The garment industry in Bangladesh employs roughly four million workers, the vast majority of them women. The sector has been a major pathway into paid employment for women who might otherwise have few options (Heath and Mobarak, 2015). But wages are low and working conditions are often difficult; garment workers in Bangladesh face safety violations (culminating in the tragic collapse of the Rana Plaza in 2013, which killed over 1100 workers), and high rates of physical and sexual abuse (Boudreau et al, 2022).
One contributing factor to poor working conditions could be that workers do not have complete information about working conditions when choosing workplaces, which could lead to inefficient matching between workers and firms and/or to firms making suboptimal investments in working conditions.
We implemented a cluster RCT to study the impact of information provision on workers, using a hybrid format with initial in-person information and follow-up via text messages and a hotline. Specifically, we began the project with a large representative survey in which we surveyed workers about working conditions and wages in their factories. We then collated this information into a scorecard that provided this crowd-source information in a color-coded scorecard with information about working conditions and wages at factories nearby (within 5 km) to experimental workers. Another treatment was a vacancy pamphlet wherein we surveyed HR managers about upcoming hiring and then provided workers information about job opening and how to apply. We then provided the scorecard, vacancy pamphlet, or both to workers, and assessed the impact of treatment on outcomes such as their beliefs about working conditions and labor outcomes.
Workers were overly optimistic about the quality of their jobs at baseline. In particular, in the control group, the average respondent believes that their factory scores a 4.17 on a scale of 1 to 5 compared to other factories within a 60-minute walk of them, where 1 is much worse; 2 is somewhat worse; 3 is about the same; 4 is somewhat better; and 5 is much better . One concern about these reports may be that workers either do not understand the scale we gave or do not actually have any sense of the distribution of working conditions across factories. However, though the scoring signals the presence of overconfidence, it also appears to reveal true information: respondents working in factories in the top quantile of the distribution of the working conditions score are approximately four times as likely to report that their factory is “much better” than other factories, compared to those in the lowest quantile.
For workers in our first wave of treatment (treated in September 2023), we find that the scorecard treatment revises these beliefs downward by 0.087 points on the 5-point scale. This effect is 0.18 points among workers whose factories scored below the median on the overall working conditions measure on the scorecard, where the gap between belief and reality was largest at baseline.
At baseline, workers also appeared to be misinformed about the salary distribution in other factories. The average worker was also overly confident about their factory’s relative pay at baseline, reporting an average score of 4.06. There is again a small overall effect of the scorecard treatment—a downward revision of 0.072 points—but a much larger effect on those whose factories scored below the median for working conditions.
How might workers respond to this new information? In the scorecard arm, workers’ newly more accurate beliefs did not translate into more mobility or better wages and working conditions. It appears the scorecard did not provide them with actionable information.
By contrast, workers in the vacancy arm were more likely to move, leading ultimately to wages that are 7-11% higher than those of the control group, without compromising working conditions. It appears that having a better sense of alternative options did allow them to find better jobs.
Interestingly, workers who received both the scorecard and vacancies pamphlet seemed to suffer from information overload; this joint treatment was less effective at correcting their beliefs than either treatment individually. This finding provides the first policy implication of our project, which is to be mindful of how much information participants can absorb, even if careful attention is paid to the design of information products.
In an environment where many workers lack information about jobs, we document that providing information about job vacancies can help workers’ careers. There is, therefore, a rationale for policymakers to facilitate the production and spread of this information to workers.
Because workers whose factories are below the median in terms of working conditions had much greater changes in beliefs, we focus on their behaviour here.
This project experimentally investigates which information and search frictions in Bangladesh’s labor market contribute to inefficient matching between workers and firms. Specifically, we test whether experimentally providing information about job characteristics (wages and working conditions), job openings, or both, will affect outcomes that are important indicators of productivity, such as the wages and working conditions they face and their mobility.
This blog is funded by the WEE-Connect initiative in support of the study “Using Text Messages to Provide Garment Workers Access to Job Information”.
