UC Berkeley

The present thesis uses original survey data from a national-scale government job training program to provide empirical microeconomic evidence on the functioning of firms, labor markets, and program targeting in low income countries. The program, apprenticeship training in informal sector Ghana, and the corresponding multi-tiered evaluation are ongoing. In the first two chapters, I utilize the random match between treatment apprentices and training firms to estimate the effects of access to labor on firms and of firm quality on apprentices. In the final chapter, I study the unusual implementation scenario in which we observe a sample of apprentices selected for the program by government officials alongside the entire pool of applicants eligible for the larger randomized controlled trial. Each of the three essays addresses, more or less directly, private enterprise development in low-income countries, an area of development economics that I believe is deeply in need of more research.

Small firms in developing countries are typically modeled as facing a frictionless market for workers, characterized by low search costs, full information, and a lack of regulation. In the first chapter, we report the results of a field experiment documenting that firms find it costly to hire workers on the open market, that the marginal revenue product of labor is positive and quite large in small firms, and that there is substantial heterogeneity in these returns as a function of (unobserved) worker ability. We study the impact of the program that randomly placed unemployed young people as apprentices with small firms in Ghana. The program provided a novel worker screening technology to firms (in addition to simply reducing search costs), as (voluntary) participation included non-monetary costs for unemployed young people applying to the program. We find that firms that were offered apprentices by the program hired and retained them for at least six months (the end of our study window). Secondly, treatment firms experience increases in revenues and profits of about seven to ten percent per assigned apprentice. Together, these findings suggest the presence of economically significant search costs in our context. Moreover, revenue and profit gains are particularly large for firms treated with high cognitive ability apprentices. This result highlights the importance of worker screening in firms’ hiring decisions, and echoes the widespread use of a sophisticated bond posting mechanism to hire apprentices in our baseline labor market. A simple model in which productivity differences associated with worker ability necessitate costly screening can predict the impacts of our program. In sum, our findings have implications for our basic understanding of labor markets in low-income settings and in particular suggest that high youth unemployment in developing economies is the result, at least in part, of substantial labor market frictions.

Apprenticeships and other types of on-the-job training are important sources of human capital development in low-income countries. In the program we study, as in most of West Africa, this training is typically undertaken by very small firms and by firm owners with relatively little formal education.

In the second chapter, I study firm-level determinants of apprenticeship training quality in this context using the same program feature exploited in the first chapter: that apprentices were randomly matched with small firms conditional on their stated preferences. I find that larger and more profitable firms cause better apprentice performance, as measured by firm owner reported apprentice competency on ten craft-specific skills. Firm owner performance on a cognitive test and past experience training apprentices are also positively associated with apprentice competency. In an effort to understand the channels by which these firm-level characteristics drive apprentice performance, we also investigate the relationship between baseline firm size, baseline firm sales, baseline firm profits, whether the firm employed any non-family workers at baseline, firm owner cognitive performance, and firm owner training experience on apprentice time use and apprentice attendance. We find broadly positive trends. These characteristics are associated with more working hours, better attendance, and less time spent doing errands unrelated to the craft.

In the final chapter, I study the direct selection by district officials of a portion of the ap- prentices who would ultimately be offered a space in the training program. These “priority” applicants were guaranteed admission, while the remainder of eligible applicants entered a randomization that assigned them to treatment or control status (after which randomized treatment and priority applicants were invited to participate in placement meetings that led to random assignment to firms, as noted above). I find that district officials select better educated applicants, who come from wealthier and better educated families, and who have more relatives working in local government, relative to the entire pool of eligible applicants. The results are strongest in the most urban region of the country, Greater Accra. Using the same firm-level follow-up data employed in chapters one and two, I estimate the relationship between priority status and apprentice-level outcomes. Priority applicants are more likely to attend a placement meeting, and ultimately to enroll in apprenticeship training through the program. They do not, however, appear to perform better once in the apprenticeship. Point estimates on their craft-specific skill competency are negative.

The essays in this dissertation, and particularly chapters two and three, will benefit from additional planned data collection. An independent assessment of apprentice skill is scheduled for later this year, and an apprentice-level survey on labor market outcomes for 2016/2017. Additional firm-level data collection with the entire sample we study in chapter one will take place this year as well.