The Impact of COVID-19 on Labor Market Outcomes: Evidence from High Frequency Phone Survey from The Gambia1 First Quarterly Report (March 2021) Executive Summary The report describes the effect of the COVID-19 pandemic on labor market outcomes for the Gambia and finds evidence that initial effects of the pandemic on employment were large. Significant contractions in employment levels were observed at the beginning of the pandemic as shown in data from the first round of the High Frequency Survey (HFS) collected in August 2020. Recent data from the HFS however suggests improvements in labor market outcomes overtime. For instance, data from the second and third round of the survey show that employment rates have increased in September and December 2020 compared to levels observed in August. This is perhaps illustrative of the variation in the extent of the lockdown policies instituted to contain the pandemic. Such policies have since been relaxed in The Gambia. The effect of the COVID-19 pandemic on labor market outcomes appears to be heterogeneous across space, gender, and sector. Although, employment rates are typically lower in rural areas compared to urban areas the share of households who lost their jobs during the pandemic is higher in the urban areas and around the capital city. Across gender, although typically employment rates are higher among males compared to females, the effect of the pandemic on employment of females is larger than males albeit statistically insignificant. Poorer households and those employed in the services sectors such as tourism and personnel services and the agricultural are disproportionately affected and hence more vulnerable. The report uses data from the 2018 Gambia Labor Force Survey (GLFS) and the 3 waves of the ongoing High Frequency Survey (HFS). First, it highlights the state of the pandemic in the country and the government’s response to contain the pandemic. Second , the report utilizes data from the first three waves of the High Frequency Phone Survey, collected between August and December 2020 to describe changes in labor market outcomes during the pandemic. Finally, the report combines the two sources of data to tease out the impact of the COVID-19 pandemic on labor market activities- in particular, employment levels. 1. Introduction The world has been impacted by a new Coronavirus disease (COVID-19). The disease has since been declared a pandemic by the WHO. As of February 24th, 2021, 111,593,583 cases and 2,475,020 deaths due to virus have been recorded in 223 countries (WHO, 2020)2. Globally, the Americas confirmed the largest numbers of cases, followed by Europe, Asia, and Africa. The first 1 This report was prepared by Tijan L. Bah with guidance from Moritz Meyer and Sering Touray. Corresponding author: stouray@worldbank.org 2 https://www.who.int/emergencies/diseases/novel-coronavirus-2019 – accessed on February 24th. 2021. 1 confirmed COVID-19 case in Africa was registered back in February 2020. As of February 24th, 2021, the African continent registered about 3,959,489 confirmed cases and 107,5653. The Gambia registered its first case of COVID-19 on 17th March 20204 when a female travelling back from the UK tested positive for the virus. On March 23rd, the first case of fatality was registered5. In response to the outbreak, the Gambia government enacted several measures. In agreement with its neighbor, Senegal, the Gambia border was closed on March 23 rd 20206. Furthermore, on March 28th, 2020, the government declared a State of Public Emergency, ordering the closure of schools and universities, places of worship and non-essential businesses, restricting public gatherings to not more than 10 people, and limiting passengers on public transportation7. As of February, 24th2021, The Gambia registered 4,640 cases, 147 total deaths, 4,089 recoveries and 423 active COVID-19 cases8. The daily number of new cases increased cumulatively overtime, reaching a peak of 248 confirmed cases on August 23rd, 2020, while the number of deaths peaked at 14 deaths on August 19th, 2020 -see panel a) of Figure 1. Over time, the number of COVID-19 tests conducted by the Ministry of Health and its stakeholders increased, totaling 45,919 tests as of February 24th. 2021. Recently, the number of COVID-19 cases have started increasing (see panel b) of Figure 1) signally a second wave of infections amidst a lax in compliance with WHO guidelines and an eminent need for another lockdown. Figure 1: COVID-19 cases in The Gambia a) Cumulative COVID-19 cases b) New confirmed COVID-19 cases 5000 4500 4000 300 Confirmed cases 3500 250 3000 Confirmed cases 2500 200 2000 150 1500 1000 100 500 50 0 3/17/2020 4/17/2020 5/17/2020 6/17/2020 7/17/2020 8/17/2020 9/17/2020 10/17/2020 11/17/2020 12/17/2020 1/17/2021 2/17/2021 0 4/17/2… 7/17/2… 3/17/2… 5/17/2… 6/17/2… 8/17/2… 9/17/2… 10/17/… 11/17/… 12/17/… 1/17/2… 2/17/2… Source: Calculations based data from Ministry of Health Situational Report. 3 COVID19_Cases (arcgis.com) – accessed on February 24th 2021. 4 https://www.chronicle.gm/gambia-confirms-first-case-of-coronavirus/ 5 https://standard.gm/gambia-registers-first-COVID-19-death/ 6 https://www.chronicle.gm/gambia-finally-closes-land-borders-and-fights-as-COVID-19-fear-intensifies/ 7 https://www.chronicle.gm/gambia-declares-state-of-emergency-to-curb-coronavirus/ 8 http://www.moh.gov.gm/ – accessed on February 24th 2021. 2 The aim of this report is to analyze the impact of the COVID-19 pandemic on labor market outcomes in the Gambia. For most households in The Gambia, returns from labor market activities form large shares of household income. According to the IHS 2015 survey, about 72 percent of yearly income comes from labor market (salaries and enterprise activities). As such, disruptions in labor markets due to the COVID-19 pandemic are likely to significantly affect household well-being and poverty rates in The Gambia. Using data from a high frequency phone survey administered by the Gambia Bureau of Statistics in collaboration with the World Bank, we examine the extent to which the COVID-19 pandemic has affected labor market activities of households in The Gambia. The surveys which are representative at national, rural, urban (Banjul and Kanifing Municipality) and other urban areas levels cover about 1,500 households sampled from the 2018 Gambia Labor Force Survey across the country. The rest of the report is organized as follows. Section 2 describes the background information on labor market outcomes before COVIDI-19 by drawing on data from the 2018 Gambia Labor Force Survey (GLFS). Section 3 uses the High Frequency Survey (HFS) data to describe the labor market outcomes during the COVID-19 pandemic. Section 4 combines both data sources to estimate the impact of COVID-19 on employment. Section 5 concludes with recommendations. 2. Labor market outcomes during COVID-19 This section covers labor market outcomes during COVID-19. We use data from the 3 waves of an ongoing High Frequency Survey (HFS). In each wave, the same set of about 1,500 households are contacted by telephone. The households are drawn from the 2018 Gambia Labor Force Survey (GLFS) and are representative at national, rural, capital city region and other urban levels. The number of households covered in August-September, September- October, and November – December 2020 was 1,576, 1,508, and 1,486, respectively. In each household, the head of the household serves as the main respondent. Box 1: Design and Implementation of the High Frequency Survey (HFS) The HFS sample is drawn from the Gambia Labor Force Survey (GLFS) collected in 2018. The GLFS data set has 5,987 households of which 5,531 (92 percent) have valid phone numbers. All households with valid phone numbers formed the population of households to be sampled. The main variable of interests is the proportion of people employed in each household. The variable is identified because of its reliability and the potential impacts of the pandemic on employment status. The sample is stratified at three levels: Banjul and Kanifing, other urban area and the rural area. The Banjul and Kanifing stratum have 1,197 households with valid numbers, while other urban and rural areas strata have 1,448 and 2,886, respectively. Each household in each stratum has equal chance of being sampled. The minimum targeted sample size is 1,534 households. The final sample is distributed across the three strata, with 414 households from Banjul and Kanifing, 544 3 households from other urban area and 566 households from the rural area. In each household, the most knowledgeable member (mostly the household head) is surveyed. 9 Households selected for the phone survey are positively selected in terms of wealth (proxied by asset index). However, this selection is only significant for the rural and other urban areas. See Figure 12 and Table 2 in the Appendix. Data from recent rounds of the HFS indicate an improvement in employment rates from levels observed at the beginning of the pandemic. Overall, across all three rounds of data, about 66 percent of respondents reported working in the last 7 days before the interview10. A significant variation is reported across time and space. In the months of August and September, about 62 percent were working, rising to 67 percent in October and further 71 percent in November and December 2020- see panel a) of Figure 2. These suggest an increase in the percentage of people working 11. Furthermore, 59 percent of people reported working in the Banjul and Kanifing regions compared to 68 percent in other urban areas and 71 percent in the rural areas. Although other factors cannot be ruled out, the COVID-19 pandemic continues to be among the major factors affecting employment in recent months. For instance, as shown in panel b) of Figure 2, the main reasons for not working during the pandemic were business or government closed (31 percent), temporarily absent (23 percent), seasonality of work (9 percent), Illness (7 percent) and reduction of staff due to less business (5 percent). Figure 2: Employment during the Pandemic a) Employment during the pandemic. b) Reason for not working 72 71 Other 19 70 Need to care for Ill relative 1 68 Percent employed 67 66 66 Ill 7 64 Retired 5 62 62 Temporarily absent 23 60 Reduction of staff 5 58 Business/Gov't closed 31 56 Aug-Sept Sept-Oct Nov-Dec All waves Seasonal worker 9 Survey wave 0 10 20 Percent 30 40 Source: own calculations HFS data, 2020. 9 Households who were unreachable were continuously phone if the survey was ongoing. 10 This was elicited from responses to the question: “Last 7 days, did you do any work for pay, do any kind of business, farming or other activity to generate income, even if only for one hour?” 11 As already mentioned, and documented in World Bank (2020), part of the increase might be due to seasonality. In Section 4, we control for this seasonality. 4 Although agriculture and trade are among the main sectors as in the LFS data, changes in sector of employment were observed across waves attributable to either seasonality and/or the effect of lock down policies. Panel a) of Figure 3 below shows that the main activity done by respondents include personal services (25 percent), buying and selling (24 percent), agriculture (24), construction (9 percent) and transportation (7 percent). Across three waves, the shares of employees in agriculture and personal services increased while those employed in the construction, transportation, buying and selling declined- see panel b) of Figure 3 below. While seasonality is likely to influence employment in sectors such as construction, buying and selling and transportation especially during the rainy season, lockdown measures introduced to limit the spread of the virus are also likely to drive the results. Across space, agriculture employs more people in the rural areas (46 percent) compared to 26 percent in other urban areas and 5 percent in the Kanifing and Banjul areas. Conversely, the buying and selling activities are concentrated in Kanifing and Banjul (42 percent) and other urban areas (22 percent) compared to 11 percent in the rural areas. Figure 3: Sector of Employment a) Sector of Employment overall. b) Sector of Employment across rounds of survey 100 Personal services 25 90 20 23 29 Public Administration 4 80 4 2 5 70 9 5 4 Professional activities 4 7 4 60 5 Transportation 7 50 33 22 22 40 Buying and Selling 24 9 9 30 1 0 9 3 3 20 1 Construction 9 5 10 26 24 17 Electricity water and Gas 1 0 Aug-Sept Sept-Oct Nov-Dec Mining & Manufacturing 4 Agriculture Mining & Manufacturing Electricity water and Gas Construction Agriculture 24 Buying and Selling Transportation 0 5 10 15 20 25 30 Professional activities Public Administration Percent of employed Personal services Source: own calculations HFS data, 2020. In terms of economic activity, majority of workers worked in their own businesses, employed by someone else and family farm, or raising livestock. Over time, the share of people working in own business declined from 60 percent in August to 51 percent in September and 48 percent in December- see panel a) of Figure 4 below. On the other hand, the share of workers employed 5 by someone else increased from 26 percent in August to 32 percent in December. Similarly, activities in the agricultural sector increased from 8 percent in August to 15 percent in September and 17 percent in December. The increase reflects the seasonality of the agriculture- August is typically the time between weeding and harvesting. Moreover, across the different strata, the share of workers in own business is larger in the Kanifing and Banjul (60 percent) compared to the other urban areas (50 percent) and rural areas (47 percent) - see panel b) of Figure 4 below. Conversely and consistently, the agriculture activity is concentrated in the rural areas and other urban areas. Figure 4: Economic Activity a) Economic Activity across waves. b) Economic Activity across Space Nov-Dec 48 2 18 32 0 Rural 47 1 30 22 0 Sept-Oct 51 2 15 32 1 Urban 50 2 12 35 1 Aug-Sept 60 5 8 26 1 0 20 40 60 80 100 Kanifing and Banjul 60 43 33 0 In your own business 0 20 40 60 80 100 In a business operated by a household or family member In your own business In a family farm or raising family livestock In a business operated by a household or family member In a family farm or raising family livestock As an employee for someone else As an employee for someone else As an apprentice, trainee, intern As an apprentice, trainee, intern Source: own calculations HFS data, 2020. Compared to pre-COVID levels, employment levels during the pandemic were lower by about 20 percent. Using retrospective responses, we can compare the employment levels pre-COVID- 19 and during the pandemic. Households who were reported to be out of work were further asked whether they were working before COVID-19. Using this information and if households who reported to be working at the time of the survey were also working before the pandemic, we can construct the share of households who have lost jobs during the pandemic12. On average, about 86 percent of the respondents were working before the COVID-19 pandemic compared to 66 percent during the pandemic, suggesting a 20 percent reduction in employment rates- see panel a) of Figure 5 below. Since the share of people working vary across time, a comparison reveals that in August, about 62 percent were working compared to 88 percent before the 12 This is plausible as 94 percent of those who are currently working reported that they were working in the same job before the pandemic. 6 pandemic. Moreover, in September, 67 percent reported working versus 86 percent pre-COVID- 19. Similarly, for the months of November and December, about 71 percent reported working compared to 86 percent before the COVID-19 pandemic. The above statistics highlight that, while the percentage of people working are still lower than pre-pandemic, a recovery is on the way. Across space, the comparison highlights differences in employment levels pre-COVID-19 and during the pandemic. In the Banjul and Kanifing area, about 59 percent are currently working compared to 84 percent before the start of the pandemic. Moreover, for the rest of the urban areas, 68 percent reported working compared to 85 percent before the pandemic. Finally, for those in the rural areas, 90 percent were working before the pandemic compared to 71 percent during the pandemic - see panel b) of Figure 5 below. More stringent lockdown policies particularly in and around the capital city is potentially the main driver of the steeper decline in employment rates compared to other rates. In rural areas on the other hand, the seasonality of agriculture is likely the main factor influencing employment rates. Figure 5: Percentage employed before and during COVID-19 pandemic a) Across Waves b) Across Space 100 100 88 90 86 85 86 90 84 85 86 90 80 80 71 71 68 66 70 67 66 70 62 59 60 Percent 60 Percent 50 50 40 40 30 30 20 20 10 10 0 0 Banjul and Urban Rural All waves Aug-Sept Sept-Oct Nov-Dec All waves Kanifing Before COVID-19 During COVID-19 Before COVID-19 During COVID-19 Source: own calculations HFS data, 2020. Notes: Pre COVID-19 (February/March 2020) is based on retrospective question on whether the respondent was employed before the pandemic. Using data from the 2018 LFS as a pre-COVID benchmark also shows lower employment levels during the pandemic. Assuming that all household heads are the same across the different surveys and the household head was the respondent in the HFS. According to the LFS, 2018 (pre- COVID-19 pandemic), 73 percent were employed at the time of the survey. Data from the first wave of the HFS collected between August and September 2020 indicates a lower level of employment of 64 percent. In subsequent waves, the employment levels increased to 70 percent in October and 75 percent in the December wave – see panel a) of Figure 6. The statistics suggests 7 that while employment rates dropped significantly at the start of the pandemic, it subsequently recovered to a higher level during the last wave. It is worth highlighting that there are seasonal effects that is not captured by this raw comparison between pre and during COVID-1913. A similar comparison across space show that employment levels pre COVID-19 and during the pandemic are heterogeneous. Current employment levels are lower than pre-COVID employment levels in both the Kanifing and Banjul (63 versus 75 percent) and other urban areas (65 versus 80 percent). However, the average percentage employed in the rural area pre- pandemic is higher than during the pandemic (73 versus 65 percent). The comparison of employment rates across time and space reveals another striking difference. In panel b) of Figure 6 below, it can be observed that the subsequent employment rates in the second and third waves, are higher for other urban and rural areas except for the capital city region (70 versus 75 percent). This breakdown further highlights the importance both timing of the survey and regional differences. Figure 6: Percentage employed before and during COVID-19 pandemic a) Across Waves b) Across Space 78 100 100 100 76 90 76 80 78 80 75 74 73 70 69 71 70 63 65 60 72 57 60 Percent employed 70 70 50 40 68 30 66 64 20 64 10 62 0 Caital city region Urban Rural 60 Before COVID-19 (LFS, 2018) Aug-Sept 58 Before COVID-19Aug-Sept Sept-Oct Nov-Dec Sept-Oct Nov-Dec Source: own calculations HFS data, 2020. Notes: Pre COVID-19 (February/March 2020) is based on the reported employment status in the GLFS, 2018. In terms of gender comparison, females experienced a slightly steeper decline in employment levels from pre-COVID levels than males. Before the pandemic, 78 percent of males were employed compared to 52 percent of females. Data from the first wave of the HFS indicate that by August 2020, 58 percent of males and 41 percent of females were employed- representing a 10 pp and 11 pp decline from pre-COVID levels, respectively. Recent data from subsequent waves 13 The LFS 2018 took place between July and September, when agricultural season is at peak and construction is low compared to the months of August to December (during the HFS). See World Bank (2020) 8 of the HFS also indicate that while employment levels of males have recovered to pre-COVID levels, employment levels of females remain lower than pre-COVID levels. The employment rates of males (females) increased to 74 (45) percent in September and 79 (55) percent in December 2020. Albeit descriptive, these results are consistent with findings from similar studies. The decline in labor market participation of females relative to males during the pandemic has often been attributed to additional responsibilities of females to care for younger children who are out of school due to school closures; as well as the sick and elderly members of their households resulting in the substitution of labor supply with homecare. See Kalenkokoski and Pabilonia (2020) and Cugagna and Romero (2021). The comparison also reveals that on average, 19 percent of household heads lost their jobs (those who were working in 2018 but were not working during the survey) with some heterogeneity across sector of employment, gender, welfare distribution and space14. The share of those who lost their jobs declined from 23 percent in August to 19 percent in September and 14 percent in December 2020. Panel a) of Figure 7 below shows that the percentage of those who lost their jobs vary across gender, space, poverty, and education levels. Females are slightly more likely to lose their jobs than males (20 versus 18 percent). Furthermore, households belonging to the top 20 richest percent are more likely to lose their jobs than the poorest 20 percent. Moreover, those residing in other urban areas (26 percent) are more likely to lose their jobs compared to those residing in Banjul and Kanifing (19 percent) and rural areas (14 percent). One source of vulnerability for household heads who are more likely to lose their jobs appears to stem from their sector of employment. For instance, across gender, females are more likely to be employed in sectors adversely affected by the pandemic such as tourism. Similar differences are also observed in urban-rural; and rich-poor dichotomies. Finally, about 43 percent of these household heads have medium level of education (higher secondary) while 34 percent has low level of education and the remaining 23 percent acquired high level of education. Households employed in the tourism sector and the personal services appear to be most vulnerable to job losses. Among those who lost their jobs, about 33 percent were employed in the tourism sector with a focus on informal service in urban areas15- – see panel b) of Figure 7. This is followed by those working in the personal services (health, education, culture, sport, and domestic work) representing 24 percent of the job loss. The large share of households employed in these sectors (nearly half of the workforce) illustrates the high vulnerability of Gambian households to the negative effects of the COVID-19 pandemic. Other sectors such as agriculture and construction activities also experienced high job losses- 15 percent and 9 percent of job losses, respectively. 14 A comparison between the 2018 LFS and the HFS shows that the distribution of sector of employment is similar across both surveys suggesting that job losses observed from the HFS data is reflective of the typical labor market in The Gambia. 15 Buying and selling, repair of goods, hotels, and restaurants. 9 Figure 7: Job Losses during the pandemic a) Across Waves b) Across Space 50 35 33 43 45 30 40 Percent employed 24 34 25 35 in percent 30 26 20 24 15 25 23 21 15 18 19 20 9 14 10 7 15 10 5 4 5 3 10 1 5 0 0 Source: own calculations HFS data, 2020. Evidence of transition across sectors is also observed among households during the pandemic – 42 percent of households who transitioned into employment at some point during the pandemic gained employment in a new sector. It can be observed from – see panel a) of Figure 8 below that overall, of the 21 percent who lost their jobs, 46 percent of them transit into employment between the first and the second wave. Of those who transit into employment, about 58 percent went back to the same sector they were working pre COVID-19 pandemic and the remaining 42 percent moved to new sectors. Indeed, there are differences across sectors- see panel b) of Figure 8. On the one hand, the personnel services (32 percent), buying and selling (25 percent), and agriculture (14 percent) attracted the largest numbers of new workers. On the other hand, electricity, and water services (100 percent), public administration (96 percent), construction (94 percent), personnel services (82 percent) professional activities (76 percent) retained most of their workers. 10 Figure 8: Job and Sector transition a) Job loss and sector transition b) Job Transition Across Sectors 50 100 9 46 90 14 11 80 39 45 42 70 60 4 45 76 82 50 100 94 73 68 96 40 40 30 12 57 35 20 15 10 4 18 11 24 6 1 11 6 10 4 7 4 0 30 in percent 25 21 20 15 10 Agriculture Mining & Manufacturing Electricity water and Gas Construction 5 Buying and Selling Transportation 0 Professional activities Public Administration Job loss Transit Transit to a different sector Personal services Source: own calculations HFS data, 2020. 3. COVID-19 and Employment The previous discussions show that labor market outcomes during the pandemic have declined when compared to the pre COVID-19 era. However, this basic comparison does not consider other confounding factors such as seasonality for example. In this section, we take this into account to understand the causal impact of the COVID-19 pandemic on our outcome of interest. Following Kalenkokoski and Pabilonia (2020), we estimate the following difference- in-difference- in-difference model: Yit = +  Wi +  COVIDt +  Wi * COVIDt +  Xit +  Seasont +  Yeari + it (1) Where Yit is outcome of interest (employment for example), Wi is vector of fixed controls of interest such as gender, strata, education levels, COVIDt is a dummy equal 1 for sample belonging to the first two waves (assuming COVID-19 restrictions were in place), and zero otherwise, Seasont is a dummy equal 1 for sample belonging to either the LFS sample or first 2 waves. This control will capture the seasonality of labor market outcomes for months July, August, and September. Finally, the Year is dummy equal 1 if year is 2020 and zero otherwise. The coefficients of interest are , which capture the overall impact of the COVID-19 pandemic, and the vector  the triple difference estimates capturing the impact of the pandemic on gender of respondents and the strata residence (Kanifing and Banjul, other urban and rural areas). The 11 models are estimated with population weights and standard errors are clustered at the household-year pair level. Table 1 below present results from the equation 1 above. The dependent variable is a dummy variable equal 1 for respondents who are employed and 0 otherwise. Estimation is restricted to household heads who are in the working age group (between 14 and 64 years old). Table 1: COVID-19 and Employment (1) (2) (3) Employed Employed Employed Female -0.292*** -0.286*** -0.306*** (0.033) (0.045) (0.045) Other urban 0.004 0.004 0.036 (0.034) (0.039) (0.038) Rural -0.075** -0.182*** -0.099* (0.034) (0.049) (0.052) Year 2020 0.797*** 0.798*** 0.443*** (0.033) (0.034) (0.069) COVID-19 -0.891*** -0.934*** -0.519*** (0.043) (0.056) (0.092) Season 0.799*** 0.826*** 0.472*** (0.027) (0.031) (0.067) COVID-19 X Female -0.010 0.019 (0.064) (0.064) COVID-19 X Urban -0.012 -0.040 (0.064) (0.065) COVID-19 X Rural 0.212*** 0.140* (0.067) (0.071) Asset Index 0.557*** (0.094) COVID-19 X Poorest 20 percent 0.181*** (0.057) COVID-19 X Richest 20 percent -0.124** (0.052) Sample Size 3197 3197 3197 Dep. Var Mean 0.67 0.67 0.67 Dep. Var Mean omitted category 0.724 0.724 0.724 Notes: The dependent variable is dummy equal 1 for employed individuals and zero otherwise. The sample is restricted to household heads who are in the working age population group. Regressions are weighted and standard errors (in parenthesis) are clustered at household-year level. The omitted group is the Banjul and Kanifing area. * p < 0.10, ** p < 0.05, *** p < 0.01 Source: The Gambia Labor Force Survey, 2018 and High Frequency Survey, 2020 The results show overall negative impact of COVID-19 on employment rates. The employment rate of females is significantly lower than their male counterparts. Additionally, employment 12 rates are 7.5 percent lower in rural areas than the Kanifing and Banjul regions and other urban areas. In column 2 of Table 1, the COVID-19 restriction variable is interacted with gender and strata is shown in the second column. The results suggests that, the impact of COVID-19 on employment is not significantly different across males and females. However, the impact of the pandemic restrictions affected those residing in the capital and urban areas more than those in the rural areas. This is consistent with the fact that the restrictions had more impact on other sectors than the agricultural sector, which is the main employing sector in the rural areas. Proxied by an asset index, considering differences in wealth across households also shows interesting variations in the effect of the COVID-19 pandemic on employment levels- with richer households being less likely to be employed during the pandemic. In column 3 of Table 2, we introduced the household asset index as additional control variable. To understand the heterogeneous impact of COVID-19 across households based on wealth levels, we interacted the poorest 20 percent and richest 20 percent households and the COVID-19 restriction variable. The results suggest that on average, richer households are more likely to be employed. However, the COVID-19 pandemic restrictions affected the richest households more than the poorest households. Households in the top 20 percent richest households are 12 percent less likely to be employed due to the pandemic. Conversely, the 20 percent poorest households are 18 percent more likely to be employed during the pandemic. This is consistent with the findings that the most rural households are less affected in terms of labor market outcomes. 16 4. Conclusion In this report, we provide descriptive and causal evidence of the impact of the COVID-19 restrictions on labor market outcomes in the Gambia. We utilized two main sources of data sets: the Gambia Labor Force Survey, 2018 and the High Frequency Phone Survey, 2020. Descriptively, the evidence suggests that employment rates were lower compared to the pre COVID-19 pandemic. However, there are heterogeneous differences across space and time. Consistently, employment rates were lower in rural areas compared to urban areas and are higher for males compared to females. Over time, employment rates increase from September 2020 to subsequent months. Furthermore, this is consistent with the findings that during the COVID-19 pandemic restrictions, some households lost their jobs with most of these households being concentrated in the urban and the capital city regions. The most vulnerable sectors include the tourism sector and personnel services and the agricultural sectors. The causal evidence suggests that the COVID-19 pandemic has disproportionately affected employment levels. Overall, during the restrictions, households were less likely to be employed especially those residing in the Banjul and Kanifing areas compared to those in the rural areas. 16 Evidence suggests that rural households complained about the impact of the implemented COVID-19 policies on their day-to-day activities (Bah et. al, 2021). 13 Similarly, during the COVID-19 restrictions, richer households were less likely to be employed than poorer households. References Bah, Tijan L; Batista Catia; Gubert Flore, and McKenzie, David. 2021. “How has COVID-19 Affected Intention to Migrate via the Backway to Europe and to a Neighboring African Country? Survey Evidence and a Salience Experiment in The Gambia”. Forthcoming. Cucagna, Maria Emilia; Romero Haaker, Francisco Javier. 2021. “The Gendered Impacts of COVID- 19 on the Labor Market: Evidence from High-Frequency Phone Surveys in Latin America and the Caribbean (English)”. Gender innovation lab policy brief Washington, D.C.: World Bank Group. Kalenkoski, Marie Charlene and Pabilonia, Sabrina Wulff. 2020. “Initial Impact of the COVID -19 Pandemic on the Employment and Hours of Self Employed Coupled and Single Workers by Gender and Parental Status”. IZA Discussion Paper No. 13443. World Bank. 2020. “The Gambia Economic Update: Preserving the Gains.” Washington, DC: World 14 Appendix A1. Background on Labor Market Outcomes This section provides a pre-COVID background of the labor market outcomes of the country. It utilizes data from the 2018 Gambia Labor Force Survey (GLFS), which is a nationally representative survey of 5,086 households drawn from across the country. The survey took place between July and September 2018. The analyses focus on the main outcomes of interests: unemployment rates, employment status, sector of employment, education, and wages. Additionally, it highlights gender and regional disparities in labor market activities. Although the working-age group represents more than half of the population, slightly more than a third are employed- mostly self-employed. According the 2018 GLFS, the Gambia has a population of 2.3 million people with about 54 percent within the working-age group (above 14 but less than 64 years). Among the working population, only 34 percent are employed. More than half (53 percent) of those working are self-employed (working on their own account without employees), about 45 percent are employees and the remaining 1 percent are employers. There is heterogeneity in labor market activities across gender and geographic areas. On one hand, among the employed, 64 percent are males while the remaining 36 percent are females- see panel a) of Figure 9. The employed population are mostly composed of those who are residing in urban areas (66 percent) compared to 34 percent in rural areas- see panel b) of Figure 9. Overall, 19 percent of labor force is estimated to be unemployed- with higher unemployment rates among females and rural dwellers. The unemployment rate among females is 20 percent compared to 18 percent for males- see panel a) of Figure 9. Across space, rural areas record 30 percent unemployment rate compared to 10 percent in the urban areas- see panel b) of Figure 9. There is further variation in unemployment rates at the level of LGAs. Basse LGA records the highest unemployment rate of 57 percent while the Kanifing LGA recorded the lowest rate of 6 percent. In terms of labor force participation status, about 47 percent of the working age population were inactive. This inactive rate varies across gender and region with females having been less active with about 57 percent versus 36 percent for males. Similarly, the inactive rate is lower in rural areas (42 percent) compared to the urban areas (50 percent). 15 Figure 9: Labor Force Participation a) Labor Force Participation across Rural/Urban. b) Labor Force Participation across Gender 60% 60% 57% Percent of working age population Percent of working age population 51% 50% 50% 46% 46% 42% 39% 40% 40% 35% 36% 30% 30% 28% 30% 23% 20% 20% 18% 19% 20% 10% 10% 10% 0% 0% Employed Unemployed Inactive Employed Unemployed Inactive Urban Rural Male Female Overall Source: own calculations GLFS, 2018. Although sectors of employment vary across regions, agriculture and trade employ the largest share of workers. The top five sectors of employment include agriculture, forestry, and fishing (13 percent), wholesale and retail trade (11 percent), construction (8 percent), and education (7 percent). However, in rural areas, agriculture is the most common economic activity (employing about 14 percent of workers) whereas in urban areas it employs 7 percent of workers. Similarly, across gender, females are more likely to be employed in agriculture whereas males dominate the construction sector. There is a strong seasonal pattern of employment across space (rural/urban) and sectors of employment (World Bank, 2020). More than half of the working-age population (60 percent) attended some form of educational institution (conventional and/or Madrassah). Of these, 23 percent completed primary education, 27 percent completed lower secondary, 35 percent completed upper secondary, and only 4 percent attained higher education. Furthermore, females are more likely to have only completed lower education levels than males - see panel a) of Figure 10. About 33 percent of females attained up to upper secondary education compared to 38 percent for males. Similarly, 5 percent of males attained higher levels of education versus 3 percent for females. How different are educational attainments across areas of residence? In rural areas, about 49 percent of the population never attended conventional or Madrassah education compared to 34 percent in the urban areas- see panel b) of Figure 10. These differences in school attendance are further reflected in the levels of educational attainments. Those in the rural areas are more likely to have lower educational attainment than those in urban areas- for instance, the percentage of the working population with only primary education is 42 percent in rural areas compared to 22 percent in urban areas. 16 Figure 10: Educational Attainment a) Educational Attainment across Gender b) Educational Attainment across Rural/Urban. Don’t know Don’t know HIGHER HIGHER DIPLOMA DIPLOMA VOCATIONAL CERTIFICATE VOCATIONAL CERTIFICATE UPPER SECONDARY UPPER SECONDARY LOWER SECONDARY LOWER SECONDARY PRIMARY PRIMARY ECE ECE 0% 10% 20% 30% 40% 0% 10% 20% 30% 40% 50% Percent Percent National Female Male National Rural Urban Source: own calculations GLFS, 2018. Majority of workers- particularly females fall in the lower end of the earning distribution. Data on earnings collected in the 2018 LFS17 albeit in intervals provides some insights into the nature of the earnings distribution. The distribution suggests that 17 percent of workers earn less than 2000 GMD per month with a higher concentration in urban areas - see panel a) of Figure 11. Most of the workers (40 percent) earn between 2000 and 3500 GMD, followed those who earn between 3501 and 5000 GMD (23 percent). Finally, about 9 percent of workers earn more than 7500 GMD. A gender comparison shows majority of the females fall in the lower end of the earning distribution than their male counterparts - see panel b) of Figure 11. That is about 20 percent of the females earn less than 2000 GMD compared to 15 percent of the males. On the other hand, 10 percent of males earn more than 7500 GMD compared to 7 percent of the females. 17 Responses were obtained from 87 percent or workers who reported to earn income every month. 17 Figure 11: Earnings a) Earnings across Rural/Urban. b) Earnings across Gender 45% 45% 42% 39% 40% 39% 40% 40% 40% 35% 35% 30% 30% 25% 23% Percent Percent 23% 23% 20% 25% 25% 19% 20% 17% 17% 20% 15% 17% 13% 11% 11% 10% 9% 15% 11% 11% 11% 10% 9% 15% 7% 10% 7% 10% 5% 4% 5% 2% 5% 0% 0% Monthly salary (in GMD) Monthly salary (in GMD) Urban Rural National Male Female National Source: own calculations GLFS, 2018. A2. Comparison of HFS, 2021 and GLFS, 2018 Figure 12: Asset index comparison by samples 18 Source: Calculations based on GHFS, 2020 and the GLFS, 2018. Notes: The blue line represents the Non-HFS sample and the gray line represents the HFS sample. The p-values for test of differences are 0.9794, 0.0001, 0.0169, 0.0001, respectively. All graphs are weighted. Table 2: Summary Statistics: Statistical Differences between HFS and GLFS (1) (2) (3) HFS Sample Non-HFS Sample T-test (1-2) Mean Mean p-value Panel A: Banjul and Kanifing strata (household head/household characteristics) Age 45.363 46.298 0.203 Female 0.299 0.246 0.182 Household size 7.054 6.608 0.412 Asset index 0.627 0.622 0.971 Working-age population 0.893 0.864 0.184 Employment status 0.736 0.722 0.661 HH has external migrants 0.196 0.158 0.131 Has attended school 0.992 0.998 0.649 Migrated internally 0.553 0.505 0.276 Unemployed 0.035 0.037 0.535 Observations 385 906 1291 Panel B: Other urban areas strata (household head/household characteristics) Age 47.302 47.120 0.277 Female 0.157 0.202 0.098 Household size 8.659 8.066 0.369 Asset index 0.571 0.548 0.000*** Working-age population 0.899 0.883 0.398 Employment status 0.764 0.696 0.001** HH has external migrants 0.139 0.161 0.309 Has attended school 0.999 0.993 0.721 Migrated internally 0.433 0.422 0.017* Unemployed 0.072 0.105 0.011* Observations 542 1108 1650 Panel C: Rural areas strata (household head/household characteristics) Age 49.037 49.947 0.985 Household size 10.459 10.520 0.148 Asset index 0.485 0.466 0.007** Working-age population 0.865 0.826 0.068 Employment status 0.533 0.543 0.001*** HH has external migrants 0.139 0.163 0.490 Has attended school 1.000 0.996 0.045* Migrated internally 0.250 0.229 0.000*** Unemployed 0.259 0.229 0.195 Observations 493 2602 3095 Notes: HFS sample represents households or individuals surveyed in the high frequency survey while LFS sample are those who are not part of this sample but surveyed in LFS 2018. * p < 0.1, ** p < 0.05, *** p < 0.01. Source: The Gambia Labor Force Survey, 2018 19