Worker satisfaction depends critically on how expectations are set by firms. Our study revealed that migrant workers experienced large losses in subjective wellbeing when randomized improvements in hostel living conditions were more modest than expected. How does housing quality affect the subjective wellbeing and workplace outcomes of migrant garment workers?
Improving living conditions may increase subjective wellbeing. However, this is not always the case. For instance, increases in subjective wellbeing can erode over time. Furthermore, subjective wellbeing is often determined by expectation-based reference points. It is unclear whether improved living conditions, without the right expectations, can indeed improve worker satisfaction, and in turn, workplace outcomes.
We used a randomized controlled trial to investigate how improved living conditions in hostels affect migrant workers’ satisfaction and turnover. The hostels we studied were employer-managed. In two phases, hostel management was transferred to a local NGO specializing in worker welfare with specific experience managing migrant worker hostels. Hostels were randomized into either phase 1 or phase 2 of the transfer process. There was a gap of approximately five months between phases, during which phase 1 hostels were under the new (NGO) management and phase 2 hostels were still being managed by the employer. At the end of this five-month gap, we surveyed a random sample of workers from all hostels to study differences in living conditions and the subjective wellbeing of workers generated by the change in management. The firm’s administrative data also allowed us to track retention of migrant workers.
Survey enumerators’ blinded evaluations of the hostels found that treatment improved living conditions in several key dimensions (particularly related to cleanliness and safety).
Workers reported being less satisfied with their living situation, their job, and their salary, and reported substantial decreases in subjective wellbeing (measured via Cantril’s Ladder and Kessler’s depression-anxiety scale) as a result of treatment.
Impacts on worker turnover, measured in the firm’s administrative data, echo the above pattern of results. There was an initial increase in retention in the first month of treatment, which quickly disappeared and gave way to (imprecisely estimated) negative impacts for the remainder of the study period.
A second sample of “joiners” received the same treatment related to improved living conditions as did the original sample, but were not exposed to the expectation manipulation that occurred in the lead-up to the phase 1 transfer. These workers experienced higher satisfaction and weak increases in subjective wellbeing as a result of treatment.
Our results show that when the actual improvements in living conditions have not measured up to workers’ high expectations, they can cause declines in satisfaction and increases in worker separation. These results are important for policymakers in low-income country contexts, emphasizing the crucial role that properly setting expectations – and implementing policy that lives up to those expectations – can play in determining the success or failure of policies. The political economy of policy making often necessitates that the potential benefits of proposed policies be widely disseminated, and the potential costs hidden, so that policies are most effectively “sold” to the public and its elected representatives. Our work points out that doing this comes at an inherent cost: the more a policy is oversold, the less likely it is that its effects will live up to expectations. If the gap between expectations and reality is large enough, even objectively successful programs may fall prey to reference dependence, and subjective wellbeing may decline.
This does not necessarily imply that policymakers will benefit from setting expectations low. If gains and losses relative to a reference point result in asymmetric changes in utility, it is likely that setting expectations extremely low would have only modest returns in terms of impacts on subjective wellbeing. Benchmarking expectations to the most likely policy outcome (with perhaps, at most, a slight undersell) could be roughly optimal in a world with implementation uncertainty.
Image credits: Nayantara Parikh
Subscribe to the Good Business Lab Mailing List to receive updates on our current initiatives, progress reports, upcoming projects and our quarterly newsletter.