BRIEF - V Poverty and Equity Global Practice COVID-19 Monitoring Survey in Poor and Slum Areas of Dhaka and Chittagong Bangladesh Labor Market Situation as of Round 3 JA NUA RY 13 – F E BR UA RY 27, 20 21 To track the impacts of the COVID-19 crisis on labor markets and household coping strategies, a rapid phone survey was implemented on a representative sample of households living in poor and slum areas of Dhaka and Chittagong City Corporations (CCs). This brief, the fifth in the series, summarizes results from the first, second, and third rounds of the rapid phone survey, conducted from June 10 to July 10, 2020, from September 2 to October 11, 2020, and from January 13 to February 27, 2021, respectively. Eighty-eight percent of respondents interviewed in the first and second survey rounds were reached in the third round (see annex 1 for details of the survey design and response rates). This brief focuses on how the labor-market situation evolved between the rounds a year after the onset of the crisis.1 Main Labor market information collected between ary 2021 were women. Despite improvements, January and February 2021 showed continued around 24.5 percent of those who had lost messages improvements in employment in poor and slum their employment due to the pandemic were areas of Dhaka and Chittagong CCs, compared out of the labor force by February 2021. Impor- to June 2020. The share of adults reporting tantly, 82 percent of those cases were women. work in the week preceding the interview increased by 16 percentage points, from 50 to Around 4 in 10 of those who started to work 66 percent, between June 2020 and February again after being impacted by COVID-19 2021. On average, the share of respondents regained jobs by switching their occupation. actively seeking jobs and absent from work For those who lost their employment due fell to less than 2 percent by February 2021.2 to COVID-19 and started work again, only 59 Employment increased in both cities, with the percent returned to the same occupation. gains in Chittagong large enough to reach pre- Workers who are living in Dhaka, those in slum COVID-19 employment levels. areas, and women, were more likely to return to the same occupation. Yet, about 76 percent The share of respondents employed recovered of those who recovered their employment for both genders, though the employment lev- remained in the same employment category el of women in Dhaka still lag far below their (daily, wage, or self-employed worker). pre-COVID-19 levels, and inactivity rates for women have risen. Pre-COVID-19 employment While revenues from self-employment have shares,3 though not strictly comparable, sug- not recovered fully, median earnings for sala- gest that employment for both men and wom- ried and wage workers seem to have bounced en in Chittagong has recovered to pre-pan- back to levels commonly reported before demic levels. In contrast, women in Dhaka the pandemic. For monthly wage and sala- were still about 8 percentage points below ried workers, median earnings in June 2020 their employment levels prior to COVID-19. were 20 percent below the earnings reported In addition, inactivity rates among women in- before the pandemic. By February 2021, me- creased from 62 to 67 percent between June dian earnings were 3 percent lower than the 2020 and February 2021. Family responsibili- pre-COVID-19 amounts. A similar recovery is ties (e.g., housewife, caretaker) were the main seen for daily workers’ wages. However, for the factor cited by women who were unemployed self-employed and business owners4, gains or absent from their jobs in round 1 and then have been slower. Median revenues for these left the labor force altogether by round 3. workers in June 2020 were about half the pre- COVID-19 amounts, and in February 2021, only About 2 in 3 adults who lost jobs due to recovered to the level 20 percent below the COVID-19 and were still not working by Febru- pre-crisis values.5 1 Previous briefs can be found here. 4 Due to the nature of the survey, it is likely these are small 2 Figures based on respondents interviewed in the three business owners. rounds. Comparisons using cross-sectional data on all re- 5 Comparisons based on values for usual earnings and rev- spondents show very similar shares. enue need to be interpreted with care, as respondents’ 3 Employment is calculated as the share of adults who re- recollection of usual amounts may differ from the actual ported work in the past week over all adults. amounts received in the week or month preceding the in- 2 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice Findings Figure 1. Labor market status in the week preceding the survey (% of respondents) 100% 31 31 32 30 28 29 32 33 30 29 30 34 36 34 36 75% 2 3 2 8 6 8 2 7 6 4 6 10 3 9 12 12 5 10 9 8 50% 66 69 68 61 61 61 62 63 63 57 25% 50 52 49 52 49 0% R1 Jun-20 R2 Sep-20 R3 Jan-21 R1 Jun-20 R2 Sep-20 R3 Jan-21 R1 Jun-20 R2 Sep-20 R3 Jan-21 R1 Jun-20 R2 Sep-20 R3 Jan-21 R1 Jun-20 R2 Sep-20 R3 Jan-21 All Dhaka Chittagong Slum Non-slum Working Unemployed Absent Inactive Note: Figures based on all adults interviewed in the three rounds. ‘Inactive’ describes respondents out of the labor force. ‘Unemployed’ are those who are actively searching for jobs. Temporarily ‘absent’ describes those who are not looking for jobs because they expect to go back to their original employment. Labor market information collected between unemployment translated into employment January and February 2021 showed continued because of those who exited the labor force improvements in employment in poor and since June. Among those adults who were slum areas of Dhaka and Chittagong CCs, unemployed in June 2020, around 82 percent compared to June 2020 (Figure 1). The of them had resumed working by February percentage of adults reporting work in the 2021 (table 1), but 15.5 percent had exited the week preceding the interviews increased labor force. Sixty-nine percent of those who by 16 percentage points, from 50 percent in were employed but were absent from their June 2020 to 66 percent in February 2021. jobs in June 2020 reported actively working The gains in employment were accompanied again by February 2021, but 29 percent exited by a reduction in the share of respondents the labor force. In addition, 6 percent of unemployed or absent from work, and no respondents who were actively working in change in the share of adults out of the labor June 2020 had exited the labor force by the force. On average, the share of respondents third survey round. These labor market exits actively seeking jobs and absent from work were offset by new entrants (17.6 percent fell to less than 2 percent by February 2021. of adults who were inactive in June 2020), Meanwhile, the share of adults out of the labor keeping the overall inactivity rates relatively force remained at around 32 percent.6 constant across the three survey rounds. The employment transition data between Employment increased in both cities, large- June 2020 and February 2021 shows that ly coming from reductions in unemployment not all the reductions of absenteeism and and absenteeism. Between June 2020 and February 2021, the percentage of adults re- porting work increased by 17 percentage terview. In addition, the values in Takas are obtained from points in Dhaka, while in Chittagong this share simple questions that can be asked in a phone survey. They may not align with figures from more detailed and face-to- grew by 13 percentage points. By February face surveys. 2021, about 69 percent of adults in Dhaka and 6 Figures based on respondents interviewed in the three 62 percent of adults in Chittagong reported rounds. Comparisons using cross-sectional data on all re- working in the week preceding the interview. spondents show very similar shares. 3 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice Table 1. Labor status transitions between June-July 2020 and January-February 2021 (% of adults) Round 3 (Jan-Feb 2021) Unemployed Absent Working Inactive Total searching from work Round 1 Working 92.1 0.4 1.5 6.0 100 (Jun-Jul 2020) Unemployed searching 82.1 0.4 2.0 15.5 100 Absent from work 68.6 0.6 1.5 29.4 100 Inactive 17.6 0.6 0.3 81.5 100 Total 65.9 0.5 1.2 32.4 100 Note: Table refers to all adults interviewed in the three rounds. ‘Inactive’ describes respondents out of the labor force. ‘Unemployed’ are those who are actively searching for jobs. Temporarily ‘absent’ describes those who are not looking for jobs because they expect to go back to their original employment. Before COVID-19, about 74 and 62 percent of The share of respondents employed has respondents in Dhaka and Chittagong, respec- increased for both genders, though the share tively, were engaged in an income-generating of women employed in Dhaka still lags far activity in the month preceding their baseline below their pre-COVID-19 level. The share surveys.7 8 For both cities, the gains in em- of men working increased by 18 percentage ployment between rounds 1 and 3 came from points between June 2020 and February 2021 large reductions in unemployment and absen- (from 74 to 92 percent). In contrast, the share teeism. The percentage of respondents absent of women working rose 13 percentage points from their jobs declined from 6 to 1 percent in from 20 to 33 percent over the same period Dhaka, and from 10 to 1 percent in Chittagong. (figure 2). Between June 2020 and February Unemployment fell 8 percentage points in 2021, employment in Chittagong increased Chittagong and 12 percentage points in Dhaka, by 19 and 6 percentage points for males and reaching less than 1 percent in both cities. females, respectively. In Dhaka over the same period, employment increased by 17 and 18 Both slum and non-slum areas showed im- percentage points for males and females, provements in employment. The percentage respectively. Figure 3 compares employment of adults working increased from 52 to 68 across time for the same respondents across percent in slums and from 49 to 63 percent cities and by gender. in non-slum poor areas between June 2020 and February 2021. The share of unemployed Inactivity rates for women have also risen respondents decreased to less than 1 percent since June 2020. Improvements in employment in both areas, while the share out of the la- mask different transitions in and out of the la- bor force remained around the same levels, bor force by gender (appendix table 1). Between compared to the first round of the survey. Ap- June 2020 and February 2021, about 4 percent pendix figure 1 confirms that the employment of men who were unemployed, and 19 percent recovery has been seen across slum and non- of men absent from work exited the labor force, slum areas within both cities. while 65 percent of those inactive in June start- ed work. In contrast, 34 percent of unemployed women and 41 percent of women absent from work in June 2020 exited the labor force by Feb- 7 Baseline surveys took place between July and September ruary 2021. Only 10 percent of women inactive 2018 in Dhaka and between September and October 2019 in June 2020 subsequently entered the labor in Chittagong. force. Thus, overall inactivity rates increased for 8 Note that the reference periods in the baseline and first women from 62 to 67 percent over this peri- follow-up surveys are different (30 days versus 7 days, re- spectively); thus, an exact employment change cannot be od. For males, the main reported reasons for calculated. Since the employment rate with a longer refer- leaving the labor force were old age or sickness ence period would be higher, it is reasonable to conclude unrelated to COVID-19. Among females, the that employment levels in Chittagong were returning to main reason for becoming inactive was family pre-COVID-19 levels by the time of round 3. responsibilities (e.g., housewife, caretaker). 4 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice Figure 2. Employment status by gender (%) 100% 6 6 7 2 8 4 1 5 11 75% 62 63 67 50% 86 92 74 1 9 8 25% 9 29 33 20 0% R1 Jun-20 R2 Sep-20 R3 Jan-21 R1 Jun-20 R2 Sep-20 R3 Jan-21 Male Female Working Unemployed Absent Inactive Note: ‘Inactive’ describes respondents out of the labor force. ‘Unemployed’ are those who are actively searching for jobs. Temporarily ‘absent’ describes those who are not looking for jobs because they expect to go back to their original employment. Figure 3. Percentage of adults working, by city and gender 100 96 93 87 88 86 90 76 75 72 50 47 39 29 27 27 25 25 21 18 0 Pre COVID 2018 R1 Jun-20 R2 Sep-20 R3 Jan-21 Pre COVID 2018 R1 Jun-20 R2 Sep-20 R3 Jan-21 Pre COVID 2019 R1 Jun-20 R2 Sep-20 R3 Jan-21 Pre COVID 2019 R1 Jun-20 R2 Sep-20 R3 Jan-21 Dhaka males Dhaka females Chittagong males Chittagong females Note: Pre-COVID-19 baseline data refer to employment in the past 30 days. Round 1, 2, and 3 data refer to employment in the past 7 days. Baseline data for Chittagong was collected between September and October 2019. Baseline data for Dhaka was collected between July and September 2018. Among adults who lost jobs due to COVID-19 respondents were significantly less likely and were still not working by February 2021, to be employed by round 3 (appendix table 2 out of 3 were women. About 73 percent 3). Most occupations showed significant in- of respondents who had lost their jobs due creases in employment (conditional on place to COVID-19 in June 2020 were working by of residence, gender, and age).9 February 2021 (appendix table 2). However, around 24.5 percent of those who had lost About 4 in 10 of those who gained employment their employment by round 1 were out of the after being impacted by COVID-19 changed labor force by round 3, while the remaining their occupation. A comparison of occupations were either absent or unemployed. Impor- indicates that, for those working in both round tantly, 66 percent of those still not working 1 and round 3, 77 percent were working in the were women, and 82 percent of those who were out of the labor force were women. A 9 The occupation variable in this regression refers to the regression analysis indicates that impacted occupation of the adult respondent before job loss due to workers living in non-slum areas and female COVID-19. Figure 4. Share of workers reporting the same occupation in round 3 as in round 1, by round 1 labor market status and occupation 100% 75% 50% 25% 0% Drivers/ Garment Construction Retail/sales Porter/ Maid/ Professional/ Own account - transport worker worker worker day laborer servant skilled retail/trade All adults working in June 2020 Adults not working in June 2020 Note: Figure shows the percentage of workers who remained in the same occupation in February 2021 (survey round 3) as in June 2020 (round 1), across a range of occupations. Green bars indicate percentages for respondents who reported working in June 2020. Orange bars show results for those who reported not working in June 2020. Thus, the orange bars represent workers who initially lost their employment due to COVID-19, then began working again by February 2021. Figure 5. Median earnings across rounds, by employment category, in Takas a. Wage and salaried workers b. Self-employed revenues 15,000 10,000 9,700 12,000 9,000 10,000 8,000 7,000 500 500 429 400 Regular R1 R2 R3 Regular R1 R2 R3 Regular R1 R2 R3 (Jun-Jul) (Sep-Oct) (Jan-Feb) (Jun-Jul) (Sep-Oct) (Jan-Feb) (Jun-Jul) (Sep-Oct) (Jan-Feb) Monthly Daily Monthly Note: Figures based on cross-sectional data. same occupation. However, for those who lost and wage/salaried workers across rounds. The their employment due to COVID-19 and then figures indicate a considerable improvement started work again, only 59 percent returned to compared to June 2020. Median earnings have the same occupation. A conditional regression reached amounts closer to the amounts reported indicates that workers living in Dhaka, those in before COVID-19. For monthly wage and salaried slum areas, and women were more likely to re- workers, median earnings in June 2020 were 20 turn to the same occupation. Workers engaged percent below the earnings reported before the in garments, retail and sales, and day labor pandemic. By February 2021, median earnings before losing their jobs due to COVID-19 were for these groups were just 3 percent lower than significantly more likely to change occupation the pre-COVID-19 amounts. A similar recovery is (Figure 4).10 However, about 76 percent of those seen for daily workers’ wages. However, for the who recovered their employment remained in self-employed and business owners, gains have the same employment category (daily, wage, or been slower. These groups’ median revenues in self-employed worker). June 2020 were about half the usual amounts, and in February 2021 their revenues remained While revenues from self-employment have not 20 percent below the usual pre-crisis values.11 recovered fully, median earnings for salaried and wage workers seem to have bounced back 11 Comparisons based on values for usual earnings and reve- to levels reported before the pandemic. Figure nue need to be interpreted with care, as respondents’ recol- 5 shows median earnings for self-employed lection of usual amounts may differ from the actual amounts received in the week or month preceding the interview. In addition, the values in Takas are obtained from simple ques- 10 Due to small sample sizes, there is no clear pattern re- tions that can be asked in a phone survey. They may not align garding job-switchers’ new occupations. with figures from more detailed and face-to-face surveys. 6 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice AP P ENDIX 1. The monitoring survey built on baseline sur- The monitoring survey in Chittagong is a veys conducted before the COVID-19 crisis. follow-up of the CITY (Chittagong Low Income Survey The monitoring survey sample for Dhaka is a Area Inclusion and PoverTY) survey carried Details follow-up of the DIGNITY (Dhaka low Income out in Chittagong City Corporation following area GeNder, Inclusion, and poverTY) survey, the same sampling strategy as in the DIGNITY which was representative of low-income areas survey. Data was collected from 1,289 and slums of the Dhaka City Corporations and individuals across 805 households between an additional low-income site from the Great- September and October 2019. er Dhaka Statistical Metropolitan Area, follow- ing a two-stage stratification design. The pri- For the monitoring survey, a representative mary sampling units were selected during the sub-sample of 1,500 households out of a first stage using probability proportional to total 2,107 baseline households was targeted. size (PPS), stratified by the poverty headcount The recontact rate was 1,483 households ratio estimated using small-area techniques. (99.5 percent). In the first tracking survey, All the households in the selected enumera- 1,483 out of the 3,665 adults surveyed in tion areas were listed during the second stage, baseline were covered. The first tracking from which 20 households were selected for survey was conducted between June 10 interviewing based on demographic stratifi- and July 10, 2020. The second took place cation. The second level of stratification was between September 2 and October 11, 2020. defined as follows: (i) households with both The third tracking survey, conducted from working-age male and female members; (ii) January 13 to February 27, 2021, aimed to households with only a working-age female; reach all respondents interviewed in the (iii) households with only a working-age male. first tracking survey. The final recontact rate Households were randomly selected from was 88 percent. The main reason for non- each stratum with the predetermined ratio of response was the inability of interviewers 16:3:1.12 The DIGNITY survey, administered be- to reach respondents by phone. The number tween July and September 2018, collected in- of observations lost due to attrition was 178 formation from 2,376 individuals across 1,302 respondents. Analysis of selection indicates households. that missing respondents were significantly more likely to live in Chittagong. However, the numbers of observations missing across key categories (area and gender) are too small to 12 Kotikula, A., Hill, R., and Raza, W.A. 2019. What Works for Working Women? Understanding Female Labor Force Participation in affect inferences at that level. Urban Bangladesh. Washington, DC: World Bank. AP P ENDIX Appendix table 1. Labor status transitions between June-July 2020 and January-February TAB LES 2021 (% of adults), by gender Males Round 3 (Jan-Feb 2021) Unemployed Absent Working Inactive Total searching from work Round 1 Working 95 1 2 2 100 (Jun-Jul 2020) Unemployed searching 95 1 0 4 100 Absent from work 78 1 3 19 100 Inactive 65 4 3 29 100 Total 92 1 2 6 100 7 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice Females Round 3 (Jan-Feb 2021) Unemployed Absent Working Inactive Total searching from work Round 1 Working 76 0 0 24 100 (Jun-Jul 2020) Unemployed searching 62 0 4 34 100 Absent from work 59 0 0 41 100 Inactive 10 0 0 90 100 Total 33 0 0 67 100 Note: Table refers to all adults interviewed in both rounds 1 and 3. ‘Inactive’ describes respondents out of the labor force. ‘Unemployed’ are those who are actively searching for jobs. Temporarily ‘absent’ describes those who are not looking for jobs because they expect to go back to their original employment. Appendix table 2. Labor status transitions between June-July 2020 and January-February 2021: Adults who stopped actively working during round 1 due to COVID-19 (% of adults) Round 3 (Jan-Feb 2021) Unemployed Absent Working Inactive Total searching from work Round 1 Unemployed searching 84 0 2 14 100 (June-July 2020) Absent from work 71 1 0 28 100 Inactive 59 3 2 36 100 Total 73 1 1 25 100 Appendix figure 1. Percentage of adults working, by city and area 100 81 75 74 71 68 70 68 67 65 59 59 56 54 56 52 50 48 41 25 0 Pre COVID 2018 R1 Jun-20 R2 Sep-20 R3 Jan-21 Pre COVID 2018 R1 Jun-20 R2 Sep-20 R3 Jan-21 Pre COVID 2019 R1 Jun-20 R2 Sep-20 R3 Jan-21 Pre COVID 2019 R1 Jun-20 R2 Sep-20 R3 Jan-21 Dhaka Slum Dhaka Non-slum Chittagong Slum Chittagong Non-slum 8 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice Appendix table 3. Linear probability model for being employed in round 3 for those who stopped actively working due to COVID-19 in round 1 Base With occupation VARIABLES (1) (2)       Dummy Dhaka 0.124* 0.043   (0.068) (0.060) Dummy Slum 0.179*** 0.127*   (0.064) (0.067) Dummy Female -0.286*** -0.334***   (0.066) (0.086) Age 0.018*** -0.000   (0.002) (0.003) Occupation     Driver   0.742***     (0.146) Garment worker   0.510***     (0.149) Transport worker   0.779***     (0.136) Construction worker   0.769***     (0.173) Retail/sales worker   0.763***     (0.193) Porter/day laborer   0.897***     (0.111) Cleaning/housemaid   0.792***     (0.167) Wage other   0.802***     (0.197) Professional/skilled   0.912***     (0.171) Own account - retail/trade   0.797***     (0.140) Own account - other   0.826***     (0.156)       Observations 333 332 R-squared 0.713 0.789 Constant NO NO Note: Robust standard errors in parentheses. Variables used as controls were measured in round 1. Occupation refers to activity before losing employment. *** p<0.01, ** p<0.05, * p<0.1     9 | COVID-19 monitoring in Dhaka and Chittagong Poverty and Equity Global Practice