POVERTY THE WORLD BANK REDUCTION AND ECONOMIC MANAGEMENT NETWORK (PREM) Economic Premise MARCH 2010 · Number 7 53745 The Impact of the Financial Crisis on Poverty and Income Distribution: Insights from Simulations in Selected Countries Bilal Habib, Ambar Narayan, Sergio Olivieri, and Carolina Sanchez As the financial crisis has spread through the world, the lack of real-time data has made it difficult to track its impact in developing countries. We use a microsimulation approach to assess the poverty and distributional effects of the crisis. In Bangladesh, Mexico, and the Philippines, we find increases in both the level and the depth of aggregate poverty. Income shocks are relatively large in the middle (and, in Mexico, the bottom) parts of the income distribution. We also find that characteristics of people who become poor because of the crisis are different from those of both chronically poor people and the general population. Our findings will be useful for policy makers wishing to identify leading monitoring indicators to track the impact of macroeconomic shocks and to design policies that protect vulnerable groups. As the financial crisis has spread through the world, its im- Aggregate Poverty and Inequality Impacts pacts on poverty and income distribution in developing countries have proved difficult to track because real-time The model predicts some increases in the level and depth data are lacking. Measuring the poverty and distributional of poverty as a result of the crisis in all three countries, with impacts of the crisis is important not only for designing ap- the extent of increase largely depending on the size of the propriate policy responses, but also for informing policies to macroeconomic impact in the country.1 In Bangladesh and protect poor people in the event of future economic shocks. the Philippines, the crisis has led to a slowdown in GDP Given that household data needed to measure impacts in real growth, which is expected to raise the poverty rate in 2010 time are not available in most countries, ex ante methods that by 1.2 and 1.5 percentage points, respectively, compared extrapolate impacts from pre-crisis data are useful tools. with what would have occurred without the crisis. The lower We build on the model described in Bourguignon, Bus- rate of poverty reduction resulting from the crisis translates solo, and Pereira da Silva (2008) and in Ferreira et al. (2008) to approximately 1.4 and 2.0 million additional poor people to simulate the impacts of the crisis in several countries. (See in the Philippines and Bangladesh, respectively, in 2010. In box 1 for details on the methodology.) For this purpose, we Mexico, GDP actually contracted by nearly 7 percent in make comparisons between scenarios with and without crisis 2009, and it is projected to grow by only 3 percent in 2010. for the same year, or in pre- and post-crisis years. Results for The cumulative impact is projected to raise the poverty rate Bangladesh, Mexico, and the Philippines are summarized in by nearly 4 percentage points in Mexico between 2008 and this note. 2010. 1 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise Box 1. Methodological Approach and Caveats To estimate how macroeconomic impacts would translate into a result of the crisis. We have applied this method to Bangla- lower incomes for households, we use a microsimulation approach desh, Mexico, and the Philippines; and similar exercises are on- that superimposes macroeconomic projections on behavioral mod- going for Egypt, Mongolia, and Poland. We find that the crisis els built on pre-crisis household data. The macroeconomic shocks has had important impacts on poverty levels and depth and on are transmitted to households through the labor market in the form income distribution. of loss of (or cross-sectoral shifts in) employment and labor earn- A number of caveats apply to these exercises. First, the countries ings, and loss of nonlabor income (including remittances from mi- studied may not be representative of the global impact. Specif- grants). The model is loosely based on the approaches described ically, because the model focuses on labor income and interna- in Bourginon, Bussolo, and Pereira da Silva (2008) and Ferreira tional remittances as transmission channels of impact, it may et al. (2008), but simplified in the sense that it is not linked to a not fully capture impacts in countries that are heavily dependent general equilibrium model that is not readily available for most on commodity exports (for example, some African countries) or developing countries. Instead, the simplified model takes the those where wealth shocks play a large role (for example, some macroeconomic projections for a specific country and year as countries in Eastern Europe). However, we do find that results given, and extrapolates the microeconomic snapshot of future im- from other simulation exercises, such as for Latvia, are in line pacts from these projections. with the ones presented here. To measure impacts of the crisis, comparisons are made either Second, the fact that the microsimulations are based on past between scenarios with and without crisis for the same year, or data that reflect the preexisting structure of labor markets and between pre-crisis and post-crisis years. Beyond impacts on household incomes necessarily implies that these structural re- poverty measures, the results also shed light on how the impacts lationships are assumed to remain constant over the time period are distributed among sectors, regions, and income groups; and for which the projections are made. what are the characteristics of those who would become poor as In contrast, the crisis has no significant impact on aggre- Impacts on the middle of the distribution can be attrib- gate inequality indexes in any of the three countries. The ag- uted to significant employment shocks to urban areas and gregate numbers, however, mask larger changes in the particularly the manufacturing and export sectors that em- underlying distribution of income and consumption. Some ploy a large number of workers in middle-income house- regions and income groups within each country suffer more holds. Urban households suffer more losses, on average, than losses than others, depending on which sectors and income do rural households in all countries because the manufactur- sources are more likely to be affected in the country. ing sector that bears the brunt of the losses in employment and labor earnings typically is concentrated in urban areas. Distributional Impacts: Who Is Affected The urban-rural gap is widest in the Philippines (with a 6 and How percent decline in the per capita income of urban house- holds, compared with a 3 percent decline for rural house- In all three countries, the impacts are relatively large in the holds). In contrast, impacts among the poor and those in rural middle (and, in Mexico, the bottom) parts of the income areas are more closely associated with lower (international) distribution.2 Between 15 and 20 percent of households in remittance flows and a decline in agricultural income. the fourth to seventh deciles of the income distribution in The significant changes predicted for the middle of the Mexico and the Philippines suffer per capita income losses distribution do not lead to higher inequality in Bangladesh that push them to a lower income decile.3 This number is and the Philippines, as measured by aggregate indexes such 10 percent in the case of Bangladesh. In Mexico, where the as the Gini coefficient. This is because overall inequality crisis has been more severe, significant impacts are also likely worsens only when the losses are disproportionately high among those at the bottom of the distribution. The poorest among the poor. Only in Mexico, where a larger shock is 20 percent of Mexican households suffers an average loss in projected at the lower end of the distribution, is there a pro- per capita income of about 8 percent, compared with 5 per- jected increase in aggregate inequality. cent for the entire population--even after existing safety- Beyond these common patterns, there is considerable net transfers that benefit many of the extremely poor people variation in how impacts are distributed within countries are taken into account. because the relative importance of different transmission 2 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise Figure 1. Changes in Household Incomes between No-Crisis and Figure 2. Growth Incidence Curves: Percent Change in per Capita With-Crisis Scenarios, 2010 Income/Consumption between No-Crisis and With-Crisis Scenarios a. Entire population a. Bangladesh 0 0 ­2 ­1 ­4 ­2 percent change ­3 ­6 ­4 ­8 Bangladesh Mexico Philippines ­5 country ­6 a b. Crisis-vulnerable or newly poor households ­7 0 ­8 ­20 0 10 20 30 40 50 60 70 80 90 100 percentile ­40 ­60 b. Mexico Bangladesh Mexico Philippines 0 country ­2 labor income nonlabor income ­4 percent change Source: Authors' calculations. ­6 Note: For Mexico, the change is from pre-crisis to post-crisis (2008 to 2010). ­8 a. Households projected either to fall into or to remain in poverty (rather ­10 than exit from it) as a consequence of the crisis. ­12 ­14 channels of impact varies from country to country. Approx- ­16 imately 90 percent of the average loss in per capita house- hold income in Mexico and the Philippines results from a 0 10 20 30 40 50 60 70 80 90 100 fall in labor income, compared with 50 percent of the aver- percentile age loss in Bangladesh (where losses in remittances are equally important) (figure 1a). Losses resulting from lower c. Philippines remittances tend to be more concentrated among better-off 0 households within urban and rural areas than losses in labor ­1 income, mainly because the initial (pre-crisis) distribution ­2 percent change of remittances was skewed toward the better-off households ­3 at the beginning. The combination of these factors generates distributional ­4 patterns that are different across urban and rural areas in dif- ­5 ferent countries. The Philippines is characterized by a high ­6 level of urbanization and a concentration of manufacturing and service sectors in urban areas, along with a relatively lim- ­7 ited role for remittances (figure 2). These translate into ­8 much larger income losses for urban areas, compared with 0 10 20 30 40 50 60 70 80 90 100 rural areas, but a relatively equitable distribution of losses percentile within each area. In contrast, significant distributional changes within urban and rural areas of Bangladesh and urban rural Source: Authors' calculations. Mexico are predicted, but with important differences (figure Note: For Mexico, the change is from pre-crisis to post-crisis (2008 to 2). In Bangladesh, relatively better-off households suffer the 2010); for the other countries, the change is from no-crisis to with-crisis greatest losses, mainly because of lower remittances resulting scenarios for the same year (2010). 3 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK www.worldbank.org/economicpremise from the crisis--with the largest losses predicted for the top because of the influence these groups can have on public 40 percent of the urban and rural areas alike. In Mexico, the perceptions and, consequently, on policy choices. losses are high among the poor because of a substantial pro- jected fall in labor income. Notes The "Crisis Vulnerable": A New Challenge 1. These results are consistent with the impact of the crisis on global poverty estimated by Chen and Ravallion (2009). To understand who are affected most by the crisis, it is useful 2. This is not surprising, given the literature on the vul- to examine the characteristics of households that are pro- nerability of the "middle class." For example, Ravallion jected to be poor as a result of the crisis, but who otherwise (2009) argues that middle-income groups in developing would not have been poor. These "crisis-vulnerable" or countries are more exposed to crisis shocks than is the rest "newly poor" households are quite different from chronically of the population. poor households and from the general population. On aver- 3. Each decile is equivalent to 10 percent of the country's age, the newly poor appear to be more skilled and urban population, when households are ranked in terms of per than the chronically poor, but less so than the general pop- capita income or consumption. The bottom and top deciles ulation. They also are more likely to be economically active are the poorest and richest 10 percent of the population, re- than the chronically poor, indicating that the crisis would spectively. have had a sizable effect on the number of "working poor." As shown for all three countries in figure 1.b, newly poor About the Authors households suffer much larger income losses (25­50 per- cent) than average households (3­5 percent). The sources All authors are part of the Poverty Reduction and Equity Group of these income losses are significant: in Bangladesh, a loss in PREM. Ambar Narayan and Carolina Sanchez are senior in remittances appears to be a good leading indicator of economists in the unit; Bilal Habib and Sergio Olivieri are poverty; in Mexico and the Philippines, "new" poverty is consultants. To learn more about the Poverty Reduction and Eq- driven almost entirely by losses in labor income (figure 1b). uity Group, please visit http://www.worldbank.org/poverty. References Relevance for Policy Bourguignon, F., M. Bussolo, and L. Pereira da Silva. 2008. "Introduction: The results have a number of potential uses for policy design Evaluating the Impact of Macroeconomic Policies on Poverty and In- and monitoring. First, they help identify a list of possible come Distribution." In The Impact of Macroeconomic Policies on Poverty "leading indicators" that can be monitored rapidly to gauge and Income Distribution, ed. F. Bourguignon, M. Bussolo, and L. Pereira the likely welfare impacts of a crisis, in the absence of real- da Silva, 1­23. Washington DC: World Bank. Chen, S., and M. Ravallion. 2009. "The Impact of the Global Financial Crisis time information on household income/consumption. Al- on the World's Poorest." World Bank Development Research Group. though the appropriate choice of indicators would depend http://www.voxeu.org/index.php?q=node/3520. on the main transmission channels in a specific country, Ferreira, F., P. Leite, L. Pereira da Silva, and P. Picchetti. 2008. "Can the Dis- manufacturing employment, wages, remittance flows, and tributional Impacts of Macroeconomic Shocks Be Predicted? A Com- relative food prices emerge as potential candidates for mon- parison on Top-Down Macro-Micro Models with Historical Data for itoring in all three countries. Second, the identification of Brazil." In The Impact of Macroeconomic Policies on Poverty and Income Distribution, ed. F. Bourguignon, M. Bussolo, and L. Pereira da Silva, the people likely to suffer the most losses can inform the de- 119­75. Washington DC: World Bank sign of policy responses. In this context, the fact that the Habib, B., A. Narayan, S. Olivieri, and C. Sanchez-Paramo, 2010a. "Assessing newly poor and the chronically poor differ significantly in ex ante the Poverty and Distributional Impact of the Global Crisis in a their characteristics suggests that expanding existing safety- Developing Country: A Micro-Simulation Approach with Application net programs (targeted toward the chronically poor) to mit- to Bangladesh." Policy Research Working Paper 5238, World Bank, Washington, DC. igate the losses of the newly poor may not be effective or ------. 2010b. "Assessing ex ante the Poverty and Distributional Impact practical. Instead, (new) interventions that address increased of the Global Crisis in the Philippines: A Micro-Simulation Approach." levels of vulnerability and that protect households against Unpublished manuscript. Poverty Reduction and Equity Group, World risk may be required. Bank, Washington, DC. The results also can have useful political economy impli- Ravallion, M. 2009. "The Developing World's Bulging (But Vulnerable) cations: the substantial losses suffered by the urban middle- Middle Class." Policy Research Working Paper 4816, World Bank, Wash- ington, DC. income and poor households, for example, are important The Economic Premise note series is intended to summarize good practices and key policy findings on topics related to economic policy. It is produced by the Poverty Reduction and Economic Management (PREM) Network Vice-Presidency of the World Bank. The views expressed here are those of the authors and do not necessarily reflect those of the World Bank. The notes are available at www.worldbank.org/economicpremise.