Research

Working Papers

How Much Can You Make? Misprediction and Biased Memory in Gig Jobs [Draft]

Abstract: Flexibility is an increasingly prominent feature of many jobs. In the gig economy, workers can choose their work hours and face wages that vary across hours and weeks. This increased complexity adds challenges to predicting and understanding job outcomes. Incomplete information or behavioral biases can then lead to inaccurate beliefs about pay and labor supply. We test this hypothesis by collecting novel survey data on 454 delivery and ride share gig workers in the United States. Comparing gig workers’ beliefs with data on their actual job performance, we find they overestimate their predictions (43%) and their recalls (31%) of weekly pay, despite it being reported prominently in their earnings statements. Furthermore, gig workers underestimate expenses and overestimate hours worked. The results are consistent with selective recall: when forming and updating their beliefs in noisy environments, workers overweight past high-paying periods. We then examine how biased beliefs affect labor market decisions. We derive predictions from a behavioral labor supply model and test them using survey data and a randomized de-biasing intervention. We find that job choices and labor supply decisions are significantly affected by mistaken beliefs in flexible gig jobs.

Can Laws Change Social Norms? Evidence from Domestic Workers in Brazil [Draft]

Abstract: We examine a 2013 constitutional amendment in Brazil that ensured equal labor rights for domestic workers. The amendment did not include any enforcement mechanism and increased the costs of formalizing domestic workers. However, the majority of the law was not enforceable for over two years. At the same time, the amendment was accompanied by intense media coverage and public interest. Our hypothesis is that this changed social norms: employing an informal domestic worker became perceived more negatively by both employers and employees. Based on an event study framework, we find that the formalization rate of domestic workers increased by between 2 and 4 percentage points after the amendment. Further, more domestic workers started being paid the minimum wage. To test whether changing social norms is a likely mechanism for our findings, we estimate a series of triple difference estimators, using variation in factors likely to amplify the effect of changing social norms. We first find that domestic workers that live with their employer formalized at much higher rates during this period. In addition, domestic workers have larger formality treatment effects when surrounded by more domestic workers and by more formalization events in their neighborhood.

Selected Work in Progress

Does the Structure of the Hiring Process Impact Gender and Race Gaps? Evidence from Brazil (with Priscila De Oliveira)

Abstract: We explore how the structure of the screening and hiring process can influence the gender and race wage gaps. We use proprietary data from a HR Tech company that assists large companies with worker recruitment in Brazil. Our dataset contains detailed information on candidates' performance at all stages of the screening process. The first question that we aim to address is: at which stages of the process do gender and race gaps emerge, and when are they the largest? The next step is to compare screening strategies across firms and how they affect the gender and race gaps. For instance, does requiring group interviews result in larger or smaller gaps in job offers? Are hiring decisions affected by the number of people involved? We will also compare firms where HR plays a larger role in recruitment with firms where area managers make the majority of decisions. Finally, we will study whether interviewers might be biased towards candidates more similar to themselves, and whether gender and race gaps vary depending on the recruiters' identities.

Finding the Perfect Hire: Screening Strategies and Job Match Quality (with Priscila de Oliveira)

Abstract: This paper examines how different screening criteria affect work performance. We use proprietary data from a HR Tech company that assists large companies with worker recruitment in Brazil. We will combine this dataset with post-hiring job performance data. Our goal will be to estimate the implicit weights that firms apply to different worker characteristics when they make hiring decisions. We analyze the consequences of these choices in terms of productivity and diversity. The next step is to identify screening strategies that can maximize the quality of the employer-employee match. Relative to this benchmark, firms may focus too much on certain applicant characteristics that are not very predictive of performance. If this hypothesis is confirmed, it may motivate changes to screening strategies. Finally, we will study when suitable matches are unlikely to be selected during the recruitment process.