Poverty Projections and Distributional Impacts of the COVID-19 Outbreak in Armenia, Azerbaijan and Georgia1 Last Update: 13 May 2020 1 This note was prepared by the South Caucasus Poverty, Equity and Gender Team from The World Bank. Corresponding authors are Natsuko Kiso Nozaki (nkiso@worldbank.org), Alan Fuchs (afuchs@worldbank.org), Dhiraj Sharma (dsharma5@worldbank.org) and Maria Fernanda Gonzalez Icaza (mgonzalezicaza@worldbank.org). 1 Contents I. Key Takeaways ................................................................................................................................... 3 II. Data sources and Methodology.......................................................................................................... 8 (1) Macro Simulations ........................................................................................................................ 8 (2) Micro Simulations ........................................................................................................................ 9 III. Main Results ...................................................................................................................................... 11 Results for Armenia ............................................................................................................................ 11 Results for Azerbaijan......................................................................................................................... 15 Results for Georgia ............................................................................................................................. 20 IV. Conclusion ......................................................................................................................................... 24 V. Annex ................................................................................................................................................. 26 A.1. Real GDP growth rate from MFMod database used in Macro simulations ................................ 26 A.2. Methodology for Nowcasting household’s consumption aggregates using GDP ratios ............. 27 A.3. Methodology for Micro simulations ........................................................................................... 28 A.5. Assumptions for extended Micro simulations on subsectors of employment ............................. 31 A.6. Additional welfare and distributional micro-level analysis in Armenia ..................................... 32 A.7. Additional welfare and distributional micro-level analysis in Azerbaijan .................................. 35 A.8. Additional welfare and distributional micro-level analysis in Georgia ...................................... 37 2 I. Key Takeaways This note summarizes the results of micro- and macro-economic simulations to assess the short-term impacts of the COVID-19 outbreak on poverty and distributional outcomes in Armenia, Azerbaijan and Georgia. The simulations project that poverty could increase between 2.2 to 5.2 percentage points in Armenia, 1.5 to 3.2 percentage points in Azerbaijan, and 2.2 to 3.6 percentage points in Georgia, compared to a counterfactual scenario in 2020, without the COVID-19 crisis. In all cases, inequality is expected to rise along poverty indicators. Income losses associated with unemployment are the main drivers of poverty increases in the three countries. Households in Yerevan, Baku City and Tbilisi, and other urban centers are most likely to be impacted in the short-term, along with households relying on incomes from service activities. The analysis leverages the latest available household microdata in each country,2 under two approaches. (1) Macro-Simulations: Forecasted real growth rates of GDP are used to nowcast household consumption and to predict the welfare status of the households in 2020, under: (a) counterfactual scenario without COVID-19), and (b) assuming different macroeconomic shocks derived from the pandemic (baseline and pessimistic scenarios).3 The approach assumes that all households are impacted equally from the crisis, with no distributional effects. (2) Micro-Simulations. Different transmission channels from COVID-19 to household welfare are estimated, including: (a) the risks of unemployment, (b) losses in wage incomes among the employed, and (c) reduced remittances inflows. The analysis assigns different shocks to each sectors of economic activity, and it simulates the corresponding losses in household welfare from survey microdata. The microsimulations incorporate the distributional impacts of COVID-19. Nonetheless, they rely on a partial equilibrium approach. All simulations represent short-term economic impacts of the COVID-19. They do not capture indirect economic impacts of the pandemic. At this moment, they do not incorporate the potential benefits of new social protection schemes and other policy responses to COVID-19. The dynamics, duration, and economic channels of the pandemic will ultimately determine its effects on poverty and inequality. Results will evolve as new data and knowledge becomes available to assess the full impact of the pandemic. 2 The 2018 Integrated Living Conditions Survey (ILCS) for Armenia; the 2015 Azerbaijan Monitoring Survey for Social Welfare (AMSSW); and the 2018 round of Household Income and Expenditure Survey (HIES) for Georgia. 3 Macroeconomic projections are taken from the most relevant sources: MTI’s MFMod database, the World Economic Outlook (April 2020), and projections by the MTI country teams. 3 The main results of the simulations are summarized in Table 1 and Figures 1. Table 1. Estimated poverty indicators, 2020 COVID-19 COVID-19 Counter COVID-19: Poverty Index Macrosimulation: Macrosimulation: factual Microsimulation Baseline scenario Pessimistic scenario Armenia Headcount Ratio 9.0 11.2 12.9 14.2 Gap 1.9 2.6 2.9 3.9 Severity 0.6 0.9 1.0 2.6 Azerbaijan Headcount Ratio 4.0 5.5 6.2 7.2 Gap 0.7 1.0 1.1 1.5 Severity 0.2 0.3 0.3 0.5 Georgia Headcount Ratio 13.3 15.5 16.9 16.1 Gap 3.8 4.6 5.0 4.8 Severity 1.7 2.0 2.2 2.2 Sources: Based on data from national household surveys: ILCS 2018 (Armenia), AMSSW 2015 (Azerbaijan), and HIES 2018 (Georgia); macroeconomic forecasts from the MFMod database (as of February and April 2020), WEO (April 2020) and MTI (May 2020). Impacts on poverty Armenia. The Macrosimulations suggest that poverty (defined by the lower-middle income class poverty line of USD 3.20 PPP) could increase in the range of 2 to 4 percentage points in 2020, relative to a counterfactual scenario. The microsimulations yield higher increases in the poverty rate, poverty gap, and poverty severity, compared to the macro-level exercises. Azerbaijan. Poverty is expected to increase by 1.5, 2.2, and 3.2 percentage points in 2020, under the baseline macrosimulation, macrosimulations under a more pessimistic outlook, and the microsimulations exercises, respectively. The outbreak not only affects the poverty rate but also on the intensity of poverty reflected in the rise in poverty gap and severity indices, with greater impact under the micro-simulation. Georgia. The pessimistic macroeconomic scenario in Georgia (assuming negative growth rate of -7.5) results in the highest poverty increase of 3.6 percentage points, with respect to the counterfactual of 13.3 poverty rate. Micro simulations project a more modest increase of 2.2 percentage points in poverty (USD 3.20 PPP 2011 poverty line). 4 Figure 1. Summary of Projected Poverty Indices Panel A. Effect of COVID-19 on projected poverty rates in Armenia 15% (Poverty line of USD 3.20 per day, 2011 PPP) 14.2% 14% 13.0% 12.9% 13% Headcount poverty rate 12% 11.2% 10.5% 11% 10% 10.5% 9% 8% 9.0% 7% 6% 5% 2018 2019n 2020f Pre-shock poverty rates Counterfactural COVID-19 Macro-simulation: Baseline scenario COVID-19 Macro-simulation: Pessimistic scenario COVID-19 Micro-simulation Panel B. Effect of COVID-19 on projected poverty rates in Azerbaijan (Poverty line of USD 5.50 per day, 2011 PPP) 8% 7.2% 7% 6.2% 6% 5.5% 5.6% 5.6% 5.3% % of Population 5% 4.6% 5.5% 4% 4.0% 3% 2% 1% 0% 2015 2016e 2017e 2018e 2019e 2020f Pre-shock poverty rates Counterfactual COVID-19 Macro-simulation: Baseline COVID-19 Micro-simulation COVID-19 Macro-simulation: Pessimistic 5 Panel C. Effect of COVID-19 on projected poverty rates in Georgia (Poverty line of USD 3.20 per day, 2011 PPP) 18% 16.9% 17% 16.1% Headcount poverty rate 15.7% 16% 15.5% 15% 13.9% 14% 13% 13.9% 13.3% 12% 11% 10% 2018 2019n 2020f Pre-shock poverty rates Counterfactural COVID-19 Macro-simulation: Baseline scenario COVID-19 Macro-simulation: Pessimistic scenario COVID-19 Micro-simulation Sources: Based on microdata from national household surveys: ILCS 2018 (Armenia), AMSSW 2015 (Azerbaijan), and HIES 2018 (Georgia); macroeconomic forecasts from the MFMod database, WEO (April 2020) and MTI (May 2020). Notes: e: Estimated, n: Nowcasted, f: Forecasted. The counterfactual scenario is based on business-as-usual GDP forecasts from MFMod (as of February 2020). Other country-specific findings Armenia - The effects on the national poverty rate would be most significant, with a 7.2 percentage points increase forecasted in the microsimulation model. - The microsimulations show important heterogeneity. Secondary cities suffer the highest poverty increases (7.2 percentage point increase, compared to 5.1 percentage points at the national level). Yerevan and Sjunik marz face higher shares of income loss. Rural areas face the lowest impact (6 percent of average income loss, compared to the national average of 9 percent). The distributional impact of income losses due to COVID-19 is regressive, with lower-income households facing higher relative losses. - More disaggregated microeconomic simulations—assuming heterogenous shocks across 21 subsectors of employment—suggest that the largest source of poverty increases comes from unemployment in retail and tourism (using the poverty line of USD 3.20 PPP 2011). However, unemployment in retail, construction and manufacturing is more relevant for extreme impoverishment (USD 1.90 PPP 2011 poverty line). 6 - Over 470 thousand Armenians could suffer downward mobility, by falling to a lower-welfare group. 234 thousand people who were not poor before the crisis could become impoverished as a result of the economic impacts of COVID-19. Azerbaijan - The overall shock in the services sector (including layoffs and decrease in wage) contributes the most to the increase in poverty under the micro simulation, followed by the increase in unemployment across all sectors. - There are distributional impacts observed across three dimensions in terms of income loss as a share of household’s total income: (1) gap across poverty status, with poor households experiencing more than twice as large loss compared to the non-poor (21.4 percent versus 10.2 percent respectively), (2) spatial gap – with larger share of loss among the urban households (concentrated in Baku and Absheron region) compared to rural counterpart (12 percent in urban versus 9.7 percent in rural), and (3) social economic status – with larger loss among the non-beneficiaries of targeted social assistance (TSA) compared to TSA beneficiaries (10.1 percent versus 7.8 percent). - Projections are based on the 2015 AMSSW, which is the latest household survey available to the World Bank for the analysis. Arguably, economic structure - including employment and income sources - may be outdated to be used as a base for projection. Given that reliable data are the foundation of effective analytical work and evidence-based policy making, the World Bank will continue to make efforts to stepping up engagement with the State Statistical Office and the line ministries to coordinate on the data collection initiative and efforts to enhance statistical capacity building. Georgia - The risks of unemployment are the main driver of poverty increases under the micro simulations. - On average, income losses from COVID-19 represent 8.0 percent of household incomes in Georgia. However, the shares of income losses do not show a clear distributional pattern across percentiles of the population. Income losses are highest in Tbilisi and lowest in rural areas. - Residents of Tbilisi, households with larger number of children, and households not reporting pension incomes will most likely face higher probabilities of unemployment. - Half a million Georgians are at risk of suffering downward mobility as a result of COVID-19 in 2020. Over 200 thousand people who were nonpoor before the crisis could become impoverished. - The extended micro simulation model results in similar poverty changes to the 3-sectoral microeconomic shocks: international poverty increases by 3.3 percentage points and the national absolute poverty rate increases by 3.8. 7 II. Data sources and Methodology The analysis uses the latest available survey data in Armenia, Azerbaijan and Georgia to project the welfare of households in 2020 and estimate the impact of COVID-19 crisis on poverty. The most recent survey used for each country is: 2018 round of Integrated Living Conditions Survey (ILCS) for Armenia; the 2015 Azerbaijan Monitoring Survey for Social Welfare (AMSSW); and 2018 round of Household Income and Expenditure Survey (HIES) for Georgia. (1) Macro Simulations The projected real growth rate of GDP per capita is used to adjust the household consumption aggregate in 20194 and to predict the welfare status of the households 2020. To assess the impact of COVID-19 on economic growth, we estimate a 2020 business-as-usual or counterfactual scenario, by leveraging pre-crisis estimations of GDP growth produced by MTI prior to the COVID-19 outbreak (February 25, 2020). Second, ex-post economic growth forecasts are used to predict the marginal effects of the pandemic on household welfare. Growth forecasts are taken from several available sources, including MTI’s database (MFMod) (latest forecasts dated April 13, 2020),5 other forecasts from MTI country teams, and the World Economic Outlook (April 2020). Table 2 summarizes the GDP growth forecasts and the sources of data. For all countries, a “baseline” scenario of GDP growth informs the main post-COVID macro-simulations. For Armenia, the baseline scenario was provided by the country economists, projecting negative growth rate of -2.8 percent in 2020. For Georgia, the baseline is taken from IMF projections, at -4.0 percent GDP growth in 2020.6 For Azerbaijan, a baseline scenario of -1.0 percent growth in 2020 is considered. For all three countries, additional forecasts by country economists are used in a more “pessimistic” or upper-bound scenario of the economic impacts of the pandemic. In contrast to the micro-simulation approach, this approach assumes that all households are impacted equally from the crisis with no distributional impact. A pass-through rate (from income growth to poverty changes) of 100 percent is assumed in all simulations. 4 Consumption aggregates in 2019 are nowcasted using the latest household survey available in each country. Methodological details are described in the Annex. 5 Data are available from \\gpvfile\GPV\Knowledge_Learning\Pov Projection\Central Team\MFM-allvintages.dta. The methodology and assumptions incorporated into the macro projection in MFMod database are unknown to the authors. Since our poverty projections are based on these macro forecasts, estimates will change with the updates on these macro figures. 6 World Economic Outlook, April 2020: The Great Lockdown. (https://www.imf.org/en/Publications/WEO/Issues/2020/04/14/weo-april-2020). 8 Table 2. Assumed macroeconomic scenarios and comparators Assumed GDP forecasts Comparators: Pessimistic Baseline GDP forecasts from World scenario scenario Economic Outlook Armenia [-6.5%, -2.8%] -1.5% Azerbaijan [-4.4%, -1.0%] -2.2% Georgia [-7.5%, -4.0%] -4.0% Source: Forecasts from the country teams as of May 5, 2020, and World Economic Outlook (April 2020). Note: Azerbaijan’s baseline scenario is taken from the MFMoD database (forecasts as of April 13, 2020). (2) Micro Simulations The impacts of COVID-19 are also simulated from a microeconomic perspective, by identifying possible transmission channels to household welfare, through: unemployment, labor incomes, and other sources of household incomes.7 After nowcasting the household welfare aggregates to 2020,8 this approach assumes heterogenous shock parameters for three sectors of economic activity (agriculture, industry and services), and it calculates the corresponding losses in household welfare. The detailed methodology is included in Appendix A.3. The main assumptions of the micro-models are: Box 1. Assumptions in Micro-level simulations 1. Loss of labor incomes 1.a. Unemployment shock. A share of workers faces reduced incomes due to layoffs or loss of jobs: • Both hired employees and self-employed workers can be potentially laid-off. • The unemployment shock in the model is randomly assigned across all workers of one sector. • Workers in service activities face a probability of .30 of losing their jobs. The corresponding probability for industry is .10. Agricultural workers do not become unemployed as a result of COVID-19. • Workers losing their job suffer a 100 percent wage income loss. 1.b. Reduced wage incomes. Workers who remain employed may also face partial losses to their labor incomes, resulting from declining salaries, reduced hours worked, sickness, etc. Active workers face: • 30 percent decrease in wage incomes in the service sector; • 20 percent decrease in wage incomes from industry; and 7 The note follows the framework proposed in the note published in April 2020 by the World Bank’s Poverty & Equity Global Practice, Poverty and Distributional Impacts of COVID-19: Potential Channels of Impact and Mitigating Policies. 8 A standard nowcasting procedure is applied to “update” the latest available household survey data to 2020. Thi s procedure yields a counterfactual scenario in 2020, that is consistent with the pre-COVID macroeconomic models. 9 • 10 percent decrease in wage incomes from agriculture. 2. Declines in other household incomes. • Household incomes are expected to fall due to declining remittances inflows, as the global and national economies contract. Remittances incomes are assumed to fall by 30 percent.9 • Household incomes from agricultural sales—when available in the data for Georgia and Armenia—are assumed to fall by the same proportion as agricultural wages (e.g. 10 percent). All shocks are assumed to last for 3 months of 2020. Table 3. Summary of assumptions in Micro-level simulations Fall in other household Probability of unemployment Declined labor incomes Scenario incomes Agricultural Services Industry Agriculture Services Industry Agriculture Remittances sales COVID- 19 Micro- .30 .10 0 30% 20% 10% 30% 10% simulation Source: Author’s. Notes: Workers assigned to unemployment lose 100 percent of their wage incomes from that sector. The effects on agricultural sales is only calculated for Armenia and Georgia. Labor incomes consider reported wages from employed and self- employed workers. The microeconomic approach does capture the distributional impacts of COVID-19. However, results should be interpreted as a lower-bound scenario, as they only account for the short-term impacts of COVID- 19, by assuming that household incomes and employment will be affected for three months of 2020 only. In the case of Georgia and Armenia, an extended microeconomic analysis is performed to assess more disaggregated effects of COVID-19 on different subsectors of the economy. The household surveys are used to identify and to shock 21 subsectors of employment, including tourism activities.10 Assumptions are summarized in Appendix A.5. The ultimate impact of the pandemic will depend on its severity and duration of the pandemic which may be different from the assumptions of our model. Thus, the eventual economic impact may be different from the results reported here and this analysis should be considered as illustrative based on the preliminary evidence. 9 According to The World Bank Remittance Report, 2020, global remittance is predicted to decline by 20 percent globally, and 28 percent for ECA region. In the simulation, the rate of contraction is rounded up to 30 percent. 10 The 21 subsectors are classified following NACE Rev. 2. 10 III. Main Results This section shows poverty projections for the three countries and presents potential distributional impacts under the micro simulation approach. Poverty rates are estimated against the international poverty lines of USD 5.5 for Azerbaijan and USD 3.2 for Armenia and Georgia, all in 2011 PPP terms. Results for Armenia The Macrosimulations suggest that poverty (defined by the lower-middle income class poverty line of USD 3.20 PPP) could increase in the range of 2 to 4 percentage points in 2020, relative to the counterfactual scenario (Figure 2). The microsimulations yield higher increases in the poverty rate, poverty gap, and poverty severity, compared to the macro-level exercises. According to the micro simulation, the poverty gap could double, and poverty severity could quadruple following the welfare and distributional effects of COVID-19 (Figure 3). Figure 2. Effect of COVID-19 on projected poverty rates in Armenia Effect of COVID-19 on projected poverty rates in Armenia 15% (Poverty line of USD 3.20 per day, 2011 PPP) 14.2% 14% 13.0% 12.9% 13% Headcount poverty rate 12% 11.2% 11% 10.5% 10% 10.5% 9% 9.0% 8% 7% 6% 5% 2018 2019n 2020f Pre-shock poverty rates Counterfactural COVID-19 Macro-simulation: Baseline scenario COVID-19 Macro-simulation: Pessimistic scenario COVID-19 Micro-simulation Sources: ILCS 2018, MFMod database and MTI projections. Notes: n = nowcasted, f = forecasted. The counterfactual scenario is based on business-as-usual GDP forecasts from MFMod (as of February 2020). The baseline and pessimistic macro simulations assume GDP growth of -2.8 percent and -6.5 percent in 2020, respectively. 11 Figure 3. Projected Effect of COVID-19 on Depth of Poverty in 2020 Panel A. Poverty Gap under different simulations, Armenia 2020 4.5 3.9 4.0 3.5 2.9 3.0 2.6 2.5 1.9 2.0 1.5 1.0 0.5 0.0 Counterfactual COVID-19 Macrosimulation: COVID-19 Macrosimulation: COVID-19: Microsimulation Baseline scenario Pessimistic scenario Panel B. Poverty Severity under different simulations, Armenia 2020 3.0 2.6 2.5 2.0 1.5 1.0 1.0 0.9 0.6 0.5 0.0 Counterfactual COVID-19 Macrosimulation: COVID-19 Macrosimulation: COVID-19: Microsimulation Baseline scenario Pessimistic scenario Sources: ILCS 2018, MFMod database and MTI projections. Notes: n = nowcasted, f = forecast. The counterfactual scenario is based on business-as-usual GDP forecasts from MFMod (as of February 2020). The baseline and pessimistic macro simulations assume GDP growth of -2.8 percent and -6.5 percent in 2020, respectively. Poverty indicators based on the poverty line of USD 3.20 (PPP 2011). The distributional impact of income losses due to COVID-19 is somewhat regressive, with lower-income households facing higher relative losses (Figure 4). Applying the micro simulation model suggests that increased unemployment is the most significant source of poverty increases, followed by the loss of wage incomes. Employment in services is most affected (Figure 5). 12 Figure 4: Distributional incidence of income losses related to COVID-19 25% Share of income losses in Armenia 20% Share of income loss 15% 10% 5% 0% 19 94 10 13 16 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 97 1 4 7 100 Percentile of household per capita consumption Sources: Microsimulations based on the ILCS 2018 and the MFMod database. Note: Welfare percentile is based on the projected consumption aggregate under the micro simulation for 2020. Figure 5. Effects of micro-simulation on poverty rates Poverty projection by type of shocks Armenia 2020, Micro Simulation 14% 12.6% 11.7% 12% 10.4% 9.6% 9.7% 10% 9.1% 9.1% 8% 6% 4% 2% 0% Increase in Wage decrease Shock to Drop in Agriculture Industry Services unemployment agricultural remittances sales Type of shock Shock by sector of employment* Sources: Microsimulations based on ILCS 2018, MFMod database and MTI projections. Notes: All poverty rates are based on the poverty line of USD 3.20 (PPP 2011). Shocks follow the assumptions presented on Table 2. Effects are not mutually exclusive. Households may be affected by multiple types of shocks. The shocks by sector of employment incorporate the channels of unemployment and wage losses for workers in each sector. 13 The microsimulations also reveal heterogenous effects on income and poverty across location and population groups. Evidence in Annex A.6. suggests that secondary urban centers suffer the highest poverty increases (7.2 percentage point increase, compared to 5.1 percentage points at the national level). Yerevan and Sjunik marz face the highest shares of income loss. Rural areas face the lowest impact (6 percent of average income loss, compared to the national average of 9 percent). The simulations also point at the relevant role of social protection policies in shielding against the economic consequences of the pandemic. For example, preliminary evidence suggests that households receiving pensions and receiving the family benefit may observe lower increases in poverty rates, compared to households not receiving pensions or family benefit (Figure A.6.b). Table 4 shows that 234 thousand Armenians who were not poor before the crisis could become impoverished as a result of the economic impacts of COVID-19. Over 470 thousand Armenians could suffer downward mobility. People in vulnerable households and the middle class could fall to poverty—including extreme poverty in some cases—after the economic loses from the pandemic (Figure 6). Table 4. Number of people suffering impoverishment or downward mobility Impoverishment Downward Mobility Unemployment 109,623 220,032 Income loss in wages 96,334 171,208 Income loss in agriculture 8,544 12,374 Income loss in remittances 22,934 52,248 Combined effect 233,892 473,407 Sources: Microsimulations based on the ILCS 2018, MTI projections, and MFMod database (as of February 2020). Notes: Impoverishment = People living in households that are nonpoor under business-as-usual, but who become poor after the employment and income shocks of COVID-19. Downward mobility = People living in households that transition to a lower welfare group as a result of the employment and income shocks from COVID-19. Figure 6. Transition Matrix of welfare status in Armenia (Number of people) Welfare status after the negative shock Welfare status at baseline Extreme poor Poor Moderate Poor Vulnerable Middle-class All Extreme poor 35,795 35,795 Poor 26,718 197,615 224,333 Moderate Poor 12,290 118,072 779,833 910,194 Vulnerable 5,121 9,698 215,069 956,303 1,186,191 Middle-class 1,757 782 1,465 82,436 434,233 520,673 All 81,681 326,166 996,367 1,038,739 434,233 2,877,185 Sources: Microsimulations based on the ILCS 2018, MTI projections and MFMod database. Notes: The welfare groups are defined as extreme poor (USD 10.00) Sources: AMSSW 2015 and database from MFMod (available from \\gpvfile\GPV\Knowledge_Learning\Pov Projection\Central Team\MFM-allvintages.dta). Notes: All poverty lines are expressed in USD 2011 PPP. 19 Results for Georgia The pessimistic macroeconomic scenario in Georgia (assuming negative growth rate of -7.5) results in the highest poverty increase of 3.6 percentage points, with respect to the counterfactual of 13.3 percent poverty rate in 2020. The micro simulation exercise projects a more modest increase of 2.2 percentage points in poverty (Figure 12). Figure 12. Poverty effects of COVID-19 in Georgia Effect of COVID-19 on projected poverty rates in Georgia (Poverty line of USD 3.20 per day, 2011 PPP) 18% 16.9% 17% 16.1% Headcount poverty rate 15.7% 16% 15.5% 15% 13.9% 14% 13% 13.9% 13.3% 12% 11% 10% 2018 2019n 2020f Pre-shock poverty rates Counterfactural COVID-19 Macro-simulation: Baseline scenario COVID-19 Macro-simulation: Pessimistic scenario COVID-19 Micro-simulation Sources: HIES 2018, MFMod database and projections from MTI. Notes: n = nowcast, f = forecast. The counterfactual scenario is based on business-as-usual GDP forecasts from MFMod (as of February 25, 2020). The baseline and pessimistic macro simulations assume GDP growth of -4.0 percent and -7.5 percent, respectively, for Georgia in 2020. Under all simulations, the poverty gap and poverty severity indicators increase with the economic consequences linked to the pandemic (Figure 13). Results from the microsimulations are similar in magnitude to the pessimistic macro simulations. 20 Figure 13. Projected Effect of COVID-19 on Depth of Poverty in 2020 Panel A. Poverty Gap under different simulations, Georgia 2020 6.0 5.0 4.8 5.0 4.6 3.8 4.0 3.0 2.0 1.0 0.0 Counterfactual COVID-19 Macrosimulation: COVID-19 Macrosimulation: COVID-19: Microsimulation Baseline scenario Pessimistic scenario Panel B. Poverty Severity under different simulations, Georgia 2020 2.5 2.2 2.2 2.0 2.0 1.7 1.5 1.0 0.5 0.0 Counterfactual COVID-19 Macrosimulation: COVID-19 Macrosimulation: COVID-19: Microsimulation Baseline scenario Pessimistic scenario Sources: Microsimulations based on the HIES 2018 and database from MFMod. Macrosimulations based on data from the MFMod database, WEO (April 2020) and MTI projections. Notes: The macroeconomic baseline scenario assumes GDP growth rate of -4.0% in 2020. The macroeconomic pessimistic scenario assumes that GDP drops at -7.5% change rate in 2020. The microsimulations assign different sectoral income and unemployment shocks described in Table 3. On average, income losses from COVID-19 represent 8.0 percent of household incomes in Georgia (Figure A.8.a). However, the shares of income losses do not show a clear distributional pattern across percentiles of the population (Figure 14). The share of income losses is highest in Tbilisi (10.6 percent) and lowest in rural areas (5.1 percent share) (Figure A.8.a). The increased probability of job loss is the main driver of poverty increases after COVID-19. The loss of wage incomes among remaining workers is the second driver. The poverty effects through agricultural sales and remittance inflows remains more subdued (Figure 15). The employment and income shocks to the services sector yield the largest marginal increase in poverty. This is confirmed in the extended microsimulation model presented in the Appendix. Unemployment and income shocks to workers in 21 wholesale and retail trade, tourism, and construction, have the largest marginal effects on poverty and extreme poverty (Figure A.8.c). Figure 14. Distributional incidence of income losses related to COVID-19 Share of income losses in Georgia 14.0% 12.0% Share of income loss 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 1 4 7 100 Percentile Sources: Microsimulations based on the HIES 2018 and the MFMod database. Note: Welfare percentile is based on the projected consumption aggregate under the micro simulation for 2020. Figure 15. Effects of micro-simulation on poverty rates Poverty projection by type of shocks Georgia 2020, Micro Simulation 16% 15.5% 16% 15% 14.8% 15% 14.3% 14% 13.7% 14% 13.3% 13.3% 13.3% 13% 13% 12% Increase in Wage decrease Shock to Drop in Agriculture Industry Services unemployment agricultural remittances sales Type of shock Shock by sector of employment* Sources: Microsimulations based on the HIES 2018 and database from MFMod (as of February 2020). Effects are not mutually exclusive. The shocks by sector of employment incorporate only the channels of unemployment and wage losses in each sector. 22 Residents of Tbilisi, households with larger number of children, and households not reporting pension incomes will most likely face higher probabilities of unemployed. Interestingly, households reporting income from social assistance in the HIES 2018, and those estimated to be TSA-eligible observe the lowest increase in the poverty rate (Figure A.8.b). Half a million Georgians are at risk of suffering downward mobility, transitioning to a lower-welfare group as a result of COVID-19 in 2020. Over 200 thousand people who were nonpoor before the crisis could become impoverished (Figure 16 and Table 6) Higher unemployment resulting from the pandemic would be the main driver of impoverishment. Figure 16. Transition Matrix of welfare status in Georgia (Number of people) Welfare status after the negative shock Welfare status at baseline Extreme Moderate Middle- Poor Vulnerable All poor Poor class Extreme poor 134,913 134,913 Poor 36,135 324,842 360,977 Moderate Poor 3,864 95,967 812,968 912,800 Vulnerable 422.6382 2,539 202,199 1,131,458 1,336,618 Middle-class 580.4675 3626.929 159,168 820,949 984,325 All 175,335 423,929 1,018,794 1,290,626 820,949 3,729,633 Sources: Microsimulations based on the HIES 2018 and MFMod database (as of February 2020). Notes: The welfare groups are defined as extreme poor (