylor, and Rob Vos (eds.). Liberalización, Desigualdad y Pobreza: América Latina y el Caribe en los 90. Buenos Aires.
INE (2001). Censo Nacional de Población y Vivienda. CD-ROM. La Paz.
INE (various issues). Información Estadística. Online Data Base. Instituto Nacional de Estadística. La Paz. http://www.ine.gov.bo.
INE-UDAPE (2000). Canasta Básica de Alimentos y Líneas de Pobreza para el Área Rural de Bolivia. Mimeo.Instituto Nacional de Estadística and Unidad de Análisis de Políticas Sociales y Económicas. La Paz.
INE-UDAPE (2002). Mapa de Pobreza 2001. Instituto Nacional de Estadística and Unidad de Análisis de Políticas Sociales y Económicas. La Paz. Mimeo.
Jenkins, Stephen P., and Peter J. Lambert (1997). Three I’s of Poverty Curves, with an Analysis of UK Poverty Trends. Oxford Economic Papers 49(3): 317–327.
Jenkins, Stephen P., and Peter J. Lambert (1998). Three I’s of Poverty Curves and Poverty Dominance: Tips for Poverty Analysis. In: Daniel J. Slottje (ed.) Research on Economic Inequality. Volume 8: 39–48. Stamford.
Jiménez, Wilson, and Fernando Landa (2004). ¿Bolivia Tuvo un Crecimiento “Pro-Pobre” entre los Años 1993 y 2002? Mimeo. Unidad de Análisis de Políticas Sociales y Económicas. La Paz.
Jiménez, Wilson, and Ernesto Yañez (1997). Pobreza en las Ciudades en Bolivia: Análisis de la Heterogenidad de la Pobreza 1990–1995. Working Paper 52/97. Unidad de Análisis de Políticas Sociales. La Paz.
Kasen, Stephan (2004). In Search of the Holy Grail. How to Achieve Pro-poor Growth. In: Bertil Tungodden and Nicolas Stern (eds.), Towards Pro-poor Policies. Proceedings from the ABCDE Europe Conference. Washington, D.C.
39
Marcoux, Alain (1998). The Feminization of Poverty: Claims, Facts, and Data Needs. Population and Development Review 24(1): 131–139.
National Research Council (1995). Measuring Poverty: a New Approach. Washington, D.C.
Pereira, Rodney, and Wilson Jimenez (1998). Políticas Macroeconómicas, Pobreza y Equidad en Bolivia. In: Enrique Ganuza, Lance Taylor, and Samuel Morley (eds.). Política Macroeconómica y Pobreza en América Latina y el Caribe. Madrid.
Psacharopoulos, George, Samuel Morley, Ariel Fiszbein, Haeduck Lee, and William C. Wood (1993). Poverty and Income Distribution in Latin America and the Caribbean: The Story of the 1980s. World Bank Technical Paper No. 351. World Bank. Washington D.C.
Ravallion, Martin (1998). Poverty Lines in Theory and Practice. Living Standards Measurement Study Working Paper No. 133, World Bank. Washington D.C.
Sahn, David, and David Stifel (2000). Poverty Comparison over Time and across Countries in Africa. World Development 28(12): 2123–2155.
Sahn, David, and David Stifel (2003). Exploring Alternative Measures of Welfare in the Absence of Expenditure Data. Review of Income and Wealth 49(4): 463–489.
Son, Hwa Hyun (2003). Approaches to Defining and Measuring Pro-Poor Growth. Mimeo. World Bank.
Székely, Miguel, Nora Lustig, Martin Cumpa, and José Antonio Mejía (2000). Do We Know How Much Poverty There Is? IDB Working Paper No. 437. Inter-American Development Bank. Washington, D.C.
Tannuri-Pianto, Maria, Donald M. Pianto, and Omar Arias (2004). Rural-Urban Migration and Human Capital in Bolivia. Mimeo. World Bank Bolivia Poverty Assessment. Washington D.C.
UDAPSO (1995). Bolivia: 1995 Poverty Profile. Mimeo. Unidad de Análisis de Políticas Sociales. La Paz.
Vos, Rob, Haeduck Lee, and José Antonio Mejía (1998). Structural Adjustment and Poverty. In: Pitou van Dijck (ed.), The Bolivian Experiment: Structural Adjustment and Poverty Alleviation. Amsterdam.
Watts, Harold W. (1968). An Economic Definition of Poverty. In: Daniel P. Moynihan (ed.). On Understanding Poverty. New York.
White, Halbert (1980). A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity. Econometrica 48(4): 817–838.
Wodon, Quentin (2000). Poverty and Policy in Latin America and the Caribbean. World Bank Technical Paper 467. World Bank. Washington, D.C.
Wolff, Edward Nathan (1997). Economics of Poverty, Inequality and Discrimination. Cincinnati, Ohio.
40
World Bank (1996). Bolivia: Poverty, Equity, and Income: Selected Policies for Expanding Earning Opportunities for the Poor. Volume 2: Background Papers. Washington, D.C.
World Bank (2000). Bolivia: Poverty Diagnostic 2000. Washington, D.C.
WBI (2003). Poverty Manual. World Bank Institute. Poverty Analysis Initiative. Washington D.C. http://www.worldbank.org/wbi/povertyanalysis/manual/
Zheng, Buhong (1993). An Axiomatic Characterization of the Watts Poverty Index. Economic Letters 42(1): 81–86.
Box A1 — A Brief Overview of Poverty and Inequality Measures
(a) Poverty Measures
There are three dimensions of poverty which have to be captured by the poverty measures (sometimes labeled the three I’s of poverty): (a) the incidence or extent of poverty (how many individuals are poor?), (b) the intensity or depth of poverty (how poor are the poor?), and (c) the inequality or severity of poverty (how is poverty distributed among the poor?) (Jenkins and Lambert 1997, 1998).
The most widely used approach to analyze the so-called three I’s of poverty is the family of poverty indices proposed by Foster et al. (1984). Let )( zyp i ≤ denote an indicator function that takes the value of 1 if income iy is less than the poverty line z, and 0 otherwise, and let
⎭⎬⎫
⎩⎨⎧ −
=≤⋅−
= 0,max)(z
yzzyp
zyz
g ii
iyi . (A1)
denote the relative poverty gap. The poverty index family of Foster et al. (1984) can then be written as
( )∑=
⋅=I
iyig
IP
1
1 λλ , (A2)
where λ measures the degree of poverty aversion. For 0=λ , equation (A2) takes into account only the incidence of poverty and simplifies to the poverty headcount index
∑=
≤=I
ii zyp
IP
1
0 )(1 (A2a)
which measures the proportion of the population which receives incomes below the poverty line. To provide information on the intensity of poverty, we set 1=λ and arrive at the poverty gap index
∑=
⋅=I
iyig
IP
1
1 1 . (A2b)
It simply represents the sum of all poverty gaps divided by the total population and measures how much would have to be transferred from each individual (including the poor) to the poor to bring their incomes up to the poverty line.47 In the calculation of 1P , each relative poverty gap (and, thus, each poor) is equally weighted. To additionally penalize the inequality of poverty requires the assignment of higher weights to greater poverty gaps (and, thus, to poorer poor). In other words, the degree of poverty aversion has to be set to 1>λ . The most widely-used approach is to take the
47 However, this figure represents only the minimum cost of eliminating poverty as the transfers would have to be
perfectly targeted and would have to be collected and distributed perfectly efficiently.
44
value of the poverty gap itself as the corresponding weight, which leads to the squared poverty gap index
( )∑=
⋅=I
iyig
IP
1
22 1 . (A2c)
(b) Inequality Measures
The most widely used inequality measure48 is the Gini coefficient
,)1(2
11 1∑∑= =
−⋅−⋅⋅⋅
=I
i
I
jji yy
IIyG (A3)
which is defined as the average difference between every pair of incomes divided by two times the mean income y and represents two times the area between the 45° line and the Lorenz curve.49 Due to its intuitive graphical interpretation, it has become by far the most commonly used inequality measure. However, the main disadvantage of Gini coefficient is that it places rather arbitrary weights to income transfers that occur in different parts of the income distribution. The distance between two income units depends only on their rank ordering but not on their income difference. As a result, assuming a bell-shaped income distribution, the Gini coefficient is relative insensitive to transfers within the group of low-ranking or within the group of high-ranking income recipients.
The Atkinson (1970) index measures the social welfare loss associated with a certain level of income inequality compared to a situation where the same total income is equally distributed. Assuming that the underlying social welfare function has a constant degree of relative inequality aversion ζ, it is given by
( )
⎪⎪⎪
⎩
⎪⎪⎪
⎨
⎧
=⎟⎟⎠
⎞⎜⎜⎝
⎛−
≠⎥⎥⎦
⎤
⎢⎢⎣
⎡⎟⎟⎠
⎞⎜⎜⎝
⎛⋅−
=
∏
∑
=
−
=
−
.11
111
1
1
11
1
1
ζ
ζζ
ζζ
ifyy
ifyy
IAII
i
i
I
i
i
(A4)
Its implicit weighting scheme to income transfers depends on the difference in the marginal social utility between the conceding and the receiving income unit, and can be controlled for by the choice of ζ. This parameter reflects the relative sensitivity to redistribution from the rich to the not-so-rich vis-à-vis to redistribution from the not-so-poor to the poor. The higher the value of ζ, the more sensitive is the Atkinson index to income transfers at the lower tail of the income distribution.
48 For a more detailed description of the theoretical foundations and the underlying axioms of inequality measures see
Cowell (1995, 2000), and Wolff (1997). 49 The Lorenz curve plots the cumulative share of the population (depicted on the x-axis) against the corresponding
cumulative share of total income (depicted on the y-axis), when the individual income units are ranked in ascending order of their incomes.
45
Box A2 — The Construction of the Bolivian Poverty Lines
According to Ravallion (1998), “a credible measure of poverty can be a powerful instrument for focusing the attention of policy makers on the living conditions of the poor.” A controversial issue in this respect is the construction of an “objective” poverty line.50 The most commonly accepted methodology, which was also adopted by Bolivia, is the cost-of-basic-needs approach:
(1) For well-defined geographical entities, time-invariant basic food baskets are defined which reflect (a) the average nutritional requirements of adults – e.g., 2135 kilocalories per day in the case of Bolivia – and (b) the local eating habits of a reference group – e.g., the middle quintile of the income distribution in the case of Bolivia. The extreme poverty lines (líneas de indigencia) are obtained by valuing each item of the basic food baskets with its average local price paid by the reference group, and they are updated using the information of disaggregated local Consumer Price Indices (CPI).
(2) To obtain moderate poverty lines (líneas de pobreza) which additionally include the cost of non-nutritional basic needs, local Engel coefficients (average expenditure shares devoted to food) of the reference group are estimated and their inverse is multiplied by the extreme poverty lines.
In Bolivia, the construction of poverty lines was compounded by data constraints. The national CPI is estimated on the basis of market prices of only four cities – La Paz, El Alto, Cochabamba, and Santa Cruz – while no price data are available for the rest of the country. As a result, the cost-of-basic-needs approach had to be modified. For the urban areas, a two-step procedure was applied. First, the poverty lines for La Paz, El Alto, Cochabamba, and Santa Cruz were constructed using local basic food baskets defined on the basis of the Encuesta de Presupuestos Familiares of 1990 and price data taken from the local CPI. Second, the poverty lines were extrapolated to the urban areas of the other departments using ad-hoc adjustment factors (Gray-Molina et al. 1999). For the rural areas, a single basic food basket was defined on the basis of the Encuesta de la Evaluación del Fondo de Inversión Social (EVI-FIS) of 1997. The absence of a rural price data required to derive a “CPI proxy” from the EVI-FIS and update it on the basis of the expenditure modules of the Bolivian LSMS (INE-UDAPE 2000).
50 See WBI (2003) for a detailed description and evaluation of different methodologies to construct poverty lines.
46
Box A3 — The Watts Index and Its Axiomatic Properties
The Watts (1968) index is defined as
∫ ⋅−=0
0
)(logP
dz
yW ξξ , (A5)
where ξ is the cumulative population share if the analysis units are ranked in ascending order of their incomes. The Watts index is the only poverty measure that gives the absolute amount of social welfare loss due to poverty, is additive decomposable,51 and satisfies all standard axioms of poverty measurement (Zheng 1993, 1997):
1) Focus Axiom: The poverty measure is only a function of the poor and, thus, invariant to income changes of the non-poor.
2) Symmetry Axiom (= Anonymity Axiom): Re-labeling the income recipients should not affect the poverty measure.
3) Population Principle: The poverty measure is invariant if two or more identical populations are pooled.
4) Monotonicity Axiom: The poverty measure increases when a poor person gets poorer.
5) Transfer Axiom: A transfer from poorer poor to richer poor increases the poverty measure.
6) Monotonicity Sensitivity Axiom: The poorer an individual is, the larger is the increase in the poverty measure due to this income transfer.
7) Transfer Sensitivity Axiom: The increase in the poverty measure is higher if a poor person with a larger distance to the poverty line makes this transfer.
51 That is, the poverty measure for a population can be written as a weighted average of the poverty measures of a set
of mutually exclusive and collectively exhaustive subpopulations.
47
– Tables –
Table A1 — Sample Means of the Variables Taken from the Living Standard Measurement Surveys Total Bolivia Deparatmental Capitals Other Urban Areas Rural Areas
Notes: a Ratio of hh members aged between 15 and 65 to all hh members. – b Gas, kerosene or electricity.
Source: Own calculations.
51
Table A2 — Sample Means of the Variables Taken from the Demographic and Health Surveys Total Bolivia Deparatmental Capitals Other Urban Areas Rural Areas
Notes: Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
55
Table A4 — Spatial Disaggregation of the Squared Poverty Gap in Bolivia, 1989 to 2002
Notes: Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
56
Table A5 — Disaggregation of the Poverty Gap in Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line 1989 1994 1999 2002 1989 1994 1999 2002
Total 45.45 41.89 32.53 32.94 27.53 25.21 15.73 15.32 (0.35) (0.25) (0.34) (0.22) By Hh Size
(0.43) (0.53) (0.43) (0.44) Notes: Poverty indices are calculated using mixed income-expenditure data. Standard errors of the poverty indices in brackets
(only applicable to those based on simulated data). a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations
57
Table A6 — Disaggregation of the Poverty Gap in the Departmental Capitals of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line 1989 1994 1999 2002 1989 1994 1999 2002
Total 32.92 25.74 21.02 24.37 15.29 9.58 8.00 9.79 By Hh Size
Notes: Poverty indices are calculated using income data. a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations.
58
Table A7 — Disaggregation of the Poverty Gap in Other Urban Areas of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line 1989 1994 1999 2002 1989 1994 1999 2002 Total 51.31 44.68 34.70 32.88 34.10 27.02 13.97 13.10 (0.92) (0.69) (0.90) (0.63) By Hh Size
(1.12) (1.12) (1.19) (1.16) Notes: Poverty indices are calculated using income data. Standard errors of the poverty indices in brackets (only applicable to those
based on simulated data). a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations.
59
Table A8 — Disaggregation of the Poverty Gap in Rural Areas of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line 1989 1994 1999 2002 1989 1994 1999 2002 Total 58.30 60.90 47.71 44.86 39.13 43.33 27.37 23.88 (0.50) (0.34) (0.57) (0.38) By Hh Size
(0.55) (0.65) (0.62) (0.67) Notes: Poverty indices are calculated expenditures data. Standard errors of the poverty indices in brackets (only applicable to those
based on simulated data). a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations.
60
Table A9 — Disaggregation of the Squared Poverty Gap in Bolivia by Household Characteristics, 1989 to 2002
(0.39) (0.42) (0.33) (0.30) Notes: Poverty indices are calculated using mixed income-expenditure data. Standard errors of the poverty indices in brackets
(only applicable to those based on simulated data). a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations.
61
Table A10 — Disaggregation of the Squared Poverty Gap in the Departmental Capitals of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 19.96 14.16 11.60 14.00 8.05 4.51 3.94 5.13 By Hh Size
Notes: Poverty indices are calculated using income data. a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations.
62
Table A11 — Disaggregation of the Squared Poverty Gap in Other Urban Areas of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 37.28 31.38 21.12 19.82 22.52 17.17 8.01 6.95 (0.83) (0.58) (0.71) (0.49) By Hh Size
(1.08) (1.02) (1.00) (0.96) Notes: Poverty indices are calculated using income data. Standard errors of the poverty indices in brackets (only applicable to those
based on simulated data). a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations
63
Table A12 — Disaggregation of the Squared Poverty Gap in Rural Areas of Bolivia by Household Characteristics, 1989 to 2002
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Total 42.21 45.83 31.85 28.53 24.59 28.84 15.65 12.94 (0.49) (0.33) (0.47) (0.34) By Hh Size
(0.54) (0.58) (0.52) (0.54) Notes: Poverty indices are calculated using expenditure data. Standard errors of the poverty indices in brackets (only applicable to
those based on simulated data). a Women aged between 15 and 49 and their husbands and partners. b In the case of DHS: Husband or partner of the oldest woman aged between 15 and 49. If she is single, this woman herself. In the case of LSMS: Household head. c Women aged between 15 and 49.
Source: Own calculations
64
Table A13 — Adjusted Spatial Disaggregation of the Poverty Gap in Bolivia, 1989 to 2002
Notes: Only poverty indices based on simulated data changed relative to Table A2.3. Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
Table A14 — Adjusted Spatial Disaggregation of the Squared Poverty Gap in Bolivia, 1989 to 2002
Notes: Only poverty indices based on simulated data changed relative to Table A2.4. Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
65
Table A15 — Influence of Adult Equivalent Scales on the Poverty Gap Disaggregated by Household Size
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Bolivia Total 28.29 26.42 15.92 15.25 13.27 13.25 5.92 5.12 (0.34) (0.24) (0.27) (0.19) By Hh Size
(0.95) (0.70) (0.88) (0.73) Notes: Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural
areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
66
Table A16 — Influence of Adult Equivalent Scales on the Squared Poverty Gap Disaggregated by Household Size
Moderate Poverty Line Extreme Poverty Line
1989 1994 1999 2002 1989 1994 1999 2002
Bolivia Total 17.00 16.42 8.68 7.99 6.88 7.23 2.95 2.59 (0.25) (0.18) (0.18) (0.13) By Hh Size
(0.79) (0.63) (0.61) (0.57) Notes: Poverty indices are calculated using income data for departmental capitals and other urban areas, expenditure data for rural
areas, and mixed income-expenditure data for total Bolivia. Standard errors of the poverty indices in brackets (only applicable to those based on simulated data).
Source: Own calculations.
67
Table A17 — Asset Endowment Among Extremely Poor, Moderately Poor and Non-poor (in Percent), 1994 and 1998
1994 1998 Extremely
Poor Moderately
Poor Non-poor Extremely
Poor Moderately
Poor Non-poor
Tangible Assets Telephone 0.02 0.29 37.58 0.40 2.13 67.81 Radio 72.93 79.67 99.59 73.19 82.63 98.31 Television 21.55 42.39 99.57 21.85 51.90 99.31 Fridge 4.57 11.58 77.13 4.46 12.82 84.28 House 79.91 72.17 53.70 77.45 66.27 62.57 Plot of Agricultural Land 54.95 39.07 0.67 53.36 32.34 0.50 In-house Access to Electricity 36.51 55.28 99.93 37.04 62.82 99.95 In-house Access to Public Water 22.72 40.27 97.49 31.25 54.53 98.30 Use of Other (Non-open) Water Source
Notes: a Gas, kerosene, and electricity. – b Husbands and partners of women aged between 15 and 49. – c Women aged between 15 and 49.
Source: Own calculations.
68
– Figures –
69
Figure A1 — Growth Incidence Curve for Bolivia, 1989 to 1999
-2
0
2
4
6
8
10
12
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
Figure A2 — Growth Incidence Curve for the Departmental Capitals of Bolivia,
1989 to 1999
-2
0
2
4
6
8
10
12
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
70
Figure A3 — Growth Incidence Curve for Other Urban Areas of Bolivia, 1989 to 1999
-2
0
2
4
6
8
10
12
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest %Growth Rates in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
Figure A4 — Growth Incidence Curve for Rural Areas of Bolivia, 1989 to 1999
-2
0
2
4
6
8
10
12
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
71
Figure A5 — Growth Incidence Curve for Bolivia, 1999 to 2002
-12
-8
-4
0
4
8
12
16
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
P0modP0
ex
Figure A6 — Growth Incidence Curve for the Departmental Capitals of Bolivia,
1999 to 2002
-12
-8
-4
0
4
8
12
16
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
P0modP0
ex
72
Figure A7 — Growth Incidence Curve for Other Urban Areas of Bolivia, 1999 to 2002
-12
-8
-4
0
4
8
12
16
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
P0modP0
ex
Figure A8 — Growth Incidence Curve for Rural Areas of Bolivia, 1999 to 2002
-12
-8
-4
0
4
8
12
16
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
P0modP0
ex
73
Figure A9 — Growth Incidence Curve for Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
Figure A10 — Growth Incidence Curve for the Departmental Capitals of Bolivia, 1989 to
2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
74
Figure A11 — Growth Incidence Curve for Other Urban Areas of Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest %Growth Rates in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
Figure A12 — Growth Incidence Curve for Rural Areas of Bolivia, 1989 to 2002
-2
0
2
4
6
8
0 10 20 30 40 50 60 70 80 90 100
Growth Incidence Curve Mean of Growth Rates for Poorest % Growth Rate in Mean
Annual Growth Rate %
Percentiles
–2
P0modP0
ex
75
Annex 2 – The CGE Model
The model employed in the policy analysis is a dynamic real-financial CGE model which combines neoclassical and structuralist characteristics, but does not account for Keynesian multiplier effects. The production structure and product market conditions, for example, correspond with standard neoclassical theory. As an important structuralist element, the segmentation of labor markets observable in Bolivia is taken into account. In addition, the savings and investment behavior of different economic agents is modeled explicitly via the specification of a financial market, which allows for the existence of credit rationing. In the following, the major components of the modeling framework will be described in a non-technical manner. A full mathematical documentation can be found in Wiebelt (2004).
Production and Trade
The model distinguishes 12 sectors (see Table A1) which produce a characteristic but not necessarily homogenous good. Rather, it is assumed for exporting sectors that, e.g. due to quality differences, domestically sold and exported goods are not identical. This is modeled by means of a Constant Elasticity of Transformation (CET) function. The exceptions are mining and oil&gas, where exports are assumed to be exogenously determined by world market conditions or by long-term contracts as in the case of gas exports to Brazil. Domestically produced and imported goods of the same category are also treated as different, which is modeled by means of a Constant Elasticity of Substitution (CES) function (Armington assumption). Finally, some sectors (utilities, construction, public services) produce pure non-tradables. This rather strong differentiation in production allows, for instance, to capture in a realistic way the impact external shocks may have on the earning opportunities of different households.
A distinctive feature of the model is the explicit treatment of traditional agriculture and (urban) informal services as informal production sectors, where most of Bolivia’s poor earn their living. Workers in these sectors are considered self-employed; they for the most part rely on their own labor inputs and use only small amounts of capital. This implies that, over one year, supply is almost constant for a given number of workers and given factor productivities; and if demand slackens, adjustment will mainly run through a fall in prices and incomes of those employed in these sectors. By contrast, formal sectors tend to produce with modern, more capital-intensive techniques and, like the government, hire skilled and unskilled workers, which provides them with greater adjustment flexibility on the supply side. Throughout the formal economy, primary factors are combined via CES production functions, while the production technology of the two informal sectors is represented by a Cobb-Douglas function to account for the fact that labor can fairly easily substitute for the very basic capital goods used in these sectors. Both formal and informal sectors use intermediate inputs in fixed proportions to production.
76
Table A1 — Classification of the CGE Model
tActivities/Goods and Services Production Factors Economic Agents
Traditional agriculture Modern agriculture Oil& gas Mining Consumer goods Intermediate goods Capital goods Utilities Construction Informal services Formal services Public services
labor Smallholder labor Urban informal labor Corporate (formal) capital Employers’ capital Urban informals’ capital Smallholders’ capital Public (infrastructure) capital
To capture the reality of Bolivian employment and to keep track in a detailed manner of the poor’s main income flows, the model assumes a high degree of labor market segmentation (see Table A1). Beside the self-employed labor of smallholders and urban informals, two types of unskilled labor (agricultural and non-agricultural) as well as skilled labor are distinguished. Labor markets are linked via rural-rural and rural-urban migration. While the former involves smallholders becoming hired workers in modern agriculture, the latter involves the absorption of smallholders by the urban informal sector. Along the lines of the Harris-Todaro model, the decision to migrate depends on wage differentials. In the urban labor market, the limited possibilities of informal workers to enter the formal workforce are taken into account by assuming that despite an existing wage differential migration is constrained. The informal sector then absorbs all those who fail to obtain formal employment at the prevailing wage. The model does allow for underemployment in the sense that people are stuck in low-paid informal sector jobs, but not for open unemployment of unskilled labor, which appears to be an accurate characterization of the Bolivian labor market except for recession years where rates of open unemployment tend to rise to non-negligible levels. Wage adjustments also ensure that all other labor markets clear.
The model also assumes segmented capital markets, with a distinction made between unincorporated and corporate capital. Three household groups (smallholders, urban informals, and employers) own unincorporated capital. While smallholders and urban informals invest almost exclusively in traditional agriculture and informal services, respectively, employers receive capital income from all formal sectors with the exception of utilities. Corporate capital, by contrast, is owned by private and public enterprises, which invest in all formal sectors and retain the respective factor income. Finally, the model separates public infrastructure capital, which is assumed to affect the level of sectoral production. This is specified by means of a CES function where public capital and aggregate private value added
77
enter as arguments. Thus, by determining its investment focus, the government can influence the income generation possibilities in different sectors and regions.
Income and Expenditures
The model identifies six representative private households groups, which are basically characterized by their distinct factor endowments (Table A1). This is justified because factor income is the single-most important income source in Bolivia given the low degree of redistribution. In addition, workers and the self-employed are disaggregated regionally as their earning possibilities and consumption patterns tend to vary between regions. Four of the six household groups (smallholders, urban informals, and agricultural and non-agricultural workers) can be considered as poor. Depending on factor endowments, households receive labor or capital income as well as (net) interest payments on financial assets. Moreover, they receive transfer income from the state and from relatives living abroad. They use their gross income to pay for taxes and consumption as well as for savings. The allocation of private consumption expenditures on different goods is modeled employing a Linear Expenditure System (LES), where poorer households devote a larger budget share to price-independent subsistence consumption than do richer households.
The government finances its current and capital expenditures out of direct and indirect tax revenues, operating surpluses of public enterprises, and capital inflows from abroad. Private and public enterprises receive capital income, subsidies and net interest payments on financial assets; they use this income to pay corporate taxes and to save in the form of retained earnings. Since financial institutions are assumed to act as mere intermediaries, their current transactions (interest payments) are also allocated to the two kinds of enterprises. Finally, the rest of the world imports and exports goods from and to Bolivia, undertakes direct and portfolio investments in the country, and provides development aid.
Financial markets
The specification of the model’s financial sector is based on Tobin’s portfolio-theoretic framework, where the interaction of stocks and flows plays a decisive role. Starting from the beginning-of-period stocks of assets and liabilities, financial markets match the savings and investment decisions of all economic agents over the period, comprising the accumulation of both physical and financial assets and liabilities. The financial markets handle simultaneously the flows arising from savings and financial accumulation, and those arising from the reshuffling of existing portfolios due to changes in asset returns. For the latter, it is assumed that individual agents have only limited possibilities to substitute among different assets, which is captured by CES functions. A further characteristic of the financial sector is that specific economic agents, e.g. smallholders, may be constrained in their access to credit, which is clearly the case for most of Bolivia’s informal producers. This is modeled by determining bank credit to the respective agent residually after all other agents’ credit demand is satisfied.
The identification of stocks in the model makes it possible to account for the revaluation of assets and liabilities, which is of great importance in the highly dollarized Bolivian economy where the value of most domestic assets is at least partially indexed to movements in the exchange rate. Together with the accumulation occurring over the period, these revaluations determine the end-of-period stocks of assets, liabilities and net wealth for each economic agent.
78
Dynamics
An important feature of the model is its recursive-dynamic nature, which means that the model is solved for a sequence of static equilibria connected through capital accumulation and labor growth. The dynamics of the model are based on assumptions concerning exogenous growth rates for different variables such as labor supply and government expenditures, as well as the endogenous savings and investment behavior of economic agents. A general advantage of the dynamic specification is the possibility to generate a medium to long run growth path. Moreover, structural change over time can be analyzed. Finally, dynamic effects running through the financial sector can be captured. For example, it can be investigated how the debt relief granted under the HIPC initiative reduces Bolivia’s debt service.
Implementation of the model
In using the model for policy simulations, 1997 was chosen as the base year, for two different reasons. First, crucial data, in particular an Input-Output Table , are available for that year. Second, 1997 appears to be a fairly “normal” year for the Bolivian economy in the sense that no major shocks occurred, rendering it an appropriate benchmark against which to evaluate counterfactual simulations.
In specifying the model numerically, the first step was to compile the (real and financial) transactions between the sectors, production factors and economic agents identified in Table A1 in a Social Accounting Matrix (SAM) for 1997 (see Thiele and Piazolo 2003). The SAM provides the statistical backbone for the calibration of the model. From the information given in the SAM, many parameters, such as tax and subsidy rates, can readily be calculated. Other parameters, such as trade elasticities and income elasticities of private demand, have to be taken from external sources. Here, the choice of parameters is based on the stylized facts known from the existing empirical literature and on what is known about Bolivia’s economic structure, not on specific estimations performed for Bolivia. Armington elasticities, for instance, are assumed to be considerably higher for agriculture than for intermediate and capital goods, where import substitution is only possible to a limited extent because Bolivia’s own production of these goods is very small and of low sophistication compared to the relevant import substitutes.
In a final step, the calibrated model was updated so as to generate a fairly smooth growth path over ten years. To achieve this, the first two years have to be left out of the simulation analysis because the dynamics of the model only stabilize after three years.
Link between the model and household data
The CGE model is linked to household survey data in order to obtain detailed results on the poverty and distributional impact of the simulated policies. The starting point for linking survey data and the CGE model is household income, split up into (1) individual factor incomes, (2) household net interest income and transfers from abroad, and (3) household public transfers including pensions. These components of household income can be identified in the CGE model (see above) as well as in the household survey.52
Households receive factor income from different sources, i.e. the individuals of a household may earn different factor incomes. The household head may be self-employed (e.g. urban informal in the CGE model) and his/her spouse may be employed as a worker (e.g. unskilled worker in the CGE model). The link between CGE and the survey is simply sequential: each
52 The household survey used is the 1999 MECOVI. This survey is to be preferred over the 1997 employment
survey as it contains more detailed and more reliable information on household incomes.
79
individual factor income in the household survey is scaled up or down according to the CGE results for the eight production factors owned by households (Table A1). This is how changes in real factor prices in the CGE model affect the distribution of income.
The remaining two components of household income and the changes therein are given by household group in the CGE model. These changes from the CGE model are applied to the survey information at the household level. The household types in the survey are classified according to the occupation of the household head, in line with the classification used in the SAM.
Annex 3 – Description of Policy Simulations Decisions by the government, commercial banks and the Central Bank provide the policy framework for domestic activities. The main domestic policy instruments included in the model are listed in Table A2. The effectiveness of domestic policies depends on external events, such as changing world market prices for exports and/or imports and changing international interest rates, as well as the rest of the world’s decisions about private and public capital flows to Bolivia. Moreover, various parameters, such as sectoral factor productivities and factor substitution elasticities, and institutional characteristics, such as the process by which wages are determined, affect the behavior of economic agents and thus their response to policy reforms.
Table A2 — Domestic Policy Variables and External Parameters
Government Banking System Rest of the World
Income/corporate taxes
Export subsidies
Import tariffs
Excise taxes
Production subsidies
Value added taxes
Transfers to households and enterprises
Real government consumption
Real government investment
(e.g. in infrastructure)
Central Bank
Minimum reserves (in relation to imports)
Central Bank interest rate
Nominal exchange rate
Commercial banks
Access to credit
Flexibility in credit allocation
Development aid (including grant element of concessional lending)
Foreign portfolio investment
Foreign direct investment
Net credit to government
Debt relief (HIPC)
Foreign interest rate
Factor income from abroad
Remittances from abroad
World prices for exports
World prices for imports
Various of the policy variables and parameters identified in the model have been used for the analysis of shocks and policies presented in Chapter 3. The remainder of this annex will provide a short description of how the simulation experiments underlying this analysis were performed.
External Shocks
80
In order to gauge the impact of the El Niño phenomenon, a simulation is undertaken which assumes that total factor productivity in the two agricultural sectors decreases every three years so as to produce a reduction of agricultural output in the order of 3 percent, which roughly represents the impact of an El Niño of medium severity. A similar scenario is also considered in the Bolivian PRSP.
A negative Terms-of-trade shock is modeled as a 10-percent decrease in the world market prices for agricultural and mining exports (except for oil&gas where prices are assumed to be fixed by contracts).
To capture the impact of the Brazilian crisis and the completion of the capitalization process, both portfolio and foreign direct investment flows are reduced in line with the fall that actually happened.
The simulation of the combined effect of external shocks simply involves the simultaneous change of parameters.
Macro Policies
To investigate whether exchange rate policy might contribute to higher competitiveness of the Bolivian economy, a simulation with a higher yearly depreciation of the Boliviano in the crawling peg regime is compared to the base run.
The real depreciation required in case of a negative shock is simulated as an increase in the minimum reserves the Central Bank holds to cover imports. This results in an endogenous reduction of Central Bank credit to the private banking system, which in turn means lower credit supply for non-financial institutions.
Structural Reforms
A less severe segmentation of urban labor markets is modeled by assuming that urban informals are allowed to migrate into the formal unskilled labor market. The extent of the migration is calibrated so as to reduce the wage differential between the two labor markets by roughly 50 percent.
The two alternative tax reforms are both simulated by increasing income tax rates for all household groups except smallholders and urban informals. In one option, the tax increase is financed by a reduction of current and/or capital expenditures, in the other option indirect (value added and excise) tax rates are reduced so as to arrive at a revenue-neutral tax reform.
Natural Resource Policies
The gas contracts imply in the model that export volumes of the oil&gas sector are raised exogenously. One simulation assumes that as a response to higher tax revenues from gas government consumption expenditures are also raised exogenously so as to keep public savings roughly constant. In an alternative scenario, government consumption expenditures are kept constant, and public savings are allowed to adjust.
Targeted Interventions in Favor of the Poor
Improved access of smallholders to credit is modeled via two mechanisms. First, the credit constraint for smallholders is relaxed by assuming that their credit is no longer determined residually by the banking system after all other agents’ demand has been satisfied, but rather according to rentability criteria. Second, substitution elasticities of portfolio selection are increased for banks, which implies that their credit allocation becomes more sensitive to differences in sectoral rentabilities. To assess how this might impact on smallholders’ ability to invest it is additionally assumed that a positive temporary terms-of-trade shock in the form of higher export prices raises traditional agriculture’s rentability.
81
Public investment tailored to the needs of smallholders is modeled by increasing the substitution elasticity in the CES function that combines public capital and aggregate private value added.
A pro-poor industrial policy is alternatively simulated for modern agriculture and for the consumer goods sector. In both cases, this involves the introduction of export subsidies as a means to increase competitiveness.
A transfer program is simply modeled as an increase in direct government payments to poor household groups. The programs are either financed by a decrease in government consumption or by a decrease in public investment.
82
Annex 4 – Simulation Results
Table A3 — Baseline Scenario
Period Indicator
0 1 2 3 4 5 6 7 8 9 10
Real GDP Growth 4.7 4.7 4.7 4.7 4.7 4.7 4.7 4.7 4.7 4.7
