In this project, the research team studied whether worker turnover contributes to the misallocation of talent in low-income countries. To address this question, the team experimentally evaluated the impacts of offering financial incentives for worker retention in a female-dominated occupation within the nascent garment manufacturing industry in Ethiopia.
The central hypothesis was that many young female workers who quit their jobs did so before spending enough time to learn the true quality of the job match, which for some workers may have been high. Standard economic theory views worker turnover as a mechanism through which economies achieve efficient talent allocation: new hires are initially uncertain about match quality and quit once they become convinced the match is poor (Jovanovic, 1979). However, the project identified several reasons why quitting decisions may occur before match quality is fully revealed. First, workers did not internalize the substantial costs that high turnover imposed on firms, including recruitment costs, production disruptions, and the loss of training investments. In contexts where access to informal employment is relatively easy, this created an asymmetry between the cost of turnover borne by workers and firms. Second, early signals about match quality were often misleading. Workers frequently adjusted over time to workplace disamenities that were initially perceived as severe, yet underestimated this adjustment due to projection bias (Loewenstein et al., 2003). As a result, early leavers—who constituted the bulk of turnover (Jovanovic, 1979; Donovan et al., 2019)—often quit sooner than was optimal. These dynamics were particularly pronounced for young female workers who had recently migrated and were only weakly attached to the labor market, leading to noisy priors about job match quality.
The project made three main contributions to the literature. First, it contributed to research on the allocation of talent by demonstrating that labor turnover—typically associated with improved allocative efficiency—can also be driven by suboptimal quitting decisions that destroy good job matches. This finding was especially relevant for young women from low-income backgrounds, who face significant barriers to securing well-matched employment (Hsieh et al., 2019). Second, the project contributed to the literature on worker retention in developing countries, which has documented high turnover rates, particularly in blue-collar occupations (Blattman and Dercon, 2018; Donovan et al., 2019). Third, it contributed to the structural behavioral economics literature (DellaVigna, 2019) by experimentally identifying and quantifying a bias in belief formation in a real-world labor market setting, complementing prior evidence from laboratory and observational studies (Loewenstein et al., 2003).
The research was conducted in the Hawassa Industrial Park, a flagship government-led development project in Ethiopia that hosts foreign garment manufacturers. As garment production was a new activity in the area, most workers were new to the industry. Factory-floor positions were overwhelmingly held by young women. The research team partnered with a firm in the park that agreed to randomize elements of its payment scheme. Prior to the main experiment, the team collected baseline data on workers and implemented a pilot study.
In the experiment, the team randomized the timing of retention bonuses offered to workers. One group received a bonus after three months of tenure, incentivizing completion of the initial two-month training period. A second group received a bonus after seven months, incentivizing sustained engagement in production work. The primary empirical test examined whether the delayed bonus generated higher long-term retention beyond the seventh month of tenure relative to the early bonus. To support interpretation and rule out alternative mechanisms—such as rising search costs or job-specific skill accumulation—the team collected detailed data on worker beliefs. The study also assessed whether workers in the delayed bonus group exhibited higher earnings and greater job satisfaction one year after the bonus offer, providing evidence on whether premature turnover had destroyed good matches. Finally, the project analyzed how the delayed bonus affected quitting behavior across different levels of the productivity distribution.
The findings of this completed project are of direct relevance to firms operating in high-turnover environments and to policymakers in Ethiopia and other low-income countries. The research team implemented a comprehensive dissemination and impact strategy to ensure that results informed firm practices and policy discussions. The project also included substantial investments in local capacity building.