Do Factory Jobs Improve Welfare? Experimental Evidence from Ethiopia

Abstract

This study explores the impact of a light-touch job-facilitation intervention that supported young female job seekers during the application process for factory work in a newly constructed industrial park in Ethiopia. Using data from a panel of 687 job seekers and randomized access to the support intervention, the study finds that treated applicants are more likely to be employed and have higher earnings and savings eight months after baseline, although these impacts are short-lived. Four years later, the effects on employment and income largely dissipated. The results suggest that young women face significant barriers to engaging in factory work in the short run that a simple job-facilitation intervention can help overcome. In the long term, however, these jobs do not offer a better alternative than other income-generating opportunities.

1. Introduction

A large body of empirical evidence suggests that stable wage work in formal firms is a preferable path to prosperity for most labor market entrants and that job seekers queue for wage work in developing countries (Blattman and Dercon 2018; Contreras, Roberto, and Esteban 2017; Meghir, Narita, and Robin 2015; Weiss 1980). These jobs are often considered better than most poor people's alternatives because they offer a higher degree of economic security (Roy 2004). Formal firms, especially in the industrial sector, often pay higher salaries than informal employers because of efficiency wages, exporter wage premia, firm competition for scarce skills, and union bargaining (Bernard and Jensen, 1999 ); El Badaoui, Strobel, and Walsh 2008; Verhoogen 2008). More recent evidence indicates that industrial jobs can lead to welfare improvements, especially for female workers (e.g., Getahun and Villanger 2018; Suzuki, Mano, and Abebe 2018). There is also evidence suggesting that industrial work can have knock-on effects on women's empowerment, demand for own and children's education, household bargaining power, and perceived quality of life (Heath and Mobarak 2015; Jensen 2012; Kabeer 2002).

Against this background, Ethiopian policy makers have prioritized large-scale industrial development centered on attracting foreign capital investments in export-oriented and labor-intensive industries that are concentrated in industrial parks to drive job creation (NPC 2015).¹ Jobs in industrial parks offer wage-employment opportunities especially for young women with relatively low educational attainment, limited job-search experience, and often no history of formal employment (Meyer et al. 2021). To study the welfare impacts of these jobs, this study randomized access to job-application support in partnership with three firms in one of the newly built industrial parks, Bole Lemi.² In doing so, it tests the hypothesis that, by supporting and facilitating the job application process for young women who seek entry-level production line positions, the intervention reduces job search barriers and improves labor market outcomes.

More specifically, the intervention is designed to overcome spatial and informational frictions during job search that can trap young women in unemployment or low paying self-employment activities locally, a concern that will increasingly become critical as cities expand. Following urbanization in Nigeria, Animashaun and Emediegwu (2023) show that aggregate household labor supply declined impacting women more adversely compared to men. In Ethiopia, Abebe et al. (2021) find that the individual's likelihood of formal employment is negatively related to their home's distance from the city center and that an intervention subsidizing transport costs can reverse the tendency to engage in unproductive self-employment in favor of formal wage employment. That study also found that young unemployed people spent close to 16 percent of their limited income on search-related activities in Addis Ababa. Deserranno (2019) shows that financial incentives increase the applicant pool for health-care positions in rural Uganda. However, local recruiters may not have a good understanding of the prohibitive role of application cost. Abebe, Caria, and Ortiz-Ospina (2021) show that firms significantly underestimate the positive impact of job-application incentives in improving both the quantity and quality of applicants. The findings show that subsidizing applications is particularly beneficial for young women who stand to gain the most from the employment opportunities.

Furthermore, young women may not necessarily search actively and in a targeted way―instead, they tend to primarily use informal search channels and are likely detached from the labor market. Job boards, where job advertisements for formal jobs are commonly posted, mostly cater to those with college-level education and rarely contain information suitable for less educated and experienced women and youth (Franklin 2018). As a result, the labor market this paper explores is highly segmented by gender and education levels, largely excluding uneducated youth and women from formal employment.³

Similarly, lack of information about available job opportunities also precludes effective job search. For example, Ashraf et al. (2020) find that making career benefits salient on recruitment posters increases the quality of applicants for a health-care position in Zambia. In the study sample, only 15 percent of job seekers had heard of the Bole Lemi industrial park, about 75 percent had no prior factory experience, and only 1.7 percent had any experience working in the industrial park at baseline. Lacking job-specific experience, job seekers may be unfamiliar with basic yet essential aspects of the job-search process including the job-application and interview process. As a consequence, the median applicant in this study made no job applications for a formal wage-paying job in the four weeks leading up to the baseline survey. Current understanding of which policies can effectively help young women secure formal jobs remains limited. Yet tackling the barriers that prevent young women from participating in the formal labor market is a first-order problem for many low-income countries (Dinkelman and Ngai, 2022).

This study's research examines the impacts of a job-facilitation intervention eight months and four years after the initial job screening and application phase. The study finds that the extra support during the application process increases individuals’ likelihood of being employed eight months after the treatment. Higher employment levels in wage work, in particular factory work, appear to be the driving force behind this result. This finding suggests that young women face significant barriers to labor-market participation and that a simple, one-off facilitation, support improves their transition to formal employment in the short run. The intervention also raises reported monthly income by nearly 30 percent. Higher income in the treatment group is accompanied by a 16 percent increase in nonfood expenditures and a 57 percent increase in savings. However, the study also finds an adverse impact on health outcomes and a downward adjustment of job applicants’ expectations and perceptions regarding the quality of factory work in response to the treatment.

Many of these effects were short-lived. In the long run—that is, four years after the initial job screening—those who had received job-application support were more likely to have had factory work experience and factory employment experience, but there were no longer any differences in earnings, expenditures, and savings; mainly because the control group caught up with the treatment group in terms of finding alternative employment opportunities. For example, average weekly number of hours worked eventually converged—67 hours for treatment and 66 hours for control individuals. Likewise, the difference in average monthly earnings between women in the treatment (1,474 birr) and control (1,350 birr) groups is no longer significant.⁴

The results from the four-year follow-up survey also show that the adverse health impacts for women in the treatment group disappear over time and that factory work is considered no more physically demanding or unhealthy than other types of jobs individuals engage in. The downward adjustment of perceptions and expectations regarding the quality of factory work, however, is maintained in the long run. This is consistent with the fact that only 12.2 percent of young women who were assigned to the treatment group and started factory jobs were still working in factories four years later. These findings suggest that factory jobs may fall short of providing a wage premium and offer only limited potential for upward mobility in the long run.

This paper makes three important contributions to the literature. The first contribution is that, to the best of the authors’ knowledge, this is the first study to systematically evaluate job-facilitation support in the context of industrial parks in low-income countries. Industrial parks host firms in modern production facilities that are often located near each other but far from urban centers and residential areas. These features of industrial parks have several implications for job searches, employment, and worker welfare. First, the relative isolation of industrial parks from city centers creates a disconnect between job opportunities and job seekers. Inadequate access to information and to transportation infrastructure exacerbates the effects of this spatial disconnect. As described earlier, only one in eight job seekers in the study sample had ever heard of the Bole Lemi industrial park, and only 1.7 percent had any experience working in the industrial park at baseline. In the presence of such informational and spatial frictions, job seekers’ beliefs about their own labor-market prospects and job attributes may be inaccurate (Banerjee and Sequeira 2020).⁵ By providing job seekers with information and transportation to industrial parks, the intervention addresses these barriers to an effective job search. Second, industrial park jobs are in export-oriented sectors, commonly in garments and textiles, that have seen growing scrutiny and intense buyer audits to ensure fair employment practices and workplace compliance. Yet it remains unclear whether industrial park jobs are of better quality than alternative employment options. Third, factories in industrial parks predominantly employ young migrants in low-skill, low-wage positions with limited unionization and access to local safety nets (Meyer et al. 2021).

The second contribution of the study is linked to the literature on active labor-market policies, especially work that explore search frictions in low-income countries (Abebe, Caria, and Ortiz-Ospina 2021; Abel et al. 2019; Franklin 2018). In growing cities, spatial and informational frictions are critical for the allocation of limited job opportunities. As a result, job search is often a self-managed process constrained by the availability of information and affordable transportation options―to acquire information about job opportunities, job seekers commonly commute to city centers to visit employers, work sites, or general-purpose job boards where vacancy information is physically displayed (Franklin 2018). Typically, applications are also submitted in person and involve commuting to submit CVs and go to interviews, written exams, and medical tests. Young women often face challenges related to accessing affordable and safe transportation options to make such commutes. This paper tests whether light-touch job-application support can help ease search frictions and improve the labor-market prospects of inexperienced young women. The study finds that job-search frictions deter female job seekers’ entry into industrial jobs, but that the search frictions do not generate long-term negative impacts on their labor-market outcomes. This result is largely consistent with emerging evidence showing that lowering search cost increases job-search intensity and improves employability in the short run, but has no long-lasting impacts (Abebe et al. 2021).

The third contribution is that this study provides a direct test of whether jobs in industrial parks provide young women with a reliable pathway to economic empowerment in a setting in which severe gender gaps in economic outcomes persist. As more industrial parks are planned and constructed in the country, understanding, and documenting the impact of these new manufacturing jobs on the livelihoods and empowerment of young women—and the associated economic and social impacts on young women's households—is crucial for shaping more inclusive industrialization policies.

This study is most closely related to the work of Blattman and Dercon (2018) and Blattman, Dercon, and Franklin (2022), who compare the one- and five-year impacts of offering young job seekers in Ethiopia factory work or a cash grant versus a control group. This study differs from theirs in that they offer all job seekers in their wage work treatment arm a factory job. This study randomizes light-touch job-application support, which, we believe, makes the study more applicable to the actual choice set of policy makers and offers a greater scalability potential. In essence, this study examines search frictions that prevent job seekers from obtaining industrial jobs. The focus of Blattman and Dercon (2018) and Blattman, Dercon, and Franklin’ (2022) is on studying matching frictions and overcoming liquidity constraints. In addition, the applicant population in this study is composed entirely of women with about 70 percent migrants, whereas their sample contains both men and women and mostly nonmigrants.⁶ Finally, this study is situated within the context of industrial-park factory jobs offered by foreign-owned exporting firms in the garment and textile sector in an urban environment, whereas they sample workers from four domestic and one foreign firm in five sectors in four administrative regions. Despite these differences, the key finding that factory jobs are not economically more beneficial than alternative employment opportunities in the long run is consistent with their results.

The remainder of the paper is organized as follows. Section 2 discusses the context of factory work in Ethiopia. Section 3 describes the research design, data, and estimation strategy. Section 4 presents and discusses the impact estimates of the intervention on a range of outcomes related to employment, earnings, and health. Section 5 provides concluding remarks.

2. Industrial Parks and Factory Jobs in Ethiopia

Despite accelerated progress toward economic development and poverty reduction, Ethiopia remains among the poorest countries in the world. Its rapidly growing population requires large-scale job creation to meet the demand for economic opportunities. Informal work (own account and family employment) dominates employment, with only 13.5 percent of the workforce classified as wage workers (ILO 2018). With over 70 percent of the population engaged in the agricultural sector, structural transformation has been slow (Bezawagaw et al. 2018). The country's growing population of landless youth in rural areas motivates the government's plan to transition from reliance on agriculture to manufacturing, particularly of tradable goods. Growth in export-oriented industrial production is one way to absorb large portions of youth looking for employment (FDRE 2010; NPC 2015). The construction of industrial parks is the main policy tool that the Ethiopian government uses to achieve this goal.⁷

There is significant evidence from other contexts that this strategy may pay off. Formal employment, particularly in export-oriented sectors, can offer workers and in particular young women better working conditions, greater workplace safety, and higher wages (El Badaoui, Strobel, and Walsh 2008; Verhoogen 2008, Lopez-Acevedo and Robertson 2016; McCaig and Pavcnik 2018; Tanaka 2020). There are several ways in which the export sector supports job growth and helps to improve working conditions. First, exporting fuels worker reallocation from informal enterprises to more productive firms that potentially offer higher monetary and nonmonetary remuneration. McCaig and Pavcnik (2018), for example, find that a drastic decline in U.S. tariffs on Vietnamese exports in 2001 led to a significant increase in formal-firm employment, particularly for younger workers from exporting provinces. Fukase (2013) shows that the tariff decline was also associated with comparatively greater wage growth among unskilled workers in provinces with greater exposure to export. Similarly, following textiles and apparel exporting, Lopez-Acevedo and Robertson (2016) find that low-skilled women across several countries in South Asia received higher wages in industrial employment compared to agriculture.

Second, supporting export-oriented sectors by place-based policies, such as industrial parks, can attract foreign direct investment that improves employment and wages. Lu, Wang, and Zhu (2019) and Zheng et al. (2017), for example, show that industrial parks attracted a new wave of foreign investment that stimulated local production and generated significant spillover effects in China. Further, both studies independently show that firms in special economic zones generated more employment and paid higher wages compared to firms outside the zones. Similarly, trade policies that promote economic integration can facilitate the entry of greenfield investments. McCaig, Pavcnik and Wong (2022) look at the longer-term cumulative effect of U.S. tariff cuts on firm entry and changes in employment in Vietnam. The authors find that the tariff cuts had long lasting impacts on labor demand and employment particularly in industries that experienced the largest decline in U.S. tariffs.

Third, export-oriented firms operate in globally competitive supply chains and face increasingly more stringent labor standards and labor-compliance requirements. To meet these standards, export-oriented firms often improve working conditions and worker compensation. Tanaka (2020) finds that exporting had significant positive effect in improving the working conditions of garment workers in Myanmar even when there was no impact on wages.

The African experience with export-oriented industrial parks, however, has been mixed (Bräutigam and Xiaoyang 2011; Farole 2011). Even where industrial parks have successfully attracted investment, concerns remain over the quality of employment, labor rights, and issues of sustainability. Labor-intensive industries that are typically found in industrial parks often produce low-value commodities with thin profit margins. With narrow margins, the international competitiveness of these industries mainly lies in minimizing production costs along the value chain by adopting measures that potentially diminish the quality of employment. Well-organized firms in industrial parks also possess market power that enables them to use collective action and jointly set wages and control labor mobility within the park, practices that are detrimental to workers’ welfare.

In Ethiopia, the first industrial park, Bole Lemi, was completed in 2014.⁸ Located on the outskirts of the capital, Addis Ababa, Bole Lemi hosts companies engaged in garment and leather goods production. These companies are all foreign-owned and produce almost exclusively for the export market. At the initiation of the present study, 10 firms were operating in Bole Lemi industrial park.⁹ Most firms in the industrial park had just begun production and were hiring workers in great numbers to gradually increase capacity. These firms were primarily hiring female employees and thus creating a path for many of these young women to enter the formal labor market for the first time. In Ethiopia—where women are at an inherent disadvantage because of social norms, unequal institutions, and other barriers to labor-market participation—the light-manufacturing sector has the potential to offer them a path out of poverty and into economic independence. Emerging evidence suggests that industrial employment generates welfare improvements for female workers in the country's modern horticultural sector (Getahun and Villanger 2018; Suzuki et al. 2018). Moreover, evidence from other developing countries shows that low-skill manufacturing jobs can promote gender equity and improve quality of life for female workers, as well as their children's health and educational outcomes (Hewett and Aminb 2000; Nicita and Razzaz 2003).

Nevertheless, feminization of labor can have unintended consequences. For example, Ghosh (2009) and Seguino (2000a) show that East Asia's export-oriented industrialization was highly dependent on gender-based wage inequalities. Furthermore, segmentation of women into these export-intensive industries characterized by high price elasticity restricts their bargaining power. These forces allowed firms to keep wages artificially low and widened the gender wage gap (Berik, Rodgers, and Zveglich 2004; Chamarbagwala 2006; Pitt, Rosenzweig, and Hassan 2012; Seguino 2000b).

3. Research Design, Data, and Empirical Strategy

To uncover the short- and longer-run impacts of factory work on a series of welfare indicators, this study evaluates an intervention that used a relatively light-touch approach to support and facilitate the job-application process at firms located in the Bole Lemi industrial park for a randomly selected sample of eligible female job seekers.

3.1. Research Design

During the initial phase of the evaluation, the research team collected hiring plans from each participating firm and used this information to target interested job candidates. All partnering firms insisted on hiring exclusively women in a certain age range for the factory floor jobs considered for this study.¹⁰ The factory positions were advertised through various channels, including posting advertisements in public places, passing out flyers in highly frequented areas of Addis Ababa, coordinating with youth associations, and using other forms of community mobilization. Unemployed individuals who had registered with their local woreda were also contacted directly.¹¹

During the recruitment process, potential candidates who were interested in factory employment were asked to bring identification and documentation of their qualifications to the nearest screening center, which were set up in several woreda offices across the three sub-cities of Addis Ababa. Trained enumerators staffed these screening centers during regular working hours to confirm that job applicants fulfilled the formal eligibility criteria for the advertised positions. Applicants with incomplete documentation or those who did not meet all the firms’ eligibility criteria were screened out from the study. All applicants who met the eligibility criteria and could produce sufficient documentation were selected into the sample and asked to stay for the baseline survey.

Immediately after the baseline survey, study participants were randomized into treatment and control groups using a lottery, with two-thirds of applicants in the treatment group and one-third in the control group. Each treatment applicant was assigned to a firm for an interview. Since each participating industrial park firm had set its own hiring criteria, there were multiple distinct applicant groups.¹² Treatment applicants who were eligible for multiple firms were randomly assigned to one of the firms for which they were eligible.

Once applicants were assigned to a firm, the enumerators formally invited treatment individuals for the job interview and offered transportation to the Bole Lemi industrial park on the day of the interview.¹³ Treatment and control individuals learned about their treatment status after completing the baseline survey but before they left the screening center. On the day of the interview, enumerators confirmed the treatment status before providing transport to the firm. After the interview process was completed, all candidates were offered transportation back to the screening center. The intervention did not prevent individuals in the control group from adopting alternative and commonly used pathways to applying for factory employment in any of the Bole Lemi industrial park firms.

Of those in the treatment group, 63.6 percent reported at the screening center on the day of the interview and were provided with transport to the industrial park. Firms conducted interviews with job candidates according to their own hiring procedures and decided whether to offer applicants the factory job. For some firms, this meant that candidates were required to complete additional stages in the hiring procedure such as a medical exam or an aptitude test. Nevertheless, the eligibility verification at the screening centers before baseline was effective as nearly all candidates were selected.¹⁴ Successful applicants then had the opportunity to accept or decline the job offer. Of those who were interviewed and who were offered a position, 48.3 percent accepted the job offer.¹⁵

Applicants who accepted the job offer typically also had to go through an onboarding process at the firm, which mainly consisted of training that lasted for several days or weeks depending on the firm. This training focused mostly on technical skills related to fabric handling, stitching, pressing, packing, and quality control.

3.2. Data

Baseline data collection took place at screening centers directly after the eligibility checks but before treatment assignment to minimize attrition and ensure full comparability. All eligible job applicants who agreed to take part in the study were interviewed at baseline, and information from 935 respondents was collected between June and August 2016.

Approximately eight months and four years after the intervention, an effort was made to re-interview all baseline respondents face-to-face. A total of 827 baseline respondents were successfully tracked in the short-term follow-up survey that took place between January and March 2017—an attrition rate of nearly 12 percent. The long-term follow-up survey took place between February and April 2020, and 741 baseline respondents were interviewed—an attrition rate of about 21 percent. The 687 respondents interviewed in all three survey waves are at the core of the impact analysis. The attrition rate for this strictly balanced sample is nearly 27 percent.

Losing more than one quarter of respondents over a period of four years is of potential concern. Whether attrition systematically differs according to treatment status and whether it varies with any observable individual characteristics is tested in this paper. Table S1.1 in the supplementary online appendix, indicates that treatment is not correlated with attrition of respondents and that the sample composition did not meaningfully change between baseline and the two follow-up surveys because of attrition. In subsequent analysis, the study further accounts for attrition by weighting each observation by the inverse of its predicted probability of panel inclusion (being tracked at follow-up) using a leave-one-out logit regression in the first stage. It also applies several bounding approaches to examine the sensitivity of the findings to different assumptions regarding missing data caused by attrition.¹⁶

Table 1 indicates that study participants randomized into the control group were on average 25 years old at baseline, one year older than those in the treatment group. Approximately 30 percent of control respondents were married, and slightly more than one-quarter were mothers at baseline. The number of migrants among applicants for factory work was high; nearly 70 percent of respondents had migrated to Addis Ababa from elsewhere in the country. On average, respondents had more than nine years of completed formal education, and nearly 14 percent of the control sample was enrolled in some form of educational program at baseline.¹⁷ Except for age and reported motherhood, table 1 does not show any systematic differences across a wide range of observable characteristics between the treatment and control groups, which suggests that random allocation of study participants worked well. An F-test cannot reject the joint hypothesis that each coefficient obtained from a regression of treatment on all variables listed in table 1 is equal to 0. To attenuate the effect of potential biases and improve precision, the educational and demographic variables are used as controls in the estimating equation.

To compare the study participants to a nationally representative sample of urban job seekers, this paper also uses data from the Urban Employment Unemployment Survey (UEUS), which was conducted at about the same time as our baseline survey. Column 1 of table S1.2 in the supplementary online appendix is based on the complete UEUS sample collected from across all urban areas in Ethiopia. Column 2 restricts the sample to Addis Ababa, and column 3 restricts this further to those who had the age and education profile that match our own survey sample. This accounts for approximately 20 percent of the Addis Ababa sample in the UEUS. While there are some differences between the study sample and the nationally representative sample of job seekers, the differences are hardly economically meaningful.

3.3. Empirical Strategy

There is considerable self-selection at various stages during the hiring and onboarding process for factory work. Not all eligible applicants in the treatment group participated in the job interviews, not all applicants who passed the interview successfully took the job, and not all applicants who started working in the factories stayed on the job until follow-up data was collected. Table S1.3 in the supplementary online appendix presents the selection processes at various stages of the experiment―how different sets of baseline characteristics correlate with selection into attending the initial job interview, accepting a job offer for factory work, and continued employment in the factory job four years after the treatment. Columns 1 to 4 show that migrants were much more likely to attend the job interview among those applicants randomized into the treatment group. This is consistent with this study's finding from qualitative interviews, which suggested that migrants’ primary reason for leaving their hometowns was to look for employment opportunities followed by a desire to pursue their education. Yet columns 5 through 8 indicate that migrants are no more likely to start a job than nonmigrants. Age and earning history appear to be the two main variables that predict the decision to start a job―younger and poorer applicants are more likely to start a job. The former mirrors firms’ preference to select younger workers, and the latter may reflect that financial stress is a strong motivator for applicants to take up this job, perhaps any job, when offered. Columns 9 to 12 show that none of the observable characteristics at baseline are correlated with the likelihood of still working on the job approximately four years after the treatment. While about a third of respondents were working on the job eight months after the treatment, this is true for only 12 percent of these respondents four years later.

The study primarily focuses on estimating the intention-to-treat (ITT) impact of the job-facilitation intervention. The ITT impacts are estimated using an ANCOVA specification for the two follow-up surveys separately:

where |${y}_{it}$| is an outcome measured for respondent i at the eight-month (⁠|$t = 1$|⁠) or four-year (⁠|$t = 2$|⁠) follow-up. The variable |$trea{t}_i\ $|is an indicator equal to 1 if the respondent was assigned to the treatment group and 0 otherwise. |${X}_{i0}$| includes a set of control variables measured at baseline including age, a dummy for marital status, a dummy for motherhood, a dummy for migration status, the highest grade completed, and evaluation duration, measured as number of days between baseline and follow-up surveys. The specifications also include month-of-survey dummies to control for the possibility of COVID-19 impacts confounding the results.¹⁸  |${y}_{i0}$| is the outcome measured at baseline. The results section's focus is on |${\beta }_2$|⁠, which measures the treatment impact, and does not report |${\beta }_1$| or |${\beta }_3$|⁠. To account for attrition, each observation is weighted by the inverse of its predicted probability of continuing in the study. All reported control group means are also adjusted using these weights.

This study tests whether treatment had any impact on a wide range of outcome variables measured in the two follow-ups. For each family of outcome variables, sharpened q-values are constructed to control for multiple hypothesis testing (Benjamini, Krieger. and Yekutieli 2006). Sharpened q-values are designed to control for false discovery rates when simultaneously testing for an effect of treatment on several set of outcome variables that may be correlated. As a further robustness test, a standardized index is created for each family of outcomes, and the treatment results on these aggregates are presented in table S1.4 in the supplementary online appendix.

Aside from the ANCOVA specification, the results from a simple mean comparison of the outcome |${y}_{it}\ $|between the treatment and control groups for the two follow-ups (labeled “ITT difference” in the tables) are shown. The sharpened q-values for these results are also computed.

4. Results

The results are split into short-run (eight-month) and long-run (four-year) impacts, which are summarized in tables 2 through 4. The set of outcomes considered includes variables that capture employment, income, expenditures, health, and expectations and perceptions about factory work.

4.1. Impacts on Employment

The immediate objective of the facilitation intervention was to ease the logistical and procedural hurdles during the factory job application process. Table 2 shows that the likelihood of having had a factory job is more than 50 percent greater than for the control group in the short run, indicating that the job-facilitation intervention was successful in bridging logistical constraints on seeking employment in the industrial park. More importantly, the intervention was successful in affecting the employment status eight months after the treatment. Table 2 indicates that the treatment group is nearly 25 percent more likely to be employed, largely driven by wage, particularly factory work. The amount of time that individuals were employed in the six months leading up to the first follow-up survey also increases.

Consistent with the hypothesis that industrial work offers steadier hours, table 2 shows that the intervention is associated with roughly five more hours of work per week in the treatment sample. The impact on self-employment is close to zero, suggesting that the factory jobs represented an addition to the set of job opportunities available to the target group and did not merely crowd out other income-generating activities in the short run.

The results from the four-year follow-up indicate that the higher levels of employment in factory work observed in the short run persist but that overall employment effects dissipate. The last two columns of table 2 show that treatment individuals are no more likely to have higher rates of current employment, wage employment, or self-employment than those in the control group. As indicated in the last two rows, the number of months employed in the six months before the second follow-up survey and weekly hours of work are similar between the two groups as well. The long-run results are mostly driven by the fact that the control group caught up to the treatment group by finding alternative employment opportunities. Four years after the intervention, average weekly labor hours converged—67 hours for the treatment and 66 hours for the control individuals.

4.2. Impact on Earnings, Expenditures, and Savings

In the short run, helping job applicants obtain factory work significantly increases their income (first row of table 3). Earnings received in the last four weeks were about 200 birr higher in the treatment group, which is an increase of nearly 33 percent over the mean of the control group at baseline. Consistent with the higher income that the treatment group earned, table 3 shows that the intervention is associated with larger nonfood expenditures and higher savings in the treatment group. These findings contrast with a key result in Blattman and Dercon (2018), who do not find significant income effects from offering factory jobs. These results are also not artifacts of multiple hypotheses testing as shown by low q-values throughout the table.

That said, four years after the intervention, the study finds no statistically significant treatment impacts on earnings, expenditures, or savings. The earnings gains that the treatment individuals enjoyed over the control group at the first follow-up are erased by the four-year follow-up (fig. S1.1). The reason for this may be that the employment rate of the two groups converges in the long run (see table 2). Given the substantive entry-level pay difference between factory work and alternative jobs observed in the short run, the lack of long-run earning impacts may result from infrequent wage adjustment or slow wage growth in industrial parks.

4.3. Impacts on Health

Industrial work can potentially expose workers to health risks. In the short run, the study finds that a standardized strenuosity score, which measures difficulty with daily (and work-related) activities, increases for the treated group (table 4).¹⁹ Respondents were also asked to assess the healthiness of factory jobs.²⁰ The estimates presented in the second row of table 4 suggest that the treatment group is more likely to consider factory work unhealthy. These results mirror the adverse health impacts that Blattman and Dercon (2018) reported.²¹ The radar graph in fig. S1.2 decomposes the index and plots impacts and corresponding confidence intervals for each of the activities in isolation. Perhaps unsurprisingly, the two activities most closely associated with and required for factory work (working on feet and standing at work bench) are driving the results on the strenuosity score.

Four years after the intervention, the adverse health impacts disappear, which may be explained by the following. First, women experiencing negative impacts in the short run may have left factory employment. Second, the negative health impacts may be temporary. There is evidence corroborating the second explanation. The study finds that the likelihood of current factory employment at four-year follow-up is not correlated with reported health status at the eight-month follow-up. Table 2 showed that individuals in the treatment group are more likely to be employed in factories, and table 4 indicates that their current jobs are no more physically demanding or unhealthy than those of the control group. Figure S1.3 also presents the same decomposition of the health index using the radar graph for long-run impacts. The graph supports the conclusion that health problems uncovered in the short run did not persist over time. This result is consistent with that of Blattman, Dercon, and Franklin (2022), who also find that negative health impacts detected one year after the intervention decay over time.

4.4. Impacts on Expectations and Perceptions

One possible explanation for the high rate of worker turnover and the short duration of workers’ spells in factories is an ex-ante mismatch between applicants’ job-related expectations and the reality of factory work. Recent qualitative studies that explore worker and manager interactions in the textile, garment, and leather manufacturing industries find that expectation mismatch is a key driver of worker turnover in Ethiopia (Barrett and Baumann-Pauly 2019; Hardy and Hauge 2019). The present study tests whether treatment group participants adjusted their expectations regarding factory work after having experienced this type of work.

The results are shown in the lower panel of table 4 and suggest that treatment group participants lowered their expectations regarding the earning potential of factory work and adjusted their expectations downward even more in the long run. They were also less likely to view factory work as a permanent job. Again, this decline is more pronounced in the long run. Although treatment applicants had significantly more negative perceptions of the job quality and healthiness of factory work in the short run, their views do not differ from those in the control group in the long run, partly because the control group also developed more negative views about factory work themselves. This finding is plausible if one accepts that information about factory jobs spreads, for example, through social networks or public media over time.²²

Intuitively, these findings are not surprising. The short-run health impacts that workers experience are likely to make factory work less preferable compared to alternatives in the long term (Blattman and Dercon 2018). By the very nature of the organizational hierarchy in factories, upward mobility of young workers’ careers in the garment industry is limited. This industry employs many relatively poorly educated, low-skilled workers who typically remain at the bottom of the organizational pyramid. Even with tenure, the chances of career progression to positions beyond line management or supervisory levels are minimal. The recruitment pool for these positions differs, and college graduates from the outside commonly fill openings. Further, employment alternatives, such as the prospect of being employed as a domestic worker in the Gulf or the Middle East, are attractive outside options (Hardy and Hauge 2019).²³ Female workers face demands on their time related to childcare and family responsibilities, as well as potentially constraining social norms, once they marry, that can effectively end careers in the garment industry. Therefore, many workers will consider jobs in labor-intensive manufacturing industries as temporary or at least nonpermanent.

4.5. Discussion

Tying together these results, it seems evident that factory jobs in the present context are not attractive enough to convince large numbers of job seekers to interview for these positions. Additionally, even when applicants accept factory positions, workers are less likely to stay on the job for a longer period of time, reducing the potential for a job-facilitation intervention to have lasting effects in the long run. This finding resonates well with previous work on this topic and serves as a reminder that worker turnover can generate a misallocation of talent and lead to welfare loss. In particular, those who quit early may not have had sufficient time to learn about important aspects of the job and the job's match with their own preferences, talents, and expectations. Often, the alternative is to engage in informal or casual work, which may not necessarily be welfare enhancing in the long run. Data on wages in different occupations underpin this argument. In fig. S1.4, the average wage earned from factory work is compared to the average wage rate earned in any other occupation reported in each of the three survey waves. Not only is the reported wage rate from factory work higher in each wave, but the spread between wages from factory work and other occupations also increases over time. Along the same lines, Meyer, Krkoska, and Maaskant (2021) also find that wages in industrial work compare favorably to wages in the local Ethiopian labor market where industrial parks are located.

This study also documents that job seekers attach a significant premium to job security. The sample respondents report that they prefer a casual job over permanent employment only if the casual job offers a wage equivalent to 179 percent of the corresponding wage from a permanent work on average. This suggests that workers are willing to accept a considerably lower wage to engage in permanent and formal jobs, but these jobs are scarce in developing countries (Abebe et al. 2021). In the four-year follow-up, for example, about 36 percent of treatment and control group respondents who quit factory work remained unemployed, and among those who are employed, more than half worked in informal jobs including food and beverage, petty trading, and domestic work.

Yet, even in the absence of better alternatives, a staggering turnover rate of factory jobs is observed in the sample. To begin with, 52 percent of those job seekers in the treatment group who were offered positions in Bole Lemi, did not accept the job offer. According to the first follow-up survey in 2017, the most important reasons named by respondents who rejected factory positions were the salary offered, working conditions and travel time, transport costs, and concerns around health and safety (see figs. S1.5 and S1.6 in the supplementary online appendix for more detail).

The study finds that the bulk of turnover happens in the first few weeks of recruitment and tends to level off over time (fig. S1.7). In the first follow-up, around 65 percent of respondents who initially accepted factory job offers and started working in Bole Lemi quit within seven months. This number increased to nearly 88 percent in the long-term follow-up.²⁴ Different wage and nonwage job characteristics dictate job-exit decisions. Among those who started, but then quit employment in Bole Lemi, the level of monetary compensation was the most frequently mentioned factor that contributed to their exit decision. However, illness and disability also played a decisive role, and so did the strenuosity connected to the work, childcare obligations, and the desire to search for better jobs (see fig. S1.6 in the supplementary online appendix for more detail).

For firms, worker quitting imposes significant costs in terms of investment in training, recruitment effort, failure to meet production targets, and reduced worker morale. Especially quitting among newly hired employees represents an important challenge related to worker turnover faced by firms in both developed and developing countries (Donovan, Lu, and Schoellman 2023). Yet firms in the study's context appear reluctant to adjust employment policies and compensation to retain newly recruited workers. This may be explained by the following. First, entry-level workers are relatively easy to replace, and the recruitment and training costs as well as the disruption costs are not large enough to initiate a policy change (Blattman, Dercon, and Franklin 2022). Second, firms may look at the initial time workers spend in the factory as an implicit trial phase that helps to improve the match between workers and the job. Third, any financial or nonfinancial retention incentive firms introduce to reduce turnover among entry-level workers must be extended to incumbent workers, thereby significantly increasing overall labor cost. Distinguishing between these varied explanations for firm behavior related to hiring, onboarding, and retaining workers is beyond the scope of this study.

4.6. Robustness Checks

Multiple hypothesis testing is a possible concern when testing numerous outcomes at the same time. In addition to the sharpened q-values the paper reports in the main results tables, this concern is addressed by aggregating related outcome variables into four distinct index families. Each index is then transformed into a z-score to generate standardized indices, one for each family of outcomes.²⁵ Following this procedure, all the outcome variables presented in tables 2, 3, and 4 are combined to generate indices on employment, income and spending, health, and expectations and perceptions. Table S1.4 reports the simple ITT differences and ANCOVA treatment effects for the indices. The intervention affects each of the four indices, but these effects are observed only in the short run. There are no long-run effects with the notable exception of the expectation and perception index.

The level of attrition in the balanced sample is another concern that is addressed in different ways. First, the study applies inverse probability weighting throughout for the main results. Second, the short-run results are re-computed using all observations available without restricting the sample to the full panel respondents only. That is, information from all 827 baseline respondents who were interviewed in the first follow-up is used. In doing so, the results are qualitatively the same as the impacts reported in the main regression tables, and the estimated coefficients are very close in magnitude. Only the ITT estimate on savings is no longer significant, but the adverse effect on the perception of factory jobs as permanent employment now is. Finally, a series of bounding approaches are implemented to check the sensitivity of the results to different assumptions on the nature of missing data due to attrition. This mainly relies on two approaches and focuses on the main results―impacts on employability and earnings at the first follow-up. For employment-related indicators, the method of bounding developed by Lee (2009) is used. For earnings, missing data is imputed using a similar procedure to what is commonly done in the related literature (Abebe et al. 2021; Blattman, Fiala, and Martinez 2014; Karlan and Vadivia 2011). This method offers a test of the sensitivity of the results to potential systematic attrition, where attritors in the control and treatment groups are assumed to exhibit different patterns that lead to overestimation (lower bounds) or underestimation (upper bounds) of treatment effects. The bounding results suggest that the main results are robust to attrition effects. These results are not reported, but they are available upon request.

5. Conclusion

New economic opportunities such as factory jobs in a growing manufacturing sector are thought to provide a potential avenue out of poverty for many people in economies undergoing structural transformation. This study facilitated access to such factory jobs at the Bole Lemi industrial park on the outskirts of Addis Ababa to study their welfare impacts. To do so, a randomly selected sample of female job seekers are provided logistical support during the hiring process to help them overcome entry barriers to formal employment that women often face. These barriers tend to be especially constraining for those with weak networks and limited labor market experience, which often characterizes young women at the beginning of their professional careers, in particular if they are migrants. The study estimates the impact of the intervention eight months and four years after the initial job screening and application phase.

The results indicate that, in the short run, a light-touch job-facilitation intervention can have large impacts on applicants’ success rates in finding employment and in improving earnings, although adverse health impacts accompany these effects. In the long term, applicants in the control group seemingly catch up and find other employment opportunities, which erases the earnings advantage of those in the treatment group but also cancels out the adverse health impacts. Moreover, negative perceptions of factory work increase in both groups. Part of this is likely due to information updating as more potential workers learn about the peculiarities of factory work. However, the absence of sustained growth in earnings raises questions about long-term career prospects for workers and about talent acquisition, retention, and productivity for firms, which have important implications for long-term industrial development.

Data Availability Statement

The data underlying this article are published in the World Bank’s microdata library https://microdata.worldbank.org/index.php/catalog/6199. Replication files can be accessed at https://github.com/girumabe/Factory-Jobs.

Funding

The work is supported by the World Bank Umbrella Fund for Gender Equality (UFGE).

Conflict of Interest

The authors declare that there is no conflict of interest related to this research work. The authors do not stand to benefit, either financially or professionally, from reporting either null, positive or negative impacts from the experiment in this research.

Author Biography

Girum Abebe (corresponding author) is a senior economist at the World Bank, Washington, DC, USA; his email address is [email protected]. Niklas Buehren is a senior economist at the World Bank, Washington, DC, USA; his email address is [email protected]. Markus Goldstein is a lead economist at the World Bank, Washington, DC, USA; his email address is [email protected]. The authors are grateful to the World Bank Umbrella Fund for Gender Equality for funding the research. They thank Taylor Van Salisbury, Adiam Hagos Hailemicheal and Endale Gebre Gebremedhine for excellent research assistance and the Ethiopian Development Research Institute (EDRI) for expert field implementation. This paper is an output of the World Bank Africa Gender Innovation Lab. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development / World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. A supplementary online appendix for this article can be found at The World Bank Economic Review website.

Footnotes

References