WPS6078 Policy Research Working Paper 6078 Impacts of Large-Scale Expansion of Biofuels on Global Poverty and Income Distribution Caesar B. Cororaton Govinda R. Timilsina The World Bank Development Research Group Environment and Energy Team June 2012 Policy Research Working Paper 6078 Abstract This paper analyzes the impact of large-scale expansion of commodities. The increased prices would cause higher biofuels on the global income distribution and poverty. food prices, especially in developing countries. Moreover, A global computable general equilibrium model is used wages of unskilled rural labor would also increase, to simulate the effects of the expansion of biofuels on which slows down the rural to urban migration in many resource allocation, commodity prices, factor prices and developing countries. The study also shows that the household income. A second model based on world- effects on poverty vary across regions; it increases in wide household surveys uses these results to calculate the South Asia and Sub-Saharan Africa, whereas it decreases impacts on poverty and global income inequality. The in Latin America. At the global level, the expansion of study finds that the large-scale expansion of biofuels leads biofuels increases poverty slightly. to an increase in production and prices of agricultural This paper is a product of the Environment and Energy Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at gtimilsina@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Impacts of Large-Scale Expansion of Biofuels on Global Poverty and Income Distribution Caesar B. Cororaton, Virginia Polytechnic University Govinda R. Timilsina, Development Research Group, World Bank Key Words: Biofuels, income inequality, poverty, general equilibrium modeling JEL Classification: C68, D31, Q16, Q43 Sector: Energy Impacts of Large-Scale Expansion of Biofuels on Global Poverty and Income Distribution1 Caesar B. Cororaton, Virginia Polytechnic University Govinda R. Timilsina, Development Research Group, World Bank Introduction Increased oil price volatility and concerns about climate change and long-term energy supply have led several countries to set targets for promoting biofuels (Boeters et al, 2008). This resulted in rapid growth of biofuel production over the last several years. However, the growth of biofuels had some unintended consequences. Food prices went up to a record high level and culminated in the global food crises in early 2008. Although there are several factors driving food prices up, the growth of biofuels has played a role (IFPRI, 2008; FAO 2008; Ivanic and Martin, 2008; and Keyzer et al, 2008; Hochman et al. 2011). Increased agricultural prices are favorable to rural households whose economic base is farming; but they hurt the urban poor because higher agricultural prices lead to higher food prices. At the aggregate level higher agricultural prices benefit countries that are net exporters of agricultural commodities, while the reverse effects are expected for countries which are net food importers. While there are a number of studies that analyze the aggregate economic effects of biofuels (IFPRI, 2008; FAO, 2008; Ivanic and Martin, 2008; and Mitchell, 2008; Al-Riffai et al. 2010; Timilsina et al. 2012) there are few studies that look into the income distribution and poverty effects of biofuels at the global level. De Hoyos and Medvedev (2009), examine the poverty effects of increased biofuel production using global CGE model. However, their model neither has a land-use module nor an explicit representation of biofuels sectors. Runge and Senaur (2007) also examine impacts of biofuels promoting policies on food prices and poverty and find that policies to promote ethanol have adverse impact on food prices and therefore on poverty especially in developing countries. On the other hand, some existing studies report the 1 Corresponding author: Govinda R. Timilsina, Senior Research Economist, Development Research Group, The World Bank. 1818 H Street, NW. Washington, DC 20433, Tel: 1 202 473 2767. Fax: 1 202 522 1151. E-mail: gtimilsina@worldbank.org. The views and findings presented here should not be attributed to the World Bank. We thank Denis Medvedev and Donald Larson for their valuable comments and Simon Mevel for research assistance. We acknowledge the Knowledge for Change Program (KCP) Trust Fund for the financial support. 2 opposite results. For example, using a CGE for Mozambique, Arndt et al. (2008) find that biofuels would have favorable effects on growth and income distribution. They show that the welfare and distributional effects of higher production of sugar cane to feed into ethanol industries are positive; the benefits would be larger if the expansion sugarcane is materialized through smallholders than through large plantations as the former employs more unskilled labor, provides higher rents to their land and also provide higher wage. Similarly, applying a CGE model, Arndt et al. (2011) show that cultivation of jatropha feedstock reduces poverty and increases GDP in Mozambique. A similar finding is reported by Hertel (2009) where developing countries with significant agricultural self-employed poverty population benefit from higher factor returns following increased production of biofuels. However, the positive impacts of biofuels on poverty reduction depend on how biofuel policies are designed. For example, Habib- Mintz (2010) finds that jatropha-based biofuel development in Tanzania would not help reduce poverty and food insecurity unless strong regulatory frameworks for land, investment management, and rural development are in place. This paper aims to analyze the effects of large-scale expansion of biofuels on the global income distribution, and poverty. The model used in our study explicitly represents both biofuels and land-use changes, which are critical to fully assess overall impacts of biofuels. Our global computable general equilibrium (CGE) model includes four types of biofuels: sugar ethanol, corn ethanol, grains ethanol and biodiesel. The land use module we developed uses a multi-level constant elasticity of transformation (CET) nest of various uses of land: forestry, pasture, agricultural crops (disaggregated further to grains, oilseeds, sugar, rice, wheat and other crops). With these developments, we are able to capture the fuel vs. food demand for agricultural commodities, land reallocation, and land rent. To assess the distributional and poverty impacts, we feed the CGE results to a Global Income Distribution Dynamics (GIDD) model, a model that is based on detailed household surveys, regularly carried out by the World Bank. We find that large-scale expansion of biofuels would improve wages for unskilled labor and land rent. However, it would also increase food prices, especially in low-income countries where food share in overall household expenditure is higher. The effects on poverty are negative (i.e., increase in poverty) in some countries, while positive in others. Globally, these differences balanced out to a slight increase in poverty. 3 The paper is organized as follows. The second section presents the analytical framework followed by the key data used for the study. Section Four presents definitions of the scenarios examined in the paper followed by discussion of the simulation results in Section Five. Finally, we present key conclusions of the study. The Analytical Framework As mentioned earlier, the study first uses a multi-sector, multi-country, recursive dynamic computable general equilibrium model to measure the impacts of large-scale expansion of biofuels. The results of the CGE model are then fed into a global income distribution dynamics (GIDD) model to capture impacts on income distribution and poverty. 2.1 The Global CGE Model The global CGE model developed for this study is different from existing CGE models in two aspects. First, it explicitly represents biofuel sector (ethanol and biodiesel) and secondly it represents land-use sector bringing land as a factor of production in the agriculture sector and also represents farmer (household) behavior for land allocations for various products. We present a very brief description of the model here. Detailed description of the model is available in Timilsina et al. (2010). The model represents the behavior of the production sector using nested CES functions. At the top nest, gross output is a Leontief aggregate of the value added bundle and an aggregate intermediate goods bundle. The aggregate intermediate good bundle excludes energy which is combined with capital and included in the value added bundle. The value added bundle includes non-capital factors of production (skilled and unskilled labor, land, and sector-specific factor if it exists) and capital/energy composite factor of production. Labor is mobile across all sectors within a country/region. There are no movements of labor across countries/regions. Capital is segregated by vintages: old and new. New capital is capital equipment installed at the beginning of the period, while old capital is capital equipment of more than a year age since installation. It is assumed that old capital has lower substitution elasticities than new capital. Countries with higher savings rates have a higher share of new capital. Furthermore, new capital is perfectly 4 mobile across production sectors while old capital has low sectoral mobility and is released using an upward sloping supply curve. The energy bundle is decomposed into liquid fuel, coal, gas and electricity. Liquid fuel is further divided into gasoline-ethanol composite, diesel-biodiesel composite and other petroleum products. The model allows substitution between gasoline and ethanol and also between diesel and biodiesel. There is one representative household in each region/country. Incomes generated from production are assumed to accrue to this single household. Disposable income of households is allocated between expenditures on goods and services, and savings. Private expenditures of goods and services are defined over consumer goods and are derived as the first order conditions of utility maximization, where the utility function is specified as constant difference in elasticities (CDE). The allocation of national income is across the following major expenditure items: public expenditures, private expenditures, and investment expenditures. Of the three, the largest is private expenditures, which involve household expenditures. In the model land is specified as a multi-level CET nest of various uses of land. At the top level, land is supplied to forestry, pasture, and agriculture crops. Land used in agricultural crops is divided into various uses. At the second nest level, land used in agricultural crops is divided to grains and oilseeds, sugar, rice, and other crops. At the third level, land used in grains and oilseeds is further disaggregated to corn, wheat, other coarse grains, and oilseeds. Thus, there are nine separate demand functions for land, each with a corresponding land supply functions. Each pair of demand for and supply of land determines a land market that is cleared by a land price. This disaggregation is critical in the biofuel analysis because it captures the differences in land returns, which drive the allocation of land to various uses. To implement the land module, land is decomposed into 18 agro-ecological zones (AEZs). In terms of modeling international trade, the model allows imperfect transformation of output and supply across markets of destination: exports and domestic markets. The transformation is through a two-level nested constant elasticity of transformation (CET) structure. At the first level, output is allocated between domestic sales and aggregate exports. At the second level, aggregate exports are allocated across foreign markets. The specification of the model is flexible because the degree of transformation is determined by the value of the elasticity 5 of transformation at each level. The model assumes product differentiation by region of origin through the Armington assumption. This assumption is embedded in the model through a two- level nested constant elasticity of substitution (CES) structure. At the first level the aggregate Armington demand is allocated between goods produced domestically and aggregate imports. At the second level, aggregate import is further disaggregated across trading partners. The degree of product differentiation depends upon the magnitude of the substitution elasticity at each level. The closure rules used in the model include: (a) savings of households are endogenous and are affected by demographic factors through a saving function; (b) government revenues are endogenous but government expenditures are a fixed share of nominal gross domestic product; (c) government balance is fixed through a uniform shift in the household direct tax; (d) investment is savings driven; and (e) current account is exogenous, thus foreign savings is fixed. Changes in foreign trade are balanced through changes in the real exchange rate. There are three factors that drive the dynamics of the model: (a) exogenous growth in population and labor (based on the population projections of the United Nations); (b) capital accumulation (based on capital stock at previous period, investment in the current period and capital depreciation); (c) factor productivity/efficiency parameters which are spread almost throughout the model. 2.2 The GIDD Model The second model is the Global Income Distribution Dynamics, or the GIDD model (Bussolo, de Hoyos, and Medvedev, 2008), which utilizes the results of the CGE to simulate the effects on global income distribution and poverty. The GIDD model uses household survey data of 116 countries, which represent about 90 percent of the world population. The GIDD model projects household survey data using three sets of ex-ante macroeconomic information: (a) changes in demographic composition which consist of projection of population by age and by educational attainment; (b) movement of labor between agriculture and non-agriculture; and (c) economic growth. Figure 1 illustrates how the GIDD model incorporates these three changes to adjust the data in the household surveys in the base year to some year in the future. 6 Figure 1: with the Overall Framework of the GIDD Model Box1 - Population Projection Box 2 - Education Projection by Age Groups by education groups Box 3 - New Population Shares (Recalibrated Sampling Weights in Household Survey) Box 4 - Simulation Results from Global CGE model Box 5 - Simulated Household Survey Source: Bussolo, de Hoyos, and Medvedev ( 2008 ) Box 1 in the figure contains population projection from the United Nations for nearly 200 countries from 2000 to 2090 in five-year intervals. The population projections are disaggregated into age groups and gender. Box 2 contains the projection of educational attainment based on the population projection in Box 1. The education projection in Box 2 is arrived at using changes in the demographic structure over time. The basic idea in projecting education into the future based on population projection is that the average educational attainment of the population changes through the “pipeline effect� as the population ages. This means that the old and unskilled today will be replaced by the young and more educated skilled individuals as the population age advances. As a result, the overall skill endowment of the population in time t+1 increases as the educational attainment improves. Information in Box 1 and Box 2 are used in Box 3 to recalibrate the base year weights in the household survey to a new set of weights consistent with the population projection in Box1 and the education attainment projection in Box 2. The basic idea is to search for a new set of sampling weights in the household survey that is consistent with the projected population and 7 education. Bussolo, de Hoyos and Medvedev (2008) provide a detailed discussion of the adjustment process, some of the key elements of those discussion are also reflected below. Let the old sampling weights be M N (1) P   wm,n  Wi n i 'm m 1 n 1 where in and im are identity column vectors, n is the number of observations in the sample, m is a vector of individual-level characteristics targeted in the GIDD process and W are the weights. The sum of all weights, W, is equal to the total population, P. In the present version of the GIDD, the individual-level characteristics are age and education. The row sums yield the totals of the population sub-groups, which are given by N (2) Pm   wm,n  Wi n n 1 Equation 2 is true for all m. The new sampling weights will incorporate the projected population and education, which is given by  N (3) Pm   am,n wm,n  ( AW )i n . n 1 Equation 3 is true for all m. The matrix A=[am,n] is a matrix of multipliers which will  ensure that the m constraints on the future structure of population P are satisfied and (A.W) is the hadamard product. This system has (m∙ n -1) variables with m constraints. It is therefore underetermined. In the GIDD model, this problem is addressed through optimization by minimizing the distance between the original matrix W and the final matrix (A.W).2 The model imposes the condition that am,n  an . With this additional condition, the distance function is  D   0.5  an  1  2 (4) n 2 As an alternative method, the GIDD model addresses this problem by adding equations to make the system exactly identified. The equations added are restrictions that the multipliers must be equal for each subgroup m. However, Bussolo et al (2008) have observed that this process can result in flawed results especially if the sampling units are sufficiently dispersed across the m sub-groups. 8 and the constraint in (3) is simplified as  N N (5) Pm   am,n wm,n   an  wm,n n 1 n 1 Equation 4 is minimized subject to the constraints in equation 5. The first order conditions are: _ M (6) a n  1   �m wm,n m 1  N _ (7) Pm   a n wm,n n 1 These conditions can be written in matrix form as follows I  W '  A   i n  W    0  �   P  (8)       The solution is  A  0 W '(WW ')1  i n  (9) �         (WW ') (WW ')1   P    Equation will yield a simpler expression for Λ 1    (10) �  WW '  P  Wi n    The matrix that needed to be inverted has a dimension of (m∙n). This reduces the dimension of the problem. Once the values for Λ are known, the first order conditions in equation 5 can be used to obtain a solution for the A matrix. The above recalibration process changes the educational endowments of the population in some year in the future, which also changes the labor supply by age and skill groups in the CGE model in Box 4. The CGE incorporates expansion of biofuels policy shocks and simulates the effects into the future on key economic variables such as real per capita GDP and per capita consumption, consumer price index of agriculture and non-agriculture commodities, labor 9 movement between rural and urban and between agriculture and non-agriculture sectors, and changes in wages of various types of labor. These simulated economic effects are used in the GIDD model in Box 5 together with the new set of recalibrated weights in Box 3. The GIDD model uses all this information to calculate the income distribution and poverty effects of large- scale expansion in biofuels in some year in the future. In projecting the data in the household survey into the future using the simulated results of key economic variables from the CGE and the recalibrated new sampling weights, two other processes are undertaken in the GIDD model. Detailed discussion of the processes is also given in Bussolo, de Hoyos and Medvedev (2008). The first process involves a movement of labor from the shrinking sector to the expanding sector. Workers that will be moved based on individual characteristics that are inputted into a probit function. For example, the probability of observing individual j working in non-agriculture (NA) is (11) Pr  NAj  1  P  X j , Z j  where Xj, and Zj are vectors of personal and household characteristics of individual j, respectively. The vector of coefficients in equation 11 is βp. Given this set of coefficients and the personal and households characteristics, workers are then ordered based on probability score calculated using equation 10. Workers with higher probability to be in non-agriculture are moved out of agriculture up to a point where the predicted share of workers by sector (a macro constraint) is satisfied. Once the labor movement takes place, the second process involves adjusting income of those who have moved. This income assignment to the “new entrants� in the expanding sector is done through a Mincer equation in agriculture (A) and non-agriculture (NA). (12) ln Y  j , s  X j β s  � i ,s where s = (A,NA). The “new entrants� will carry their personal endowments Xj and their residual εj to sector where they move. However, those who have moved from agriculture to non- agriculture will be paid with prices βNA. In order to incorporate the variances in the distribution 10 of unobservables between agriculture and non-agriculture, it is necessary to rescale their residuals as in the following equation (13) ln Y  j , NA  X j β NA  � * j where ε*j = εj,A·(σε,NA/ σε,A) and σε,s the standard deviation of the distribution of the residuals in sector s. The new income generated in the microsimulation usually does not match consistently with the income generated in CGE simulation. The GIDD model applies several steps to adjust factor returns by skill type and sector, and average income per capita based on the results of the CGE model. Let [ys,l] be the initial distribution of earnings of labor type l in sector s in the macro data. Also, let a series of wage gaps (s+l - 1) as y s ,l (14) g s ,l  1 y1,1 where y1,1 is the average labor earnings of unskilled workers in agriculture. The macro data will have wage premiums which may or may not be consistent the wage premium in the micro data. Let the wage premiums in the micro data be [g‟s,l]. In the GIDD model, the counterfactual wage gaps are calculated as follows ˆ g s ,l (15) g s' ,l  g s' ,l ˆ g s,s The adjustments thus far do not guarantee that the average change in per capita income/consumption generated by the microsimulation is consistent with the average change in the CGE model. To ensure that the average change is the same across the two models, the following adjustment carried out in the GIDD ˆ y (16) y'  y' ˆ y 11 Data The data required for the study can be divided into two groups: (i) data needed for the global CGE model and (ii) data needed for GIDD model. A detailed discussion on the data for the global CGE model is available in Timilsina et al. (2010), so in this section we focus data required for GIDD model. The main sources of data in the GIDD model include: (a) the dataset assembled for the production of the World Bank World Development Report (WDR) for developing countries, which are drawn largely from the Living Standards and Measurement Study (LSMS) and the African Institute for Sustainability and Peace (ISP)-Poverty monitoring group; (b) the Europe and Central Asia (ECA) databank and the different World Bank sources for Eastern Europe countries, and (c) the Luxembourg Income Studies (LIS) database for most of the developed countries. There are two versions of the GIDD model. The original version includes data for the year 2000, while the second includes updates for the year 2005. In the original version, nationally representative household surveys of various countries nearest to 2000 were chosen. For household surveys of countries valued not in year 2000, the following adjustments were applied in order to “value� them in 2000. Local consumer price index (CPI) in the country was used to adjust income and consumption to 2000 domestic values. Furthermore, the values in the household surveys were converted to international dollars in year 2000 using the Purchasing Power Parity (PPP) factor obtained from the Penn World Tables. A correction factor was also applied to the population weights to make the population level consistent to year 2000. In the updated GIDD model, the values for 2000 were updated to 2005 using updated population and 2005PPP factor. Also, the mean income at the country level was adjusted to match the latest poverty rates reported in POVCAL3. Thus, the poverty rates at the country level are consistent with the 2005PPP poverty estimates in Chen and Ravallion (2008). The GIDD database covers all regions in the world. Eastern Europe and Central Asia is 100 percent covered; Latin America 98 percent; South Asia 98 percent; East Asia and Pacific 96 3 POVCAL is a tool developed at the World Bank for calculating poverty and inequality indices. 12 percent; High Income Countries 79 percent; Sub-Saharan African 74 percent; and Middle East and North Africa 70 percent (Ackah, et al, 2008). Definition of Scenarios We developed three scenarios and analyzed over the period 2009-2020.The first scenario is the Business as Usual (BaU) scenario. This is the baseline scenario which incorporates a number of assumptions. The first set of assumptions is on the world prices of three sources of energy (coal, oil, and natural gas) which are exogenous variables in the model, whose values were derived from the projections calculated outside the model. Table 1 shows the price indexes of these three sources of energy used under the BaU. The price of coal has the highest increase in 2004-20 (161 percent), followed by the price of natural gas (85 percent), and by the price of oil (113.7). There was an oil price spike in 2008 and 2009, but it stabilizes thereafter. Table 1: World price index of energy Year World price of coal World price of oil World price of natural gas 2004 100.0 100.0 100.0 2005 101.0 104.9 105.4 2006 102.1 110.0 111.1 2007 103.1 115.4 117.0 2008 104.2 121.0 123.3 2009 105.2 126.9 130.0 2010 106.3 133.1 137.0 2011 107.4 139.6 144.4 2012 108.5 146.4 152.1 2013 109.6 153.6 160.3 2014 110.8 161.1 168.9 2015 111.9 168.9 178.0 2020 114.6 168.9 178.4 In 2004-2015 world prices of these commodities are exogenous, but after 2015 they are endogenous The second set of assumptions under the BaU scenario is the growth rates of gross domestic product (GDP) of the different countries and regions in the model. The GDP growth rates were based on the growth projections made by the World Bank. In Table 2 one can observe that China has the highest annual growth in real GDP followed by India. Developing countries have relatively higher real GDP growth than developed countries. 13 Table 2: Assumptions on GDP and population growth GDP growth, % Population (million), growth (%) Ave. 2005- Countries/Regions 2005 2020 2020 /a/ 2004 2020 Growth /a/ China 10.1 7.2 8.4 1,303 1,431 0.59 Japan 1.9 0.8 1.2 128 123 -0.24 Indonesia 5.7 4.5 5.0 218 259 1.09 Malaysia 5.0 4.3 4.8 25 31 1.39 Thailand 4.5 3.5 3.9 64 69 0.52 India 9.2 6.4 7.4 1,080 1,292 1.13 Canada 3.1 1.5 1.9 32 34 0.39 United States 3.1 2.1 2.1 294 330 0.73 Argentina 9.2 4.3 5.2 38 44 0.89 Brazil 2.9 3.5 3.7 184 217 1.03 France 1.7 1.2 1.5 61 63 0.27 Germany 0.9 1.3 1.6 83 79 -0.29 Italy 0.2 0.8 1.0 58 54 -0.42 Spain 3.6 2.2 2.7 43 42 -0.03 United Kingdom 1.8 1.8 2.0 60 60 0.05 Russia 6.4 4.9 5.6 144 131 -0.57 South Africa 5.0 4.3 4.6 46 50 0.43 Rest European Union and EFTA b/ 2.9 2.2 2.6 196 194 -0.09 Rest of Latin America and Caribbean 4.5 3.4 3.8 326 401 1.30 Australia and New Zealand 2.8 2.2 2.6 24 27 0.67 Rest of East Asia and Pacific 5.0 6.0 6.4 343 404 1.02 Rest of South Asia 6.7 4.8 5.4 398 527 1.78 Rest of Europe and Central Asia 7.3 5.7 6.1 227 247 0.51 Middle East and North Africa 5.2 3.7 4.1 360 469 1.66 Rest Sub-Saharan Africa 6.2 5.7 6.0 690 942 1.97 /a/ Growth growth /b/ EFTA - European Free Trade Area Source: World Bank‟s Development Data Group The third set of assumptions under the BaU scenario is on the population in each of the countries and regions in the model also shown in Table 2. Growth rates of population were based on the population growth projection of the United Nations. One can observe that developed countries have generally declining population growth rates. The fourth set of assumptions under the BaU scenario is that existing biofuel policies (e.g. already implemented mandates, subsidies and import duties) will continue throughout the study horizon (2010-2020). Biofuel penetration in the BAU scenario, which is defined as the 14 share of biofuels in the total liquid fuel consumption in the road transportation is presented in Table 3. The second scenario is the announced biofuel targets (AT) scenario. This scenario considers the implementation of biofuel use targets consistent with what countries already have announced. Table 3 shows shares of biofuels under the AT scenario in 2020. There are notable increases in the biofuel penetration as compared to that in the BAU case for India, Thailand, France, Germany, Italy, Spain, United Kingdom, and in the rest of European Union. On the other hand, biofuel penetrations for Brazil, United States and Malaysia are not much different from that in the BaU scenario as existing policies and incentives, if continued will be sufficient to meet the AT scenario. Table 3: Share of biofuels in total liquid fuel demand for transportation (biofuel penetration) Biofuel penetration BaU AT ET Countries/regions 2009 2020 2020 2020 China 2.0 2.4 3.7 7.3 Japan 0.4 1.0 1.0 1.2 Indonesia 1.8 3.3 5.0 10.0 Malaysia 1.8 5.1 5.1 5.1 Thailand 1.3 2.8 5.2 10.4 India 3.1 4.9 16.7 16.7 Canada 1.0 2.5 4.1 8.2 United States 3.4 8.3 8.3 8.3 Argentina 1.4 3.3 5.0 10.0 Brazil 9.5 18.8 18.8 19.0 France 1.5 4.5 10.0 20.0 Germany 2.2 5.9 10.0 20.0 Italy 0.9 2.5 10.0 20.0 Spain 0.8 2.3 10.0 20.0 United Kingdom 0.3 1.0 10.0 20.0 Russia 2.7 5.3 5.3 5.3 South Africa 3.1 5.3 5.3 5.3 Rest European Union and EFTA 0.5 1.5 10.0 20.0 Rest of Latin America and Caribbean 0.8 2.1 2.1 3.0 Australia and New Zealand 0.4 1.0 1.2 2.5 Rest of East Asia and Pacific 0.4 1.0 1.5 2.5 Rest of South Asia 0.5 0.9 0.9 0.9 Rest of Europe and Central Asia 0.8 1.9 1.8 1.9 Middle East and North Africa 0.1 0.2 0.1 0.1 Rest Sub-Saharan Africa 1.3 3.4 3.4 3.4 Mandates for cellulosic ethanol are not included. 15 The third scenario is the enhanced biofuel targets (ET) scenario. This scenario generally considers a doubling of the announced targets, keeping the timing of the implementation of the targets unchanged. In India, however, we retain the AT target level because it is already very high in 2020 (see Table 3). Simulation Results 5.1 Selected CGE Results The effects of expansion of biofuels on world prices of feedstock are presented in Figure 2. Under the AT scenario where biofuels targets announced by various countries are assumed to be fully implemented by 2020, the price of sugar, the main biofuel feedstock globally, exhibits the highest increase, more than 7 percent from the BaU scenario. If the targets were doubled (i.e., ET scenario) the increase in sugar price from the BaU scenario would be around 10 percent in 2020. In contrast, the increase in prices of other commodities would be less than 2 percent under AT scenario and less than 4 percent under ET scenario. Figure 2: World prices of feedstock in 2020 (% change from the BaU Scenario) 2.2 Other grains 0.9 3.6 Corn 1.1 9.8 Sugar 7.4 3.0 Oilseeds 1.3 ET AT 2.3 Wheat 1.0 0 2 4 6 8 10 12 percent 16 Higher prices of these crops results in higher food prices as illustrated in Figure 3. For developing countries food CPI increases by 0.5 percent in 2020 in AT, but for developed countries the increase is only 0.2 percent4. Food CPI increases by 1.1 percent in ET in developing countries in 2020, but only by 0.6 percent in developed countries5.The increase in food CPI in developing countries is relatively higher than in developed countries because the share of food expenditure in total household income is higher in developing countries than in developed countries. Figure 3: Food consumer price index in 2020 (% change from the BAU scenarios) 1.1 Developing 0.5 0.6 Developed 0.2 ET AT 0.0 0.2 0.4 0.6 0.8 1.0 1.2 percent Among the developing countries, East Asian countries have the highest increase in food CPI both in AT and ET relative to BaU as shown in Figure 4. The increase in food CPI in Latin American and African countries is also notable relative to the increase in developed countries. 4 Food CPI refers to composite prices of agricultural goods and processed foods, where agricultural goods comprise paddy rice, wheat, oil seeds, sugar cane and sugar beet, corn, other cereal grains, vegetables, fruits and crops, and livestock. 5 These price changes are for 2020, but there is gradual increase in food CPI every year after the expansion in biofuels which is implemented starting from 2009. 17 The differences in the effects on food CPI across regions/countries are due to the variations in the share of food expenditure in total household income. Figure 4: Food price increase at country or regional level (% change in food CPI from the BaU Scenario) Rest Sub-Saharan Africa Middle East and North Africa Rest of Europe and Central Asia Rest of South Asia Rest of East Asia and Pacific Australia and New Zealand Rest of Latin America and Caribbean Rest European Union and EFTA South Africa Russia United Kingdom ET AT Spain Italy Germany France Brazil Argentina United States Canada India Thailand Malaysia Indonesia Japan China 0.0 0.5 1.0 1.5 2.0 2.5 18 The effects on real per capita GDP vary considerably across countries and regions as Figure 5 shows. While Thailand shows relatively higher increase in food prices in Figure 4, it is partly compensated by the positive increase in real per capita GDP in both scenarios. Similar pattern is observed in Indonesia. But this is not the case in Sub-Saharan Africa where poverty incidence is the highest, in Middle East and North African, and in Russia. In these regions, food prices increase while real per capita GDP declines. The decline in real capital GDP is due to the decline in output of mining and service sectors. In India, the decline in real per capita GDP is due to the high increase in food prices. Figure 5: Percentage Change in Real Per Capita GDP from the BaU Scenario Rest Sub-Saharan Africa Middle East and North Africa Rest of Europe and Central Asia Rest of South Asia Rest of East Asia and Pacific Australia and New Zealand Rest of Latin America and Caribbean Rest European Union and EFTA South Africa Russia United Kingdom ET AT Spain Italy Germany France Brazil Argentina United States Canada India Thailand Malaysia Indonesia Japan China -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 19 An expansion in biofuels leads to larger demand for feedstock thereby resulting in higher demand for factors used in feedstock production. Prices of these factors are expected to increase. In the CGE model, there are four types of labor: skilled urban labor, skilled rural labor, unskilled urban labor, and unskilled rural labor. In discussing the results on factor prices, wages of various labor skills are expressed as ratios relative to the wage of unskilled rural labor and are compared to the baseline. The effects on the wage ratios of the various labor types in AT are presented in Figure 6. Except for the United States, wages in all countries and regions decline relative to the wage of unskilled rural labor, which implies relatively higher wage for unskilled rural labor. The increase in the wage for unskilled rural labor causes movement of unskilled urban labor to the rural sector. The decline in wages of various labor skills relative to the wage of unskilled rural labor is generally higher in developing countries than in developed countries. This is because feedstock production in developing countries is relatively intensive in the use of unskilled rural labor. The highest increase in the relative wage of unskilled rural labor is in India, Middle East and North Africa, Sub-Saharan Africa, Brazil, Argentina, Thailand, Indonesia, and the rest of Latin America. Similar pattern of wage effects is observed in ET in Figure 7. 20 Figure 6: Wage Relative to Rural Unskilled Wage (Percentage Change in AT from BaU) Rest Sub-Saharan Africa Middle East and North Africa Rest of Europe and Central Asia Rest of South Asia Rest of East Asia and Pacific Australia and New Zealand Rest of Latin America and Caribbean Rest European Union and EFTA South Africa Russia Skilled urban wage United Kingdom Spain Skilled rural wage Italy Unskilled urban wage Germany France Brazil Argentina United States Canada India Thailand Malaysia Indonesia Japan China -2.0 -1.5 -1.0 -0.5 0.0 0.5 21 Figure 7: Wage Relative to Rural Unskilled Wage (Percentage Change in ET from BaU) Rest Sub-Saharan Africa Middle East and North Africa Rest of Europe and Central Asia Rest of South Asia Rest of East Asia and Pacific Australia and New Zealand Rest of Latin America and Caribbean Rest European Union and EFTA South Africa Skilled urban wage Russia Skilled rural wage United Kingdom Unskilled urban wage Spain Italy Germany France Brazil Argentina United States Canada India Thailand Malaysia Indonesia Japan China -4.5 -4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 In developing countries, the relatively higher wage for unskilled rural labor reduces the migration out of agriculture. There are no movements of unskilled labor in developed countries, but there are notable shifts in developing countries as shown in Figure 8a and b. In both scenarios there is a decline in unskilled urban labor relative to the baseline and a corresponding increase in unskilled rural labor in developing countries with expansion in biofuels. This is in 22 response to the higher relative wage for unskilled rural labor. In AT, the largest movement of unskilled labor towards agriculture is observed in India and Middle East and North Africa, but there are also labor movement in Brazil, Argentina, Thailand and Indonesia. Figure 8: Movement of Unskilled Labor (Percentage Change from BaU) (a) AT Scenario Rest Sub-Saharan Africa Middle East and North Africa Rest of South Asia Rest of East Asia and Pacific Rest of Latin America and Caribbean South Africa Brazil Argentina Urban unskilled labor India Rural unskilled labor Thailand Malaysia Indonesia China -0.2 -0.1 -0.1 0.0 0.1 0.1 0.2 23 (b) ET Scenario Rest Sub-Saharan Africa Middle East and North Africa Rest of South Asia Rest of East Asia and Pacific Rest of Latin America and Caribbean South Africa Brazil Argentina Urban unskilled labor India Rural unskilled labor Thailand Malaysia Indonesia China -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 The expansion of biofuels does not only increase the demand for feedstock, but also it increases the returns to land. Table 4 shows the comparison between household labor income (skilled and unskilled) and land income that also occurs to land owning or agriculture or rural households. In some countries and regions, household labor income declines because not all wages increase – we have observed higher wages on unskilled rural labor relative to wages of all skills. But in all countries and regions, land income has increased. The highest increase in household land income is observed in European countries because these are the countries having relatively higher targets (i.e., 10%) for biofuels. The increase is substantial under the ET scenario because biofuels targets are doubled under this scenario. Developing countries also see higher household land income. The interaction between biofuels and land use is critical in the analysis 24 because without taking these interactions into the analysis, the poverty and income distribution effects of biofuels may be overestimated. Table 4: Impacts on Household Labor and Land Income in 2020 (Percentage Change from BaU) Land Income Labor Income AT/BaU ET/BaU AT/BaU ET/BaU China 0.8 2.6 0.0 -0.1 Japan 1.6 4.6 0.0 0.1 Indonesia 1.0 3.5 0.0 0.0 Malaysia 0.6 1.5 -0.1 -0.2 Thailand 1.9 6.7 0.2 0.5 India 3.7 4.7 0.0 0.3 Canada 2.0 5.9 0.0 -0.1 United States 1.4 3.9 -0.1 -0.4 Argentina 1.8 5.0 0.1 0.2 Brazil 3.3 7.7 0.7 1.1 France 11.6 37.3 0.3 0.6 Germany 4.9 15.6 0.1 0.2 Italy 6.2 17.1 0.2 0.3 Spain 5.0 14.8 0.2 0.5 United Kingdom 9.6 27.1 0.1 0.2 Russia 0.8 2.1 -0.4 -0.9 South Africa 1.2 3.0 0.1 0.2 Rest European Union and EFTA 2.9 7.9 0.0 0.1 Rest of Latin America and Caribbean 0.9 2.7 -0.1 -0.4 Australia and New Zealand 1.2 3.3 0.1 0.1 Rest of East Asia and Pacific 0.7 1.7 0.1 0.2 Rest of South Asia 0.6 1.5 0.0 -0.1 Rest of Europe and Central Asia 1.2 3.2 0.1 0.2 Middle East and North Africa 1.3 3.3 -0.7 -1.6 Rest Sub-Saharan Africa 0.4 1.1 -0.3 -0.6 Relatively higher wages and demand for unskilled rural labor because of increased biofuel production have favorable effects on agriculture and rural households in developing countries. Higher land income also favorably affects them, although the impact is higher for richer farmers than for poorer farmers with less farm holdings. But food consumer prices increase as biofuel production and food production compete for feedstock. The increase in food prices is higher in developing countries than in developed countries. The next section will 25 discuss how these effects will net out and affect on poverty and income distribution in developing countries. 5.2 Effects on Global Poverty and Income Distribution The CGE results on key economic variables were incorporated into the GIDD model to simulate the distributional and poverty effects of expansion in biofuels. In the poverty analysis, two poverty threshold levels were applied: $1.25 per day and $2.50 per day. In Table 5, we present the GINI coefficient, the poverty headcount, and the poverty incidence of major regions in BaU in 2005 and 2020. Table 5: Poverty and income distribution in BaU GINI Population Poverty Headcount (million) Poverty Incidence (%) Regions Coefficient (million) Poor-1 /a/ Poor-2 /b/ Poor -1 Poor-2 BaU 2005 /c/ East Asia 0.4190 1,805 280 898 15.49 49.73 Industrial Countries 0.3905 711 - 3 0.00 0.42 East Europe and Central Asia 0.3911 449 34 85 7.51 18.95 Latin America 0.6055 492 48 115 9.70 23.40 Middle East 0.3972 205 8 56 3.98 27.47 South Asia 0.2893 1,439 538 1,231 37.41 85.56 Sub-Sahara Africa 0.5584 445 234 355 52.54 79.73 ALL 0.7017 5,546 1,141 2,743 20.58 49.46 BaU 2020 East Asia 0.38443 2,014 111.7 605.1 5.54 30.04 Industrial Countries 0.40234 744 - 3.0 0.00 0.40 East Europe and Central Asia 0.36617 453 14.3 26.9 3.15 5.93 Latin America 0.57415 591 49.5 137.3 8.38 23.24 Middle East 0.40380 257 0.4 22.5 0.16 8.75 South Asia 0.31384 1,735 211.7 1,131.3 12.21 65.21 Sub-Sahara Africa 0.55093 567 176.4 356.0 31.09 62.74 ALL 0.65466 6,361 564.1 2,282.0 8.87 35.87 /a/ Poor-1 threshold is $1.25 (PPP) per day /b/ Poor-2 threshold is $2.50 (PPP) per day /c/ Business as usual The GDP growth projection incorporated in the BaU scenario results in falling poverty incidence, poverty headcount and GINI coefficient in all regions between 2005 and 2020. In the $1.25 per day poverty threshold, the global poverty incidence declines from 20.58 percent in 2005 to 8.87 percent in 2020. In the $2.50 per day poverty threshold, the global poverty incidence declines from 49.46 percent in 2005 to 35.87 percent in 2020. The GINI coefficient 26 also declines from 0.7017 in 2005 to 0.6547 in 2020, indicating declining income inequality. The GINI coefficients are also significantly different across regions. The poverty effects of expansion biofuels at the regional level are presented in Figure 9 for the $1.25 per day poverty threshold. Figure 9: Regional Poverty Effects (Change from BaU in 2020 ($1.25/day) ALL 6,849 5,810 Sub-Saharan Africa 1,613 537 South Asia 5,619 5,434 Middle East -9 -4 Latin America -255 -120 ET AT East Europe and Central Asia -58 -20 Industrial Countries 0 0 East Asia -62 -17 -1,000 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 thousand The poverty results are mixed across regions and countries. There is an increase in global poverty headcount both in AT and ET relative to the baseline. With the $1.25 per day poverty threshold the increase in global poverty headcount is 5.8 million in AT and 6.8 million in ET. The increase comes largely from South Asia (5.6 million in AT and 5.4 million in ET) and Sub- Saharan Africa (537 thousand in AT and 1.6 million in ET). The poverty effects of higher biofuel demand using the $2.50 PPP per day poverty line is presented in Figure 10. Relative to the lower poverty line, there is a substantial increase in the 27 number of poor. Our results indicate about 42 million people will be below the poverty line. The increase largely comes from East Asia. This comes from China as we show in Table 6. Figure 10: Regional Poverty Effects (Change from BaU in 2020 ($2.50/day) ALL 42,503 41,995 Sub-Saharan Africa 1,664 707 South Asia 6,443 6,051 Middle East 283 187 Latin America -558 ET AT -289 East Europe and Central Asia -8 -4 Industrial Countries 0 0 East Asia 34,679 35,342 -10,000 0 10,000 20,000 30,000 40,000 50,000 thousand Figure 11 shows the effects of the expansion in biofuels on income distribution. Note that the results shown are the difference in the GINI coefficient in 2020 between the biofuel scenarios (AT and ET) and BaU. The change in the GINI coefficient is small across major regions. But overall, there is a slight reduction in the GINI coefficient. The GINI coefficients in Sub-Saharan Africa and Latin America decline, while they increase in the rest of the regions. 28 Figure 11: Change in the GINI Coefficient (Change from BaU in 2020) ALL Sub-Saharan Africa South Asia Middle East Latin America East Europe and Central Asia Industrial Countries ET AT East Asia -0.300 -0.200 -0.100 0.000 0.100 0.200 0.300 Table 6 presents the poverty results at the country level. There is higher number of poor in Sub-Saharan Africa, especially in Nigeria, Tanzania, Kenya, Mali and Senegal, but lower in South Africa. Similar patterns of effects are observed in both biofuel scenarios, AT and ET and in both poverty lines. Using the $1.25 PPP per day poverty line, the increase in the number of poor in both AT and ET largely comes from South Asia. Within the region, the increase in poverty comes from India. There is a slight increase in poverty in Pakistan and Bangladesh. Using the $2.5 PPP per day poverty line, the number of poor is slightly higher. There is declining number of poor in Latin America in both AT and ET and in both poverty limes. The reduction largely comes from Brazil and to some extent from El Salvador. In East Asia, there is slight reduction in the number of poor using the $1.25 PPP per day poverty line, but there is considerable increase using the $2.50 PPP per day poverty line for both biofuel scenarios, AT and ET. The increase comes largely from China. But in the rest of the region, there is reduction in the number of poor in Indonesia, Vietnam, Philippines, and Thailand. Similarly, there is also lower poverty 29 in East Europea and Central Asia. In the Middle East, the poverty effects mainly come from Yemen. There is reduction in poverty using the $1.25 per day threshold but there is higher poverty using the $2.50 per day poverty line. Conclusions More than 40 countries around the world have set targets for further promotion of biofuels. However, given the present commercially deployed technologies, which compete heavily for raw materials used in food production, a large-scale expansion of biofuels would put pressure on food supply and prices which would, in turn, have implications on poverty and global income distribution. Using a global computable general equilibrium model and a global income distribution dynamics model, this study analyzes the distributional and poverty effects of large-scale expansion in biofuels. The results from the computable general equilibrium model indicate that relatively large-scale expansion of biofuels leads to a bit higher world prices of feedstock (e.g., sugar, corn, oilseeds, wheat, and other grains) which translates to higher food prices. The jump in food prices is higher in developing countries than in developed countries. The impact on real per capita GDP of the large-scale biofuel expansion varies across the countries/regions. It is found to increase in countries where biofuel industry has already advanced, such as Thailand, Brazil, Argentina, and Indonesia. On the other hand, per capita GDP declines in countries which have ambitious targets for the near future but current level of production is very small. Countries such as, India, Sub-Saharan Africa, Middle East and North African regions, Russia, and China would bear the highest losses in their per capita GDP. The large-scale expansion of biofuels leads to a bit higher wages of unskilled rural labor relative to wages of the other labor types, skilled urban, skilled rural, and unskilled urban in both developed and developing countries. The magnitude is, however, bigger in the latter. The positive wage effects on unskilled rural labor reduce migration out of agriculture. This is because production of feedstock in developing countries is relatively intensive in the use of unskilled rural labor. Furthermore, production of biofuels causes an increase in household land income in rural areas. 30 This study also finds that large-scale expansion of biofuels leads to a slight increase in the number of poor people if the poverty line used is $1.25 PPP per day. The increase largely comes from South Asia (India) and Sub-Saharan Africa. A large number of countries in Sub- Saharan Africa would realize an increase in poverty due to the large-scale expansion of biofuels. If the poverty line is drawn at $2.50 PPP per day, the number of people below the poverty line would increase significantly. The GINI coefficient in Sub-Saharan Africa and Latin America would drop, whereas it would increase in the rest of the regions. At the global level, the GINI coefficient would drop, although the magnitude is small. While the effects of higher biofuel mandates on rural unskilled workers are positive in terms of higher relative wages and labor movement from urban to rural, the increase in food prices especially in developing countries is a major concern. Thus, since the increase in food prices is largely due to dependence of the current biofuel technology on food crops as raw materials in biofuel production, from the perspective of global poverty the development of a biofuel technology that uses less food crops is critical. References Al-Riffai, P., B. Dimaranan, and D. Laborde, 2010. "Global Trade and Environmental Impact Study of the EU Biofuels Mandate", Report for the European Commission, DG TRADE, ATLASS Consortium. Ackah, C., Bussolo, R. de Hoyos, and D. Medvedev. 2008. “A New Dataset on Global Income Distribution�. Washington, DC: World Bank. Arndt, C., R. Benfica, F. Tarp, J. Thurlow, and R. Uaiene. 2008. “Biofuels, Poverty and Growth: A Computable General Equilibrium Analysis of Mozambique�. IFPRI Discussion Paper 803. Washington, D.C. International Food Policy Research Institute. Beghin, J., Timilsina, G, van der Mensbrugghe, D. and Mevel, S. 2010. “The Impacts of Biofuels Targets on Land-Use Change and Food Supply: A Global CGE Assessment� Unpublished World Bank manuscript. Birur, D, Hertel, T., and Tyner, W. 2008. “Impact of Biofuel Production on World Agricultural Markets: A Computable General Equilibrium Analysis�. GTAP Working Paper No. 53. Boeters, S. Veenendaal,P. van Leeven, N. Rojas-Ramagoza, H. 2008. “The Potential for Biofuels alongside the EU-ETS�. Paper for Presentation at the Eleventh Annual GTAP Conference „Future of Global Economy‟, Helsinki, June 12-14. Bussolo, M., R. de Hoyos, and D. Medvedev. 2008. “Economic Growth and Income 31 Distribution: Linking Macroeconomic Models with Household Survey Data at the Global Level�. Washington, DC: World Bank. Bussolo, M., R. de Hoyos, D. Medvedev, and D. van der Mensbrugghe. 2008. “Global Climate Change and its Distributional Impacts�. Washington, DC: World Bank. Chen, S. and M. Ravallion. 2008. "The developing world is poorer than we thought, but no less successful in the fight against poverty", World Bank Policy Working Paper 4703. Cline, W. 2007. Global Warming and Agriculture: Impact Estimates by Country. Washington, DC: Center for Global Development and Peterson Institute for International Economics. de Gorter, H. and D. Just. 2010. “The Social Costs and Benefits of Biofuels: The Intersection of Environmental. Energy and Agricultural Policy�. Applied Economic Perspectives and Policy. 32 (1): 4-32. de Hoyos, R. 2008. “Global Income Distribution Dynamics (GIDD)�. Washington, DC: World Bank. de Hoyos, R, and D. Medvedev. 2011. “Poverty Effects of Higher Food Prices: A Global Perspective,� Review of Development Economics, forthcoming. Food and Agricultural Organization. 2008. “Soaring Food Prices: Facts, Perspective, Impacts and Action Required�. High-Level Conference On World Food Security. FAO, Rome. Habib-Mintz, N. 2010. Biofuel investment in Tanzania: Omissions in implementation. Energy Policy 38, pp. 3985–3997. Hertel, T. 2009. “Analyzing the Poverty Impacts of Biofuel Mandates�. Transition to a Bioeconomy: Global Trade and Policy Issues Conference, Washington DC. Hertel, T., Tyner, W., and Birur, D. 2008. “Biofuels for All? Understanding the Global Impacts of Multinational Mandates�. GTAP Working Paper No. 51. Hahn, R., and C. Cecot. 2009. “The Benefits and Costs of Ethanol: An Evaluation of the Government‟s Analysis�. Journal of Regulatory Economics. 35 (3): 275-295. Gal Hochman, G., Rajagopal, D., Timilsina, G.R. and Zilberman, D. 2011. The Role of Inventory Adjustments in Quantifying Factors Causing Food Price Inflation, World Bank Policy Working Paper, No. 5744, World Bank, Washington, DC. Horridge, M. 2008. “SplitCom: Programs to disaggregate a GTAP sector�. Melbourne, Australia: Centre of Policy Studies, Monash University. IMF Commodity Prices. (http://www.imf.org/external/np/res/commod/index.asp) International Food Policy Research Institute. 2008. “Higher Food Prices: The What, Who, and How Proposed Action�.IFPRI Policy Brief, IFPRI, Washington D.C. Ivanic, M. and Martin, W. 2008. “Implications for Higher Global Food Prices for Poverty in Low Income Countries�. Development Research Group, Trade Group. World Bank, Washington D.C Keeney R., and Hertel, T. 2008. „The Indirect Land Use Impacts of U.S. Biofuel Policies: The Importance of Acreage, Yield, and Bilateral Trade Responses� GTAP Working Paper No. 32 52. Keyzer, M., Merbis, M., Nubé, M., and van Wesenbeeck, C., 2008. “Food, Feed, and Fuel: When Competition Starts to Bite�. SOW-VU Brief, Center for Food Studies, VU University, Amsterdam. Mevel, S. 2008. “Inserting Biofuels into the GTAP Database�. Washington, D.C.: World Bank. Mevel, S. 2008. “Splitting the GTAP „other cereal grains‟ (GRO) sector into Corn and Other Cereal Grains�. Washington, D.C.: World Bank. Mitchell, D. 2008. “A Note on Rising Food Prices�. Development Prospects Group. World Bank: Washington, D.C. June. Runge, C., and B. Senauer. 2007. “How Biofuels Could Starve the Poor�. Foreign Affairs. 86 (3): 41-53. Taheripour F., Birur D., Hertel T., and Tyner W. 2008. “Introducing Liquid Biofuels into the GTAP Data Base�. GTAP Research Memorandum No.11. Taylor, J., and P. Van Doren. 2007. “The Ethanol Boondoggle�. The Milken Institute Review, First Quarter. Timilsina, G.R., Beghin, J.C., van der Mensbrugghe, D. and Mevel, S. 2010. “The Impacts of Biofuel Targets on Land-Use Change and Food Supply: A Global CGE Assessment.� Policy Research Working Paper 5513. The World Bank. Washington, DC. Timilsina, G.R., Beghin, J.C., van der Mensbrugghe, D. and Mevel, S. (2012), The Impacts of Biofuels Targets on Land-Use Change and Food Supply: A Global CGE Assessment, Agriculture Economics, Vol. 43, pp. 313-330. van der Mensbrugghe, D. 2009. “The ENVironmental Impact and Sustainability Applied General Equilibrium (ENVISAGE) Model�. Washington, D.C.: World Bank. 33 Appendix Table 6: Impacts on Poverty by Country (Change from the BaU Scenario) Unit: Thousand AT/a/ ET /b/ AT ET Countries/Regions Poor-1 /c/ Poor-2 /d/ Poor-1 Poor-2 Countries/Regions Poor-1 Poor-2 Poor-1 Poor-2 Sub-Sahara Africa 537 707 1,613 1,664 East Asia -17 35,342 -62 34,679 Comoros 0 0 0 0 China 0 35,684 0 35,684 Lesotho 0 0 0 0 Mongolia 0 0 0 0 Malawi 0 0 0 0 Malaysia 0 0 0 0 Niger 0 0 0 0 Papua New Guinea 0 0 0 0 Rwanda 0 0 0 0 Indonesia 0 -206 -28 -759 Sierra Leone 0 0 0 0 Cambodia -13 -3 -29 -2 Zambia 0 0 0 0 Philippines 0 -27 0 -86 Burundi 8 45 17 59 Thailand -4 -23 -5 -53 Benin 8 33 23 62 Vietnam 0 -83 0 -105 Burkina Faso 19 36 42 60 Côte d'Ivoire 7 23 9 49 Latin America -120 -289 -255 -558 Cameroon 7 38 20 82 Bolivia -1 0 5 1 Ghana 35 25 91 72 Brazil -34 -182 -155 -325 Guinea 9 13 38 37 Chile 0 -1 5 5 Kenya 0 28 0 398 Colombia -7 -7 1 7 Madagascar 31 44 54 60 Costa Rica 0 0 -2 -4 Mali 14 41 41 93 Dominican Republic 0 -5 0 -13 Mauritania 3 2 6 3 Ecuador 1 0 10 1 Nigeria 314 294 1,066 455 Guatemala -64 0 -67 5 Senegal 20 60 47 90 Guyana 0 0 0 0 Tanzania 57 55 164 196 Honduras 0 -6 -1 -7 Uganda 6 5 12 12 Haiti 0 0 0 12 South Africa 0 -35 -17 -64 Jamaica 0 -2 0 -4 Mexico 0 0 0 0 East Europe and Central Asia -20 -4 -67 -8 Nicaragua -4 4 -2 -3 Bosnia & Herzegovina 0 0 0 0 Panama -5 0 -16 2 Czech Republic 0 0 0 0 Peru -3 -11 -3 -33 Slovak Republic 0 0 0 0 Paraguay -2 -2 -13 -3 Turkmenistan 0 0 0 0 El Salvador -1 -76 -4 -191 Albania 0 0 0 0 Venezuela, Rep. Bol. 0 0 -12 -8 Armenia 0 0 0 0 Azerbaijan -1 -3 -1 -7 Middle East -4 187 -9 283 Bulgaria 0 0 0 0 Egypt 0 0 0 0 Estonia 0 0 0 0 Iran, I.R. of 0 0 0 0 Georgia 0 0 0 0 Tunisia 0 0 0 0 Hungary -11 1 -58 2 Jordan 0 0 0 1 Kazakhstan 0 0 0 0 Morocco 0 -20 0 20 Kyrgyz Republic 0 3 0 0 Yemen, Republic of -4 207 -9 263 Lithuania -7 0 -8 0 Moldova -1 0 -1 0 South Asia 5,434 6,051 5,619 6,443 Macedonia, FYR 0 0 0 0 Bangladesh -11 13 13 54 Poland 0 0 0 0 India 5,383 6,001 5,440 6,167 Romania 0 -2 0 -3 Sri Lanka -1 -8 -1 -17 Russia 0 0 0 0 Nepal -8 28 54 38 Tajikistan 0 -4 0 -4 Pakistan 71 17 113 200 Turkey 0 0 0 0 Ukraine 0 0 0 -4 Uzbekistan 0 0 0 9 /a/ Announced biofuel targets /c/ Poor-1 threshold is $1.25 (PPP) per day /b/ Enhanced biofuel targets /d/ Poor-2 threshold is $2.50 (PPP) per day 34