Poverty and Inequality Monitoring: Latin America and the Caribbean Safeguarding Against a Reversal in Social Gains During the Economic Crisis in Brazil Executive Summary  Brazil has accomplished impressive reductions in poverty and inequality between 2004 and 2014 as a result of rapidly growing formal employment, higher real wages and redistributive social assistance programs such as Bolsa Família. With labor income as the major source of income of the poor and vulnerable households, the current economic crisis poses a serious threat to the sustainability of the gains in poverty and inequality reduction. As in the 2008-2009 financial crisis, Brazil’s social assistance and safety net system has a critical role in safeguarding the social gains achieved so far by preventing more Brazilians from falling into poverty. Yet the expansion of the budget for the social safety net system is hampered by the challenging fiscal consolidation environment in Brazil.  This note summarizes the findings of the analysis carried out regarding the poverty and inequality impacts of the ongoing economic crisis in Brazil in 2016 and 2017. The first objective is to get an estimate of the extent to which the deteriorating macroeconomic conditions and shrinking labor markets in Brazil will impact on poverty and inequality. The second objective is to generate a detailed profile of the “new poor� associated with the crisis. The third and final objective is to get estimates of the additional budget needed for the Bolsa Família Program to effectively mitigate the poverty impacts of the crisis and protect the past achievements of Brazil in poverty and inequality reduction until the necessary structural reforms take place in Brazil and the engines of growth are reignited.  The analysis is repeated for two scenarios regarding the changes in real GDP in 2016 and 2017 for the purpose of providing a sufficiently narrow zone for policy decisions anticipating the adverse impacts of the crisis on poverty. The distributional impact of each scenario is evaluated first under the assumption of no changes in the real budget of Bolsa Família and second after allowing for an increase in the real Bolsa Família budget, which allows for an increase in the coverage of the program to the new poor based on the program’s current level of “real� benefits and eligibility rules.  The results of the microsimulation analysis suggest that indicators about inequality and poverty will increase in 2016 and remain high in 2017. In scenario 1, the Gini index measuring inequality is predicted to increase from 0.515 to 0.522 in 2017 as is the poverty headcount ratio (at the R$140 poverty line) which will increase from 8.7% to 9.8% representing an increase in the number of poor by 2.5 million people. In the more pessimistic scenario 2, the Gini increases to 0.524 in 2017 and the higher poverty headcount ratio of 10.3% in 2017 represents an increase in the number of poor by 3.6 million people. However, poverty rates will increase mainly in the urban areas and less in the rural areas (where poverty rates are higher to begin with). The analysis also reveals that the people falling below the poverty line as a consequence of the crisis are likely to be slightly younger in age, skilled, located in urban areas, located in the southeast, previously working in the service sector, and white.  The depth and duration of the current economic crisis in Brazil gives rise to the opportunity to expand the role of Bolsa Família from an effective redistribution program to a true safety net program that is sufficiently flexible to expand its coverage to the “new poor� households generated by the crisis. The analysis in this note suggests that an increased budget (in real terms) of about 4.73% (R$1.25 billion) and 6.9% (R$1.82 billion) from the 2015 budget of R$26.4 billion, would be a very effective way of targeting scarce financial resources to the most needy among the “new poor� households generated by the crisis. In nominal terms, or in 2017 Reais, the estimate of the budget required in 2017 can be calculated by multiplying the 2015 budget (R$26.4 billion) with the estimated increase in the “real� Bolsa Família budget (e.g., 1.0473 under scenario 1) and the expected inflation rate between 2015 and 2017 (e.g., 10% or 1.10). Using these specific values, the estimate of the nominal budget required in 2017 is R$30.41 billion.  The distribution of the additional Bolsa Família budget to the newly eligible households among the “new poor� can prevent the extreme poverty rate in Brazil from increasing beyond the level of 2015, though its impact on preventing the overall poverty rate from increasing is not as dramatic. It is important to bear in mind that the preceding estimates of the additional budget needed for Bolsa Família are derived based on the program’s current level of real benefits and eligibility rules and assuming annual adjustments in the nominal budget in par with the annual inflation rate so at to maintain the purchasing power of the benefits constant over time. Delays in adjusting the nominal value of the transfers of Bolsa Família in par with the prevailing inflation rate are likely to lead to higher poverty rates than those estimated in this note (all else equal). As the Brazil Systematic Country Diagnostic (2016) highlights, in spite of the limited fiscal space in the medium run, there is ample scope to expand funding for the most progressive elements of social policy, through reallocations from poorly targeted social transfers and through improvements in the efficiency of spending. 1. Background and Motivation1 determinant of poverty reduction and shared prosperity. For the poorest Brazilians, however, social B etween 2004 and 2014, more than 28.6 transfers have been more important than labor million Brazilians have escaped poverty. markets in the past decade. Fifty eight percent of the Yet, Brazil remains one of the most decline in extreme poverty in Brazil between 2004 unequal countries in the world. The reduction in and 2014 was due to changes in non-labor income poverty is an achievement of regional significance, (mainly transfers from the Bolsa Família conditional representing almost 50 percent of the reduction in cash transfer program) (see Figure 2). poverty in the whole Latin American and Caribbean (LAC) region (Figure 1). Brazil also experienced a Figure 1: Progress in Poverty and Inequality rapid decline in inequality over the past decade, Reduction in Brazil with the Gini coefficient of household incomes falling from 0.57 to 0.52 in 2014. To a large extent, it was due to a policy of social inclusion in the context of a booming economy, fueled by favorable external conditions. Brazil’s achievements were also of historical significance, in that it was the first time in the history of Brazil that a sustained reduction in poverty and inequality had been achieved. Nevertheless, even after the reduction in poverty and inequality, Brazil remains one of the most unequal countries in the world, with a Gini coefficient higher than in most countries except Colombia and Honduras in Latin America and Caribbean and a few countries in sub-Saharan Africa. Source: Calculations based on the National Household Labor markets drove shared prosperity, while Sample Survey (PNAD - Pesquisa Nacional de Amostra de transfers helped reduce extreme poverty. The Domicílios) 2004–2015. road to prosperity for the majority of poorer Brazilians has been through a formal sector job. In Favorable external conditions have played a this regard, Brazil is similar to other middle-income critical role in shaping labor market outcomes in countries, where labor earnings represent the Brazil. The commodity price boom prompted largest share of income among the B40, and hence significant real exchange rate appreciation and this the performance of the labor market is a key in turn encouraged the growth of non-tradable 2 domestic services. Rising job opportunities for low- poverty rate in 2015 increasing to 8.7% (from 7.4% skilled workers in these sectors led to rising incomes, in 2014). At the same time, inequality as measured which in turn fed back into growing demand for by the Gini index, seems to have stabilized between goods and services such as housing, durable goods, 2014 and 2015 at 0.52 points (see Figure 1). The and retail or transportation. fast increase in unemployment and the fall in real wages is likely to lead to rising poverty in 2016 and Figure 2: Sources of Reductions in Poverty, possibly in 2017. Extreme Poverty and Inequality, 2004-2014 Given the depth of the recession, Brazil’s social safety net system can play a key role in preventing more Brazilians from falling into poverty. Increases in the budget for social assistance and in particular for the Bolsa Família program can be instrumental for avoiding more severe losses to the social gains achieved in the last decade. Yet the expansion of the budget for the social safety net system is hampered by the challenging fiscal consolidation environment. This policy note summarizes the findings of the analysis carried out regarding the poverty and inequality impacts of the ongoing economic crisis in 2016 and 2017. The first objective is to get an Source: Calculations using changes in poverty and changes in estimate of the extent to which the deteriorating income by source between PNAD 2004 and 2014 Note: Following Brazil's legal age for adulthood, the macroeconomic conditions and the weak labor component "share of adults" refers to members of housheholds markets in Brazil will impact on poverty and of ages 18 and above. inequality in 2016 and in 2017. The second objective is to generate a detailed profile of the “new poor� As of the second half of 2015, Brazil is in the associated with the crisis. The third and final midst of a deep recession, with the economy objective is to get estimates of the additional budget contracting 3.8 percent in 2015 and a similar needed for the Bolsa Família Program to effectively contraction expected in 2016 (with GDP mitigate the poverty impacts of the crisis and protect contracting by 3.4 percent). The budget deficit the past achievements of Brazil in poverty and exceeded 10 percent of GDP last year, and is inequality reduction until the necessary reforms take projected to remain high in the context of political place in Brazil and the engines of growth are infighting preventing fiscal reforms. Inflation is well reignited. above the target range, reflecting increases in regulated prices and the pass-through from currency depreciation. 2. Methodology With labor income as the major source of income of the poor and vulnerable households there is a The microsimulation model employed for the serious threat to the sustainability of the gains in analysis of the poverty and distributional impacts poverty and inequality reduction. The recession of the crisis in Brazil, is based on the way resulted in a loss of 1.6 million formal sector jobs in macroeconomic shocks are transmitted to the labor 2015. Consequently, unemployment has surged, market through losses in employment and lower from 4.3 in December 2014 to 11.8 percent in labor earnings.2 The model combines macro level October 2016. Real wages are also contracting, information on the projected growth of output, with the average monthly real wage falling 4.2 employment, population and the labor force percent in 2015. Estimates from the recently participation, with micro level information on labor released 2015 PNAD, collected in October 2015, force status, sector of employment, labor and non- reveal that the trend in the decline of poverty has labor income, and basic job characteristics. reversed with extreme poverty increasing to 3.4% in 2015 (from 2.8% in 2014) and the overall 3 The microsimulation model involves three basic means of a earnings function (Mincer, 1974). The steps: (1) the estimation of a baseline model; (2) model parameters are estimated based on published simulation analysis; and (3) impact assessment data on macroeconomic variables for the baseline (see Figure 3 for a visual summary). The first step year 2015, and on household level data from the uses household- and individual-level information to 2015 PNAD, officially released to the public domain model labor market behavior. Labor force and in November 2016. After removing observations with employment status are divided into five categories missing income and non-usual household members in total, and are modeled as functions of household (such as maids and their family members), the number and individual characteristics.3 Parameters are of observations used for this analysis is roughly estimated by means of a multinomial logit model 350,000. The ADePT simulation software used for occupational choice as in Ferreira et al. (2008). ensures the consistency among the macro variables in Labor earnings for all employed individuals are the baseline year and the individual aggregates then modeled as a function of individual and job from the micro part of the model. characteristics, and parameters are estimated by Figure 3: The Basic Steps in the Microsimulation Model Source: Authors’ work based on Olivieri et al. (2014). The projected values of the main macroeconomic period in the agricultural and service sectors (by and demographic variables in the model are key 2.5% and 2.8%, respectively). Between 2016 and ingredients for the simulation analysis. The 2017, it is expected that positive growth rates will growth rate in GDP between 2015 and reappear for the economy overall and each of the 2016/2017 in the country overall and in each of three key sectors. In scenario 2, which is more the three sectors is obtained from the Fall 2016 pessimistic, it is assumed that the decline in real issue of the Brazil Macro-Poverty Outlook (MPO) GDP between 2015 and 2016 will the slightly of the World Bank (see Table 1). Specifically, the greater than in the baseline scenario 1, i.e., GDP Brazil MPO forecasts that real GDP for the declining by 3.7%, instead of -3.4%, whereas the Brazilian economy overall will decline by 3.4% decline in GDP between 2016 and 2017 continues between 2015 and 2016 and increase by 0.5% to be negative at -1.0%, instead of the increase of between 2016 and 2017. The MPO also predicts 0.5% in the base case scenario 1. In scenario 2, the that the manufacturing sector will shrink at an sector-specific declines in GDP are derived by annual rate of 3.6% between 2015 and 2016, as rescaling the change between 2015 and 2016 well as a decline in real GDP over the same and between 2016 and 2017. 4 Table 1. Projected annual GDP change in scenario 1 and 2. Scenario 1 Scenario 2 2015-2016 2016-2017 2015-2016 2016-2017 Total -3.4% 0.5% -3.7% -1.0% Agriculture -2.5% 0.8% -2.7% -1.6% Manufacturing -3.6% 0.8% -3.9% -1.6% Service -2.8% 0.6% -3.0% -1.2% Source: World Bank staff estimates. Based on the parameters estimated for the household non-labor income, which is held constant baseline model with 2015 data, the second step in real terms. The new per capita household income simulates the process by which the projected also helps to gauge the increase in the Bolsa changes in the macroeconomic variables for 2016 Família budget required to mitigate the poverty and 2017– such as projected changes in and inequality impacts of the crisis. aggregate GDP, aggregate employment, the labor force participation rate– are translated into The poverty impacts of the crisis are estimated changes in the employment status (and in the sector based on the unofficial poverty lines of R$70 of employment), and changes in the labor income and R$140 (in June 2011 prices). The at the individual or micro level for 2016 and Government of Brazil measures poverty rates using 2017. Specifically, total employment is estimated the administrative lines of R$70 per capita per to change by -1.7% in 2016 and by 0.6% in month and R$140 per capita per month (in June 2017, by applying the sector specific employment- 2011 prices) based on the thresholds used to output elasticities of 0.01 for agriculture, 0.69 for determine eligibility in the Brasil Sem Miséria Plan manufacturing, and 1.16 for services on the and the Bolsa Família Program. In April 2014, the predicted changes in real GDP in each sector, Government of Brazil revised the eligibility respectively. Combined with an assumption on how threshold to R$77 for extreme poverty and R$154 the aggregate labor force participation rate may for poverty, while as of June 29, 2016, the change with the decline in GDP, the aggregate eligibility threshold for extreme poverty is R$85 unemployment rate is derived as a residual. At the and R$170 for poverty. In spite of the recent micro level, all working-age individuals are increase in the threshold for eligibility, the R$70 reassigned across alternative occupations or and R$140 thresholds (adjusted for inflation) unemployment to match the projected aggregate continue to be used as the implicit poverty lines for employment changes in specific economic sectors the estimation of poverty in Brazil and as the and total employment at the national level. For this yardsticks for monitoring of the Brasil sem Miseria purpose, the occupational choice model estimated plan.4 In addition to the poor the vulnerable are in the first step for this computation is used. Once defined as those with incomes between R$141 and individuals are reassigned to new occupational R$290.5 status in 2016 or 2017, his/her new income is The microsimulation analysis consisting of the estimated based on the parameters of the baseline three main steps summarized above are earnings model. repeated for different projected values (under The third and final step of the microsimulation each scenario) of the main macroeconomic and exercise is to assess the poverty and distributional demographic variables between 2016 and 2017. impact of the crisis by generating the new income This note summarizes the poverty and distributional distribution associated with the projected changes impacts associated with two macroeconomic of the key macroeconomic aggregates. The “new� scenarios regarding the projected declines in income of individuals and/or households is employment and increases in unemployment in the calculated by adding up the individual “new� context of declining aggregate economic activity labor income estimated in step two above and (see Table 2 below). 6 During the last decade, per 5 capita GDP and per capita household income have or the labor participation rate, in 2015 and 2016 been growing along similar paths. It is conceivable is also based on PNAD Continua. It is assumed that that the current economic crisis breaks this pattern. the labor participation rate in 2016 and 2017 will be approximately the same compared to that in The population size in 2016 and 2017 is 2015 (actually only 0.1% higher) after accounting adjusted based on the actual rate of population for growth in the population from year to year change between 2015 and 2016 (using PNAD and discouraged worker effects associated with Continua). The fraction of the working age the crisis.7 population (16 years and older) in the labor force, Table 2. Comparison Between Scenario 1 and Scenario 2 Employment change since 2015 Scenario Year Annual Unempl. Labor Total Agriculture Manufacturing Service GDP Rate Force Change Particip. Rate 1 2016 -3.4% 11.2% 63.6% -1.7% -1.6% -9.6% -0.2% 2017 0.5% 11.8% 63.6% 0.6% 0.0% 0.6% 0.7% 2 2016 -3.7% 11.2% 63.6% -1.7% -1.6% -9.6% -0.2% 2017 -1.0% 13.3% 63.6% -1.2% -0.0% -1.1% -1.4% Source: World Bank staff estimates Note: Numbers in grey cells are set exogenously whereas numbers in cells with no shade are determined as residuals. The distributional impact of each one of these as pressures at the lower part of the wage scenarios is evaluated first under the assumption distribution ease off during the crisis period. To the of no change in the budget of Bolsa Família and extent there is a drift back to informality and the second after allowing for an increase in the Bolsa minimum wage becomes less binding, the same skill Família budget and the coverage of the program. distribution in the population during the crisis period The poverty and distributional impacts under these may be associated with very different relative two different scenarios without and with changes in earnings outcomes in 2016-2017 than in 2014, and the Bolsa Família budget provide a sufficiently possibly higher poverty and inequality. Taking into narrow zone for policy decisions anticipating the account these caveats associated with the adverse impacts of the crisis on poverty. microsimulation model, the estimates of poverty and inequality discussed for scenarios 1 and 2 may Caveats actually provide only a lower bound of the poverty As already mentioned, the microsimulations rely on a and inequality impacts of the crisis. In addition, the number of untested assumptions necessary to make model assumes that the factors of production (labor microeconomic data consistent with macroeconomic and capital) are immobile across space, and that the projections.8 Key among these is the assumption that rate of change in GDP is the same as the rate of the structural relationships summarized by the change in household income. Finally, prices are held parameters of the regression equations estimated in constant throughout the analysis.10 the baseline year 2015 remain unchanged in 2016 and 2017. In other words, the functional relationships that determine either employment in a 3. Poverty and inequality impacts specific sector or the wage earned by an individual According to the results of the microsimulation are assumed to be remain unaffected by the analysis, poverty headcount ratios will rise in continuation of the crisis in 2016 or the turnaround in 2016 and remain high in 2017 (see Figure 4). In real GDP growth in 2017. This assumes, for instance, scenario 1, the number of extreme poor will increase that the rather unusual decline in the skilled to by 1.5 million people, from 6.8 million in 2015 to unskilled wage premium observed between 2002 8.3 million in 2016 (and to 8.5 million in 2017), and 2014 continues during the crisis period.9 It is raising the extreme poverty headcount ratio from conceivable that the skills premium increases again 6 3.4% in 2015 to 4.1% in 2016 and to 4.2% in in 2017, which represents an increase in the number 2017. The number of moderate poor will increase of extreme poor between 2015 and 2017 by 2.6 by 2.3 million from 17.3 million in 2015 to 19.6 million people, whereas the higher moderate million in 2016 (and to 19.8 million in 2017). This poverty headcount ratio of 10.3% in 2017 pushes up the poverty headcount ratio from 8.7% to represents an increase in the number of moderate 9.7% in 2016. In the more pessimistic scenario 2, the poor by 3.6 million people (between 2015 and continuing rise in the extreme poverty rate results in 2017). a higher extreme poverty headcount ratio of 4.6% Figure 4. Poverty headcount ratio Scenario 1 Scenario 2 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) However, poverty rates will increase more in the level in 2017, rural areas in 2017 will have urban areas and less in the rural areas (Figure 5). headcount ratios only slightly higher than the ones as In contrast to urban areas where poverty headcount of 2015. ratios will rise in 2016 and remain at the higher Figure 5. Poverty headcount ratio in urban and rural areas Scenario 1 Scenario 2 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014). The rise in poverty during the crisis in Brazil will the predicted increase in inequality appears to be be accompanied by an increase in income independent of the index of inequality used (e.g., inequality in the country (see Figure 6). Moreover, Gini index or Theil index).11 7 Figure 6. Income inequality Scenario 1 Scenario 2 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014). 4. Transition Matrices and the Profile of the composed of the exact same households in the “New Poor� bottom part of the distribution of income in 2015. Many of the new members of the lower part of the As a consequence of the crisis, households at the income distribution in 2016/2017 are individuals lower part of the distribution of income in either who in 2015 reported not receiving any non-labor 2016 or 2017 will have lower income compared income (including cash transfers from social to households at the lower part of the income assistance programs such as Bolsa Família). With the distribution in 2015. Irrespective of the crisis continuation of the crisis in 2016 these individuals scenario analyzed, the Growth Incidence Curves lose their jobs and thus their primary source of (GIC) in Figure 7 below reveal that households income, i.e., income from labor.12 As the analysis located at the lower part (bottom 10%) of the below demonstrates, an increase in the budget of distribution of income in 2016 or 2017 will have the Bolsa Família program that would allow for an significantly lower income than households at the increase in the program’s coverage, could be a very corresponding part of the income distribution in effective means of mitigating the adverse impacts of 2015. This is even more apparent in the growth the crisis on income and welfare. incidence curves obtained under the more pessimistic scenario 2. As a result of the crisis, the bottom part of the distribution in 2016 or 2017 is not necessarily Figure 7: Growth Incidence Curves Scenario 1 Scenario 2 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) Note: The horizontal axis is the percentile of per capita household income as of 2016 or 2017. Tables 3a and 3b provide a more detailed picture which the crisis reaches its peak, 11.8 million people of the transitions of individuals into and out of will move one or more steps down the ladder by poverty and vulnerability between 2015 and 2016 (12 million in scenario 2), whereas only 0.2 2016/17. Between 2015 and 2016, the year in million people will move up (Table 3a). 8 Table 3a: Transitions into and out of Poverty and Vulnerability (2015 vs 2016) Scenario 1 Scenario 2 0.2 million move up 0.2 million move up 11.8 million move down 12.0 million move down Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) Note: Numbers in table are the numbers of individuals in thousands. Between 2015 and 2017, the differences between vulnerable status in 2015, 2.9% (0.840M) are scenario 1 and the more pessimistic scenario 2, are estimated to fall into moderate poverty in 2017. more apparent (Table 3b). In scenario 1, 5.8 million Under either scenario, the majority of the people people will move one or more steps down the estimated to fall into extreme poverty in 2017 are income ladder by 2017, whereas hardly anyone will originating from an income level in 2015 that is move up the income ladder. The lower number of above the poverty threshold of R$140. people moving down the income ladder in scenario 1 compared to the number of people moving down The microsimulation analysis is also able to shed the income ladder between 2015 and 2016, is due light on the profile of the “new poor� during this to the fact that by 2017, scenario 1, predicts a economic crisis. Based on the poverty status, using modest increase in real GDP. Among the extreme the R$140 poverty threshold, at the baseline and poor as of 2015, 99.98% will remain in extreme projected years, individuals can be classified into: 1) poverty in 2017, while 3.1% of moderate poor new poor, i.e., individuals who were not poor in (0.322M), 2.0% of the vulnerable (0.656M), and 2015 but become poor in 2017; 2) structurally poor, 0.5% of middle-class (0.836M) are estimated to fall i.e., individuals who were poor in 2015 and remain into the extreme poverty status. Among those in in poverty in 2017; and 3) non-poor, i.e., households who were not poor in 2015 nor in 2017. Table 3b: Transitions into and out of Poverty and Vulnerability (2015 vs 2017) Scenario 1 Scenario 2 0.0 million move up 0.0 million move up 9 5.8 million move down 7.9 million move down Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) Note: Numbers in table are the numbers of individuals in thousands The analysis suggests that the people falling the “new poor� is almost as high as in non-poor. This below the poverty line are likely to be younger in implies that the current economic crisis will push into age, skilled, located in urban areas, previously poverty skilled people who would otherwise be working in the service sector, and white. Figure 8 above the poverty threshold. A similar story shows how the characteristics of household heads emerges when comparing the race of the “new vary across the new poor, structurally poor, and poor� and the structurally poor. The share of whites non-poor households, with scenario 2 (the profile is larger among the non-poor, compared to the was practically identical for scenario 1). The structurally poor. However, the “new poor� are more household heads classified as structurally poor are likely to be white than the structurally poor. The almost 9 years younger than household heads analysis also reveals the “new poor� are likely to be classified as non-poor, and the households classified located in urban areas. In a similar vein, the “new as “new poor� are three years younger than the poor� are estimated to be located primarily in the structurally poor household heads. A wider gap is Southeast while a smaller fraction of them are evident between the “new poor� and structurally located in the Northeast, where most of the poor in the proportion of skilled people. A structurally poor are located (see Figure 9).13 comparison of the share of skilled people between Finally, the majority of the “new-poor� in 2017 structurally poor and non-poor clearly indicates that consists of individuals who in 2015 were working in the structurally poor tend to be low-skilled people. the service sector (see Figure 10). However, the share of skilled people in the pool of Figure 8: Characteristics of household head and Poverty Status in 2017 (Scenario 1) Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) 10 Figure 9: Region of Residence (in 2015) and Poverty Status in 2017 (Scenario 1) Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) Figure 10: Occupational Status in 2015 and Poverty Status in 2017 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) 5. Mitigating the Poverty Impacts of the Crisis Brazil initiated these programs years ago, and other through Bolsa Família LAC countries have followed suit. Between 2000 and 2013 at least 18 countries in the region introduced Social assistance in Brazil consists of three main inclusive reforms, which sought to increase coverage pillars. First, disability benefits provide transfers to of the elderly (Rofman et al. 2014). older or disabled people known as Benefício de Prestação Continuada (BPC); second, the inclusion of Chronic poverty is addressed through the Bolsa self-employed or agricultural family workers into Família Program (PBF), the flagship CCT program social insurance institutions, such as the Rural Pension of the MDS. It provides cash transfers to poor Program (Previdência Social Rural or PSR); and third, households, conditional on school attendance and use targeted income support, such as the Bolsa Família of maternal and child health services. The program CCT program. The benefits of the social assistance was brought to scale at remarkable speed with the programs for poverty prevention in old age in Brazil number of beneficiaries going from 3.6 million to are received primarily by low-income workers, both 11.1 million families in four years (see Figure 11). As rural and urban, who move in and out of informality of 2014, the program reaches about 56 million during their working lives (Gragnolati et al. 2013). individuals or 14 million households— around a 11 quarter of Brazil’s population. Spending as a the global financial crisis 2008-2010. As Figure 11 percentage of GDP increased from less than 0.05 below highlights, the increase of the program’s percent of GDP in 2003 to over 0.5 percent in budget and coverage of eligible families during the 2013, with the increases in spending since 2011 during the global financial crisis of 2008 to 2010 mostly due to increases in the amount of benefits. contributed to the continued decline of poverty and inequality in Brazil (see Figure 1) in spite of a Increases in the Bolsa Família budget have been decline (-1.2%) in GDP per capita in 2009. an effective means of addressing the impacts of Figure 11: Bolsa Família Beneficiaries and Spending Source: MDS and World Bank LAC Social Protection Database. Changes in non-labor income received from cash budget is calculated based on the simulated changes transfer programs such as the Bolsa Família in per capita household income. In 2015, a family program are simulated by appropriate with a per capita monthly income less than R$154 is assumptions regarding eligibility criteria and eligible to receive the benefits of Bolsa Família. The targeting accuracy. Unfortunately, the 2015 PNAD amount of payment depends on the levels of per includes no specific information on the benefits capita income and the number of children, received by Bolsa Família beneficiaries. Thus the summarized in Table 4 below. marginal change in the total amount of Bolsa Família Table 4. Bolsa Família Benefit Structure in 2015 Per capita income Per capita income below R$77 R$77 – R$154 R$77 per family Yes R$35 per child (age<5) up to 5 children Yes Yes R$42 per teenager (age 15-17) up to 2 teenagers Yes Yes Additional benefit until per capita income reaches R$77 Yes Source: MDS (2015). 12 Table 5 shows how the eligibility of a family for group they can receive only payments for children Bolsa Família benefits can change between 2015 and teenagers since their per capita income in 2017 and 2017 as a consequence of the crisis. The first is estimated to be greater than R$77. The remaining group of families, denoted as (1) in Table 5, has per three groups of families (4) through (6) will lose their capita income between R$77 and R$154 in 2015 eligibility for some components of (or the whole but is projected to have income less than R$77 in package) of Bolsa Família benefits. 2017. These families become newly eligible for the basic R$77 payment and the amount needed to fill Based on the results from the microsimulations, it the gap if its per capita income remains under R$77. is possible to obtain estimates of the number of The second group, denoted by (2), consists of the families in each of the six groups in Table 5. families who were not eligible for Bolsa Família in Assuming that benefits will be provided to all 2015 but are predicted to have per capita income families that satisfy the criteria in Table 4, it is less than R$77 in 2017. These families will be able possible to calculate the additional budget to receive the full Bolsa Família package. The third (marginal budget) required by the Bolsa Família group is the families who will also become newly Program to cover the households affected in the eligible for Bolsa Família, but unlike the second years 2016 and 2017. Table 5. Changes in eligibility for Bolsa Família 2015 2017 Increase Decrease (1) R$77 - R$154 R$154 R$154 R$77 - R$154 Child/Teen (4) R$154 Child/Teen (6) R$77 - R$154 >R$154 Child/Teen Source: World Bank staff estimates based on MDS (2015) Note: The same thresholds are used in 2017 as in 2015 because all the analysis in the model is carried out in terms of 2015 prices. The increases in the Bolsa Família benefit structure and eligibility thresholds adopted in June 2016, where practically identical with the prevailing inflation rate between 2015 and 2016. Following the methodology outlined above, the This increase in the budget needed can be broken marginal change in the budget of Bolsa Família down in two parts: The R$978M that is needed to needed to extend coverage of the program to the cover the 0.522M families falling below the extreme “new poor� can be calculated (see Figure 12). poverty threshold of R$77 per capita per month, Abstracting from operational issues associated with and the R$199M that is needed to cover the the identification and targeting of the “new poor 0.232M families falling between the thresholds of households, the microsimulation analysis for scenario R$77 and R$154 per capita per month. With 1, implies that in 2017, 0.810M new families (not scenario 2, in 2017, 1.163M new families will be individuals) will be eligible for Bolsa Família eligible for Bolsa Família benefits. This entails an benefits.14 This entails an increase in the Bolsa increase in the Bolsa Família budget by R$1.82 Família budget of R$1.25 billion, or a 4.73% billion or a of 6.9% increase from the 2015 budget increase in the budget (of R$26.4 billion in 2015).15 (of R$26.4 billion in 2015). Figure 12: Marginal change in the budget and the numbers of recipients of Bolsa Família (b) Marginal Change in the Number of Bolsa (a) Marginal Change in the Bolsa Família Família recipient families Budget 13 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) The distribution of the additional Bolsa Família 3.4% to 3.5% in 2016 and 2017 whereas in budget to eligible families under the current rules scenario 2 the extreme poverty in 2017 rises to of eligibility can prevent the extreme poverty rate 3.6%). In contrast, the increase in the extreme in Brazil from increasing beyond the level of poverty rate is significantly higher in the absence of 2015. Figure 13 presents the estimated poverty any adjustment in the real budget of Bolsa Família. rates for scenarios 1 and 2 under two alternative assumptions with respect to the budget of Bolsa The Bolsa Família program provides a very Família: (i) keeping the real budget constant, which effective way of targeting scarce financial prevents coverage of the “new poor� by the resources to the poorest households among the program (without additional BF) and (ii) expanding new poor households (see Figure 14). Without the the real budget of Bolsa Família to increase additional Bolsa Família budget, the cumulative coverage of the program to the “new poor� distribution curve will shift up (blue line) due to the generated by the crisis in 2016 and 2017, assuming decline in per capita household income.16 The the current level of real benefits and eligibility rules transfer of the additional Bolsa Família budget and (with additional BF). In both scenario 1 and 2, its distribution to the “new poor� families based on expanding the real budget of Bolsa Família to cover the program’s current level of real benefits and the new poor manages to maintain the extreme eligibility rules will shift the line back to its 2015 poverty rate at about the same level as in 2015 (in position (red line overlapping with grey line scenario 1 the extreme poverty rate increases from depicting the 2015 cumulative distribution of income). Figure 13: Estimated Poverty Rates without and with increased Bolsa Família coverage (b) Scenario 2 (a) Scenario 1 14 Source: World Bank staff estimates based on PNAD 2015 and the ADePT Simulation Module (Olivieri et al. 2014) Figure 14: Changes in income distribution without Bolsa Família program constitutes the flagship and with expanded Bolsa Família coverage program of the Ministry of Social Development (MDS). Conditional cash transfer programs such as Bolsa Família do not only have a redistributive role, but also an important role at protecting the poor in times of an economic downturn. To fulfill this function, counter-cyclical (increased) budgets are required at times of crises to increase coverage of the increasing R$70 Poverty line number of poor. The analysis in this note suggests that an increased budget (in real terms) of about 4.73% R$140 Poverty line (R$1.25 billion) and 6.9% (R$1.82 billion) from the 2015 budget of R$26.4 billion, would be a very effective way of targeting scarce financial resources to the most needy among the “new poor� households generated by the crisis.17 The distribution of the additional Bolsa Família budget to the newly eligible households among the “new poor� Source: World Bank staff estimates based on PNAD can prevent the extreme poverty rate in Brazil from 2015 and the ADePT Simulation Module (Olivieri et al. increasing beyond the level of 2015, though its 2014) Note: The vertical lines in the figures above denote the impact on preventing the overall poverty rate from R$70 poverty line in 2015 Reais (=R$92.3 in 2015) increasing is not as dramatic. It is important to bear and the R$140 Poverty line in 2015 Reais (= R$184 in in mind that these estimates are derived based on 2015). the program’s current level of real benefits and eligibility rules and assuming annual adjustments in the nominal budget in par with the annual inflation 6. Policy Considerations rate so at to maintain the purchasing power of the The depth and duration of the current economic benefits constant over time. Delays in adjusting the crisis in Brazil gives rise to the opportunity to nominal value of the transfers of Bolsa Família in expand the role of Bolsa Família from an effective par with the prevailing inflation rate are likely to redistribution program to a true safety net lead to higher poverty rates than those estimated in program that is sufficiently flexible to expand its this note (all else equal). coverage to the “new poor� households generated One encouraging message emerging from this by the crisis. Brazil has managed to build one of analysis is that the fiscal adjustment currently the largest safety net systems in the world and the 15 under implementation in Brazil can be the crisis on extreme poverty is relatively low (less accomplished at virtually little or no cost to than 7% in the pessimistic growth scenario). As the poverty. Even with the depth of the current Brazil Systematic Country Diagnostic (2016) recession, the social gains Brazil made in the last highlights, in spite of the limited fiscal space in the decade do not appear likely to be reversed under medium run, there is ample scope to increase the a range of plausible assumptions. This suggests that budget for the most progressive elements of social Brazil has crossed an important threshold, and that is policy, through reallocations from entitlement an important legacy of the past decade. The programs and through improvements in the estimated increase in the budget for the Bolsa efficiency of public spending.18 Família program required to mitigate the impacts of 16 References Bourguignon, F., Bussolo, M., Pereira da Silva, L. (2008). “The impact of macroeconomic policies on poverty and income distribution: Macro-micro evaluation techniques and tools.� In F. Bourguignon, M. Bussolo, & L. Pereira da Silva (eds.), The impact of macroeconomic policies on poverty and income distribution: Macro- micro evaluation techniques and tools. Washington, D.C.: World Bank. Brazil Macro Poverty Outlook (2016: Fall) Washington D.C., The World Bank. Brazil Systematic Country Diagnostic (2016) “Retaking the Path to Inclusion, Growth, and Sustainability�, May 2016. Report No: 101431-BR, Washington D.C., The World Bank. Cabanillas, O.B., Lugo, M.A., Nielsen, H., Rodríguez-Castelán, C., & Zanetti, M.P. (2015). “Is Uruguay more resilient this time? Distributional impacts of a crisis similar to the 2001-02 Argentine crisis.� Journal of Banking and Financial Economics, 2(4), 64-90. Ferreira, F., Leite, P., Pereira da Silva, L., & Picchetti, P. (2008). “Can the distributional impacts of macroeconomic shocks be predicted? A comparison of top-down macro-micro models with historical data for Brazil.� In F. Bouruignon, M. Bussolo, & L. Pereira da Silva (eds.), The impact of macroeconomic policies on poverty and income distribution: Macro-micro evaluation techniques and tools. Washington, D.C.: World Bank. Gragnolati, Michele, Ole Hagen Jorgensen, Romera Rocha, and Anna Fruttero. 2011. Growing Old in an Older Brazil: Implications of Population Aging on Growth, Poverty, Public Finance, and Service Delivery. Washington, DC: World Bank. MDS (2015). Plano Brazil Sem Miséria: Caderno de Resultados 2011 | 2014.\ Mincer, J. (1974). Schooling, experience and earnings. New York: Columbia University Press for the National Bureau of Economic Research. Olivieri, S., Radyakin, S., Kolenikov, S., Lokshin, M., Narayan, A., & Sánchez-Páramo, C. (2014). Simulating distributional impacts of macro-dynamics: Theory and practical applications. Washington, D.C.: World Bank. Rofman, Rafael, Ignacio Apella, and Evelyn Vezza, eds. 2014. Beyond Contributory Pensions: Fourteen Experiences with Coverage Expansion in Latin America. The World Bank. http://elibrary.worldbank.org/doi/book/10.1596/978-1-4648-0390-1. Skoufias, E. R. Gukovas, and T. Scot (2016) “Variations in Participation and Employment in Brazilian Metropolitan Areas� Background paper for the Brazil Programmatic Poverty Analysis Task of the World Bank. 17 1 Note prepared by the LAC Poverty and Equity GP Team: Emmanuel Skoufias, Shohei Nakamura, and Renata Gukovas. Martin Raiser, Ricardo Paes de Barros, Pedro Olinto, Oscar Calvo-Gonzalez, and Antonio Nucifora, provided very constructive suggestions and feedback. This note is a revised and updated version of an earlier note using data from the 2014 PNAD as a baseline year. 2 See Olivieri et al. (2014) for a detailed description of the ADePT simulation module developed for microsimulations of the poverty and welfare impacts of economic crises (http://go.worldbank.org/UDTL02A390) .The ADepT crisis simulation module is based on a simplified version of the approaches developed by Bourguignon, Bussolo, and Pereira da Silva (2008) and Ferreira et al. (2008). 3 The categories are inactivity (or being out of the labor force), unemployment, and employment in the following three sectors: the primary sector (agriculture, fishing, and mining); manufacturing (including electricity, gas, and water); and services (including the construction sector). 4 See DECRETO Nº 8.794, DE 29 DE JUNHO DE 2016 5 Based on the Secretaria de Assuntos Estratégicos (SAE) the middle class consists of individuals with incomes above R$291. http://www.sae.gov.br/imprensa/sae-na-midia/governo-define-que-a-classe-mediatem-renda-entre-r-291- e-r-1-019-cidade-verde-em-24-07-2013/#ixzz35UobUtKL 6 For more details, see the companion technical note. 7 Recent work by Bank staff also suggests that during this crisis, discouraged worker effects are likely to dominate instead of the added worker effects that in past crises in Brazil typically contributed to women and other temporary workers increasing their labor force participation during the crisis period (Skoufias, et al. 2016). 8 See Olivieri et al (2014) for a more detailed discussion of the limitations of the ADePT Simulation module. 9 See Figure 1.12 in the Brazil Systematic Country Diagnostic (2016) where the evolution of the wage skill premium and inequality is presented. 10 Given that the analysis is being carried out with only one year of data from the 2015 PNAD, adjusted 2015 nominal income for general inflation between 2015 and 2016 or 2017 compared against a poverty threshold that is also adjusted by the same general inflation rate is equivalent to price changes having no real effects. Another issue related to prices is spatial disparities in the cost of living. Although Brazil does not have an official poverty line, the administrative poverty lines used by the Bolsa Família program make no adjustments for cost of living differences across regions or between urban and rural areas. Alternative simulations carried out (not reported) taking into account spatial cost of living differences in the baseline year using a spatial cost of living index derived from the 2008/9 POF survey yielded qualitatively similar estimates on the distributional impacts of the crisis in Brazil. 11 The Gini index, for example, is most sensitive to income differences at about the middle of the distribution. 12 The very large declines for the very bottom percentiles of income are primarily due to extreme values and outliers (i.e., very low but non-zero values of reported income). Reproducing the GIC excluding the bottom 1% of the income distributions yields the same general picture with the maximum decline i n income at the “new� bottom percentile of income being around -40%. 13 Southeast region includes the following States: Minas Gerais, Espirito Santo, Rio de Janeiro, and São Paulo. 14 In practice, targeting errors are always an issue of concern in the implementation of cash transfer programs such as Bolsa Família, and it is important to develop targeting mechanisms that minimize inclusion and exclusion errors. 15 Administrative records from the Bolsa Família program report a budget of R$26.4 billion in 2015. The budget calculations in this note, net out the number of families that would graduate or become ineligible from Bolsa Família because of increases in their income. It is also assumed that every family whose per capita income less than R$154 and meeting the official criteria receives the benefit. 16 The cumulative distribution functions in Figure 14 zoom-in the per capita income range between R$0 and R$250. 17The estimate of the nominal budget required in 2017 can be calculated by multiplying the 2015 budget (R$26.4 billion) with the estimated increase in the “real� Bolsa Família budget (e.g., 1.0473 under scenario 1) and the expected inflation rate between 2015 and 2017 (e.g. 10% or 1.10). Using these specific values, the estimate of the nominal budget required in 2017 is R$30.41 billion. 18 The Brazil Systematic Country Diagnostic (2016) is accessible at: http://documents.worldbank.org/curated/en/180351467995438283/Brazil-Systematic-country-diagnostic 18