COVID-19, Labor Market Shocks, and Poverty in Brazil: A Microsimulation Analysis1 Fabio Cereda, Rafael M. Rubião, and Liliana D. Sousa Poverty and Equity Global Practice, World Bank July 31, 2020 Abstract: In this note we estimate the short-term economic impact of the COVID-19 crisis on Brazilian families vis- a-vis labor shocks. The analysis, using a microsimulation model which incorporates subnational shocks from a computable general equilibrium growth model, shows that over 30 million workers in Brazil may see significant reductions in their labor income in 2020 due to the COVID-19 pandemic. Two-thirds of these workers are informal workers or own-account workers, groups without access to unemployment protection. These household shocks would reduce average per capita income by 7.6 percent, with the largest impact on the second and third quintiles of the income distribution. These income shocks are inequality- increasing: without any mitigation measures, inequality would increase by 4 percent. The country’s first line of defense, its existing unemployment insurance system, reduces the income shock to 5.3 percent. Even so, an additional 8.4 million Brazilians could fall into poverty. The policy responses announced by the government, and particularly the Auxilio Emergencial (AE) transfer, have the potential to fully absorb the labor income shock for the poorest 40 percent and reduce poverty. Yet, these results reflect annualized income, obscuring the sharp reduction in monthly income if demand shocks persist after the AE ends. Looking towards the next phase of the response, considering extensions of AE that are either less generous or more restricted provide a fiscally prudent approach for continuing to support Brazil’s most vulnerable. 1This is a background note for: World Bank. 2020. “COVID 19 in Brazil: Impacts and Policy Responses.” World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/34223 License: CC BY 3.0 IGO.. 1 Acknowledgements The authors recognize the contributions of the broad team who worked on “COVID-19 in Brazil: Impacts and Policy Responses,” in particular Marek Hanusch, Cornelius Fleischhaker, Joaquim Bento de Souza Ferreira Filho, Xavier Ciera, and Antonio Soares Martins Neto, as well as the team who has contributed to BraSim, especially Matteo Morgandi, Katharina Maria Fietz, Alison Rocha De Farias, and Jia Gao. The authors worked under the guidance of Paloma Anos Casero, Ximena Del Carpio, Rafael Munoz Moreno and Pablo Acosta. The authors are grateful for comments received from Gabriela Inchauste and Hernan Winkler. 2 Contents Section 1. Brazilian’s economic vulnerability to COVID-19 ....................................................................... 6 Section 2: Methodology .............................................................................................................................. 10 2.1 Modeling COVID-19 income shocks ............................................................................................... 11 2.2 Modeling unemployment protection ................................................................................................. 13 2.3 Modeling the Bolsa Familia queue ................................................................................................... 15 Section 3: Results........................................................................................................................................ 16 3.1 Impact on household income ............................................................................................................ 17 3.2. Impact on Poverty and Inequality .................................................................................................... 21 3.3 Caveat: Perfect Targeting of AE ....................................................................................................... 27 3.4 After the Auxilio Emergencial ends.................................................................................................. 28 Section 4: Conclusion ................................................................................................................................. 31 Annex .......................................................................................................................................................... 33 Annex 1: Poverty Headcount, by scenario .............................................................................................. 33 Annex 2: Poverty Headcount Growth, by scenario ................................................................................. 34 Annex 3: CGE Model Results................................................................................................................. 35 Annex 4: Poverty Gap, by scenario ........................................................................................................ 36 Annex 5: Poverty and population distribution, by region ....................................................................... 37 Annex 6: Official Unemployment Insurance eligibility rules ................................................................. 38 Annex 7: FGTS modelling rules and detailed assumptions .................................................................... 39 Annex 8: Average income, by definition, by disposable income quintile .............................................. 41 Annex 10: Downside Scenario................................................................................................................ 42 Annex 11: AE rules and their presence in BraSim model ...................................................................... 43 3 The first case of COVID-19 in Latin America was identified in Brazil on February 26, 2020. Since then, the virus has spread throughout the country, reaching from Brazil’s largest cities to isolated communities in the Amazon. By July 16, 2020, the number of officially recorded cases of COVID-19 in Brazil was over 2 million (the second highest in the world) with more than 76,000 deaths.2 The economic effects of the pandemic are being widely felt throughout the country as the country falls into one of the most severe recessions of its history with projections of a contraction in excess of 8% of GDP for 2020. In this note, we present initial estimates of the economic costs of this pandemic on households in Brazil. Most Brazilian households are economically vulnerable to labor income shocks in general, but there are additional risks due to the specific timing and nature of the COVID-19 crisis. There is a confluence of several factors worsening the prospects for Brazilian households. First, the pandemic hit Brazil while it was still in recovery from the 2014-16 domestic crisis. The income of its poorest 40 percent in 2019 was below its 2014 level, unemployment levels remained near crisis levels, and household debt had grown significantly. In all, more than half of Brazilian households are either poor or vulnerable to falling into poverty. The ability of these households to weather a further shock was already strained. Second, while Brazil has a wide social protection system, has made significant strides in increasing formality in its labor market, and has a relatively large and generous unemployment insurance system, two in five Brazilians live in households where the majority of income is from unprotected employment. Third is the exposure of the vulnerable population to the idiosyncratic features of the COVID-19 crisis. Two-thirds of children and youth are in this category, leaving them exposed to long term human capital shocks through school interruptions, possible increases in school dropouts, and high youth unemployment. A majority of the economically vulnerable live in high-density urban areas and with often precarious sanitary conditions while depending on informal work in sectors that have been highly impacted. At the same time, while the pandemic began in major cities, it has swiftly moved into rural areas including the Amazon, with heavy tolls in some of the most rural states in the country, home to rural poor, including traditional and indigenous peoples, and forest communities, who have low access to health care. In this note, we focus on the short-term impact of the COVID-19 crisis on household income through labor income shocks. For this, we need to consider two angles – first, which sectors will suffer job and labor income losses, and, second, how these losses affect household income. We do this by first estimating demand shocks across states and sectors using a macroeconomic model of the Brazilian economy. Given the uncertain climate, we modeled a baseline and a downside scenario. Based on these demand shocks, the model generates expected impacts on wages and wage bills. Second, we distribute these wage shocks to individual workers using a microsimulation tool, which generates estimates of the magnitude of the shock to overall family income. Along with the shock, we use the tool to model the role of the unemployment insurance system and two critical policies quickly implemented during the onset of the pandemic. The true impact of the pandemic on households depends on the social insurance policies already in place and those designed to address the shocks. This note focuses on three of these policies: the unemployment insurance system, and two important emergency income support measures that were quickly implemented by the Brazilian Government in response to the crisis - the expansion of the Bolsa Familia Program (PBF), and the Emergency Aid transfer program. The expansion of the country’s flagship social protection program PBF added a 1.2 million new families from the program’s waiting list. The Emergency Aid (Auxilio Emergencial, AE) program, a BRL 600 (equivalent to US$7.40 per day in 2011 PPP) cash transfer of three months targeting informal, own-account, and unemployed workers living in low-income households as well as PBF beneficiaries, is expected to cover between 53 to 68 million workers. As the 2 WHO (World Health Organization), Coronavirus Disease (COVID-19) Situation Report - 155 https://www.who.int/docs/default- source/coronaviruse/situation-reports/20200623-covid-19-sitrep-155.pdf?sfvrsn=ca01ebe. 4 situation is quickly evolving, including ongoing policy discussions of how to continue public support in the near future, it is important to note that the note reflects the response programs implemented as of late June 2020. The AE, for example, is likely to be extended for an additional two months. We find that without mitigation measures, this sharp economic contraction has the potential to severely impact households and push millions into poverty. This impact would be felt throughout the whole income distribution, but particularly workers in the second and third quintiles, whose income depend more on vulnerable labor and less on government transfers. In our baseline scenario, annual household income is expected to decrease by 7.6 percent. The second and third quintiles can expect the largest losses at around 14 percent, and up to 21 percent in our downside scenario. The richest quintile, on the other hand, is expected to lose only 4.6 percent of their income. Unemployment insurance protects almost all quintiles’ income, buffering 30 percent of the average income reduction, though with a smaller effect for the lowest quintile where very few are formal sector workers. The AE has the potential to absorb much of this shock and, for the poorest, represents a sizeable though temporary boost to income. With high coverage rates in the first two quintiles and its relative generosity – with payments about 6 times higher than the average PBF benefit – this program is expected to increase annualized household income by 14 percent for the poorest 20 percent and 3 percent for the second quintile. Given the notable vulnerability of millions of Brazilian households and the significant progression of the disease in the country, the economic impact of the pandemic in Brazil could be especially strong compared to other nations. The Global Economic Prospects released in June 2020 by the World Bank projects a global recession of 5.2% and a decrease of 8.0% in the Brazilian GDP.3 According to this report, the more substantial impact of the crisis in Brazil is related to the continued growth of COVID-19 cases in the country, the fall in commodity prices, low private investment due to higher uncertainty, and low public investment due to the limited government’s fiscal space. The generous transfer provided by the AE comes in good time and can, at least temporarily, mitigate the impact on lower income households. The main question that nevertheless remains is what to do after the AE ends if the crisis persists and employment does not recover? The first beneficiaries of the AE began to receive it in April – most beneficiaries will have ended their three months before September. New proposals for extending cash transfers beyond the AE are being floated by policy makers and the public alike. With estimates of up to 68 million eligible beneficiaries, the AE is expected to cost up to BRL 135 billion, 2 percent of the projected GDP for 20204. Given the fiscal costs of AE on one hand and the continuing health crisis on the other, it is likely that a program of continued income support at lower generosity levels than the AE will be maintained into the second half of 2020. The simulations suggest that, if unemployment persists, reducing the payment by half would still be enough to absorb the impact on the poorest 40 percent. Similarly, reducing the coverage of the AE by imposing stricter income eligibility could generate fiscal savings without increasing poverty. At the same time, short-term income protection is only one part of the puzzle. Brazil is currently one of the most adversely affected country in terms of case and death counts related to the pandemic. To plan future actions, it is useful to consider the three stages that countries must pass to overcome the COVID-19 3 World Bank. 2020. Global Economic Prospects, June 2020. Washington, DC: World Bank. DOI: 10.1596/978-1-4648-1553-9. License: Creative Commons Attribution CC BY 3.0 IGO. 4 World Bank. 2020. Global Economic Prospects, June 2020. Washington, DC: World Bank. DOI: 10.1596/978-1-4648-1553-9. License: Creative Commons Attribution CC BY 3.0 IGO. 5 crisis.5 Brazil is currently in the relief stage, which involves an emergency response to the health threat posed by COVID-19 and its immediate social, economic and financial impacts. Once the pandemic shows signs of being under control, the country will need to enter a restructuring stage focused on strengthening health systems for pandemic readiness, restoring human capital, and restructuring firms and sectors, debt resolutions for firms, as well as recapitalization of companies and financial institutions. Finally, the resilient recovery stage entails taking advantage of new opportunities to build a more sustainable, inclusive and resilient future in a world transformed by the pandemic. Given the expected wide-spread destruction of employment, especially for lower income workers, reactivating these workers during the next phases will need to be a priority. As the country saw after its 2014-16 crisis, employment opportunities for the poorest took significantly longer to recover than for skilled workers. In the likely event of a segmented job recovery and accelerated structural transformation, strategies will be needed to facilitate labor reallocations. This process may leave some workers dislocated and in need of prolonged income support during this process. Strong attention will be needed to the incentive-compatibility of temporary benefits and unemployment insurance, for instance by including training or job search requirements during paid time off from work. Active interventions can equip vulnerable workers with skills and information to navigate these changes. Given the scale of benefit recipients and social distancing measures, training and intermediation policies will likely need to be digitally enabled. Affordable and wide access to internet and digital literacy itself will be an essential skill. This note is organized as follows. In Section 1 we present the Brazilian context, discussing key aspects of the vulnerability of Brazilian households to the pandemic shock. In Section 2 we explain key methodological aspects of our modelling. In Section 3 we present our simulation results for both the pandemic shock and the policy-responses. The results are organized to first look at the impact on household income, then the impact on inequality and poverty, and then considering future extensions of the AE program. In Section 4 we conclude our analysis by putting our main results into perspective and discussing further resilience and vulnerabilities of Brazil when facing the COVID-19 pandemic. Section 1. Brazilian’s economic vulnerability to COVID-196 Even before the pandemic, half of Brazilians (52%) were economically vulnerable, being either already in poverty (living on less than US$5.50 per day in 2011 PPP) or at risk of falling into poverty (living on per capita income between US$5.50 to US$13 per day). This is particularly the case in the North and Northeast regions of Brazil, where in most states, between 70 and 80 percent of the population falls into this category (Figure 1). This population is mostly employed in precarious and unprotected jobs, urban, and young, including more than 7 out of 10 Brazilian children and youth7. They belong to groups expected to suffer a higher income shock. Relatively few households can weather significant labor income shocks. Particularly important is to consider that Brazil’s poorest were still recovering from the 2014-2016 crisis, as the income of the bottom 40 percent is still below the pre-crisis level. Between these years over 5.6 million Brazilians fell into poverty 5 World Bank, 2020. COVID-19 Crisis Response Approach Paper: Saving Lives, Scaling-up Impact and Getting Back on Track. Washington, DC: World Bank. 6 This section is largely derived from World Bank (2020) “COVID 19 in Brazil: Impacts and Policy Responses”. 7 The poor represent 20% of the Brazilian population and include 36% of all Brazilian children (<15 years old) and 25% of the youth (15-24). Seventy-two percent of the poor live in urban areas and 67% of those who work are in precarious jobs (informal or own-account), two groups likely to be especially exposed to the COVID-19 crisis. This profile is very similar for the economically vulnerable (those living on $5.50 to $13 per day), which represent 32% of the country’s population and 37% of all Brazilian c hildren and youth. The proportion living in urban areas is even higher, 85%. Informal or own-account workers amount to 43%, and 67% are working in the retail or services, which are expected to be the sectors most affected by the crisis. 6 (defined as living on less than $5.50 per day in 2011 PPP terms) and the inequality, measured by the Gini Index, changed its downward trend and increase significantly between 2015 and 2018. The fall in labor income from 2014 to 2019 occurred more sharply in the already vulnerable groups such as young people between 20 and 24 years old (-17.8%), and individuals with low education (-15.1%)8. Moreover, unemployment rates remain near crisis-level (Figure 2) and household debt burden is high at 45 percent of household income, reflecting increased non-mortgage debt since 2017 (Figure 3). At the same time, few Brazilians have savings: only 32 percent of individuals in Brazil declared saving in the previous year, compared to 73 percent in OECD countries, 37% in countries with similar GDP per capita, and 48% in the world.9 These factors suggest many households have little room to absorb another shock. Figure 1: Subnational economic vulnerability Figure 2: Unemployment rates (Percent of state population who is poor or (percent of labor force, 2012Q1-2020Q1) vulnerable, 2018) 35 30 25 20 15 10 5 0 2012 2013 2014 2015 2016 2017 2018 2019 2020 All Youth: 18-24 Source: World Bank estimates based on SEDLAC Source: IBGE unemployment indicators, downloaded on (World Bank and Cedlas). June 9, 2020. Note: This map shows the percentage of the state population living on less than $13/day (2011 PPP). There is an important overlap between vulnerability in income (ability to pay for food and rent) and vulnerability in living conditions (adequate housing and services). Poorer households have less access to improved sanitation, running water, and private bathrooms (Figure 4) – all important services to reduce the spread of illness. About one in five Brazilians live in slum or substandard housing, and another 32,000 are homeless. Epidemiological models find that COVID-19 is likely to spread more in high-density areas, such as slums, placing the urban poor as particularly exposed. At the same time, rural populations, including indigenous peoples, forests, and traditional communities, are also facing additional risks arising from barriers to access of basic services, including healthcare. 8 Neri, M. 2019. “A escalada da desigualdade – Qual foi o impacto da crise sobre a distribuição de renda e a pobreza?” FGV Social – Center for Social Policies. 9 Source: Relatório de Cidadania Financeira (2018). Banco Central do Brasil. Available at: . This report uses data from the Global Findex Database (World Bank) 7 Figure 3: Household debt burden Figure 4: Lack of access to adequate sanitation (Percent of household disposable income, 2007- (Percent, 2018) 20) Mortgage borrowing Non-mortgage borrowing Total debt Overall 56% 50 Poor (5.5 USD PPP11) 40 36% 30 26% 21% % 15% 20 10% 08% 03% 10 Trash Connection to Connection to Private 0 collection public water sanitation bathroom system system jun/09 abr/10 jun/14 abr/15 dez/16 jun/19 out/07 dez/11 out/12 out/17 ago/08 fev/11 ago/13 ago/18 fev/16 Source: World Bank based on Central Bank of Brazil. Source: World Bank estimates based 2018 PNADC. Figure 5: Share of population by majority The key transmission channel through which the income source Covid-19 crisis is expected to affect household (Percent, 2018) welfare is through market demand and supply shocks that translate into labor income losses. A Pensions and Pensions public transfers; and public large proportion of Brazilian households face a high 23% transfers, risk of losing their income: Two in five Brazilians 59% 31% protected rely mostly on unprotected income sources (Figure CLT (>6 months CLT (>6 months 5), defined as the population for whom a majority of protection); 27% protection); 15% household income comes from informal jobs, own- account work and formal employment with less than six months of salary protection in case of job loss.10 Informality 50% Informality and own- For the poorest 20 percent, the share of people 41% and own- unprotected account relying on unprotected income increases to half of account work; 43% work; 32% the population. Exposure to pandemic-related unemployment or All Poorest 20% labor income shocks is heterogenous, affecting Source: World Bank based on BraSim and 2018 some types of workers more than others. Informal PNADC. and own-account workers have no formal income protection mechanisms in place, whereas public sector workers and most formal private sector salary workers, called Consolidação das Leis do Trabalho (CLT) contracts, have employment protection and access to unemployment insurance, severance pay, and 10The analysis uses job and worker characteristics to simulate unemployment insurance eligibility, severance pay (multa) and employer-funded savings account (FGTS) balances as described in Section 2. Based on these amounts, we calculate how many months of protected salary each formal private-sector salary worker will have in the event of a layoff. 8 employer-funded savings accounts. At the same time, economic sectors are differentially exposed. The risk of employment interruption is higher for sectors with heavy reliance on face-to-face interactions. Low wage workers and women are more likely to be in these sectors (See Figure 6 and 7), and hence more likely to suffer the employment shock first. As a way of minimizing the impacts of the pandemic, the Government as well as private employers are changing regulations and adapting to new forms of work relations. The 2017 labor reform that regulated part-time work and the recent Benefício Emergencial de Manutenção do Emprego e da Renda (BEm) introduced flexibility for firms to suspend paid work as a strategy to reduce the amount of outright job destruction. Under the BEm wage subsidy, firms are able to temporarily reduce their labor demand, by reducing hours or imposing temporary layoffs, while protecting the employment relationship. While the BEm program is not modeled explicitly in this note, it is implicitly part of the unemployment insurance response in the results reported below since temporarily laid off workers under this program receive benefits equivalent to UI. In 2014, only 8 percent of workers were allowed to telework11 so, in response to the pandemic, emergency measures were adopted to facilitate the adoption of telework options. However, few workers and firms were ready or able to transition to telework options with telework largely limited to higher paid skilled workers.12 Figure 6: Face to face interactions by salary Figure 7: Face to face interactions by gender (Average score by income decile, formal sector) (Average and median score, formal sector) 0,64 0,58 0,56 0,58 0,54 0,54 0,54 0,51 0,5 0,47 0,46 0,45 0,47 0,46 Average f2f score Male Female Average f2f score Median f2f score Source: Avdiu, B., X. Cirera, G. Nayyar and A. Soares Martins (2020) “The Impact from COVID -19’s Social Distance Measures in Brazil. Who are the Most Vulnerable Groups?” The World Bank . Brazil possesses some notable sources of resilience, especially as compared to many other middle- income countries, which have allowed it to roll out a strong response to the economic crisis. First, Brazil benefits from also having a relatively large formal sector workforce with some unemployment protection and savings mechanisms in place. Second, Brazil has provided near-universal access to pensions and/or social security to its older population, which is also the most vulnerable to COVID-19. Third, Brazil has in place a robust infrastructure for delivery of its emergency measures, notably its beneficiary registry, 11 Mello, A., & Dal Colletto, A. (2019). Telework and its effects in Brazil. In Telework in the 21st Century. Edward Elgar Publishing. 12 We do not model telework explicitly. However, as described below, the allocation of labor shocks in this analysis are determined by reductions in demand at the sectoral level, not by necessity of social distancing. In this way, telework options do not affect the distribution of labor shocks. 9 the Cadastro Único, with 76.4 million Brazilians registered, complemented by other tools, including an existing network of NGOs supporting government actions in the slums. This existing infrastructure allowed for quick implementation of the AE and increases the likelihood of the AE benefit to reach eligible individuals. Two of the policy responses implemented by the Government are particularly important for alleviating the economic impact of the crisis on lower-income Brazilian households and are thus the focus of this analysis.13 First, the expansion of the Programa Bolsa Familia (PBF) to include families who were already eligible but were not receiving the benefit due to program budget constraints. The expansion of PBF is expected to add 1,225,000 families to the program who were in the queue. This increases the total coverage of the program by 8.6 percent to 14.26 million families at an estimated annual cost of BRL 3.1 billion (0.05 percent of GDP based on 2020 projection). The ‘Auxílio Emergencial’ (AE) is a temporary emergency program targeting families living in poverty and informal or own account workers. The program is making three monthly transfers of BRL 600, just under 60 percent of the minimum wage, to individuals with family income below half the minimum salary per capita, with a maximum value per family of BRL 1200.14 Section 2: Methodology We build on the BraSim microsimulation tool to estimate the impact of employment shocks and expanded social protection measures on the income of Brazilian households. A microsimulation model of the Brazilian population developed in 2019 by the World Bank, BraSim is a tool designed to analyze and model reforms to the tax and social protection system and how these affect equity, the efficiency of public spending, and fiscal outcomes (public spending and tax revenue). It is an incidence analysis tool that models partial equilibrium distributional implications of changes to policies. The tool does not model behavioral changes or consider general equilibrium implications. That is, the model assumes that household and individual decisions, such as employment and consumption level, are not affected by changes in policies. The tool works on a synthetic population based on the household survey that was adjusted to better approximate official tax collection, program participation rates, and the composition of the labor market, including detailed modeling of various contract types. It functions by applying the rules of each transfer program, direct and indirect taxes, pension system contributions, and employer contributions to simulate the effect of these changes on the after-fiscal policy income of each household. Following the Commitment to Equity (CEQ) methodology,15 BraSim estimates the incremental impact of different fiscal policy (taxation and benefits) on take home income. BraSim estimates four of the CEQ concepts of income: 1) gross market income is household income received from market activities prior to any taxation or subsidy16; 2) net market income is the gross market income minus direct taxes and social security contributions; 3) disposable income adds government transfers to the net market income; and 4) consumable income deducts indirect taxes from disposable income. The income definition used in this analysis is disposable income. Fiscal policy in Brazil, particularly its relatively generous non-contributory pension programs, increases the disposable income of quintiles 1 and 13 A number of other steps are being taken to reduce food insecurity, including daily monitoring of food prices and allowing distribution of food acquired by the National School Food Program to the families of school-going children while classes are suspended. Additional measures to protect vulnerable populations include distributing PPE and hygiene items; temporarily eliminating utility-cutoffs due to non-payment; and installation of hand washing stations. 14 Single mothers who qualify for the AE can receive BRL 1,200 per month. 15 Lustig, N. 2018. CEQ Handbook Estimating the Impact of Fiscal Policy on Inequality and Poverty. CEQ Institute at Tulane University and Brookings Institution Press. 16 Contributory pensions, which can be considered deferred employment income, can also be included in this definition. 10 2 relative to their gross market income (Figure 8).17 For the top 3 quintiles, the disposable income is lower than market income due to taxation, particularly of formal labor income. BraSim also models eight different labor contract types, including different private and public sector formal dependent workers, formal sector own-account workers, and informal own-account and dependent workers.18 The lowest two income deciles rely heavily on informal employment (both dependent and own-account work) while formal dependent workers are a crucial part of the middle of the income distribution (see Figure 9). Figure 8: Average income, by quintiles of gross Figure 9: Percentage of each type of labor market plus pensions income income, by wage decile Q1 Q2 Q3 Q4 Q5 (RHS Axis) 100% 1600 4500 80% 1400 4000 1200 3500 60% 3000 1000 2500 40% 800 2000 600 1500 20% 400 1000 200 500 0% 1 2 3 4 5 6 7 8 9 10 0 0 Gross market Net market Disposable Consumable CLT Public workers income plus income income income Military MEI pension SIMPLES Contractor Informal own-account Informal salary Source: World Bank estimates based on BraSim microsimulation model. Note: Figure 8 reports the average per capita monthly income of each income quintile. Due to its far higher value, quintile 5 (Q5) is reported on the right-hand side (RHS) axis. See Annex 9 for a table of average income reported in this figure. Figure 9 reports the distribution of workers by contract type, across decile of per capita disposable income. For the analysis of COVID-19 and mitigation strategies, three extensions were made to BraSim. First, a methodology was implemented to translate sector-level payroll shocks from a computable general equilibrium (CGE) model of the Brazilian economy into unemployment shocks in the BraSim population. Second, it was necessary to model the monetary value of unemployment protection available to each formal dependent worker. As job destruction in the public sector is not expected during the short-term, this was done only for private sector workers. And third, it was necessary to model the Bolsa Familia Program’s (PBF) “queue” – a population of approximately 1.5 million families that had qualified for the PBF in late 2019 but had not yet been added to the program. These are detailed below. 2.1 Modeling COVID-19 income shocks To model labor interruptions related to COVID-19, BraSim was linked with a CGE model which simulates sectoral and subnational economic impacts (see Box 1). Two GDP growth scenarios were 17 Benefício de Prestação Continuada (BPC) and rural pensions cover 35 percent of the 65 and older population and pay at least one minimum salary per month. 18 BraSim distinguishes between private sector dependent worker (CLT, Consolidação das Leis do Trabalho ), public servants, and the military, as each of these groups have different tax treatment and rights. There are three types of formal own-account workers in BraSim: 1) contractor (trabalhador autonomo), microentrepreneur (MEI, Microempreendedor Individual), and small business owner (SIMPLES). Finally, informal workers are distinguished between own-account and dependent workers. 11 modeled for 2020: a baseline growth scenario of -8 percent, and a downside scenario of -10.9 percent. For each of these scenarios, we obtained from the CGE model a different variation of wage bills and consequently, impacts with varying levels of intensity in BraSim. While we can think of them as unemployment shocks, more accurately they are reductions in labor income. For example, workers in the formal sector whose employers take advantage of the BEm program may not be technically unemployed but will see their wages reduced to unemployment insurance levels. At the same time, informal workers like street vendors may see a reduction in sales without necessarily becoming unemployed or even reducing their hours. The CGE model calculates reductions in wages across 75 cells, defined by the intersection of 15 regions and 5 productive sectors (see Annex 3). By construction, the CGE model assumes no unemployment and no labor market frictions. While this works well in the CGE framework, it does not reflect the labor dynamics in Brazil. Like most countries, Brazil has sticky wages and enforceable labor contracts, which mean that, especially in the formal sector, productivity and output shocks cannot be fully reflected in adjusted wages or hours. Instead, these shocks are more likely to translate into reduced hours (when legally feasible) or unemployment spells. We used the CGE’s sector-level wage bill shocks to generate employment or earnings interruptions in BraSim for private sector workers, excluding the primary sector. Box 1. Subnational distribution of economic projections using the TERM-BR CGE Model Based on World Bank projections, the Brazilian GDP is projected to fall by 8 percent in 2020, in the baseline scenario, and by 10.9 percent in the downside scenario. Real consumption of households is projected to fall by 15 percent in the baseline scenario, and by 25 percent in the downside scenario. As shown by the severity of the contraction of real consumption, this shock is primarily driven by a sharp reduction in families’ spending. Therefore, the sectorial and regional effects of the macroeconomic downturn will be driven by where families spend the most, especially services. In this analysis, the TERM-BR CGE model was used to distribute the projected national economic shock across the different regions and sectors of Brazil. The TERM-BR belongs to the broader model family named TERM (The Enormous Regional Model), which was originally developed by Victoria University’s Center of Policy Studies - CoPS. TERM models have been adapted to several countries since then, including Brazil, as in the work of Ferreira Filho and Horridge (2016)19 and Diniz (2019)20. The TERM-BR model used in this study is the most recent version of these models for Brazil. It is an inter-regional, bottom-up, annual recursive dynamic model with detailed regional representation, distinguishing up to 136 sectors (industries), 136 commodities and 27 regions. These regions are represented by interdependent models, one for each unit of the federation (26 states and the Federal District), which are interconnected and can trade through the goods, labor and capital markets. In this study we used the aggregated results for 15 regions crossed with 5 sectors. These are detailed in Annex 3. 19 Ferreira Filho and Horridge (2016) “Climate change impacts on agriculture and internal migrations in Brazil” Centre of Policy Studies/IMPACT Centre Working Papers g-262, Victoria University, Centre of Policy Studies/IMPACT Centre. 20 Diniz (2019) "Impactos econômicos e regionais dos investimentos em geração de energia elétrica no Brasil”. PhD Thesis – USP/Escola Superior de Agricultura “Luiz de Queiroz”. 12 Unemployment shocks are distributed across individual workers in each sector based on individual and household characteristics. This was based on a logit regression to estimate the likelihood of each worker being employed.21 Each individual within each of the 60 groups (4 sectors x 15 regions) was then ranked. To make the rank closer to reality, we apply a penalty coefficient for informal workers, decreasing their likelihood of being employed.22 Following the rank, we select a certain number of individuals to receive the unemployment shock. For the baseline scenario, the unemployment shock is equivalent to 6 months – a 50 percent loss of annual labor income. The downside scenario implies a loss of 7 months, or 58.3 percent of annual income. We apply these shocks over the individuals in each group until the total amount of income removed reaches the total CGE wage bill change estimated for each group. Given the severity of the current shock, there is no labor reallocation in the short-term in our modeling – individuals lose labor income for 6 or 7 months, depending on the scenario, and are unable to replace it through employment this year. 23 2.2 Modeling unemployment protection Private sector dependent workers (CLT) in Brazil have access to three income protection mechanisms: unemployment insurance (seguro desemprego – SD), employer-funded compulsory savings account (FGTS), and severance pay (multa), paid by the employer and corresponding to 40 percent of the FGTS account balance. Throughout this text, we use the term “unemployment insurance”, or UI, to refer to these three benefits together. For modelling SD, we first identify who is eligible and then estimate the monthly benefit amount for each eligible worker. The benefit amount is straightforward to estimate based on the current salary, following the official rules summarized in Table 1. In our model we consider only workers with job tenure of 12 months or more to be eligible to SD. In reality, if it is not the first time the worker requests the SD, he or she may be eligible to receive the benefit as long as their last job tenure was at least 6 months long (see the official rules in Annex 6). We excluded unemployment coverage for workers with less than 12 months for three reasons: 1) we do not have employment history in BraSim beyond the current job; 2) official data reveals that the majority (77.4 percent) of SD applications are from first-time applicants;24 and 3) Only 10 percent of the CLT population in BraSim has tenure 6-11 months. Following the rules of the program, if the laid-off worker has been working for 12-23 months, he or she is eligible to SD equivalent to 4 months of salary. For workers with job tenure of 24 months or more, the duration of the benefit increases to 5 months. The main difference between the law and the way SD is modelled in BraSim is for workers with job tenure between 6 and 11 months. 21 The Employment Logit model used the following covariates: age, age-squared, years of schooling, gender, race, a dummy for head of the household and the spouse, state fixed-effects, and household income and precarity indicators (roofing material, possession of fridge washing machine, TV and computer, access to internet, cable TV, and number of rooms), and main earner characteristics (years of schooling and dummies for sector of activity, except services). 22 We obtain this coefficient from a Logit model running in a panel data linking the PNADC 1 st interview to 5th interview. The model assesses the likelihood of exiting employment at the individual level using the following covariates: job informality, a dummy for head of the household and the spouse, age, age squared, gender, education, and state fixed-effects. 23 We treat the primary sector and the public sector differently. Due to its special labor regime and high rate of self-employment, for private sector wage changes we do not model unemployment shocks for some workers. Instead all workers in each cell receive the CGE estimated wage change. Based on macroeconomic projections and current legislation, we do not expect layoffs of public servants in the short term as a consequence of the pandemic hence no public sector effects were modeled. 24 DIEESE, 2017. Anuário do Sistema Público de Emprego, Trabalho e Renda 2016: Seguro-Desemprego: livro 3. Departamento Intersindical de Estatística e Estudos Socioeconômicos. São Paulo: DIEESE. (page 42) 13 In order to consider the full unemployment protection system, it was also necessary to model each worker’s FGTS account balance and, by extension, severance pay.25 One FGTS account is created for each CLT contract, and people can have multiple open accounts. Each month, CLT workers contribute with 8 percent of their monthly salary during their whole tenure period, and FGTS accounts yield an interest of 3 percent annually. People can withdraw funds from their FGTS accounts only under certain circumstances (e.g. lay-offs, sickness, investment in real estates). Recent public measures to boost consumption have allowed the withdraw of BRL 500 for all FGTS accounts without any condition. Table 1: Amount of monthly SD entitlement per wage bracket Wage Bracket Value per month Up to BRL 1,450.23 MW or 80% of Wage From BRL 1,450.24 until BRL 2,417.29 (Wage - BRL 1,450.23) x 50% + BRL1,160.80 Over BRL 2,417.29 BRL 1,643.72 Source: LEI 7.998/1990 (LEI ORDINÁRIA) 11/01/1990 BraSim has limited information regarding Figure 10: Months of CLT protected salary. Number of workers and share by quintile. job-history and no information about FGTS 15 12,16 12,30 accounts or previous withdrawals. A number Milhões of simplifying assumptions were needed to 10 7,98 estimate likely FGTS account amounts per 4,02 5 worker. We assume only one account per CLT 0,25 worker, corresponding to their current primary 0 job – thus implicitly assuming that older accounts Q1 Q2 Q3 Q4 Q5 have been already cashed out. We apply the 18% accumulation rule to each worker’s current 23% 28% 34% 35% 41% salary, which is then multiplied by their job 26% tenure and discounted by the yearly minimum 29% wage growth rate (see Annex 7). 28% 26% 25% 22% 23% Guided by administrative data as benchmarks 21% 19% 1% 21% 19% and to account for the substantial withdrawals 2% 18% 3% 33% 2% 3% permitted throughout the life of these 26% 5% 18% 21% 19% accounts, we applied an FGTS withdrawal rate 13% that decreases with age, starting at 90 percent for National Q1 Q2 Q3 Q4 Q5 workers aged 18 or under. We apply this 0 to 1 Months 1 to 3 3 to 6 6 to 9 9+ withdrawal rate only to workers who earn BRL 1,000 per month or less- which is the group for Source: World Bank estimates based on BraSim which our estimates most differed from the microsimulation model. benchmark. This is also the group who are most likely to need access to their FGTS balances for 25 Severance pay is a direct percentage of the FGTS balance. 14 short-term consumption smoothing - the poorer and the younger. After these adjustments, the estimated FGTS accounts aggregate balance roughly matched the government’s benchmark number (see Annex 8). The results show that, even with these three sources of income protection benefits, an estimated 40 percent of CLT workers would incur a loss of income if facing an unemployment spell longer than 6 months. Eighteen percent are in a position of greater vulnerability, receiving only 1 month of salary or less when combining unemployment insurance, severance pay, and employer-funded savings accounts. That is, beginning in the second month of unemployment, these individuals would have no income. This is because people whose current job tenure is of less than 12 months are unlikely to be entitled to UI. In scenarios of unemployment shocks of more than six months, formal income protection mechanisms would gradually become insufficient to cushion the impact of income loss. The presence of vulnerable CLT workers occurs throughout the entire income distribution but mostly marked in the bottom quintiles (see Figure 10). 2.3 Modeling the Bolsa Familia queue In late 2019 and early 2020, PBF’s budget constraint and continued effects of the 2014-16 crisis on the poorest generated a queue of about one and a half million families who met the eligibility criteria for inclusion in the PBF but could not be added to the program. In March 2020, as a response to the COVID-19 crisis, the Government acted to add 1.225 million families from the queue to the program. The BraSim model, relying on 2017 data, had no queue. All families who qualified for PBF in the tool were included, which reflected accurately the situation in 2017. Figure 11: Distribution of type of worker who Figure 12: Coverage (%) of Bolsa Familia before received income shocks during the creation of (Reference) and after PBF extension PBF Fila (Simulation), by decile of gross market income 76 70 63 57 31% 4042 47% 18 19 12 22% 4 2 Formal, salary worker Formal, own account Informal Reference Simulation Source: World Bank estimates based on BraSim microsimulation model. In order to update BraSim to the 2020 period, it was necessary to identify the families that had become eligible for PBF since 2017. To do this, we applied unemployment shocks to the main earner of a subset of households to simulate a loss of income - which is, in principle, a likely driver of why families applied for the Bolsa Familia benefit. To select these households, we first excluded those who were unlikely to qualify for PBF even after a labor income shock, such as families with significant nonlabor income - particularly pensions (including BPC) and child support beneficiaries. Then, using a logit model, we identified the main earners, based on household and economic characteristics, who were more similar to 15 the current PBF beneficiaries.26 An unemployment shock was then applied to the 1,225,000 main earners with the highest probability based on the logit model. Nearly half of these main earners who were reallocated to unemployment were informal workers and another 22 percent were own-account workers (Figure 11). Even with this shock, not all families became eligible. We were able to model 95 percent of the total number of new families added, achieving 1,163,000 families, or approximately 3.3 million people. Figure 12 shows that the vast majority of these newly eligible families are found in the poorest two deciles, where coverage of PBF increases by 6 percentage points. Section 3: Results The simulations suggest that between 31 and 34 million workers will suffer employment interruptions equivalent to a loss of 6 to 7 months of earnings (Figure 13). As a reference point, in February 2020, 12.3 million Brazilians were unemployed. The informal sector will be the most affected by the unemployment shocks, though the impact on formal workers will be substantial as well. About 71 percent of the informal sector, or over 17 million workers, will suffer the unemployment shock. The situation of these workers is more difficult since they do not count with any mechanism to protect their income. A third of CLTs, the largest group of formal workers, will also suffer unemployment spells, totaling about 10.8 million workers (under the baseline scenario). Figure 13: Number of unemployment shocks The microsimulation model estimates the per occupation and scenario (in millions) impact of employment interruptions or unemployment spells on annualized household 40 income. Figure 14 shows the distribution of per 35 capita disposable income in the BraSim population. 3,5 30 3,0 It ranges from an average of below BRL 300 per 25 12,4 month for the poorest quintile to close to BRL 10,8 20 3,500 for the wealthiest quintile. This is the pre- 15 crisis income used for the analysis. The first set of 10 17,6 18,5 results presented in this section measures the 5 impact on annualized average income across the 0 income quintiles and across the regions after the Baseline Downside shock is implemented. Informal Formal, salary worker Formal, own account In order to quantify the impact of these Source: World Bank estimates based on BraSim scenarios on the population, we also estimate the microsimulation model. effect on inequality and poverty. While Brazil does not have an official poverty line, we rely on two important administrative values to assess household need. We proxy extreme poverty using the Bolsa Familia eligibility, sometimes referred to as the Bolsa Familia poverty line. It corresponds to BRL 178 per month per capita (the equivalent of US$ 2.25 per day in 2011 PPP). The share of the population living on less than half a minimum salary is the default definition of poverty used in this analysis. It is an important proxy of poverty in Brazil since it is the eligibility threshold for Cadastro Único. It is approximately US$ 6.30 (2011 PPP) per day. The share of the population living under each of these thresholds in 2019 are respectively, 7.8 and 29.1 percent (Figure 15). While the simulations are based on the income distribution 26 The logit model for predicting the likelihood of the main earner in the household being in the PBF queue used the following covariates: age, age-squared, years of schooling, gender, race, number of children and adolescents in the household, state fixed- effects, and precarity indicators (roofing material, possession of fridge washing machine, TV and computer, access to internet, cable TV, and number of rooms). 16 of BraSim, the results for poverty and inequality are adjusted to use poverty and inequality in 2019 as the baseline for reporting new poor, for example.27 Figure 14. Average monthly income per Figure 15. Poverty and inequality trends, 2012- capita, by quintile 2019 4.000 35 55,5 3.500 30 55,0 54,5 Poverty rate, % 3.000 25 54,0 Gini index 2.500 20 53,5 BRL 2.000 15 53,0 52,5 1.500 10 52,0 1.000 5 51,5 500 0 51,0 2012 2013 2014 2015 2016 2017 2018 2019 - Bolsa Família 1/2 Minimum salary Gini Source: World Bank estimates based on BraSim. Source: World Bank estimates based on PNADC 2012- 2019. 3.1 Impact on household income The effect of the COVID-19 pandemic on household income in Brazil is expected to be largest in the middle of the distribution. As a consequence of employment and labor income shocks, annual household per capita income is expected to fall by 7.6 percent nationally – including by 14.9 percent in the second quintile and 13.9 percent in the third quintile (See Figure 15a).28 These results are based on annualized income, reflecting a smoothing of income over the year though obscuring the severity of the income shock (Box 2). These are the quantiles most affected by the crisis, made up families who rely largely on low and middle-skill employment and with less reliance on government transfers. The poorest and the fourth quintiles would also suffer significant shocks approaching -9 percent. The richest quintile is the least affected, with income decreasing by only 4.6 percent. This is in large part explained by the composition of workers in this quintile, including high skilled workers with access to telework options and large shares of public servants (no public sector job destruction is expected in the short-term). This disparity in income shocks is also reflected regionally. As the hardest hit regions are the North and the Center-West, with a 10 percent of income reduction, followed by the Northeast, with an income reduction of 8.2 percent (See Figure 16b). The richest regions, South and Southeast, are expected to be the least affected with an income reduction of 5.9 and 7.3 percent, respectively. Brazil’s existing unemployment insurance system is a crucial first line of defense. We estimate that the UI buffers 30 percent of the income reduction, with the shock mitigated from -7.6 to -5.3 percent nationally. However, this protection is distributed unequally, as the UI buffers the two richest quintiles’ losses by 34- 38 percent, while, for the poorest quintile, by at most 6 percent (See Figure 16a). The poorer quintiles have lower formality rates and instead depend more on informal and own-account jobs without income protection. Accordingly, the UI is also more significant for the richest regions. For instance, for the 27 These rates are calculated from PNADC 2019, while BraSim results are based on adjusted PNADC 2017 data. Therefore, for calculating the simulated absolute values, such as the number of new poor, we apply BraSim’s simulated changes in poverty rat es to the absolute PNADC 2019 poverty rates. 28 Downside scenario results are reported in Annex 10 17 Southeast region the UI reduces the magnitude of the income shock by one third, from -7.3 to -4.9 percent, while for the North region the buffering is of less than one fourth of the shock, which decreases from -10.1 to -7.7 percent (See Figure 14b). Figure 16: Income impacts of household employment shocks and the Unemployment Insurance (a) Effects of the pandemic and UI on household (b) Effects of the pandemic and UI on national household income, by quintile, baseline scenario poverty, by region, baseline scenario 0% 0% -2% -2% -4% -3,1% -4% -6% -4,6% -5,6% -5,3% -4,2% -6% -4,9% -5,2% -5,3% -8% -8,2% -7,6% -6,4% -6,1% -5,9% -10% -8,8% -8,9% -8% -6,9% -7,3% -7,7% -7,7% -10,5% -8,2% -12% -11,1% -10% -9,9% -10,1% -14% -13,9% -12% -16% -14,9% Q1 Q2 Q3 Q4 Q5 National Shock (Baseline) Shock (Baseline) + UI Shock (Baseline) Shock (Baseline) + UI Source: World Bank estimates based on BraSim microsimulation model. Box 2. Monthly income shocks A critical caveat of our analysis is that the results reported are primarily based on annualized income. The assumption of perfect income smoothing over the year can obscure the severity of the short- term impact of these income shocks. For instance, in our analysis, a negative shock of BRL 1200 BRL in one month is translated into a BRL 100 loss per month through the whole year. Therefore, the 3- months BRL 600 transfer, which amounts to BRL 1800 a year, is converted into a BRL 150 transfer per month during the whole year of 2020. But in reality, households in the lowest income groups will on average experience three months with higher than usual income during the onset of the pandemic (mostly April through July, depending on enrollment date) as a result of the AE. After these transfers end, and if employment remains weak, these same households will then experience a severe reduction in income. Hence, to simulate the problem from the households’ perspective, we also estimated the monthly impact of the crisis, i.e. the variation in income in the month the shock hits.29 29For simulating the monthly income, we removed the 13th salary and Abono Salarial from the annualized income. We then simulate the pandemic shock and policies by (1) assuming a 100 percent income drop on those affected by the unemployment shock; (2) adding the monthly unemployment insurance entitlement for those who have it; (3) and giving eligible people one BRL 600 monthly payment of the Auxílio Emergencial. 18 Figure 2.1: Monthly household income effects, Relative to their pre-pandemic income, the baseline scenario month-after-shock income of the two bottom 0% quintiles would fall by 26 percent on average – after taking into account unemployment -5% -5,1% insurance – and by 30 to 37 percent if we do not, -10% -7,9% which is a plausible scenario given that the UI is -10,1% -10,0% -15% also temporary (See Figure 2.1). If we consider -16,1%-17,3% -15,0% the Auxílio Emergencial, the situation improves -20% substantially. The AE can give the bottom three -25% -24,9% quintiles an average income boost of 78, 35 and -30% -27,1% -26,3% 14 percent in the month of the shock, and can -30,2% -35% buffer three-quarters of the monthly losses for the -40% -36,8% fourth quintile. Q1 Q2 Q3 Q4 Q5 National Shock Shock + UI Source: World Bank estimates based on BraSim microsimulation model. The second line of defense against the COVID-19 pandemic’s mass destruction of employment are the Government’s emergency mitigation measures. According to the microsimulations, the expansion of the PBF will increase the income of these affected families but, overall, has only a marginal impact on income. After the PBF extension, the income of the poorest 20 percent increases by 1.5 pp relative to its level after the shock and unemployment protection are considered. The Auxilio Emergencial will have a more significant impact on the poorest 40 percent. Assuming that the transfers are well disbursed to all eligible individuals, the benefit would cover between 52.7 to 67.7 million workers at a cost of BRL 106 to BRL 135 billion, which is equivalent to 1.6 to 2 percent of the projected GDP for 2020 (Box 3). Importantly, we use the after-shock income to identify individuals who are eligible for the AE transfer. Box 3. AE modelling through three different scenarios According to the Auxilio Emergencial rules, to be a beneficiary, individuals must meet the following requirements: 1) be over 18; 2) be unemployed or work as informal, contractor or as a microentrepreneur; 3) have family per capita income under ½ minimum salary or family total income under 3 minimum salaries. 4) cannot be recipient of BPC, Unemployment Insurance, nor pensions. 5) earned less than BRL 28,559.70 in taxable income in 2018. 6) maximum of 2 benefits per household. To simulate the program, we build three different scenarios. The first, used as the baseline for results in this note, assumes perfect targeting and perfect compliance: all eligible people will receive the benefit and that even incomes that the government cannot track would be declared as were declared in the household survey (informal income, for example). The second estimates the impact of under-coverage and assumes that 50 percent of all eligible people who are not already Bolsa Familia beneficiaries will not gain access to the benefit. The third, here called Flexible, relaxes rules 3) and 6) in which the 19 government has difficulty enforcing requirements not easily confirmed through administrative records. The result is that families may receive more than two benefits when there are more than two eligible workers in the household, and the income used for AE does not include informal or other non-traceable incomes. Table B.3.1 shows the number of beneficiaries and the cost of the AE program considering the three scenarios mentioned above, and the two estimates of the COVID-19 income shocks explained on this note. Table B.3.1: Number of beneficiaries and cost of the AEs Number of beneficiaries (in millions) Scenario Baseline AE Flexible Under-coverage Baseline 52.7 67.1 40.2 Downside 53.9 67.7 40.8 Cost of program (in billions, BRL, and as a share of 2020 GDP projection) Baseline AE Flexible Under-coverage Baseline 106 (1.6%) 133 (2%) 81 (1.2%) Downside 109 (1.6%) 135 (2%) 82 (1.2%) Source: World Bank estimates based on BraSim microsimulation model. Figure 17: Income impacts of mitigation measures: CGE - Microsimulation Household Analysis (a) Effects of the expansion of PBF and the (b) Effects of the expansion of PBF and the implementation of AE on household income, by implementation of AE on national household poverty, quintile, baseline scenario by region, baseline scenario 20% 4% 3,2% 14,0% 2% 15% 0,2% 10% 0% -2% 5% 3,0% -1,9% -2,6% -2,4% -2,6% -4% 0% -4,5% -4,1% -6% -5,0% -4,8% -3,0% -2,2% -5,2% -5% -3,1% -5,9% -4,1% -4,2% -5,6% -5,2% -8% -6,9% -6,7% -7,5% -10% -10,9% -10,5% -10% -15% Q1 Q2 Q3 Q4 Q5 National Shock (Baseline) + UI + PBF + Transfer Shock (Baseline) + UI + PBF + Transfer Source: World Bank estimates based on BraSim microsimulation model. Note: The first set of bars (Shock (Baseline) + UI+ PBF) in figures a and b reflects the change in household income after the expansion of PBF. The second set (+ Transfer) reflects the cumulative change in household income after the expansion of PBF and the AE transfer 20 The key driver behind this impact is that a payment of BRL 600 (US$7.40 per day in 2011 PPP), or BRL 1200 to families in the poorest two quintiles will represent a boost relative to their pre-shock incomes. The payment corresponds to an increase of 14 and 3 percent of the first and second quintiles’ average income (Figure 17a), relative to their pre-shock income. Annex 11 presents the impact of the transfers in the face of a more severe employment disruption scenario. In this scenario, the effect of the transfers is still positive and cancels out almost the whole effect of the pandemic on the poorest 40 percent, and it is capable of buffering 60 percent of the shock suffered by the third quintile. The regional impact of the AE also reflects the power it has to increase the income of the poorest 20 percent. Figure 17b shows that the Northeast and North regions, and especially rural areas, are those most impacted by the AE, as these also concentrate lower-income populations. 3.2. Impact on Poverty and Inequality The results show that the shocks themselves are inequality increasing, due to the higher vulnerability of lower wage workers to the unemployment shocks. Without any mitigation measures, in our baseline scenario the crisis would increase inequality, measured by the Gini index, by 4 percent. Even after taking unemployment insurance into account (UI) the Gini index would still increase from 54.7 to 56.4, a substantial 3.1 percent increment (Figure 18). To put this inequality increase in perspective, the Gini index increased by 2.8 percent between 2015 and 2016 during the domestic crisis - this was the largest one-year increase since at least the early 1990s. Figure 18: Gini index after employment shocks and the Auxilio Emergencial The Auxílio Emergencial would represent a significant – though temporary - reduction in 57 56,4 inequality in Brazil in 2020, if well implemented. 57 Even with the estimated large increases in income 56 for the lowest income quintiles, inequality would 56 55,0 55 54,7 remain between 53.4 to 54.4, depending on 55 54,0 54,1 implementation success and the size of the shock. Gini 54 For context, this would still be higher than 54 53,4 inequality had been in 2015 (52.5). In our baseline 53 53 scenario the Auxílio Emergencial is expected to 52 reduce inequality with the Gini index falling from 52 54.7 in 2019 to 53.4, a decrease of 2.4 percent. 2016 2017 2018 2019 2020 Even if we consider the imperfect targeting of the Simulated AE, the Gini index would decrease to a range of Historic + Shock (Baseline) + UI + Policies 54.0 to 54.4, considering both scenarios. These Source: World Bank estimates based on BraSim and results show that, as seen the strong impact that the PNADC. AE is estimated to have on the lower quintiles’ policies implemented to refrain the income shock may also have a temporary but substantial redistributive power. As a result of the labor income shocks shown above, between 8.4 and 11 million Brazilians would fall into poverty depending on the length of the lockdown, already considering UI. The baseline scenario results in a 13.4 percent increase in the share of people living on less than ½ minimum salary (See Figure 19a), which increases from 29.1 to 33 percent, already accounting for the unemployment benefits received by formal workers. In terms of internationally comparable poverty lines, this suggests a potential of 7.2 million new poor at the $5.50 (2011 PPP) (Box 4). The effect is more substantial in urban areas, for which the poverty would rise by 16.5 percent, from 25.1 to 30.8 percent. Rural poverty is already at higher levels and is expected to increase less, by 4.6 percent, from 53.6 to 56.7 percent. 21 Box 4. Estimates of internationally comparable poverty rate Due to methodological differences in welfare aggregates, BraSim cannot be used to directly estimate the impact of the crisis on internationally comparable poverty rates. Since the ½ minimum salary line is close to the $5.50 per day line, the BraSim results can be used to inform the selection of elasticities for estimating poverty at the $5.50 per day line. These trends imply 7.2 million new poor at the $5.50 per day International poverty line for Upper Middle-Income Countries. Figure B.4.1 Poverty rate evolution, based on $5.50 (2011 PPP) per day 25% 22,7% 22,3% 23% 21,6% 21,2% 20,1% 20,4% 21% 19,9% 19,5% 19,4% 18,7% 19% 17,7% 17% 15% 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Source: World Bank estimates based on BraSim and SEDLAC (World Bank and Cedlas). Note: Projections use point-to-point elasticity (2015-2017) with pass-through = 0.7 based on private consumption per capita in constant LCU. Figure 19: Poverty impacts of household employment shocks after Unemployment Insurance (a) Effects of the pandemic on household poverty with (b) Effects of the pandemic on national household and without UI, baseline scenario, ½ MW line poverty after UI, baseline scenario, millions of people 25% 22,7% 3,50 3,2 Milhões 3,0 20% 18,2% 16,5% 2,4 2,50 2,1 15% 13,4% 2,0 10% 1,50 1,1 1,0 1,0 0,9 5,8% 0,9 4,6% 1,0 5% ,50 0,3 0,3 0% - National Pov Urban Pov Rural Pov Center North Northeast South Southeast Shock (Baseline) Shock (Baseline) + UI West 1/2 MW Poverty Line PBF Poverty Line Source: World Bank estimates based on BraSim microsimulation model. Most of the new poor come from the Southeast and Northeast regions, which amounts for 3.2 and 2.4 million people, or 38 and 29 percent of the new poor, respectively, on our baseline scenario. Due to the striking inequalities within the country, using the same poverty line for regional analysis can be misleading. Among the poorest 40 percent of Brazil, 68 percent of them live in the North or Northeast regions. Hence, if we take into account a lower poverty line, such as the PBF Poverty line, then 44 percent 22 of the new poor would come from the Northeast, 20 percent from the North, and only 24 percent from the Southeast, despite the latter having a population five times larger than the North (See Figure 19b). Regarding the age profile of the new poor, the increase in poverty across different age groups occurs similarly in the population under 65 years old, between 3.8 pp and 5.1 pp. Among those over 65 years old, there is an increase of only 0.8 pp in poverty (See Annex 1 and 2). A lower poverty increase for the elderly reflects the coverage of the public pensions system, which makes this groups less vulnerable to market income shocks. In the absence of the emergency measures implemented by the Government, unemployment insurance on its own would prevent 3 million people from falling into poverty. Despite being more concentrated on the top quintiles, the formal unemployment insurance would reduce the pandemic effect on national poverty, especially in urban areas. After considering UI, the increase in poverty nationally would fall from 18.2 percent 13.4 percent. From the 8.4 million people who would fall into poverty after a labor shock, most are in families of informal workers, representing 77.9 percent of the total (Figure 20). Formal workers under CLTs contracts correspond to 14.9 percent after considering the social security benefits to which they are entitled. Figure 20: Composition of workers who, after Combined, the PBF expansion and the AE suffering the income shock, would fall into have the potential to decrease poverty rates poverty, baseline scenario relative to the baseline of no pandemic. The overall effect of PBF expansion reduces poverty only marginally; the population living on less than BRL 178 per month (the PBF eligibility Formal, salary 14,9% criteria and a value close to $1.90 per day) is worker 7,2% estimated to decrease by 0.2 pp, and by 0.01 pp Formal, own- if we consider the ½ minimum salary line. This account is because the generosity of PBF remains low: 60 Informal 77,9% percent of families in PBF receive less than BRL 200 per month. Moreover, an estimated 450,000 eligible families continue to wait for access to PBF, without taking into account the likely increase in families needing assistance caused by Source: World Bank estimates based on BraSim the pandemic. By contrast, the Auxílio microsimulation model. Emergencial would fully reverse the expected annual national poverty increase of 13.4 percent due to the shock, and further decrease the poverty rate relative to the baseline pre-shock poverty level by 2.3 percent (Figure 21a).30 If we consider the extreme poverty line based on PBF, the impact of the AE on the poorest individuals becomes more evident. The extreme poverty rate would decrease from 7.8 percent, in 2019, to 5.5 percent in 2020, a 30 percent reduction (Figure 21b). 30These results are based on the baseline unemployment shock derived from CGE-based projections of sectoral income losses used above. This estimate takes into account unemployment benefits and the PBF Extension. 23 Figure 21: Poverty impacts of household employment shocks and the Auxilio Emergencial (a) Effects of the Auxilio Emergencial on household (b) Effects of the Auxilio Emergencial on household poverty, baseline scenario, ½ MW line poverty, baseline scenario, PBF line 20% 40% 16,5% 30% 30% 23% 15% 13,3% 20% 11% 10% 10% 4,5% 5% 0% -10% 0% -20% -2,3% -1,7% -5% -4,0% -30% -29% -29% -30% -10% -40% National Urban Rural National Urban Rural Shock (Baseline) + UI + PBF + Transfer Shock (Baseline) + UI + PBF + Transfer Source: World Bank estimates based on BraSim microsimulation model. With these mitigation measures, the number of Brazilians living on less than one half minimum salary could fall by almost 1.4 million in our baseline scenario, or increase by about 1.1 million in our downside scenario – in both cases, these are significant improvements relative to the expected 8.4 to 11 million new poor. As discussed, this improvement is wholly driven by the Auxílio Emergencial, which is relatively large if compared to the average disposable income of the poorest 40 percent. At the same time the PBF benefit amount is not enough to bring people out of poverty, as it represents a small fraction of the income necessary to push a family above half minimum wage. While poverty rates could fall as a result of the relative generosity and wide coverage of the AE transfer, it is important to note that, for about 4.4 million Brazilians who fall into poverty after the income shock, the AE will not be sufficient to pull them back. In total 8.4 million people would fall into poverty without mitigation measures (Box 5). The mitigation measures, especially AE, will protect 4 million of these individuals from falling into poverty. In addition, they will pull 5.8 million previously poor families out of poverty, resulting in a decrease of poverty rates. Even so, 4.4 million people whose families experience a labor income shock will fall into poverty – these are the new poor. They fall into poverty note because they do not get the AE, but rather because it is insufficient to compensate them for their income loss. They are mostly urban families from the Northeast (36.2 percent) and the North (30.4 percent), the poorest regions in the country, with heavy reliance on income from informal jobs. After mitigation measures, the median family in this category will see their monthly income fall to about R$420 per person – below the half minimum wage threshold but above the extreme poverty threshold. 24 Box 5. Poverty transition after the Covid19 mitigation measures Among the 8.4 million Brazilians who fall into poverty (½ minimum salary line) after the income shock, 4.4 million do not exit this condition after receiving the AE. Even with the surge of new poor, there is a decrease in the poverty rate caused by previously poor people who cease to be so after receiving the transfers of the mitigation measures (Table B.5.1). The profile of the new poor can be seen in detail in Table B.5.2. Table B.5.1 Poverty transition during Covid19 crisis Number of People (in millions) Become poor after the Covid19 shock + UI 8.4 Stay poor after the mitigation measures 4.4 Previously poor who exit poverty after mitigation measures 5.8 Net poverty results -1.4 Source: World Bank estimates based on BraSim microsimulation model. Note: Poverty line of half minimum salary per capita Table B.5.2 New poor profile before and after the mitigation measures New poor after the New poor who remain after income shock mitigations measures Number of people (millions) 8.4 4.4 Families with informal workers (%) 79 92 Families with children (%) 67 69 Children (%) 31 34 Urban families (%) 91 91 Families in South region (%) 11 10 Families in Southeast region (%) 38 10 Families in North region (%) 11 30 Families in Northeast region (%) 30 36 Families in Center-West region (%) 10 14 Source: World Bank estimates based on BraSim microsimulation model. The mitigation measures will also decrease the poverty gap, bringing the poor closer to a decent income level. The Poverty Gap index corresponds to the mean distance of the poor’s income from the chosen poverty line. Considering the half minimum salary poverty line and starting from a level of 13.5 percent, the national poverty gap increases to 15.9 percent after the baseline scenario income shock.31 However, after the inclusion of the Auxilio Emergencial the poverty gap decreases to 11.6 percent, lower than the baseline level (see Figure 22a). If we consider the PBF poverty line (Figure 22b), the relative impact is stronger, as the national gap drops from 3.8 percent to 1.2 percent, a 68 percent decrease. This is 31We use PNADC 2019 absolute numbers along the text, but this baseline poverty gap corresponds to BraSim’s original number (which is referenced in PNADC 2017). 25 because the Auxílio Emergencial (BRL 600) is large relative to the PBF line (BRL 178) and is a reflection of the substantial effect it has on the poorest individuals. A transfer of this magnitude brings the after-AE gap of the North and Northeast, the poorest regions, close to the pre-shock gap of the richest regions – a substantial improvement. Figure 22: Poverty gap after employment shocks and the Auxilio Emergencial a) Effects of the pandemic on poverty gap (1/2 MW), b) Effects of the pandemic on poverty gap (PBF Pov. baseline scenario Line), baseline scenario 9,0% 35% 8,0% 30% 7,0% 25% 6,0% 20% 5,0% 3,8% 13,5% 4,0% 15% 11,6% 3,0% 10% 2,0% 1,2% 05% 1,0% 00% 0,0% Baseline Shock (Baseline) + UI + PBF + Transfer Baseline Shock (Baseline) + UI + PBF + Transfer Source: World Bank estimates based on BraSim microsimulation model. Though the increase in income and the decline in poverty are significant gains, the reality is that household income in the lower income quintiles will remain low relative to the rest of the population even after receiving the AE (Figure 23). Though almost half of the expected Auxílio Emergencial (AE) beneficiaries were already PBF beneficiaries (46 percent), poverty reduction associated with the AE is driven by beneficiaries who are not in the PBF. This group, composed of informal workers, unemployed, and own-account workers, lives in households with higher income. In the end, the simulations suggest that over 1 million beneficiaries of the PBF may still fall into poverty even after receiving the AE (see Figure 24). Figure 23. Average per capita income (BRL) by Figure 24: Change in number of poor for PBF quintile, before and after, baseline scenario recipients and non-recipients, with and without transfer, baseline scenario 10 3.500 8 3.000 3,25 6 2.500 Millions 4 2.000 2 5,16 1.500 0 1,04 1.000 -2,47 -2 500 -4 - Without transfer With transfer Q1 Q2 Q3 Q4 Q5 Non-PBF PBF Pre-Shock Shock + UI + Transfer Source: World Bank estimates based on BraSim. Note: Income in this figure is reported in 2017 BRL. 26 3.3 Caveat: Perfect Targeting of AE An important caveat is that the default results impose no errors of exclusion or inclusion on the Auxilio Emergencial. There are two major concerns when implementing a program of this magnitude: (1) ensuring that the benefit is paid only to those who meet the criteria (errors of inclusion); and (2) ensuring that all eligible people receive the benefit (errors of exclusion). Eligibility for the AE is automatically evaluated for the individuals who are already in the Cadastro Único (CU), which simplifies substantially the implementation of the AE. Since the CU already contains data on a large fraction of the poorest Brazilians, including all PBF beneficiaries and those who had already applied, the risk of errors of exclusion for the poorest is low. By May 2020, the government had analyzed all 52 million individuals in the CU, of which 29.7 million were approved, including 96 percent of Bolsa Familia beneficiaries.32 The risk of exclusion is for those families who were not already receiving public assistance, including those whose income has fallen substantially due to the pandemic. Millions of eligible individuals who are not in the CU have needed to register using the official website or mobile app: as of June 16, 2020, 54.5 million individuals had been screened. Digital registration can be problematic for those with lower levels of schooling or without internet access.33 To measure the potential magnitude of this issue, we performed benchmark exercises using administrative data, and tried alternative modellings of the AE considering imperfect targeting scenarios (see Box 3). A state-by-state benchmark analysis suggests low errors of exclusions, as the number of beneficiaries in our baseline scenario is similar to the official numbers for most states. We conducted a benchmark analysis based on reporting by the Brazilian Government of state-level AE coverage (see Annex 12).34 For 20 out of 27 states, our estimates of the number of beneficiaries correspond to more than 90 percent of the benchmark, showing that our model’s AE distribution approximates reality well. This is especially true for the poorest and most rural regions (North and Northeast) which also have higher coverage of Bolsa Familia, for which our model averaged 95 percent of the benchmark. Though we expect low errors of exclusion, we simulate the impacts of AE under an imperfect targeting scenario. Aligned with how the AE benefit was administered, we assumed that all families that already receive Bolsa Família and who would see an increase in monthly benefits under AE would automatically be enrolled in AE. To simulate the AE with errors of exclusion, we assumed that only 50 percent of eligible families who were not in the PBF would receive the benefit. Under this scenario, instead of falling, national poverty increases by 2.9 percent in the baseline scenario and by 7.2 percent in the downside scenario (see Figure 25). 32Casalecchi, Alessandro (2020). “Cenários para a despesa com o auxílio emergencial 2020.” Nota Técnica Nº 42 7 de Maio de 2020. Instituição Fiscal Independente. 33P&S, Politicas Publicas & Sociedade (2020) “Covid-19: Políticas Públicas e as Respostas da Sociedade. Dificuldades com aplicativo e não uso da rede de proteção atual limitam acesso ao auxílio de emergência.” Boletim 5 8 de maio de 2020. Rede de Pesquisa Solidária. Politicas Publicas & Sociedade. https://redepesquisasolidaria.org/wp-content/uploads/2020/05/boletim5.pdf 34 Portal da Transparência: http://www.portaltransparencia.gov.br/beneficios. Accessed on 06/25/2020. 27 Figure 25: Poverty rate after shock and Auxilio Emergencial with errors of exclusion a) Effects of the pandemic, UI and imperfect target AE. b) Effects of the pandemic and imperfect target AE on on poverty rate (1/2 MW), baseline and downside poverty rate growth (1/2 MW), baseline scenario scenarios 40% 20% 16,5% 35% 15% 13,3% 10% 30% Baseline Shock 4,5% 4,5% 5% 2,9% 25% Downside Shock Imperf. Target. (Baseline)) 0% Imperf. Target. (Downside) 20% -1,4% -5% 2019 Baseline Shock Shock + UI Shock + UI + National Pov Urban Pov Rural Pov Policies Shock (Baseline) + UI + PBF + Transfer (Imperf) Source: World Bank estimates based on BraSim microsimulation model. Keeping in mind the important caveat of data limitations facing all survey-based models and noting that the income used for our modeling is based on adjusted 2017 household data, our analysis suggests potential errors of inclusion. Our estimates of coverage are below 80 percent of the reported numbers for only 2 states. The most distant estimate is for Roraima (72 percent); this is explained by the recent inflow of Venezuelan immigrants to this state who are eligible to the AE but are not well represented in the survey and hence in our model. For other states, a likely reason for the lower estimates of coverage in our model versus the official coverage rates is the potential of underreporting of informal income or household composition by some individuals when applying for the program. We tried to capture this possibility in our Flexible targeting scenario, which does not consider informal income in establishing AE eligibility and allows for more than two AE benefits per household (see Box 3). As of June 18, 2020, 63.5 million35 individuals had received the Auxílio Emergencial, a number closer to the flexible targeting estimates (67.1 million) than to our baseline (52.7 million). While these errors of inclusion imply an increase in cost of 25 percent, they are not expected to affect poverty estimates. The national poverty reduction of -2.3 percent observed in the perfect targeting scenario remains constant in the flexible targeting (see Annex 2). 3.4 After the Auxilio Emergencial ends The uncertainty around the length of the pandemic’s economic effects raises concerns regarding the temporary aspect of the Auxílio Emergencial. Therefore, authorities have begun discussing the extension of the benefit for two or three more months, but with a reduced value. Many proposals have been put forward, and we concentrate on two: (1) the proposal of 2 extra payments of BRL 300, which was widely discussed in early June 2020; and (2) the proposal of 3 extra payments of BRL 500, 400, and 300, which gained traction by late June. As a first step, we perform a month-after-shock analysis of an extension of AE where AE monthly benefits are reduced from BRL 600 to BRL 300. Considering the scenario where labor income shocks persist, and the government reduces the AE to BRL 300, Figure 26 shows that, on a monthly 35Source: . Accessed on 25/06/2020. 28 basis, this amount would be sufficient to fully mitigate the impact on the poorest 40 percent, but would no longer have an impact in the third quintile. Figure 26: Effects of the AE and extension on monthly household income, after shock, UI and PBF extension, baseline scenario 90% 78,0% 80% 70% 60% 50% 40% 34,5% 25,3% 29,4% 30% 20% 14,5% 10,2% 10% 4,9% 2,8% 4,7% 0% -10% -0,6% -2,2% -4,9% -2,6% -3,6% -20% -6,1% -5,0% Q1 Q2 Q3 Q4 Q5 Rural Urban National + Transfer (600) Shock + Transfer (300) Source: World Bank estimates based on BraSim microsimulation model. This figure reports the effect on monthly income. To analyze further this extension of the AE benefit, Figure 27 shows a month-after-shock growth incidence curve of the pandemic shock, the BRL 600 AE and the BRL 300 extension. The high generosity of the BRL 600 payment fully offsets the shock for the poorest 60 percent of the population. The lower benefit amount would still be able to continue fully mitigating the impact of the crisis on the poorest 45 percent of the population. It is also noticeable, however, that a further reduction on the benefit amount may decrease its effectiveness substantially. Even so, a BRL 300 extension of the program could serve to mitigate the impact of the income shock significantly after the end of the three months of the AE. Other policy options exist, such as improving the targeting of the program to focus on households living on less than half one minimum salary. Figure 27: Month-after-shock disposable Income Growth Incidence Curve – Baseline shock, Auxílio Emergencial and BRL 300 extension 200% Average Shock Baseline Shock 150% Shock + AE (BRL 600) 100% Shock + AE (BRL 300) 50% 0% -50% 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 Percentiles of per capita consumption 29 Source: World Bank estimates based on BraSim microsimulation model. Another potential policy consideration is to restrict the eligibility to AE benefits given the significant fiscal constraints facing the country and the continued need of support for households most severely effected. Specifically, the original AE included households using two income eligibility rules: (1) income up to half a minimum salary per capita; or (2) income up to 3 minimum salaries total. In practice, the second condition allows for smaller households with higher salaries to be included in the program. One way to improve the efficiency of the program is to reduce its coverage to needier families by reducing rule’s (2) threshold. For analyzing this possibility, we iterated the microsimulation model by varying only AE’s second income condition, keeping all the rest constant. We started with a threshold of BRL 460, which is above the half-minimum salary limit (BRL 437 in 2017, which is the year BraSim is based on), and increased it to BRL 4,000, in intervals of 10. The current policy threshold of 3 minimum salaries corresponds to BRL 2,811 in BraSim and is represented by the vertical line. Figure 28 shows that it is possible to reduce the costs of the program while maintaining its positive effects on poverty reduction. The figure shows the effect of varying the second eligibility criteria, the parameter on the x-axis, on 1) the total program costs (right y-axis) and 2) poverty (left y-axis). This happens because most of the poverty gains are from the inclusion of families who are eligible under the first income condition (up to ½ minimum salary per capita). Therefore, by increasing the family income threshold from BRL 460 to BRL 4,000, the new entrants into the program are families who are excluded based on the first rule, which may be individuals living alone or smaller families with income above the poverty line.36 Figure 28: Varying income eligibility thresholds of Auxílio Emergencial – BraSim X-Axis: AE family income threshold; Right-hand-side Y-Axis: AE total cost; Left-hand-side Y-axis: Poverty rate; Vertical line is the current policy rule 80% 120 Bilhões Total Cost of the Program (3 months) 70% 100 60% 80 50% Poverty Rate 40% 60 30% 40 20% 20 10% 0% 0 460 580 700 820 940 1660 2740 3820 1060 1180 1300 1420 1540 1780 1900 2020 2140 2260 2380 2500 2620 2860 2980 3100 3220 3340 3460 3580 3700 3940 Family income threshold 1/2 MW Poverty 1 MW Poverty Cost Source: World Bank estimates based on BraSim microsimulation model 36Note that this simulation assumes perfect targeting of the program. Imperfect targeting (errors of exclusion or inclusion) may cause the program’s cost to be less sensitive to changes in the rule. 30 Table 2 reports the poverty and fiscal implications of an extension of two months of the AE under three policy options: 1) same benefit amount but removal of the family income rule of the AE eligibility criteria; 2) half benefit amount and same coverage; 3) simple extension of current AE benefits. Assuming no errors of inclusion, compared to the simple extension under the third option, the first option would reduce the cost of the extension by up to BRL 21.2 billion, or 0.3 percent of 2020 projected GDP, without reducing the program’s poverty gains substantially (see Table 2, Panel A). Under the flexible targeting scenario, which allows for errors of inclusion, savings could still reach BRL 17 billion, or 0.25 percent of the GDP. As shown above, this adjustment of the eligibility threshold would not affect poverty rates, and extreme poverty (based on PBF inclusion threshold) would also remain unaffected. Extending the AE at half its current benefit value for an additional two months without adjusting eligibility criteria could increase the costs of the program by BRL 33.2 billion (0.5 percent of GDP) and reduce poverty by 2.1 p.p. An extension of the current benefits would represent a cost increment of 1 percent of GDP and a poverty reduction of 3.8 pp. If we consider the flexible targeting scenario, the poverty gains of the AE Extension remain the same, but the cost increase is of 1.0, 0.6 and 1.2 percent of the GDP, for the three scenarios respectively. Table 2: Auxílio Emergencial and Extensions: Summary of Social and Fiscal Indicators Panel A: Perfect Targeting of AE 2-month extension 2- month extension 2- month extension Pre-Covid AE modified rules* (BRL 300) (BRL 600) (BRL 600) PBF Poverty (%) 7.8 5.5 2.3 3.8 2.3 1/2 MW Poverty (%) 29.1 28.4 24.7 26.4 24.7 Gini 0.547 0.534 0.517 0.524 0.514 Cost (BRL Billions) - 106.4 151.7 139.7 172.9 Cost (2020 GDP % share) - 1.6 2.3 2.1 2.6 Panel B: Flexible Targeting of AE: Allowing Errors of Inclusion** 2-month extension 2- month extension 2- month extension Pre-Covid AE modified rules* (BRL 300) (BRL 600) (BRL 600) PBF Poverty (%) 7.8 5.5 2.3 3.8 2.3 1/2 MW Poverty (%) 29.1 28.4 24.6 26.4 24.6 Gini 0.547 0.533 0.515 0.522 0.513 Cost (BRL Billions) - 133.5 201.0 175.7 218.0 Cost (2020 GDP % share) - 2.0 3.0 2.6 3.2 * Considers only the 1/2 MW per capita income threshold for the AE eligibility, ignoring the 3 MW familiar income allowance. ** Families are allowed to receive more than two benefits, and the income threshold ignores informal and other non-traceable incomes. Source: World Bank estimates based on BraSim microsimulation model. Section 4: Conclusion The analysis shows the severity of income shocks related to the pandemic and the extent to which two key policy responses quickly implemented by the Government have been able to buffer these shocks during the initial months of the pandemic. The pandemic is expected to inflict a widespread income 31 shock in the Brazilian economy, especially on informal workers. More than 31 million workers are expected to suffer significant labor income loss or unemployment spells, out of which two thirds are expected to be informal or own-account workers. Of those workers who will fall into poverty, 78 percent are informal. This is due to the already predominance of informal workers among the poorest quintiles, and to their lack of access to unemployment protections available for public and CLT workers. Given the large number of vulnerable individuals in Brazil, the estimated income shock could substantially increase poverty in 2020. In a worst-case scenario, poverty could grow to 34.2 percent, above the 33.8 level observed in 2017 in the aftermath of the 2014-2016 domestic crisis. The existing unemployment insurance system buffers 30 percent of the expected income reduction, and though benefits are skewed towards the two top income quintiles, it would prevent 3 million people from falling into poverty. In addition to UI, the emergency measures introduced in response to the pandemic, in particular the temporary AE benefit, are able to, on average, absorb the income shocks related to the pandemic for the lowest income quintiles and reduce poverty. Under the baseline scenario, if well implemented the AE would temporarily reduce the number of poor by 1.4 million people relative to the pre-shock levels, reducing poverty to 28.5 percent, similar to the pre-crisis rate of 2014 (28.4 percent). This improvement is due to the relative generosity of the AE benefits and their effective targeting to the poorest 40 percent of the population. The end of the AE may imply, however, that poverty can rebound in 2021 even if employment begins to recover. Furthermore, this estimate is based on annualized income and assumes perfect consumption smoothing over the year – unlikely to occur as low-income households consume high shares of income on their basic needs. As a result, once the transfer program ends, these households will revert to lower income levels and, if the pandemic economic shock persists, they may suffer significant income shortfalls over the rest of 2020. As the AE comes to an end, policy makers and the public are asking what measures can be taken next. Upwards of 60 million people are receiving the transfer. Given its wide coverage, including households with income of up to three times the minimum wage, relative generosity at almost 60 percent the monthly minimum wage, and the tight fiscal situation facing the country, follow-up measures are likely to be smaller. The low generosity of the BFP and its limited coverage is unlikely to be enough to support the economically vulnerable population after the end of the Auxílio Emergencial. A potential extension of the AE for three months of BRL 300 would reduce the income impact of the shock on households in the poorest 40 percent, increasing their income on average by 25 and 5 percent, for the first and second quintiles respectively (instead of a 26-27 percent decrease without it). Otherwise, reducing the income eligibility to the AE can provide fiscal savings without increasing poverty. In addition to new transfers, the creation of labor reallocation mechanisms, such as professional training and incentives for job search during the paid time off from work, will be necessary for economic recovery and reduction of public spending. Despite the strong response already in place, there are several key vulnerabilities worth emphasizing. First, Brazil’s high inequality underlies structural challenges that cannot be resolved in the short term, including poor quality of urban housing and services, especially in informal poor access and overcrowding are hard to solve in the short-term. Hence, measures taken to address the urban poor’s low-quality housing will not be fast enough to mitigate the pandemic pressure on public health. Second, the still worsening health condition, both in Brazil and globally, suggests that the measures taken so far to reduce the spread of the disease have not been sufficient. This casts significant uncertainty over the extent of the economic crisis and the potential for economic recovery. 32 Annex Annex 1: Poverty Headcount, by scenario37 (1) Shock (4) Shock + UI + PBF + Auxilio Emergencial (AE) (2) Shock + Unemployment Insurance (UI) (5) Shock + UI + PBF + AE Flexible Targeting38 (3) Shock + UI + PBF extend (PBF) (6) Shock + UI + PBF + AE Imp. Targeting PBF Poverty Line National Urban Rural Center West North Northeast South Southeast Age 0-14 Age 65+ Benchmark 7.8 5.7 20.6 3.3 14.1 16.2 2.6 3.6 13.7 1.0 (1) 10.1 7.9 23.3 4.9 19.0 20.2 3.4 4.9 17.9 1.2 (2) 9.7 7.5 23.1 4.5 18.2 19.7 3.3 4.6 17.2 1.2 (3) 9.6 7.4 22.8 4.5 18.0 19.4 3.2 4.5 17.0 0.9 Baseline (4) 5.5 4.0 14.4 2.2 10.9 11.5 1.7 2.5 11.1 0.2 (5) 5.5 4.0 14.4 2.2 10.9 11.5 1.7 2.5 11.1 0.2 (6) 5.5 4.1 14.5 2.3 11.0 11.6 1.7 2.5 11.2 0.2 (1) 11.5 9.3 24.5 6.4 21.5 22.1 3.9 5.9 20.2 1.3 (2) 10.8 8.6 24.2 5.5 20.2 21.3 3.7 5.3 19.0 1.3 (3) 10.6 8.5 23.9 5.5 20.0 21.0 3.6 5.3 18.8 1.0 Downside (4) 5.9 4.4 14.9 2.4 11.6 12.3 1.8 2.6 11.9 0.3 (5) 5.9 4.4 14.9 2.4 11.6 12.3 1.8 2.6 11.9 0.3 (6) 6.0 4.5 15.0 2.5 11.8 12.4 1.9 2.7 12.1 0.3 1/2 Minimum Wage Poverty Line National Urban Rural Center West North Northeast South Southeast Age 0-14 Age 65+ Benchmark 29.1 25.1 53.6 19.1 46.8 49.2 14.4 19.3 47.0 8.5 (1) 34.4 30.8 56.7 25.8 53.5 54.8 18.6 24.2 53.1 10.1 (2) 33.0 29.3 56.1 23.9 52.1 53.4 17.4 22.8 51.5 9.7 (3) 33.0 29.3 56.1 23.8 52.1 53.4 17.4 22.8 51.5 9.7 Baseline (4) 28.4 24.7 51.5 19.6 46.6 47.9 14.1 18.7 46.2 7.3 (5) 28.4 24.7 51.5 19.6 46.6 47.9 14.1 18.7 46.1 7.3 (6) 30.0 26.3 52.9 21.0 48.4 49.6 15.3 20.1 48.1 8.3 (1) 36.2 32.8 57.6 28.6 55.5 56.3 20.2 26.1 55.0 10.8 (2) 34.2 30.6 56.9 25.6 53.4 54.6 18.5 24.0 52.9 10.2 (3) 34.2 30.6 56.8 25.6 53.4 54.6 18.5 24.0 52.9 10.1 Downside (4) 29.6 26.0 52.1 21.0 47.9 49.1 15.1 19.8 47.6 7.7 (5) 29.6 26.0 52.1 21.0 47.9 49.1 15.1 19.8 47.6 7.7 (6) 31.2 27.6 53.7 22.6 49.7 50.9 16.4 21.3 49.6 8.6 Source: World Bank estimates based on BraSim microsimulation model. 37 The poverty estimates calculated by the BraSim micro-simulation model are lower than those calculated based on the PNADC. This is because in the micro-simulation model we take into account simulated income from direct government transfers (Abono Salarial, Salario Família, BPC, Bolsa Família), which are usually underreported in PNADC. 38 The Flexible Targeting corresponds to our AE eligibility upper bound estimates, which apply the income eligibility excluding income that is not reported by a third party (informal and self-employment income) and ignores the limit of 2 benefits per household. 33 Annex 2: Poverty Headcount Growth, by scenario (1) Shock (4) Shock + UI + PBF + Auxilio Emergencial (AE) (2) Shock + Unemployment Insurance (UI) (5) Shock + UI + PBF + AE Flexible Targeting (3) Shock + UI + PBF extend (PBF) (6) Shock + UI + PBF + AE Imp. Targeting PBF Poverty Line National Urban Rural Center West North Northeast South Southeast 0-14 65+ Benchmark 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 (1) 29.8 39.7 12.7 48.5 34.6 24.4 30.0 37.3 30.9 21.0 (2) 24.7 32.0 11.9 37.7 28.8 21.3 26.4 27.9 25.6 18.2 (3) 22.9 30.0 10.6 35.7 27.0 19.7 23.3 26.1 24.1 -12.8 Baseline (4) -29.3 -28.9 -30.2 -32.8 -22.9 -29.3 -35.4 -31.5 -18.7 -77.0 (5) -29.4 -28.9 -30.2 -32.8 -23.0 -29.3 -35.4 -31.5 -18.7 -77.0 (6) -28.7 -28.1 -29.8 -30.3 -22.4 -28.7 -34.9 -30.8 -18.1 -77.0 (1) 47.7 64.2 18.9 94.9 52.3 36.4 52.6 64.0 47.6 27.8 (2) 38.6 50.9 17.3 67.8 43.0 30.9 43.6 49.1 38.8 23.5 (3) 36.9 49.0 15.9 66.4 41.1 29.4 41.0 47.2 37.5 -7.5 Downside (4) -24.4 -22.3 -28.0 -26.7 -17.7 -24.5 -30.0 -26.6 -13.2 -72.2 (5) -24.4 -22.3 -28.0 -26.7 -17.7 -24.5 -30.0 -26.6 -13.2 -72.2 (6) -23.0 -20.5 -27.5 -24.3 -16.4 -23.6 -27.7 -24.6 -11.9 -72.2 1/2 Minimum Wage Poverty Line National Urban Rural Center West North Northeast South Southeast 0-14 65+ Benchmark 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 (1) 18.2 22.7 5.8 35.1 14.3 11.4 28.8 25.3 12.9 19.7 (2) 13.4 16.5 4.6 24.8 11.2 8.6 20.7 18.1 9.6 14.7 (3) 13.3 16.5 4.5 24.8 11.2 8.5 20.7 18.1 9.6 14.1 Baseline (4) -2.3 -1.7 -4.0 2.6 -0.4 -2.6 -2.2 -3.5 -1.8 -13.4 (5) -2.3 -1.7 -4.0 2.5 -0.5 -2.7 -2.2 -3.6 -1.9 -13.6 (6) 2.9 4.5 -1.4 10.0 3.3 0.8 6.2 3.9 2.3 -2.2 (1) 24.5 30.6 7.4 49.6 18.6 14.5 39.7 34.8 17.0 27.3 (2) 17.6 21.8 6.0 33.9 14.1 11.0 28.4 24.1 12.5 20.3 (3) 17.5 21.7 5.9 33.8 14.0 10.9 28.3 24.1 12.5 19.7 Downside (4) 1.7 3.4 -2.9 10.1 2.3 -0.2 4.6 2.3 1.3 -9.6 (5) 1.7 3.4 -2.9 10.0 2.3 -0.2 4.6 2.2 1.3 -9.7 (6) 7.2 9.8 0.0 18.5 6.1 3.5 13.5 10.0 5.4 1.1 Source: World Bank estimates based on BraSim microsimulation model. 34 Annex 3: CGE Model Results Baseline Scenario - Wage Bill Variation (%) Wholesale and Other CGE Regions Agriculture Industry Transport trade services 1 Rondonia -5.2 -7.9 -6.6 -6.2 -17.8 2 Amazon -6.0 -11.1 -5.9 -5.5 -17.9 3 ParaToc -4.4 -12.8 -3.9 -4.1 -16.4 4 MarPiaui -2.8 -7.3 -5.1 -5.4 -15.2 5 PernAlag -3.0 -4.7 -5.9 -5.9 -14.1 6 Bahia -3.2 -3.0 -1.9 -4.6 -13.1 7 RestNE -2.8 -11.4 -8.5 -5.5 -15.6 8 MinasG -4.5 -6.4 -6.9 -4.8 -13.4 9 RioJEspS -5.0 -8.8 -4.0 -4.2 -13.8 10 SaoPaulo -6.1 -4.3 -3.5 -4.4 -10.6 11 Parana -3.9 -3.9 -2.7 -4.2 -11.8 12 SCatRioS -2.9 -5.0 -3.1 -4.0 -11.2 13 MtGrSul -4.1 -4.6 -2.8 -4.7 -14.0 14 MtGrosso -2.4 -5.7 -1.5 -5.8 -13.5 15 Central -6.1 -7.4 -6.6 -5.6 -15.0 Downside Scenario - Wage Bill Variation (%) Wholesale and Other CGE Regions Agriculture Industry Transport trade services 1 Rondonia -7.0 -11.2 -8.6 -10.0 -23.0 2 Amazon -8.1 -14.9 -8.4 -8.9 -23.2 3 ParaToc -6.0 -16.8 -5.4 -7.3 -21.2 4 MarPiaui -4.0 -10.8 -6.6 -9.2 -19.7 5 PernAlag -4.0 -7.4 -7.6 -10.0 -17.9 6 Bahia -4.4 -5.6 -3.0 -8.0 -16.9 7 RestNE -3.7 -15.2 -10.8 -9.3 -19.8 8 MinasG -6.0 -9.5 -9.0 -8.1 -17.2 9 RioJEspS -6.7 -12.3 -5.6 -7.3 -17.8 10 SaoPaulo -8.4 -7.1 -5.4 -7.7 -13.8 11 Parana -5.6 -6.6 -4.4 -7.4 -15.4 12 SCatRioS -4.4 -7.7 -5.0 -7.2 -14.7 13 MtGrSul -5.8 -7.0 -4.0 -8.2 -18.1 14 MtGrosso -3.7 -8.4 -2.4 -9.5 -17.5 15 Central -7.9 -10.3 -8.4 -9.3 -19.2 Source: World Bank estimates based on BraSim microsimulation model. 35 Annex 4: Poverty Gap, by scenario (1) Shock + Unemployment Insurance (UI) + PBF extend (2) Shock + Unemployment Insurance (UI) + PBF extend + Auxílio Emergencial (AE) Baseline Downside Benchmark (1) (2) (1) (2) National 3.8 3.9 1.2 4.2 1.3 Urban 2.9 3.0 0.9 3.3 1.0 Rural 9.1 9.3 3.1 9.6 3.2 Center West 1.8 1.8 0.5 2.0 0.5 Poverty Gap North 5.6 6.1 2.2 6.6 2.4 (PBF) Northeast 7.2 7.8 2.4 8.3 2.6 South 1.7 1.5 0.4 1.7 0.4 Southeast 2.2 2.1 0.6 2.3 0.6 0-14 5.9 6.5 2.5 7.0 2.6 65+ 0.1 0.0 0.0 0.0 0.0 National 13.5 15.9 11.6 16.7 12.3 Urban 10.9 13.3 9.6 14.2 10.3 Rural 29.1 31.2 24.0 31.9 24.5 Center West 7.7 10.2 7.0 11.3 7.8 Poverty Gap (1/2 North 22.5 26.4 20.3 27.7 21.3 MW) Northeast 24.5 27.7 20.7 28.7 21.6 South 5.8 7.1 4.9 7.6 5.3 Southeast 8.2 10.1 7.0 10.8 7.6 0-14 22.2 25.5 19.8 26.7 20.8 65+ 1.6 1.8 1.0 1.8 1.1 Source: World Bank estimates based on BraSim microsimulation model. 36 Annex 5: Poverty and population distribution, by region Table A5.1 Number of New Poor, by region Region 1/2 MW Poverty Line PBF Poverty Line New Poor % New Poor % Center West 852,068 10 267,966 6 North 996,696 12 953,918 20 Northeast 2,400,611 29 2,073,054 44 South 934,429 11 268,812 6 Southeast 3,219,395 38 1,098,473 24 Total 8,403,199 100 4,662,223 100 Source: World Bank estimates based on BraSim microsimulation model. Figure A5.1 Proportion of population living in each region, per decile 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1 2 3 4 5 6 7 8 9 10 Disposabe Income Deciles North Northeast Southeast South Center West Source: World Bank estimates based on BraSim microsimulation model. 37 Annex 6: Official Unemployment Insurance eligibility rules According to the Brazilian Law Brazilian law (Lei 7.998/1990), the on-the-job months required for SD eligibility can be distributed freely among different employers and throughout the previous 18 (or 36) months preceding the layoff date. In our model we do not consider previous employers, but only the current one and the corresponding job tenure. This is because we only have information regarding the current employment of the individual interviewed in PNADC, i.e. we do not know about their employment history. In any case, the implications of this assumptions are not worrisome. Table A6.1 Number of times the worker Time working before layoff Duration of the benefit requested the benefit date 12-23 months (out of the past 4 months 36) 1st +24 months (out of the past 36) 5 months 9-11 months (out of the past 12) 3 months 12-23 months (out of the past 2nd 4 months 36) +24 months (out of the past 36) 5 months 6-11 months (at least 6 3 months continuously) 12-23 months (out of the past 3rd 4 months 36) +24 months (out of the past 36) 5 months 38 Annex 7: FGTS modelling rules and detailed assumptions Method: We use only the information available in BraSim for estimating this amount. The total amount available FGTSi is calculated according to the following equation: T (1) FGTSi = ∑t i[(Wi × 0.08) × DMW,t × ((1 + 0.03)Ti−1 − 1)] Where FGTSi is the amount available in worker i's FGTS account, Ti is the current job tenure of worker i, Wi is the current labor income reported, and DMW,t is the Minimum Wage discount rate applied to the wages received at the period t of the current tenure. Assumptions: (1) We assume that unemployed individuals have already cashed their FGTS accounts in the moment they were fired in the past. This need for cash may be especially true for low-income individuals, in which we are more interested. (2) We assume that the individuals’ current wage has been stable since the start of the current employment at time t-n, growing in nominal terms at the same rate as the minimum salary. While this may underestimate the income growth of higher income groups, it is not a problematic assumption for low wage workers whose earnings are anchored by the minimum wage. 70 percent of CLT workers earn 2 minimum wages or lower. (3) We assume that the individual cashed their FGTS accumulated during their active employment contract following a fixed rate that decreases arithmetically with age (See Annex X+3). In order to validate our assumptions, we compared our imputed FGTS with a benchmark based on administrative data which shows the aggregate balance of FGTS accounts per income bracket). For this group, our estimates are very high – 254 percent of the benchmark, driven by younger workers - probably because this group widely used the withdraws permitted by the government. Comparing to benchmark, simulated FGTS Balance for the poorest (income <1000) is higher by BRL 16 Billion. Table A7.1 FGTS Balance in BraSim vs. Benchmark BraSim Benchmark Income Bracket (A)/(B) (A) (B) <1000 26,722,636,341 10,510,925,596.40 254 1000-2000 101,076,225,830 110,239,468,097.17 92 2000-3000 55,106,489,065 62,388,779,114.30 88 3000-5000 50,963,924,435 59,368,687,824.91 86 5000-10000 49,774,815,482 57,947,881,254.36 86 10000-20000 31,899,668,935 41,364,267,535.01 77 >20000 17,049,490,870 37,184,826,225.28 46 Total 332,593,250,958 379,004,835,647 88 Source: World Bank estimates based on BraSim microsimulation model and FGTS. 39 To correct for this, we applied, only for this income group, a withdrawing rate that decreases with age, starting at 90 percent for workers aged 18 or less. Table A7.3 Age-regressive withdrawal rate for low income (<1000) Proportion Age Bands Age Bands Proportion Withdrawn Withdrawn <18 90.0 45-49 51.5 18-20 84.5 50-54 46.0 21-24 79.0 55-59 40.5 25-29 73.5 60-64 35.0 30-34 68.0 65-69 29.5 35-39 62.5 >69 24.0 Source: World Bank estimates based on BraSim microsimulation After these corrections, the estimated FGTS accounts aggregate balance matched the government’s benchmark number, reducing to 10.7 Billion (compared to 10.5 of the benchmark). Table A7.3 Adjusted FGTS Balance in BraSim vs. Benchmark BraSim Benchmark Income Bracket (A)/(B) (A) (B) <1000 10,722,655,555 10,510,925,596.40 102 1000-2000 101,076,225,830 110,239,468,097.17 92 2000-3000 55,106,489,065 62,388,779,114.30 88 3000-5000 50,963,924,435 59,368,687,824.91 86 5000-10000 49,774,815,482 57,947,881,254.36 86 10000-20000 31,899,668,935 41,364,267,535.01 77 >20000 17,049,490,870 37,184,826,225.28 46 Total 332,593,250,958 379,004,835,647 88 Source: World Bank estimates based on BraSim microsimulation model and FGTS. 40 Annex 8: Average income, by definition, by disposable income quintile Table A8.1 Disposable Income Quintiles Income Category Q1 Q2 Q3 Q4 Q5 (a) Gross market income plus pension 86 412 803 1,380 4,221 (b) Net market income 81 373 698 1,173 3,445 (c) Disposable income 272 459 762 1,215 3,467 (d) Consumable income 247 424 701 1,096 3,038 (e) (c/a) 314% 111% 95% 88% 82% Source: World Bank estimates based on BraSim microsimulation model. 41 Annex 10: Downside Scenario Figure A10.1 Poverty and income impacts of household employment shocks and the Auxilio Emergencial: CGE - Microsimulation Household Analysis (a) Effects of the pandemic on household (b) Effects of the pandemic on household poverty, incomedownside scenario downside scenario 0% 60% 50% -5% 40% -10% 30% -15% 20% 10% -20% 0% -25% Q1 Q2 Q3 Q4 Q5 National Shock (Downside) Shock (Downside) + UI Shock (Downside) Shock (Downside) + UI c) Effects of the Auxilio Emergencial on household d) Effects of the Auxilio Emergencial on household income and poverty, downside scenario income and poverty, downside scenario 40% 15% 35% 10% 30% 5% 25% 20% 0% 15% -5% 10% -10% 5% 0% -15% -5% -20% National Q1 Q2 Q3 Q4 Q5 Income Income Income Income Income Income Shock (Downside) + UI + PBF + Transfer Shock (Downside) Shock (Downside) + UI Source: World Bank estimates based on BraSim microsimulation model and FGTS. 42 Annex 11: AE rules and their presence in BraSim model Auxílio Emergencial rules BraSim Three monthly payments of BRL 600. ✓ Max 2 recipients per household ✓ Single mothers can receive two benefits ✓ +18 years old ✓ MEI, self-employed, informal workers, and unemployed individuals. ✓ Family per capita income of maximum ½ minimum salary or Family total income of ✓ max 3 MW Cannot be recipient of BPC, UB, INSS, nor pensions. ✓ Individuals must choose between PBF and this emergency benefit. ✓* Earned less than BRL 28,559.70 in taxable income in 2018. *We assume all eligible PBF families switch for the 3 months duration of the Program, as the PBF benefit is substantially lower Source: World Bank. 43