HOW DID COUNTRIES RESPOND TO THE COVID-19 CRISIS? EMERGING PATTERNS ON JOBS ‑RELATED POLICIES ABSTRACT This brief investigates differences in countries’ jobs-related policy responses to the COVID-19 pandemic. Four main patterns emerge. First, the type of labor policies adopted by countries varied greatly according to their income level. Low-income countries were more likely to implement public works programs but not other policies such as unemployment benefits, labor regulations, wage subsidies, training and placements policies, firm liquidity support, and cash transfers to workers. Meanwhile, countries with a more formal workforce and existing unemployment benefits systems were more likely to implement policies such as unemployment benefits and labor regulations. Second, low- and lower-middle-income countries devoted a lower share of their gross domestic product (GDP) to expenditure on new jobs- related policies. Third, conditional on countries’ income group, the magnitude of the GDP shock did not have a statistically significant correlation with the adoption of different policies. This may reflect uncertainty in the extent of the GDP shock when the policy response was determined or noise in the measure of GDP. Finally, countries that adopted more stringent COVID-19 restrictions were more likely to adopt changes to labor regulations, specifically changes in working conditions, to try to soften the blow on workers. These results suggest that the policy response to the crisis in low- and lower-middle- income countries was constrained by the lack of resources, resulting in lower-cost policies with generally limited impacts on workers. 1. INTRODUCTION This brief focuses on the policy response to the COVID-19 While previous analysis has shown that the impacts of crisis and uncovers patterns relating to various countries’ COVID-19 on the labor market were severe (Miguel and choice of jobs-related policy response as the crisis Mobarak 2021) and had major impacts on labor markets unfolded. In particular, it sheds light on the degree to (Khamis et al. 2021a, 2021b, 2021c) across developing which countries implemented different policies. Which countries, less attention has been paid to the policy jobs policies were adopted in response to the shock and response. Of the 224 countries covered in the World what determined their adoption?1 Bank’s Social Protection and Jobs (SPJ) Policy Inventory, 1 See Contreras et al. (2023). 1 all responded with more than one policy to ease the of informality in the labor market. On the other hand, COVID-19 shock.2 Yet, to the best of our knowledge, unemployment benefits were more likely implemented there is no quantitative evidence on the patterns of in countries with previous unemployment protection jobs-related policy responses in developing countries. schemes. In addition, we find that changes in the This brief focuses on policies of the labor market number of COVID-19 cases, GDP, or public debt were supply and demand side, following the classification not significantly correlated with the implementation of of Kamran et al. (2022). The categories of jobs-related unemployment benefits. Labor regulations were more policies analyzed are public works, cash transfers likely to be adjusted in countries that experienced more to workers, income tax reduction, unemployment COVID-19 cases per million of inhabitants (health shock) benefits, wage subsidies, entrepreneurship support, and a significant increase in public debt as a share of training and placement, labor regulations, and firm GDP (debt shock). liquidity support.3 Overall, these results indicate that the level of The analysis combines data from the Policy Inventory, development of a country mattered most in determining the World Bank High Frequency Phone Survey (HFPS), the implementation of the jobs-related policies. In and various data on the employment and economic addition, previously existing structures of the labor conditions from the International Labour Organization market such as the degree of informality and whether an (ILO), International Monetary Fund (IMF), and World unemployment benefits system existed were important Bank to investigate how the impacts of the COVID‑19 determinants of the implementation of policies such crisis were related to countries’ policy response. It also as unemployment benefits and labor regulations. In incorporates data on COVID-19 and mobility restrictions contrast, measures of the magnitude of the shock from the Oxford COVID-19 Government Response such as GDP, health, and debt shocks were more tracker. The data are described in further detail in weakly correlated with the policy response. These Appendix A2 and the empirical methodology in Appendix results highlight the importance of establishing social A3, which generated the regression coefficients for the protection programs in lower-income contexts that can result figures.4 be scaled up quickly during a crisis and of accurately monitoring the impact of the crisis in real time. The brief is structured as follows. The first section reports Furthermore, as countries recover from the pandemic three stylized facts. First, the type of jobs policies adopted shock, governments should decide which policies are by countries varied greatly according to their income most suitable to protect their citizens; however, the level. Specifically, low-income countries were more likely decision is constrained by their fiscal space. As of to adopt public works programs and labor regulatory 2022, low- and lower-middle-income countries carry adjustments and less likely to adopt cash transfers to significantly more debt than two years ago (World workers, unemployment benefits, and wage subsidies. Bank 2022). The results reported in this brief should be Second, high-income countries were able to devote a considered in this context of debt distress, for example, larger share of their gross domestic product (GDP) to countries that experienced higher increases in their new jobs policies. Finally, neither changes in GDP nor debt-to-GDP ratio due to the pandemic were more the number of COVID-19 cases reported by countries likely to implement labor regulations policies (which are statistically significantly correlated with the type of are usually less expensive than others). jobs policies adopted. The second section focuses on two types of jobs-related policies: unemployment benefits and labor regulations. Unemployment benefits and labor regulations were less likely to be implemented in countries with a high degree 2 The data of the World Bank COVID-19 SPJ Policy Inventory cover 224 countries from March 2020 to December 2021. 3 See Appendix A1 for a review of the classification of jobs-related policies following Kamran et al. (2022). 4 Appendix A5 provides information on the number of countries covered. Even though the main sample that we use includes 165 countries, when combining the policy inventory with other datasets on the jobs outcomes, such as employment, or with datasets relating to pre-existing jobs policies, the country coverage is reduced. We included the number of observations available in each of the figures or tables. 2 2. OVERALL FINDINGS FIGURE 1 Likelihood of adoption of different jobs policies Finding 1: The type of jobs policies adopted by low-income countries by countries varied according to their income level. 0.351*** Public works Looking at the nine different jobs-related policies from −0.509*** the labor market demand and supply side, low-income CT to workers countries were generally less likely to implement unem- −0.0399 ployment benefits, labor regulations, wage subsidies, Income tax reduction training and placements policies, firm liquidity support, −0.599*** Unemployment benefits and cash transfers to workers but more likely to imple- ment public works programs.5 −0.675*** Wage subsidies Low-income countries were more likely to implement −0.0896 Entrepreneurship support public works programs. Controlling for the stringency of COVID-19 regulations, probability of implementing −0.255** Training and placement public works programs in response to the crisis was 35.1 percentage points higher for low-income coun- −0.137* Firm liquidity support tries compared to high-income countries (Figure 1). In −0.213* contrast, after controlling for stringency, low-income Labor regulations countries were less likely to implement unemployment benefits, labor regulations, wage subsidies, training and −1.0 –0.5 0 0.5 placement, firm liquidity support, and cash transfers to Percentage points workers. Implementation of income tax reduction and Source: Authors’ calculations. entrepreneurship policies was not significantly different Note: This figure shows the coefficient and the 95% confidence for low-income countries compared with high-income interval of a dummy variable that captures low-income countries countries. (variable of interest) following Model 1 (discussed in Appendix A3) in different regressions, one for each jobs-related policy Finding 2: High-income countries spent incidence as dependent variable. The coefficient reported here more on their adopted jobs policies as a is the coefficient for the low-income country dummy versus the share of GDP. reference category of high-income countries. The regression is based on a sample of 165 countries. Differences found to be statistically significant are indicated by level: *p < 0.1, Expenditure on jobs policies, as a share of 2019 GDP in **p < 0.05, ***p < 0.01. current prices, was linked to the level of the countries’ development: low-income, lower-middle-income, and upper-middle-income countries spent less than high-in- come countries, with large and statistically significant differences in magnitude for low-income and low- er-middle-income countries compared with high-income countries. The difference in spending on jobs policies as a share of GDP between upper-middle-income and high-income countries was also large, at 1.7 percent- age points (5.6 percent versus 3.9 percent), but is not statistically significant. 5 See Appendix A4 for the descriptive statistics of jobs policies adoption by country level. 3 BOX 1. LOW COVERAGE OF EXPENDITURES DATA AS A CONSTRAINT FOR POLICY ANALYSIS The World Bank’s COVID-19 SPJ Policy Inventory reported expenditure data for 28 percent of interventions. The distribution of expenditure data varies across income groups for the countries included in the inventory. In particular, 24 percent of the policies of high-income countries reported expenditure data, 29 percent for upper-middle-income countries, 26 percent for lower-middle-income countries, and 17 percent for low-income countries. This relatively low share is due to a lack of available public information on expenditures on policy interventions across low- and middle-income countries. The lack of expenditures data and its relatively low coverage bias the estimates of the distribution of expenditure across programs reported in Figure 2. However, for some countries such as Burkina Faso, detailed expenditure data are accessible. The Emergency Response Plan is expected to cost 4.6 percent of the GDP and the expenditures have been disbursed over three years.6 While the COVID-19 SPJ Policy Inventory includes 35 policies for this country, only 4 from Burkina Faso come with this type of detailed expenditure data. The GDP share of the expenditures of these four policies is only 0.22 percent. Furthermore, these four policies make up only 5 percent of the estimated amount of the Emergency Response Plan, or 15 percent of the estimated expenditures for the Emergency Response Plan per year. Consequently, the lack of expenditures data on all policies affects a fuller assessment of countries’ responses to the pandemic. Even though the World Bank’s COVID-19 SPJ Policy Inventory fills the gap of reporting the different policies implemented across countries, future efforts could consider other strategies to improve data collection of expenditures on policy interventions. FIGURE 2 Expenditures in jobs policies as share of the GDP 2019 (simple average across countries) Low-income 1.36 countries *** 2.16 Lower-middle- income countries *** 3.90 Upper-middle- income countries n/s 5.62 High-income countries All countries 3.71 (N=142) 0 2 4 6 8 Share of GDP Source: Authors’ calculations. Note: This figure shows the simple average across countries of the expenditures in jobs policies as a share of the GDP in 2019. It includes information for 142 countries. This figure also shows the significance of the statistical test of difference across income groups with respect to high-income countries. We observe that low-income countries spent a significantly lower share of their GDP in jobs policies when compared with high-income countries (this difference is statistically significant at 1 percent). Differences found to be statistically significant are indicated by level: *p < 0.1, **p < 0.05, ***p < 0.01; n/s = not significant. 6 World Bank’s Public Expenditure and Revenue Review for Burkina Faso 2022. 4 Finding 3: Changes in GDP or the number of FIGURE 3 COVID-19 cases reported by countries were Correlation between GDP shock and jobs not significantly correlated with the type of policy adoption jobs-related policies adopted. 0.191 During the first year of the pandemic, most countries Public works experienced a negative growth shock. For a sample of −0.407 165 countries, we calculated the difference between CT to workers the estimated GDP growth of 2020 (estimated in 2019) −0.0541 Income tax reduction and the observed one in 2020, which is a measure of the GDP shock (measured in percentage points).7 0.945 Unemployment benefits Conditional on countries’ initial level of income and the stringency of the COVID-19 response, the magnitude 0.622 Wage subsidies of the GDP shock did not have a statistically significant correlation with policy adoption for any of the nine −0.054 Entrepreneurship support jobs-related policies analyzed (Figure 3).8 In terms of magnitudes, most of the coefficients are small although −0.494 Training and placement a few imply a moderately strong relationship.9 This may partly be because the extent of the GDP shock was not −0.160 Firm liquidity support known when the policy response was determined and 1.018 because GDP is measured with noise.10 Labor regulations We also explored whether the policy response was −2 −1 0 1 2 related to the size of the COVID-19 shock, as measured Coefficient by COVID-19 cases per million reported by country (in Source: Authors’ calculations. log scale).11 This indicator only includes the cumulative Note: This figure shows the coefficient and the 95% confidence confirmed cases until December 31, 2020. Like GDP interval of the GDP shock (variable of interest) following Model 1 shocks, the magnitude of the COVID-19 shock did (discussed in Appendix A3) in different regressions, one for each not have a statistically significant correlation on policy jobs-related policy incidence as dependent variable. The regression adoption, conditional on countries’ initial level of income, is based on a sample of 165 countries. except for the adoption of labor regulations—countries that reported a higher level of COVID-19 cases were benefits and labor regulations. The policies classified as significantly more likely to implement labor regulations unemployment benefits included new or modified policies. The magnitude of its effect was relatively small. policies during the pandemic; the World Bank’s An increase of 10 percent in COVID-19 cases is only SPJ Policy Inventory does not include previous associated with a 0.5 percentage point increase in the unemployment benefits policies that were instated probability of implementing labor regulations. before the pandemic. The new unemployment benefits were adopted by 39 percent of countries and labor regulations by 68 percent of countries, so 3. A CLOSE-UP LOOK AT UNEMPLOYMENT adoption was not universal across all countries. While BENEFITS AND LABOR REGULATIONS unemployment benefits were mainly implemented in high-income countries, labor regulations were We next turn specifically to the analysis on two jobs- adopted by the majority of countries in each group. related policies during the pandemic: unemployment The following sections examine in more detail how the 7 The average GDP shock was −7.89 percentage points. It can be interpreted that on average, the actual GDP growth of 2020 was lower than the estimated GDP growth for 2020 (in 2019) by 7.89 percentage points. 8 See Appendix 3 for a discussion on the methodology. 9 For example, the probability of implementing a policy of unemployment benefits was 5 percentage points lesser for countries that experienced a decrease in their GDP growth of one standard deviation. 10 In particular, higher dispersion of growth estimates due to measurement errors would be expected to be more prevalent in developing countries (Angrist, Goldberg, and Jolliffe 2021). 11 See Appendix 6, Figure A6.1. 5 adoption of these two policies relates to the magnitude FIGURE 4 of the shock, previous policies, and degree of previous Unemployment benefits adoption and initial labor market informality. labor market and policy conditions Unemployment benefits as a COVID-19 −0.559*** jobs-related policy response Low-income country (N=165) 0.276 High-income countries were more likely to adopt Stringency index (N=165) unemployment benefits (63 percent) as opposed −0.519* to low-income countries (0 percent) (see Appendix Informality (N=52) A4, Figure A4.3). Figure 4 shows the coefficients 0.597 from different linear probability models in which the Benefit incidence —poorest quintile (N=89) dependent variable is the probability of implementing unemployment benefits and the controls include income −1 0 1 2 group dummies.12 Unemployment benefits were less Percentage points likely to be implemented in low-income countries and Source: Authors’ calculations. also less likely to be the response in countries with a Note: This figure shows the coefficient and the 95% confidence high degree of informality in the labor market (51.9 interval of the variable of interest. This figure is based on separate percentage points less likely). It is important to note regressions coefficients for the initial labor market and policy that informality is correlated with income level. conditions and the adoption of unemployment benefits as dependent variable. The coefficients for income group, stringency, informality, and benefit incidence are from one specification Unemployment benefits were much more likely imple- (equation 1 in Appendix A3). mented in countries with previous unemployment pro- tection schemes (an estimated 94.5 percentage points more likely).13 This indicates the countries’ ability to scale FIGURE 5 up the unemployment benefits for COVID-19 due to the Unemployment benefits adoption and existing schemes. Unemployment benefits were 27.6 COVID‑19-related shocks percentage points more likely to be adopted when the country implemented more stringent mobility regulations, 0.036 measured using the Oxford COVID-19 stringency index, COVID-19 cases (N=165) but the effect is not statistically significant. 0.945 GDP shock (N=165) Labor regulations as a COVID-19 0.907 jobs‑related policy response Employment shock (N=91) −0.009 At the same time, countries with lower income or that Debt shock (N=165) had a higher share of the employed in an informal job were less likely to implement labor regulations −1 0 1 2 3 in response to COVID-19 (Figure 6). This mirrors the Percentage points experience of countries with unemployment benefit Source: Authors’ calculations. adoption: lower income levels and previous level of Note: This figure shows the coefficient and the 95% confidence informality make labor market policy adoption such as interval of the variable of interest. The coefficient represents the labor regulation less likely. An increase of 1 percentage estimated conditional correlation between the independent variable point in informality is associated with the decrease in the and the probability of a country adopting unemployment benefits, likelihood of implementing labor regulations policies by after controlling for stringency and income group category. This figure is based on separate regressions for different shock measures. 1.079 percentage points.14 In terms of previous policies, The regression models follow Equation 1, see Appendix A3. labor regulations were more likely to be adopted by 12 Appendix 5 has more details on samples covered. When combining the policy inventory with other datasets on the labor market outcomes, such as employment or with datasets relating to pre-existing jobs policies or conditions, the country coverage is reduced. 13 This result is generated from a regression, based on Model 2, Appendix A3. 14 An increase of 1 standard deviation in informality decreases the probability of implementing labor regulations by 25 percentage points. 6 FIGURE 6 FIGURE 7 Labor regulations changes and initial labor Labor regulations changes and market and policy conditions COVID‑19‑related shocks −0.213* −0.213* Low−income country (N=165) Low−income country (N=165) 0.402 0.402 Stringency index (N=165) Stringency index (N=165) −1.079*** −1.079*** Informality (N=52) Informality (N=52) 0.943** 0.943** Benefit incidence Benefit incidence —poorest quintile (N=89) —poorest quintile (N=89) −2 −1 0 1 2 −2 −1 0 1 2 Percentage points Percentage points Source: Authors’ calculations. Source: Authors’ calculations. Note: This figure shows the coefficient and the 95% confidence Note: This figure shows the coefficient and the 95% confidence interval of the variable of interest. This figure is based on interval of the variable of interest. This figure is based on separate separate regressions coefficients for the initial labor market regressions coefficients for the different shocks and the adoption and policy conditions and the adoption of labor regulations of labor regulations as dependent variable. The regression as dependent variable. The coefficients for income group, models follow equation 1, see Appendix A3. Differences found stringency, informality, and benefit incidence are from one to be statistically significant are indicated by level: *p < 0.1, specification (equation 1 in Appendix A3). Differences found **p < 0.05, ***p < 0.01. to be statistically significant are indicated by level: *p < 0.1, **p < 0.05, ***p < 0.01. countries that had a higher percentage of program example, the majority of the labor regulations adopted beneficiaries in the poorest quintile. were related to changes in working conditions.15 Labor regulations were more likely to be adopted by Labor regulations pertaining to working conditions were countries that experienced a larger health and public popular for countries in all income groups, accounting debt shock due to COVID-19. For countries that expe- for 36–52 percent of the changes in labor regulations. rienced an increase of 1 standard deviation in the debt In results from regression models, the relationship shock, the probability of implementing labor regulations between changing working conditions and countries’ policies increases by 4 percentage points. As with income group category was not statistically significant.16 unemployment benefits, however, GDP shocks did not Contrary to this, based on these regression models, the have a statistically significant effect on the probability general level of stringency in response to COVID-19 is of implementing labor regulations. associated with a specific type of labor regulations policy: countries that had higher than average stringency were Higher stringency and more flexible more likely to implement labor regulations in the form working conditions at the country level of work condition changes, even when accounting for the number of COVID-19 health cases.17 An increase Looking into the details of what exact labor regulations of 1 standard deviation in the stringency index is cor- were adopted (Table 1), regulating working conditions/ related with an increase of 6.5 percentage points in the methods (45 percent) was the main labor regulatory probability of implementing a labor regulations policy response across countries. In high-income countries, for pertaining to working conditions. 15 Labor regulations included regulations that promote alternate arrangements or schedules compared with the traditional working day and week. This included changes in working conditions such as flexible time, reduced time / part-time, compressed work week, telework / telecommuting (for example, shortened working hours, flexible working arrangements, including remote work). 16 See Model 3 and Model 4 for Appendix A3 for methodology. 17 For Model 3 and Model 4, controlling for number of COVID-19 cases, the estimated coefficients were similar in magnitude and weakly significant at the 90% confidence level. 7 TABLE 1 Share of countries that implemented at least one policy related to the subcategories of labor regulations (%) Subcategories for labor regulations Lower-middle- Upper-middle- All Low-income High-income income income Severance payment 4 0 0 9 4 Hiring flexibility 9 0 2 7 20 Dismissal procedures 16 12 19 19 13 Working conditions/methods 45 42 36 49 52 Leave policies 16 4 5 16 31 Remuneration 24 12 19 30 28 Labor inspector interventions 25 27 19 30 26 Other regulatory adjustment 13 4 21 12 13 N 165 26 42 43 54 Source: Authors’ calculations 4. CONCLUSIONS resources, resulting in lower-cost policies with generally This brief employs a unique dataset on jobs-related limited impacts on workers. policies and combines it with data on economic shocks and preexisting policies to analyze the differences This highlights the importance of investing in social in countries’ jobs policies response to the COVID-19 protection systems and policies before a crisis, so that pandemic, focusing on unemployment and labor existing programs can easily be scaled up if necessary. regulations. The results in this brief are conditional Future work in this area may include further analysis on correlations and not causal relationships because policy the long-term effect of the COVID-19 crisis on different adoption is a function of many unobserved factors. types of policies adopted and their impacts on different groups. In addition, future work should investigate on The main finding is that the presence of existing jobs- how the adoption of policies is related to the pace of related policies and social protection systems greatly the recovery. increased the probability that countries utilized these responses in the wake of the COVID-19 response and implemented thereafter. These results suggest that the policy response to the crisis in low- and lower- middle-income countries was constrained by the lack of 8 REFERENCES Angrist, Noam, Pinelopi Koujianou Goldberg, and Dean Jolliffe. 2021. “Why Is Growth in Developing Countries So Hard to Measure?” Journal of Economic Perspectives 35 (3): 215–42. Contreras, Ivette, Melanie Khamis, David Newhouse, and Michael Weber. 2023. “Labor Market Policy Adoption in Response to COVID-19: Evidence from a policy inventory” Mimeo. Kamran, Mareeha, Ingrid Mujica, María Belén Fonteñez, David Newhouse, Claudia Rodriguez Alas, and Michael Weber. 2022. “Exploring Two Years of Labor Market Policy Responses to COVID-19: A Global Effort to Protect Workers and Jobs.” JobsWatch COVID-19 Brief. Khamis, Melanie, Daniel Prinz, David Newhouse, Amparo Palacios-Lopez, Utz Johann Pape, and Michael Weber. 2021a. “The Early Labor Market Impacts of COVID-19 in Developing Countries: Evidence from High-Frequency Phone Surveys (English).” Policy Research Working Paper No. WPS 9510, COVID-19 (Coronavirus), World Bank Group, Washington, DC. Khamis, Melanie, Daniel Prinz, David Newhouse, Palacios-Lopez, Amparo Palacios-Lopez, Utz Johann Pape, and Michael Weber. 2021b. “Early Labor Market Impacts of COVID-19 in Developing Countries.” JobsWatch COVID-19 Brief. Khamis, Melanie, Daniel Prinz, David Newhouse, Palacios-Lopez, Amparo Palacios-Lopez, Utz Johann Pape, and Michael Weber. 2021c. “The Evolving Labor Market Impact of COVID-19 in Developing Countries.” JobsWatch COVID-19 Brief. Miguel, Edward, and Ahmed Mushfiq Mobarak. 2021. “The Economics of the Covid-19 Pandemic in Poor Countries.” NBER Working Paper 29339. World Bank. 2022. Poverty and Shared Prosperity 2022: Correcting Course. Washington, DC: World Bank. 9 APPENDIXES A1. CLASSIFICATION OF JOBS-RELATED POLICIES ACCORDING TO KAMRAN ET AL. (2022) COVID-19 SPJ Policy Inventory and Jobs Framework The COVID-19 SPJ Policy Inventory includes information on several types of jobs-related programs. For this brief, jobs-related policies are classified into four categories: (1) income support policies, which include public works programs, cash transfers to economically active persons,18 income tax reduction, unemployment benefits, and wage subsidies; (2) active labor market policies, which include entrepreneurship support and training and placement assistance; (3) firm liquidity support policies, such as tax relief for firms, credit facilities, credit guarantees, corporate tax reductions, and relief from social security contributions for firms; and (4) labor regulations which define formal work relationships. FIGURE A1.1 COVID-19 SPJ Policy Inventory and Jobs Framework Income support for the poor given in the form of cash wages, services or food Public works in exchange for work effort Income support policies Cash transfers to Non-contributory cash transfers to economically active persons without economically active persons being conditional on the beneficiaries’ actions Labor income tax deferral or reduction, change in basic personal Income tax reduction income tax credits Unemployment benefits Benefits payable to workers due to job loss paid only for a limited period Wage subsidies to employers who maintain existing jobs with or without Wage subsidies reduced work Programs that encourage the unemployed and target groups to start their Entrepreneurship support policies own business or to become self-employed, among other ALMPs Active labor Training and placement Training programs, including vocational, cash for training, workplace training, assistance and placement services including counseling and job matching, among others support policies Credit facilities and guarantees, payment facilities, taxes, social security Firm Firm liquidity support contribution for firms, utility and rent support, among other special measures regulations Labor Changes in severance payments, dismissal procedures, hiring flexibility, working Regulatory adjustments conditions, leave policies, remuneration, labor inspection, among others Source: Kamran et al. 2022. 18 Within social protection and jobs policies, cash transfer programs and public work programs are often considered as social assistance. For this brief, these are considered as labor market policies as they affect workers. The policy inventory differentiates between the typical cash transfers targeted to the household and those targeted directly at workers with a labor market policy objective. Similarly, unemployment benefits and social security contributions paid by firms are included within labor market policies. Policies regarding social security contributions targeting individuals directly are considered part of social insurance. 10 A2. DATA SOURCES This brief combines uniquely collected primary data from the World Bank COVID-19 SPJ Policy Inventory; the World Bank HFPS; supplementary data on the employment and economic situation from the ILO, World Bank Global Jobs Indicators Database (JOIN), Atlas of Social Protection Indicators of Resilience and Equity [ASPIRE]),Employing Workers Database (EW) and the IMF; and data on the extent of the COVID-19 immediate impact on health (measured as COVID-19 cases per million population) from Our World in Data and the Oxford COVID-19 Government Response tracker for the stringency index. TABLE A2.1 Data sources Variable Definition Source Policy Variable that takes the value of 1 if the country adopted the policy between March COVID-19 SPJ Policy 2020 and December 2021. Inventory There are nine jobs-related policies: (1) Cash transfers to workers, (2) Public works, (3) Unemployment benefits, (4) Training and placement assistance, (5) Wage subsidies, (6) Entrepreneurship support, (7) Income tax reduction, (8) Labor regulations, (9) Firm liquidity support. Income group There are four levels of the variable of income groups: (1) High-income, World Bank (2) low‑income, (3) Lower-middle-income, and (4) Upper-middle-income. Stringency Stringency index (average in Q1 2020). The stringency index is a composite measure COVID-19 Government index—Q1, 2020 based on nine response indicators including school closures, workplace closures, and Response Tracker travel bans, rescaled to a value from 0 to 100 (100 = strictest). If policies vary at the subnational level, the index shows the response level of the strictest subregion. Employment shock Difference in the employment to the population ratio captured by Labor Force HFPS and ILO datasets Surveys between Q4 2019 and Q4 2020 (or in the case of countries without information, we used the share of respondents who were working before the pandemic but stopped worked, included in the HFPS). Expenditures in policies Expenditures on the type of policy in US dollars (in ‘ln scale’) COVID-19 SPJ Policy Inventory Health shock COVID-19 cases per million people. Cumulative confirmed COVID-19 cases until Our World in Data (COVID-19 cases) December 31, 2020 (in ‘ln scale’) Informality Share of employed individuals (15–64 years old) working in an informal job (last JOIN available data point between 2008 and 2018—JOIN dataset) GDP shock Shock GDP percentage points. This variable captures the difference between IMF estimated 2020 GDP growth in 2019 and observed GDP growth in 2020. Debt shock Shock debt percentage points. This variable captures the difference between general IMF government gross debt in 2019 and general government gross debt in 2020 (both as a percentage of GDP). Previous unemployment Availability of an unemployment protection scheme EW benefits Benefit incidence Percent of program beneficiaries in the poorest quintile ASPIRE poorest quintile 11 A3. METHODOLOGY This brief combines uniquely collected primary data from the World Bank COVID-19 SPJ Policy Inventory; the World Bank HFPS; supplementary data on the employment and economic situation from the ILO, World Bank (JOIN, ASPIRE), and the IMF; and data on the extent of the COVID-19 immediate impact on health (measured as COVID-19 cases per million population) from Our World in Data and the Oxford COVID-19 Government Response tracker for the stringency index. The data of the World Bank COVID-19 SPJ Policy Inventory cover 224 countries from March 2020 to December 2021. In general, we estimate two equations (1 and 2) following a linear probability model. Standard errors are clustered at the country level. Both equations follow the same structure with the dependent variable as the policy adoption (a dummy variable that takes the value of 1 if the country adopted the policy in 2020 or 2021) and the right-hand side includes variables of interest with controls for income group and a stringency measure (equation 1) and additionally a measure of the degree of informality in the labor market (equation 2). Equation 2 is the regression equation that is applied to all previously existing measures of policies in the labor market (in this brief the availability of unemployment protection scheme and labor regulations) where the degree of informality in the labor market may matter. The coefficients reported in the figures in this brief are one of the variables of interest. Model 1 (1) policyc =  β1 + β2incomegroupc + β3stringencyavQ120 + β4variable of interestc + εc Model 2 (2) policyc =  β1 + β2incomegroupc + β3stringencyavQ120 + β4variable of interestc + β5informallast_av + εc The variable of interest is a measure of the shock of the COVID-19 crisis, which is COVID-19 cases (health shock), a GDP shock, a debt shock, an employment shock, informality, and benefit incidence (see Appendix A2 for definitions). Each of these variables of interest is entered separately and the coefficient is reported in the main graphs. Also, control variables for income group and stringency are added to both Model 1 and Model 2. Moreover, Model 2 also accounts for the degree of informality. The regression models, Model 3 and Model 4, have the dependent variable as 1 for working conditions/method and 0 for other labor regulations: Model 3 (3) work_conditionsc =  β1 + β2incomegroupc + β3stringencyavQ120 + εc Model 4 (4) work_conditionsc =  β1 + β2incomegroupc + β3stringencyavQ120 + β4lncasespermillionc + εc 12 A4. DESCRIPTIVE STATISTICS FIGURE A4.1 Share of countries that implemented at least one labor market policy between January 2020 and January 2022 by type 100 98 Percentage points 80 67 68 60 55 38 39 41 38 40 24 20 0 io x id ge or p ks s fit t su ui Firm ns r ta e n d er n e en t io o ct ta pp hi s is m n or la ab n s s t or y bs a nc t as ac e g a k s u u rs ie be ym pp dit du e su W or w gu L t re o m e pl nin ne w o ic pl bl c liq to re In ai m Pu ep Tr CT ne re tr U En Source: Authors’ calculations. Note: This figure includes information for 165 countries. FIGURE A4.2 Share of countries that implemented at least one jobs-related policy between March 2020 and December 2021 by type and income group A. Low-income countries (N=26) B. Lower-middle-income countries (N=42) Public works 42 Public works 36 CT to workers 27 CT to workers 64 Income tax reduction 23 Income tax reduction 24 Unemployment benefits 0 Unemployment benefits 26 Wage subsidies 12 Wage subsidies 43 Entrepreneurship support 42 Entrepreneurship support 36 Training and Training and 31 23 placement assistance placement assistance Firm liquidity support 92 Firm liquidity support 98 Labor regulations 54 Labor regulations 62 0 20 40 60 80 100 0 20 40 60 80 100 Percentage points Percentage points C. Upper-middle-income countries (N=43) D. High-income countries (N=54) Public works 21 Public works 9 CT to workers 79 CT to workers 80 Income tax reduction 58 Income tax reduction 41 Unemployment benefits 47 Unemployment benefits 63 Wage subsidies 56 Wage subsidies 83 Entrepreneurship support 40 Entrepreneurship support 44 Training and placement Training and placement 33 54 assistance assistance Firm liquidity support 100 Firm liquidity support 100 Labor regulations 70 Labor regulations 80 0 20 40 60 80 100 0 20 40 60 80 100 Percentage points Percentage points Source: Authors’ calculations. 13 FIGURE A4.3 FIGURE A4.4 Share of countries that adopted at least one Share of countries that adopted at least one policy of unemployment benefits between March policy of labor regulations between March 2020 2020 and December 2021 by income group and December 2021 by income group Low-income countries Low-income countries 0 54 (N=26) (N=26) Lower-middle-income 26 Lower-middle-income 62 countries (N=42) countries (N=42) Upper-middle-income Upper-middle-income 47 70 countries (N=43) countries (N=43) High-income countries 63 High-income countries 80 (N=54) (N=54) All countries (N=165) 39 All countries (N=165) 68 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 80 90 Percentage points Percentage points Source: Authors’ calculations. Source: Authors’ calculations. Note: This figure includes information for 165 countries. Note: This figure includes information for 165 countries. A5. DIFFERENT SAMPLES USED IN THE ANALYSIS Sample 1: 165 countries. This sample was used to estimate Figure 1 in addition to Figure 3 and Figure 5 (individual models including income level and stringency index) and Figure 4 and Figure 6 (individual models including health shock, GDP shock, debt shock). TABLE A5.1 Sample 1. Inventory + COVID-19 cases + GDP shock + public debt + stringency index Variable Mean Standard deviation Min Max Cash transfers to workers 0.6727 0.4706 0 1 Public works 0.2424 0.4299 0 1 Unemployment benefits 0.3939 0.4901 0 1 Training and placement assistance 0.3758 0.4858 0 1 Wage subsidies 0.5455 0.4994 0 1 Entrepreneurship support 0.4061 0.4926 0 1 Income tax reduction 0.3818 0.4873 0 1 Labor regulations 0.6848 0.4660 0 1 Firm liquidity support 0.9818 0.1340 0 1 High-income countries 0.3273 0.4706 0 1 14 Variable Mean Standard deviation Min Max Low-income countries 0.1576 0.3655 0 1 Lower-middle-income countries 0.2545 0.4369 0 1 Upper-middle-income countries 0.2606 0.4403 0 1 GDP shock −0.0789 0.0548 −0.4214 0.0334 Debt shock 0.1099 0.1213 −0.2778 0.7257 Stringency Index—Q1, 2020 0.2203 0.0954 0.0211 0.7450 Health shock (COVID-19 cases) 8.2504 2.2648 1.1569 11.1998 Source: Authors’ calculations. Sample 2: 91 countries. This sample was used to estimate Figure 4 and Figure 6 (individual model including employment shock). TABLE A5.2 Sample 2. Inventory + employment + stringency Label Mean Standard deviation Min Max Cash transfers to workers 0.7802 0.4164 0 1 Public works 0.2088 0.4087 0 1 Unemployment benefits 0.5934 0.4939 0 1 Training and placement assistance 0.4396 0.4991 0 1 Wage subsidies 0.6703 0.4727 0 1 Entrepreneurship support 0.4286 0.4976 0 1 Income tax reduction 0.3626 0.4834 0 1 Labor regulations 0.7802 0.4164 0 1 Firm liquidity support 0.9890 0.1048 0 1 High-income countries 0.4615 0.5013 0 1 Low-income countries 0.0549 0.2291 0 1 Lower-middle-income countries 0.1868 0.3919 0 1 Upper-middle-income countries 0.2967 0.4593 0 1 Stringency Index—Q1, 2020 24.3934 8.1300 6.5862 58.8606 Employment shock −0.0439 0.0629 −0.2948 0.0832 Source: Authors’ calculations. 15 Sample 3: 52 countries. This sample was used to estimate Figure 3 and Figure 5 (individual model including informality). TABLE A5.3 Sample 3. Inventory + informality + stringency index Label Mean Standard deviation Min Max Cash transfers to workers 0.5962 0.4955 0 1 Public works 0.3077 0.4660 0 1 Unemployment benefits 0.2308 0.4254 0 1 Training and placement assistance 0.3077 0.4660 0 1 Wage subsidies 0.4038 0.4955 0 1 Entrepreneurship support 0.3269 0.4737 0 1 Income tax reduction 0.4038 0.4955 0 1 Labor regulations 0.7885 0.4124 0 1 Firm liquidity support 0.9808 0.1387 0 1 High-income countries 0.0769 0.2691 0 1 Low-income countries 0.3077 0.4660 0 1 Lower-middle-income countries 0.2692 0.4479 0 1 Upper-middle-income countries 0.3462 0.4804 0 1 Stringency Index - Q1, 2020 0.2134 0.1198 0.0574 0.7450 Informality 0.5840 0.2301 0.0740 0.9994 Source: Authors’ calculations. Sample 4: 50 countries. This sample was used to estimate Figure 3 and Figure 5 (individual model including previous unemployment benefits). TABLE A5.4 Sample 4. Inventory + COVID-19 cases + GDP shock + public debt + stringency + informality + previous unemployment benefits Label Mean Standard deviation Min Max Cash transfers to workers 0.6200 0.4903 0 1 Public works 0.3000 0.4629 0 1 Unemployment benefits 0.2400 0.4314 0 1 Training and placement assistance 0.3200 0.4712 0 1 Wage subsidies 0.4200 0.4986 0 1 Entrepreneurship support 0.3200 0.4712 0 1 16 Label Mean Standard deviation Min Max Income tax reduction 0.4000 0.4949 0 1 Labor regulations 0.7800 0.4185 0 1 Firm liquidity support 1.0000 0.0000 1 1 High-income countries 0.0800 0.2740 0 1 Low-income countries 0.2800 0.4536 0 1 Lower-middle-income countries 0.2800 0.4536 0 1 Upper-middle-income countries 0.3600 0.4849 0 1 Stringency Index - Q1, 2020 0.2169 0.1208 0.0574 0.7450 Previous unemployment benefits 0.3400 0.4785 0 1 Informality 0.5871 0.2338 0.0740 0.9994 Source: Authors’ calculations. Sample 5: 89 countries. This sample was used to estimate Figure 3 and Figure 5 (individual model including benefit incidence in the poorest quintile). TABLE A5.5 Sample 5. Inventory + COVID-19 cases + GDP shock + public debt + stringency + benefit incidence Label Mean Standard deviation Min Max Cash transfers to workers 0.6739 0.4713 0 1 Public works 0.3261 0.4713 0 1 Unemployment benefits 0.3043 0.4627 0 1 Training and placement assistance 0.3370 0.4753 0 1 Wage subsidies 0.4457 0.4998 0 1 Entrepreneurship support 0.3804 0.4882 0 1 Income tax reduction 0.3913 0.4907 0 1 Labor regulations 0.6630 0.4753 0 1 Firm liquidity support 0.9891 0.1043 0 1 High-income countries 0.0543 0.2279 0 1 Low-income countries 0.2065 0.4070 0 1 Lower-middle-income countries 0.3696 0.4853 0 1 Upper-middle-income countries 0.3696 0.4853 0 1 Stringency Index—Q1, 2020 0.2163 0.1038 0.0211 0.7450 Benefit incidence (Poorest quintile) 0.0949 0.0860 0.0000 0.5759 Source: Authors’ calculations. 17 A6. ADDITIONAL FIGURES FIGURE A6.1 Jobs-related policies adoption and health shock (COVID-19 cases) CT to workers Public works Unemployment benefits Training and placement Wage subsidies Entrepreneurship support Income tax reduction Labor regulations Firm liquidity support −0.1 −0.05 0 0.05 0.1 Percentage points Source: Authors’ calculations. Note: This figure shows the coefficient and the 95% confidence interval of the variable of interest. This figure is based on separate regressions coefficients for the different shocks and the adoption of labor regulations as dependent variable. The regression models follow equation 1; see Appendix A3. This brief was prepared by Ivette Contreras, Melanie Khamis, David Newhouse, and Michael Weber. The production and publication of this report has been made possible through financial support from the World Bank’s Jobs Umbrella Multi-donor Trust Fund (MDTF), which is supported by the UK’s Foreign, Commonwealth & Development Office/UK AID; the Governments of Austria, Germany, Italy, and Norway; the Austrian Development Agency; and the Swedish International Development Cooperation Agency. The team is also grateful to the Poverty and Equity Global Practice and the Data for Goals group for collecting, harmonizing, and sharing the phone survey data. Aggregate indicators from the World Bank High Frequency Phone Survey (HFPS) are available at the HFPS Dashboard at https://www.worldbank.org/en/data/interactive/2020/11/11/covid-19-high-frequency-monitoring-dashboard. All Jobs Group’s publications are available for free and can be accessed through the World Bank or the Jobs and Development Partnership website. Please send all queries or feedback to Jobs Group. Join the conversation on Twitter: @WBG_Jobs #Jobs4Dev. 18