This project aims to examine whether job-quitting decisions contribute to the misallocation of talent in developing countries. Turnover decisions are not always optimal: the high rates of turnover among recent hires faced by firms in developing countries may be partly driven by workers who have not spent enough time on the job to fully understand whether they are suited to the role. Destroying a potentially productive employee-job match will have negative consequences for the both the firm and the worker. The resulting misallocation and potential reduction in labour demand might particularly affect the young female workers who make up most of the entry level workforce in Ethiopia's industrial parks.

There are several reasons to expect that an employee might quit before the suitability of their skills to the role has been fully revealed. First, workers do not feel the effects of the costs that turnover imposes on the firm: higher recruitment costs, challenges in meeting production targets and deadlines, lost training investments. On the other hand, in markets where (perceived) access to informal work opportunities is relatively easy, quitting a formal job may not be as costly for the worker. However, less demanding informal jobs may not be a good match for many individuals in the long run. Second, early signals about the suitability of an employee for a role may be misleading. Features of the work environment that are initially perceived as unpleasant may eventually stop the creation of long-term adverse effects for the employee. However, due to a phenomenon called "projection bias" (Loewenstein et al. 2003), people often underestimate their ability to adjust to new circumstances. Workers who leave the job early —which represent the bulk of the turnover problem faced by firms in both developed and developing countries (Jovanovics, 1979; Donovan et al. 2019) — may thus quit sooner than optimal.


This project will begin with a randomised controlled trial (RCT) conducted in cooperation with a large garment factory in the Hawassa Industrial Park, Ethiopia. The partner factory is a major employer and employs several thousand garment workers; mostly women aged 18 to 25. Workers are recruited from rural areas around Hawassa or have recently migrated to the city to work in the industrial park. The factory job is often their first experience with formal employment, and they are, thus, often uncertain about the demands of the job and the nature of the work environment. Turnover levels are high: up to 60% of workers leave during the first twelve weeks of employment, which imposes major costs on the firm. The high levels of turnover mean that the firm recruits between 50 and 150 new workers each month.

The RCT will randomly assign 2000 workers to either an early or late retention bonus scheme.

  • Early bonus scheme
    The early bonus incentivises completion of the initial training period, but does not incentivize workers to acquire experience on the production line.
  •  Late bonus scheme
    The late bonus incentivises workers to complete the training period and spend about 5 months working on the production line.

There are three key advantages of having the early bonus as a control condition, compared to a no-bonus control. First, the two bonuses have similar present value, which limits unintended effects due to workers' resentment of pay inequality. Second, researchers limit confounding effects related to differential access to liquidity (e.g. workers in both experimental conditions will have access to additional liquidity to face health shocks). Third, the team make sure both groups are incentivized to complete the training phase. The results will thus be independent of the skill acquisition during the training period.

The project’s experimental design is motivated by a simple conceptual framework that highlights the implications of uncertainty about the suitability of a candidate to a role on:

  • job leaving decisions;
  • the allocation of workers' talent;
  • and the firm's ability to retain its more productive employees.

As the bulk of worker compensation in this setting comes from a known, fixed salary, the key source of uncertainty workers face is related to an unpleasant aspect of the work. In particular, workers know that with time they will adjust to the conditions of the job, but do not know by how much. Each period workers decide whether to stay in the factory or to leave and earn the value of an outside option. If they stay, the employee learns the benefits to what initially appears as an unpleasant element of the role, and consequently changes their mind about whether this job is better than the outside option. The late retention bonus incentives workers to stay longer and observe the true value of the disamenity. If workers had initial misconceptions of their suitability for the work, the late retention bonus might permanently increase retention and, thus, benefit both the employer and workers.

In addition to the main experiment, the team will also conduct a monthly high frequency survey with a random subset of 150 workers in each treatment group. This survey will measure employment outcomes, work satisfaction, and beliefs. These data will enable us to compare baseline beliefs about future work satisfaction and productivity against observed developments.



Data collection is completed, and analysis is ongoing.



July 2019 - December 2023


Map - red dot for ethiopia

Firms, Farms and Labour


IZA Institute of Labour Economics logo