Poverty and Equity Global Practice Losing Livelihoods The Labor Market Impacts of COVID-19 in Bangladesh M a r i a E ug e ni a Ge noni A f sana I f fat Khan N andini K r i s hnan Net hr a Pal aniswamy Wa meq R az a Losing Livelihoods: The Labor Market Impacts of COVID-19 in Bangladesh M ar i a E ug e ni a Ge noni A f sana I f fat Khan N a ndini K r i s hnan Net hr a Pal aniswamy Wameq R az a 1 Abstract This paper provides early insights into the labor market impacts of the ongoing COVID-19 crisis in Bangladesh, with a special focus on three especially vulnerable areas: poor areas in Dhaka and Chittagong City Corporations and Cox’s Bazar district. We build on household surveys collected before the crisis and phone monitoring surveys collected after the start of the crisis to shed light on the implications of COVID-19 for employment and earnings. The findings presented here indicate substantial labor market impacts both at the extensive and intensive margin, with important variation across areas and gender, largely due to the nature of occupations affected by the crisis. The findings also point to substantial uncertainty about job prospects. Keywords: COVID-19, coronavirus, Dhaka, Chittagong, Cox’s Bazar, labor market, poverty JEL codes: D1, I15, I31, J2 J461 1 The World Bank Group. Corresponding author: M.E. Genoni (mgenoni@worldbank.org). The authors are grateful to the ed- itor and referees for their helpful comments. We also thank Claudia Berg, Luz Carazo, Joaquin Endara, Arshia Haque, and Flavio Riva for excellent research assistance and valuable inputs into survey design and implementation. We also thank Benu Bidani, Suleiman Namara, and Yutaka Yoshino for their useful comments and guidance. The Cox’s Bazar Panel Survey (CBPS) is the result of a partnership between the Yale Macmillan Center Program on Refugees, Forced Displacement, and Humanitarian Responses (Yale Macmillan PRFDHR), the Gender & Adolescence: Global Evidence (GAGE) program, the Pov- erty and Equity Global Practice of the World Bank, and the State and Peacebuilding Fund (SPF) administered by the World Bank. The SPF is a global fund to finance critical development operations and analysis in situations of fragility, conflict, and violence. The SPF is supported by Australia, Denmark, Germany, The Netherlands, Norway, Sweden, Switzerland, and The United Kingdom, as well as IBRD. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the World Bank, the Executive Directors of the World Bank, or the governments they represent. All errors and omissions are our own. 2 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Introduction The ongoing COVID-19 pandemic has creat- and their households depend on activities that ed an unprecedented crisis in Bangladesh are being directly affected by the crisis. Third, that risks erasing the substantial progress the absence of formal safety nets is expected in household incomes and poverty reduc- to exacerbate impacts, as income shocks tend tion achieved during the past decades. The to be largely managed with households’ own sharp decline in demand for manufactured resources. goods, particularly from the export-oriented Ready-Made Garments sector, is expected Information from the recently collected rep- to affect employment creation in urban ar- resentative phone surveys in poor and slum eas, an important driver of poverty reduction areas of Dhaka and Chittagong and in Cox’s in the past (Hill and Genoni, 2019). In addi- Bazar highlight the substantial labor market tion, large labor-income losses are expected impacts due to COVID-19. Job losses and tem- for households engaged in informal services porary absence are widely reported in all three and labor-intensive sectors like construction, areas, with Dhaka reporting the largest job due to slower demand and social-distancing losses, whereas in Cox’s Bazar, respondents measures. Moreover, the domestic coronavi- tended to report being employed, but tem- rus outbreak and the consequent healthcare porarily absent from work. Given the largely burden, together with related disruptions, will informal nature of the jobs held by the major- exacerbate the negative impacts on access to ity of active and temporarily absent workers services and poverty. who report themselves as being employed, it is difficult to predict how fully this currently This article combines recent panel surveys reported employment will translate into active and existing household surveys collected jobs post-lockdown. Job and monetary losses before and after the advent of COVID-19 to are accompanied by widespread uncertain- shed light on the impacts of the pandemic on ty about whether people will be able to keep households’ economic wellbeing. In particular, their jobs or keep their businesses running. the focus is on the labor market. High-den- sity slum and urban areas, as well as areas In addition, given the low rates of female labor of high localized density in Cox’s Bazar, may force participation, women appear to be dis- be particularly vulnerable to this crisis, as proportionately affected by the COVID-19 crisis lockdowns and social-distancing measures and have experienced relatively higher job loss- overlap with elevated risks of disease spread. es. In Dhaka and Chittagong, these have trans- Thus, we zoom in on these three areas to bet- lated into women leaving the labor force, while ter understand the implications of COVID-19 in Cox’s Bazar, women have been more likely to for these labor markets. look for work. In addition, in Dhaka and Chit- tagong, women who remained actively work- The analysis is motivated by the fact that im- ing experienced larger declines in earnings and pacts of the pandemic on households’ econom- more uncertainty about their job prospects. ic well-being and poverty will largely depend on how labor markets respond to the crisis, The analysis indicates that the differential im- as labor income has been the main source of pacts across areas and gender are linked to poverty reduction in the past (Hill and Genoni, workers’ occupations before the crisis. The 2019). Pre-existing vulnerabilities are therefore COVID-19 pandemic has disrupted activities to a source of concern, particularly for the urban varying degrees, and workers in certain vulner- poor. First, even before the COVID-19 crisis, a able activities have been more affected. The large share of the population – 8 in 10 Bangla- differential impacts for women are also related deshis – were poor or vulnerable to falling into to their engagement in highly impacted sectors, poverty, suggesting that income losses related such as garments and housemaid services. to COVID-19 are likely to push a large share of the population into poverty. Second, a sub- The next section describes the panel surveys stantial part of this vulnerability, particularly in and household surveys used for the analysis. urban areas, arises from the fact that incomes Subsequently, the main findings are presented. among a large share of Bangladeshi workers The final section provides some reflections. 3 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Data national-level analysis before the COVID-19 from the greater Dhaka statistical Metropolitan crisis relies on the household Income and Area, following a two-stage stratification design. sources The enumeration areas were selected during the Expenditure survey (hIEs) collected between april 2016 and March 2017 by the Bangladesh first stage using probability proportional to size, Bureau of Statistics. This is the latest official stratified by the poverty headcount ratio esti- source of household income, consumption, and mated using small-area techniques. all house- poverty data for Bangladesh (ahmed et al, 2017, holds in the selected enumeration areas were 2020). The findings post COVID-19 draw on two listed during the second stage, from which rapid panel phone surveys described below. 20 households were selected for interview- ing based on a demographic stratification. This Monitoring surveys in poor and slum areas of second level of stratification was defined as Dhaka and Chittagong City Corporations follows: (i) households with both working-age male and female members; (ii) households with To track the impacts of the COVID-19 crisis on only a working-age female; (iii) households with labor markets and household coping strategies, only a working-age male. households were ran- a rapid phone survey was implemented on a domly selected from each stratum with the pre- representative sample of households living in determined ratio of 16:3:1 (Kotikula et al, 2019). poor and slum areas of Dhaka and Chittagong The DIgnITy survey, administered between July City Corporations (CCs). The analysis presented and september 2018, collected information here summarizes results from the first round from 2,376 individuals across 1,302 households. of the rapid phone survey conducted from June 10 to July 10, 2020. The monitoring survey in Chittagong is a follow- up of the CITy (Chittagong low income area The monitoring survey built on baseline sur- Inclusion, and PoverTy) survey carried out in veys conducted before the COVID-19 crisis. The Chittagong City Corporation following the same sample for Dhaka is a follow-up of the DIgnITy sampling strategy as the DIgnITy survey. Data (Dhaka low Income area gender, Inclusion, and was collected from 1,289 individuals across poverTy) survey which was representative of 805 households between september and low-income areas and slums of the Dhaka City October 2019. figure 1 presents the location of Corporations and an additional low-income site the sampled areas in Dhaka and Chittagong. Figure 1. Location of sampled areas in Dhaka and Chittagong a. Dhaka a. Chittagong Source: authors’ rendition, based on the DIgnITy and CITy dignity_neighborhood data. 4 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice For the monitoring survey, a representative Bangladesh’s local economy started experi- sub-sample of 1,500 households out of a total encing impacts of the COVID-19 crisis in early 2,107 baseline households was targeted. The to mid-March 2020, with the country’s first recontact rate was 1,483 households (99.5 case being reported on March 7. A full coun- percent). In this first tracking survey 1,483 out trywide lockdown was in place from March of the 3,665 adults surveyed in baseline were 26 to May 28, 2020. The monitoring survey covered. It is important to note that at the measured outcomes across three periods: (i) time of the follow-up, 2.3 percent of adults During the survey period (7 days prior to sur- had moved residence from their baseline vey period between June 10 and July 10; (ii) location. The analysis includes those adults from March 25 to the time of the interview for even though they are currently located outside individuals who reported being unemployed the City Corporations. Given the small share in the week preceding the interview; and (iii) of the sample that moved, the results are not from January to March 25, 2020, for individu- affected by their inclusion. als who reported being unemployed from 25 March onwards (immediately before the lock- Table 1 presents some descriptive characteris- down started). tics by area and gender. The adults interviewed in 2020 were 35 years old on average, and 45 Cox’s Bazar rapid phone survey percent of them were female. Approximately 57 percent of the adults interviewed were the A rapid phone survey was implemented in main breadwinner of households with an aver- April-May 2020 on a representative sample of age household size of 4 people. On average, 1.4 recently displaced Rohingya households and household members generate income. Respon- their host communities in the Cox’s Bazar dis- dents from Dhaka are more likely to be living trict of Bangladesh, to track the impacts of the in slum areas than those located in Chittagong (70 versus 52 percent, respectively). Chittagong COVID-19 crisis on labor markets, wages, and households are larger (0.4 members more on household coping strategies. This survey built average) and therefore show higher dependen- on the 2019 Cox’s Bazar Panel Survey (CBPS), cy ratios. About 49 percent of slum residents which is a multi-topic survey that focused are women, on average, contrasting with a low- on socio-economic outcomes and access to er percentage in other areas (39 percent). health services. Table 1. Descriptive characteristics of adults living in poor and slum areas of Dhaka and Chittagong Non-   All Dhaka Chittagong Slum Male Female Slum Female (%) 44.8 45.3 44.2 48.7 39.4 Age (mean) 35.0 35.1 34.8 34.7 35.6 37.1 32.5 Breadwinner (%) 57.4 58.7 56.2 56.9 57.6 92.1 14.7 Household members (#) 4.3 4.1 4.5 4.2 4.4 4.2 4.3 Members who contribute 1.41 1.43 1.38 1.44 1.36 1.43 1.39 to HH earnings (#) Slum (%) 61.1 70.4 52.0 57.0 65.9 Dependency ratio (Members 15-64/Members 0.59 0.57 0.61 0.60 0.57 0.55 0.63 <15 and 65+) Observations 1483 836 647 951 493 770 713 Note: Information from Round 1 collected between June 10 and July 10, 2020. Figures are weighted. 5 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice The CBPS was designed to be representative out of the 9,045 adults surveyed in baseline of recently arrived Rohingya (displaced after were covered.2 August 2017) and Bangladeshi households re- siding in host communities in Cox’s Bazar, and Similarly to the Dhaka and Chittagong surveys, the baseline for this survey was completed in the labor module for the Cox’s Bazar survey August 2019. The CBPS survey was represen- measured outcomes across three periods: tative of two types of hosts: those with low (i) During the survey period (7 days prior to and high exposure to the Rohingya influx. To survey period in late April to mid-May); (ii) from distinguish between host communities that March 1 to early April 2020 for individuals who are more or less affected by the arrival of the report being unemployed during the survey Rohingya, the survey’s sampling strategy uses period (when a potential lockdown was under a threshold of 3-hours walking time from a discussion and gradually came into effect); campsite to define two strata for hosts: (i) and (iii) from January to early March 2020 for host communities with potentially high expo- individuals who report being unemployed from sure to the displaced Rohingya, and (ii) host March 1 onwards (when the first known cases communities with potentially low exposure. of COVID-19 were identified in Bangladesh). Table 2 summarizes key characteristics of the CBPS baseline respondents in host communi- The findings from the follow-up are present- ties (both high and low exposure). ed as a panel update on baseline Bangladeshi adults. Employment is defined as the share This first round of the rapid phone survey, one of the labor force reporting having worked at of a series of high frequency follow-up surveys least one hour in the past seven days or being to track the evolution of the COVID-19 crisis, temporarily absent from work. The labor force was conducted from 21 April-20 May 2020 (a is defined as adults over the age of 15 who month into the two-month-long COVID-19 are either currently employed or not employed lockdown). A sub-sample of 3,176 out of the but actively seeking work over the past seven 5,020 households surveyed at baseline were days. Similarly, unemployment rates are re- covered by this survey. The baseline CBPS ported as a percentage of the labor force that survey was designed to be administered has not worked in the past seven days or been to two randomly selected adults in each temporarily absent from a job but has actively household. In this first tracking survey, 3,009 looked for work in the stated recall period. Table 2: Descriptive statistics for host communities, Cox’s Bazar Panel Survey Households Adult Respondents % Household Ages Ages Ages Ages Female- headed % Age (av- women size 0-6 7-14 15-64 65+ households women erage) 50.7% 5.1 16.6% 21.6% 58.0% 3.7% 17.7% 57.4% 33 Findings Pre-crisis vulnerabilities The impacts of COVID-19 on households’ eco- a source of concern, particularly for the urban nomic well-being and poverty will largely de- poor who rely on the informal sector for work pend on how labor markets respond to the cri- and incomes.2 sis, as labor income has been the main source of Bangladesh’s poverty reduction in the past Even before the COVID-19 crisis, a large share and, on average, comprises more than 80 per- of the population – 8 in 10 Bangladeshis – cent of total household income for the poor- est 40 percent of households (Hill and Genoni, 2 The results are weighted using adjusted baseline weights 2019). Pre-existing vulnerabilities are therefore that account for non-response and selection into the inter- view based on characteristics measured at baseline. 6 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice were poor or vulnerable to falling into poverty. focuses on short-term labor-income impacts According to the HIES 2016/17 about 25 due to slower GDP growth in the fiscal year percent of the population were living in poverty 2020. The analysis considers the slowdown in and another 54 percent could be considered growth in agriculture, industry, and services, vulnerable, as they had consumption levels along with changes in inflation.5 It uses a full very close to the poverty line (between the pass-through rate to model a slowdown in official upper poverty line and twice the line).3 household real incomes and increases in the Poverty and vulnerability were high in both cost of living. Moreover, as an important share urban and rural areas (Figure 2), suggesting of household incomes are informal and not that income losses related to COVID-19 are well captured in macro-growth projections, likely to push a large share of the population the simulation separately models additional into poverty. reductions in labor incomes for daily and in- formal workers, as well as for self-employed Figure 2. Poverty and vulnerability by area workers in services and other affected sec- (% of the population) tors such as manufacturing, construction and transport. An international remittance income Poor Vulnerable Middle class 100% shock is also added. In the main scenario, in- 22% 19% 30% comes of informal workers and in affected 80% sectors are assumed to decline by 50 percent 60% (about six months of no income), and interna- 54% 55% 40% 51% tional remittances are also assumed to fall by half. A less severe scenario is also estimated, 20% 25% 27% 19% assuming that incomes and remittances de- 0% cline by only 25 percent (one quarter of the National Rural Urban year without income). Source: Authors’ calculations using HIES 2016/17. Note: Poverty defined using the official upper poverty rate. The results from this simulation suggest sub- Vulnerable households are households with per capita con- stantial reductions in per capita household sumption between the official upper poverty line and twice consumption and poverty associated with the upper poverty line. Middle class households are those with per capita consumption above twice the upper pov- the crisis. Comparing with a scenario with- erty line. out COVID-19, in 2020, average household per capita consumption would decline by an es- A micro-simulation using the HIES 2016/17 was timated 13 percent, with an estimated loss in conducted to assess the potential impacts of annual consumption of about US$ 10.7 billion the COVID-19 crisis on household per capita (Table 3). The national upper poverty rate is consumption and poverty rates in 2020, com- estimated to rise from 23 percent to 35 per- pared to a non-COVID situation.4 The simulation cent (Figure 3), with around 3 percentage points coming from the reduction in interna- 3 Poverty is defined based on the official upper poverty tional remittances. This represents approxi- line from the Bangladesh Bureau of Statistics (Ahmed et mately 21 million additional people falling into al, 2019). poverty in 2020. In a more moderate scenar- 4 This microsimulation follows a similar approach to that io where informal and self-employed workers used by Habib et al. (2010) to assess the ex-ante distribu- experience an income loss of three months tional impact of the global financial crisis on Bangladesh. Essentially, the simulation combines COVID-19-adjusted and remittances fall by 25 percent, poverty is macroeconomic growth projections for 2020, inflation pro- estimated at 31 percent nationally. jections for 2020, and the income shocks described above with pre-crisis microdata on household consumption in- labor market/consumption decisions. It does not model come and expenditure to simulate poverty headcounts the health impacts on households affected by COVID-19. and consumption distributions for 2020 under different Welfare analysis only assesses changes to income and con- scenarios. Note that this microsimulation is a short-term sumption. Other non-monetary effects, such as the impact exercise with important limitations in the context of a high of COVID-19 on human capital, are not assessed. level of uncertainty about the extent and complexity of the 5 The microsimulation considers a non-COVID GDP growth pandemic in Bangladesh. The microsimulation assumes rate of 6.7 percent and a COVID GDP growth rate of 1.6 per- that there will be no new mitigation measures, such as cent for FY20 and 1 percent for FY21 based on projections cash-transfers/assistance or adjustments in households’ from the World Bank Global Economic Prospects. 7 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Table 3. Average household consumption per capita No COVID-19 With COVID-19 Location Baseline 2019 Scenario in 2020 Scenario in 2020 Change (%) National 4,845 5,114 4,532 -13% Rural 4,276 4,502 4,011 -12% Urban 6,299 6,677 5,860 -15% Source: Authors’ calculations using HIES 2016/17. Figures in monthly Takas of 2020. Figure 3. Estimated poverty rates for 2020 COVID-19 and non-COVID-19 scenarios 35% 36% 32% 25% 23% 17% No COVID-19 Scenario With COVID-19 National Rural Urban Source: Authors’ calculations using HIES 2016/17. When comparing urban and rural areas, the compared to 41 percent of workers in rural simulation indicates that urban areas would areas (Figure 4). In urban areas of Dhaka, 3 be more severely affected. Average household in 4 workers are expected to have been di- per capita consumption in urban areas is rectly affected, in Chittagong this is true for estimated to be 15 percent lower and poverty 63 percent of workers. Dhaka and Chittagong 82 percent higher with COVID-19, relative to a divisions comprise 68 percent of all directly scenario without COVID. In contrast, average affected workers. household per capita consumption in rural areas is estimated to be 12 percent lower and The absence of formal safety nets is expected poverty 44 percent higher. to exacerbate impacts, as income shocks tend to be largely managed with households’ own These severe impacts arise in large part be- resources. According to HIES 2016/17, cop- cause, particularly in urban areas, the in- ing responses to income shocks are varied, comes of a large share of workers and their but few households rely on formal response households depend on activities directly mechanisms. A large share of poor and vul- affected by the crisis. These include daily nerable households that reported an income and self-employed workers in non-agricul- shock in the past year reduced food con- ture and salaried workers in manufacturing.6 sumption as a response, and this proportion In 2016, 2 in 3 urban workers were engaged is larger in urban areas (Figure 5). The wide- in activities directly impacted by the crisis, spread reliance on food consumption as a way to cope with shocks suggest potential neg- ative impacts on food security. Depletion of 6 This does not imply that workers in agriculture or salaried workers in non-garment sectors have not been affected. own savings and assets can also compromise However, such workers are not directly impacted by the households’ future earning potential, particu- slowdown in exports or social-distancing measures in high- larly in a protracted period of decreased in- ly dense economic areas, or they may have relatively secure comes and work. sources of employment. 8 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Figure 4. Percentage of workers in directly affected sectors, by division and area 74% 63% 66% 66% 59% 61% 58% 54% 47% 44% 45% 39% 37% 39% 41% 35% Urban Rural Barisal Chittagong Dhaka Khulna Rajshahi Rangpur Sylhet All Source: Authors’ calculations using HIES 2016/17. Figure 5. Ways to cope with income shocks among poor and vulnerable households, by area (% of households that experienced an income shock in the past year) Reduced food consumption Help from friends Government's help Used savings Looked for more work Obtained credit 80% 69% 69% 63% 60% 55% 50% 42% 38% 40% 33% 30% 20% 22% 20% 14% 12% 9% 7% 9% 8% 9% 6% 4% 0% Rural Urban Source: Authors’ calculations using HIES 2016/17. Note: Poor and vulnerable households only. COVID-19 labor-market impacts from the stopped work between March 25, when the monitoring surveys official COVID-19 lockdown was announced, and the time of the interview. In Dhaka, 1 in 4 Employment impacts so far are large in terms respondents reported not actively working in of jobs losses, absenteeism, and reduced earn- the week preceding the interview but having ings, in a context of high uncertainty about worked before March 25, 2020. In Chittagong jobs prospects this figure was 22 percent. Slum areas showed a higher share of people stopping work (26 Information from the recently collected percent) compared to non-slum poor areas (19 phone surveys in poor and slum areas of Dha- percent).7 When asked why they had stopped ka and Chittagong and in Cox’s Bazar high- working, 9 out of 10 respondents attributed light substantial labor-market impacts due to the change to COVID-19-related disruptions. COVID-19. Job and monetary losses are ac- companied by widespread uncertainty about The group of Dhaka and Chittagong respon- keeping jobs and businesses running. dents that stopped actively working is com- posed of people expecting to resume work, Job losses and temporary absence are widely searching for a new job, or exiting the labor reported in all three study areas, with Dhaka force (Figure 6). About 32 percent of adults reporting the largest job losses, whereas in who had stopped working after March 25 Cox’s Bazar, respondents tended to report were not searching for jobs, as they expect- being employed, but temporarily absent from ed to resume their previous activity. However, work. In poor and slum areas of Dhaka and Chittagong CCs, 23 percent of adults had 7 This is observed both in Dhaka and Chittagong slums. 9 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Figure 6. Dhaka and Chittagong - Employment Figure 7. Cox’s Bazar district - Labor force status among respondents who stopped ac- indicators between baseline and follow up tive work after March 25 (% of adults) Inactive Searching Temporarily absent Baseline (Mar-Aug 2019) Follow up (Apr-May 2020) 100 95.3% 24 29 88.7% 32 39 75 41 42 51.4% 50 41 42 42.1% 39 38 25 34 11.3% 27 29 24 4.7% 20 0 All Dhaka Chittagong Slum Non-Slum Labor force participation Employment rate Unemployment rate the remaining 68 percent seem to have expe- 56 percent, with the majority of these absenc- rienced a job loss, as they report exiting the es being recorded after COVID-19. Unsurpris- labor force or currently searching for jobs. Job ingly, an overwhelming majority of temporar- losses were higher in Dhaka (76 percent) than ily absent workers attributed the situation to Chittagong (59 percent). Slum areas also show COVID-related restrictions on work. higher job losses (71 percent) than non-slum areas (61 percent). It is also important to note Reported income losses were widespread that some of the respondents expecting to re- across all three areas. In Dhaka and Chittagong, sume their previous jobs may not be able to, about 80 percent of wage workers and 94 per- thus actual job losses may have been higher cent of business owners said that their earn- than these results initially suggest. ings were lower than usual. Median wages for salaried and daily workers declined by about In Cox’s Bazar district, economic lockdowns 37 percent compared to usual earnings im- were imposed early, due to the risks associ- mediately before COVID-19.8 The decline was ated with disease spread in the densely pop- higher in Dhaka (42 percent) than Chittagong ulated Rohingya camps. However, compared (33 percent) and in slum areas compared to to Dhaka and Chittagong, Cox’s Bazar district non-slum poor areas (43 and 33 percent, re- is less urbanized, with its urban areas being spectively).9 Dhaka showed wage declines located relatively far from concentrations of larger than Chittagong’s for both genders. recently displaced Rohingya. Although close to 90 percent of the Bangladeshi living in Cox’s As in Dhaka and Chittagong, those who re- Bazar reported being employed during the mained active during the lockdown in Cox’s lockdown (Figure 7), these employment rates Bazar reported reduced earnings, with urban- mask high rates of temporary absence from ized, low-exposure areas being more affect- work. Reported employment, even within the ed across all employment types (Figure 8). lockdown period, remained high (89 percent). Among wage workers, daily and weekly wage However, a large share of the labor force re- laborers faced much higher losses in income ported being employed but temporarily absent (49 percent) compared to salaried workers. from work, i.e. not actively working. Almost 2 out of 3 adults who reported being employed 8 The labor market questions for daily workers in the fol- were in fact not actively working in the 7 days low-up questionnaire were simplified due to time con- before the survey. In contrast, during the base- straints. To estimate daily wage changes for daily workers, line period (March to August 2019), temporary reported weekly hours were converted to daily hours as- absence from work among the employed was suming 8 hours of work per day and allowing a work week less than 1 percent. Rates of temporary ab- to be up to 7 days. sence from work increased from 3 percent to 9 Comparisons are statistically significant conditional on the different slum/non-slum composition of the cities. 10 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Figure 8. Cox’s Bazar district - Comparison of pre-crisis and lockdown earning levels for different employment types Baseline rates Crisis impacted rates (March-May 2020) Weekly wage labor Monthly salaries Nonwage earnings Weekly wage labor Monthly salaries Nonwage earnings earnings (last month) earnings (last month) High exposure Low exposure Monthly salaried wage workers in Cox’s Bazar have been relatively protected in terms of in- Figure 9. Dhaka and Chittagong - Expectations come losses, reporting 15-19 percent reduc- about keeping current employment next month tions across high- and low-exposure hosts. (% of adults who worked in the past week) Given the largely informal nature of the jobs Yes No Not sure held by the majority of active and temporar- 100 ily absent workers who report themselves as being employed, it is difficult to predict how 28 32 39 fully this employment will translate into active 75 45 50 jobs post-lockdown. This is reflected in the 4 widespread uncertainty that respondents from 5 3 Dhaka and Chittagong reported about keeping 2 jobs and businesses running (Figure 9). Only 58 50 2 percent of active workers thought they would be able to keep their job or activity running in 69 64 the month following the survey. In Chittagong, 58 53 25 48 69 percent of workers expected to continue working. Compared to Chittagong, Dhaka shows a much higher degree of uncertainty about employment prospects, with only 48 percent 0 of workers thinking they would keep their in- All Dhaka Chittagong Slum Non-slum come-generating activity. Slum residents show higher levels of uncertainty: 53 percent of work- ers in slums expected to remain at their jobs, In Dhaka and Chittagong, the percentage of compared to 64 percent in non-slum areas. males and females stopping work between March 25 and the interview date was 23 and Females have been disproportionately affected 24 percent, respectively. However, given their due to their overall lower participation in the low participation in the labor force, women’s labor market and their occupations employment experienced a larger reduction. The share of actual job losses among those Given the low rates of female labor force partic- stopping work was not very different by ipation, women appear to be disproportionately gender, but men were more likely to actively affected by the COVID-19 crisis and have experi- look for another job while women were more enced relatively higher job losses. In Dhaka and likely to exit the labor market. Only 2 in 10 men Chittagong, these have translated into women stopping active work exited the labor force, leaving the labor force, while in Cox’s Bazar, compared to more than 1 in 3 women in Dhaka women have been more likely to look for work. and Chittagong (Figure 10). 11 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Figure 10. Dhaka and Chittagong - Employ- Figure 11. Cox’s Bazar district - increasing ment status for those who stopped active unemployment rates by gender work after March 25, by gender (% of adults) and exposure area Unemployed Temporarily absent Inactive Baseline (Mar-Aug 2019) Follow up (Apr-May 2020) 16.1% 100 22 34 12.3% 75 32 9.2% 50 32 6.7% 6.8% 25 5.3% 47 4.7% 34 2.3% 0 Male Female Note: Inactive are those respondents who left the labor force. High exposure Low exposure High exposure Low exposure Unemployed are those who are actively searching for jobs. Temporarily absent are those who are not looking for jobs Male Female because they expect to go back to their original employment. In Cox’s Bazar, although unemployment rates In addition, in Dhaka and Chittagong, women increased across areas and gender, women who continued actively working experienced in more urban, low-exposure areas were sig- larger declines in earnings and more uncer- nificantly more likely to become unemployed tainty about their job prospects. This seems (Figure 11). However, this increase was not to be linked to their engagement in occu- driven by job losses, but by new labor force pations hard hit by the crisis. Reductions in entrants seeking jobs. Two-thirds of new la- wages for salaried and daily workers were sig- bor force entrants in Cox’s Bazar were wom- nificantly higher for women, consistent with en, largely driven from low-exposure regions their high engagement in the garment sec- (70 percent); and close to 60 percent of these tor and housemaid services, both of which entrants are secondary household members, have been severely impacted by COVID-19. i.e. the spouses or children of the household The median wage decline for women was 43 heads. These increases are likely driven by the percent, compared to 33 percent for men. Ta- high rates of temporary absence from work ble 4 shows that active working women were reported among men. mainly concentrated in the garment industry Table 4. Percentage of workers across occupations by work status in 2020 and by gender Males Worked in Worked past Stopped active work Lost job since Occupation 2020 week since March 25 March 25 Drivers 13% 16% 5% 6% Garments worker 6% 7% 6% 7% Transport worker 7% 6% 10% 13% Construction worker 8% 8% 10% 13% Retail or sales worker 3% 3% 3% 5% Porter 14% 14% 13% 13% Cleaning or housemaid 1% 1% 0% 0% Wage - other 8% 8% 8% 5% Professional skilled 6% 5% 11% 7% Own account - retail 22% 24% 18% 15% or trade Own account - other 11% 9% 16% 16% Total 100% 100% 100% 100% 12 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice   Females Worked in Worked past Stopped active work Lost job since Occupation 2020 week since March 25 March 25 Drivers 0% 0% 0% 0% Garments worker 28% 42% 16% 19% Transport worker 0% 0% 0% 0% Construction worker 1% 0% 2% 3% Retail or sales worker 0% 0% 1% 1% Porter 1% 0% 1% 1% Cleaning or housemaid 43% 28% 56% 54% Wage - other 7% 9% 6% 4% Professional skilled 4% 2% 5% 1% Own account - retail or 11% 14% 9% 10% trade Own account - other 5% 5% 5% 7% Total 100% 100% 100% 100% Note: “Wage - other” groups occupational categories of less than 2% for the total sample. (42 percent) or working as a housemaid or more affected (World Bank 2020a). In the case cleaner (28 percent). Conditional on place of of Bangladesh, as in other developing econo- residence, age, and education, females were mies where labor informality is high, a larg- 13 percent more likely to report a wage loss er share of the population has been affected. than males and experienced a 14 percent larg- The analysis for Dhaka, Chittagong, and Cox’s er wage loss than males. In addition, while 59 Bazar indicates that the differential impacts percent of men expected to remain at their across areas and gender are linked to workers’ job in the next month, this was true for 52 occupations before the crisis. The differential percent of women (Figure 12). impacts for women are also related to their engagement in highly impacted sectors, such Figure 12. Dhaka and Chittagong - Expecta- as garments and housemaid services. tions about keeping current employment next month by gender (% of adults who worked in In Dhaka and Chittagong, workers reporting the past week) job losses were engaged across different Yes No Not sure types of occupations (Table 3).10 Among men, 100 the composition of occupations for those who continue to work actively and those who 38 lost work is fairly similar, though job losses 75 47 are slightly more likely for transport and 3 50 1 construction workers and less likely for drivers (rickshaw, private cars, etc.). For women, job 25 59 52 losses are also observed across occupations, but 54 percent of the cases are housemaids or cleaners. Comparing to the share of women 0 Men Female actively working or who had worked in this occupation in 2020 indicates that females in Occupation-specific vulnerabilities are an im- housemaid services have been more affected. portant factor shaping the impacts of the crisis 10 Job losses are statistically significant different from zero Across the world, including developed coun- across all occupation groups presented in Table 1 for both tries, the COVID-19 pandemic has constrained males and females. However, note that the sample for certain occupations more than others, and some occupations is small. Occupation percentages below workers in more informal activities have been 10 percent should be interpreted with care. 13 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice The more severe impact of the crisis in poor that Dhaka’s poorer performance is probably areas of Dhaka seems to be explained by the related to the different distribution of work- different occupational composition of workers. ers across sectors, and the larger reliance on Dhaka has a larger share of respondents in oc- COVID-19-vulnerable occupations, rather than cupations that have been strongly affected (in to larger wage drops within sectors. particular male transport workers and house- maid services). Table 5 shows a multivariate In Cox’s Bazar, wage workers were more likely to regression for the probability of reporting an report temporary absence during the COVID-19 income loss and the change in income for dai- lockdown period, whereas non-wage workers ly and wage workers. The results indicate that were more likely to experience reduced in- different occupations have experienced in- comes and operational activity. A higher pro- come losses and that, conditional on the oc- portion of employed enterprise owners report cupation, the location of the worker (Dhaka or actively working during the lockdown as op- Chittagong) is not significantly correlated with posed to employed wage workers, who mostly wage declines. When comparing wage changes report being temporarily absent from jobs (Fig- across sectors and cities, Chittagong appears ure 13). About 1 in 3 respondents who report to show larger income reductions (sample siz- being employed but temporarily absent from es are too small to confirm statistically signifi- work report being daily or weekly wage labor- cant differences in many cases). This indicates ers; half are non-wage own account workers. However, actively working enterprise owners Table 5. OLS regression for experiencing are much more likely to report lower incomes a wage decline than active wage workers. In other words, for   Reported a Wage those actively working during the lockdown, wage drop change (%) overall incomes have fallen due to reduced ac- (1) (2) tivity or reduced work hours, but wage rates Drivers 0.788*** -0.260*** have remained relatively inflexible downwards. (0.046) (0.072) Garments worker 0.912*** -0.481*** Rates of temporary absence in Cox’s Bazar were (0.039) (0.061) higher for low-exposure hosts (67 percent) than Transport worker 0.853*** -0.450*** for high-exposure hosts (53 percent), and higher (0.066) (0.103) for employed men. This could potentially be Construction 0.486*** -0.162* worker (0.057) (0.089) Retail/sales 0.847*** -0.524*** Figure 13. Cox’s Bazar district - Share of worker (0.086) (0.134) employment types among hosts describing Porter/day laborer 0.597*** -0.0797 themselves as temporarily absent from work (0.042) (0.066) Monthly salaried workers Daily/weekly wage laborers Cleaning/house- 0.927*** -0.609*** Non-wage workers maid (0.065) (0.101) 100% Wage other 0.874*** -0.548*** 12% 14% (0.051) (0.080) Professional/ 0.818*** -0.347*** 80% skilled (0.067) (0.105) 35% 38% Living in Dhaka 0.0359 -0.0301 60% (0.035) (0.055) 40% Observations 532 532 R-squared 0.826 0.332 53% 49% Notes: Standard errors in parentheses. Estimates weight- 20% ed. Regressions exclude the constant so coefficients can be interpreted as conditional means for each occupation category. 0% *** p<0.01, ** p<0.05, * p<0.1 High Exposure Low Exposure 14 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice explained by the nature of jobs that these groups In addition, income losses during the lockdown are typically engaged in: population segments for active non-wage own account workers and which were more dependent on agricultural business owners vary across high- and low- and home-based income-generating activities exposure areas in Cox’s Bazar. Among non- that are prevalent in high-exposure areas were wage workers, high-exposure hosts faced able to participate in some kind of economic much lower income losses (15 percent) than activity during the lockdown. This contrasted their counterparts in low-exposure areas (43 with the situation of respondents in service percent). With high-exposure hosts more sector jobs. These jobs were less accessible dependent on agriculture and low-exposure during the period. Table 5 highlights how active hosts on industrial and service sector income sources during the lockdown (April) are occupations (World Bank, 2020), this further mostly agricultural, not only in the more rural highlights the differential impact that the high-exposure areas, but also among active lockdowns have had on economic sectors, jobs reported in the relatively more urbanized with agricultural work facing more limited low-exposure regions.11 disruptions than other activities. Table 6. Cox’ Bazar district- Top 5 jobs reported by persons actively working High-exposure hosts % Low-exposure hosts % Farmer (on own land) 14.87% Self-run agricultural activities 15.91% Agricultural day laborer 13.79% Small businessman (tongs) 10.50% Small businessman (tongs) 9.02% Agricultural day laborer 10.19% Self-run agricultural activities 8.55% Private sector employee 7.51% Rickshaw/van driver 5.47% Hens/duck rearing 5.27% Other 48% Other 51% Conclusions This paper documents some early insights The analysis shows high levels of job un- into labor-market impacts of the COVID-19 certainty, reflected by the high absenteeism pandemic in Bangladesh, with a special focus rates and dim expectations of active work- on three vulnerable areas of the country. As the ers. This makes it difficult to infer the extent crisis develops, future rounds of representative to which this crisis will translate into per- monitoring data on the same respondents will manent job losses with longer-term conse- help understand the evolving impacts and quences for poverty, food-security, and fu- potential recovery. The findings presented ture earnings. It is likely that reported rates here indicate substantial labor-market impacts of employment in these rapid phone surveys both at the extensive and intensive margin, – which derive in large part from high rates of with important variation across areas and temporary absence - are underestimating the gender, largely due to the different nature magnitude of job losses which may be real- of occupations affected by this crisis. This ized once social-distancing measures lift and variation in occupational composition explains workers attempt to rejoin their jobs. Wheth- why workers in Dhaka appear to be more er some people’s positive expectations will adversely affected than those in Chittagong, translate into re-employment of these large- why those living in more urbanized parts of ly informal wage workers post-lockdown will Cox’s Bazar district have faced larger income depend on a host of factors, such as which and job losses, and why women have borne a sectors of the local economy are prioritized disproportionate burden from the crisis.11 in the partial economic reopening, how lo- calized quarantines of neighborhoods and 11 According to baseline findings, 41 percent of hosts in areas impacts job accessibility and mobility, high-exposure areas rely on agriculture for their livelihoods, and the overall economic outlook for the ma- compared to 30 percent for hosts in low-exposure areas jor sectors of employment in the economy. (World Bank, 2020). 15 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice Currently, the consequences of the crisis for This is an important part of the story but be- the continued operations of household enter- yond the scope of this paper. In poor areas of prises and self-employed workers continue Dhaka and Chittagong, 8 in 10 adults reported to play out, with considerable uncertainty as experiencing stress or anxiety that affected to how these activities will be sustained in a their ability to carry out their day-to-day ac- context of protracted decline in earnings. tivities in the month preceding the interview. The main reasons cited were fear of the ef- High levels of uncertainly in the job market are fects of COVID-19 on self and family (56 per- also generating stress and anxiety that may cent of cases) and fear of a loss of income (41 further exacerbate health impacts associated percent of cases). with the pandemic, notably in mental health. References Ahmed, F., D. Roy, M. Yanez-Pagans, and N. Yoshida. 2017. “Design of a Multi-Stage Stratified Sample for Poverty and Welfare Monitoring with Multiple Objectives: A Bangladesh Case Study.” World Bank Research Working Paper. Washington, DC: World Bank. Ahmed, F., M. E. Genoni, D. Roy, and A. Latif. Forthcoming (2020). “Official Methodology used for Poverty Estimation Based on the Bangladesh Household Income and Expenditure survey 2016/17: A Technical Note”. Bangladesh Development Studies. Kotikula, A., R. Hill, and W.A. Raza. 2019. What Works for Working Women? Understanding Female Labor Force Participation in Urban Bangladesh. Report. Washington, DC: World Bank. Habib, Bilal, Ambar Narayan, Sergio Olivieri, and Carolina Sanchez-Paramo. 2010. “Assessing Ex Ante the Poverty and Distributional Impact of the Global Crisis on a Developing Country: A Micro-simulation approach with Application to Bangladesh,” Policy Research Working Paper 5238. Washington, DC: World Bank. Hill, Ruth, and Maria Eugenia Genoni. 2019. Bangladesh Poverty Assessment: Facing Old and New Frontiers in Poverty Reduction.  Washington, D.C.: World Bank. World Bank. 2020a. Which Jobs Are Most Vulnerable to COVID-19? What an Analysis of the European Union Reveals (English).  Research & Policy Briefs; no. 34 Washington, D.C.: World Bank Group. http:// documents.worldbank.org/curated/en/820351589209840894/ Which-Jobs-Are-Most-Vulnerable-to-COVID-19-What-an- Analysis-of-the-European-Union-Reveals World Bank. Forthcoming (2020). Cox’s Bazar Panel Survey Briefs. 16 | Labor Market Impacts of COVID-19 in Bangladesh Poverty and Equity Global Practice