Research

When facing discrimination in the labor market, some members of disadvantaged minority groups decide to hide signals of their minority status. We combine a theory of strategic hiding and two field experiments with Roma applicants and majority employers in the Czech Republic to study an important unintended consequence of hiding: Selection into hiding may skew the pool of minority applicants that employers can observe, influencing their perceptions of the minority group and preferences for discrimination in future hiring decisions. We find that selection into hiding creates a vicious cycle: Roma applicants with better education and job experience are almost twice as likely to hide their Roma identity by omitting Roma signals in resumes and are more likely to speak without a Romani accent in interviews. Employers are overconfident about their ability to detect Roma applicants and unaware of selection into hiding. As a result, selection into hiding increases employers' negative stereotypes about Roma applicants, accounting for about 40% of the gap in beliefs about Roma and non-Roma ability in our experiment, and reduces employers' willingness to review minority applications in the future. 

Should policymakers intervene to reduce social segregation? The answer depends on the effect of increased contact among those who would not seek it out on their own. We conduct a field experiment in the context of political segregation in Brazil to measure the joint distribution of aversion to cross-partisan interactions and the treatment effects of actual interactions on outgroup hostility. To do so, we randomize participants into interactions offering incentives of twice the daily minimum wage for compliance. We find evidence of a segregation trap: the most hostile partisans avoid interactions because they incorrectly believe they will be unpleasant and unproductive. Actual interactions correct these beliefs and lead to large, long-lasting reductions in hostility. Importantly, these effects are largest among partisans with high levels of baseline hostility and aversion to outgroup interactions, and lead to an increase in demand for future interactions. In a cost-benefit analysis, we find that current policy interventions targeting volunteer participants are outperformed by incentive-based interventions at any level of budget. Finally, we show that a lower-cost intervention that substitutes a video for actual interactions can produce similarly beneficial effects.

Optimal behavior often requires planning future actions when choosing what to do today. We propose that individuals’ ability or willingness to plan ahead is limited, leading to systematic decision errors. In an online experiment, participants face sequential consumption–savings problems with two periods, or equivalent one-period versions with identical information, available actions, and payment timing. Participants are 18 p.p. less likely to spend optimally in the two-period decision problem and, on average, overspend early on. In treatments where subjects actively search for information, we show that timing shapes attention: in one-quarter of sequential problems, participants ignore future actions entirely, behaving like there is no tomorrow, and even those who look ahead prioritize information about the present.

We study exploding offers by considering the strategic interaction between a low-tier firm and a set of workers within a large job market. Each worker has a private value for the firm and may receive offers from preferred top-tier firms according to an exogenous stochastic process. We show that workers receive exploding offers whenever their probability of receiving a top-tier offer is relatively small with respect to their risk aversion, and that there is a level of risk aversion beyond which a worker only receives offers that expire as soon as possible---independently of what all other firms are doing. In markets in which exploding offers are prevalent and workers' and the firm's preferences satisfy a weak form of positive correlation, workers' expected welfare is maximized if and only if exploding offers are banned. However, any minimal offer length that falls short of banning exploding offers may lead to workers falling through the cracks - competitive workers are not approached by the strategic firm but end up receiving no top offer. 

Expectations are a critical factor in determining actions in a great variety of economic interactions. Hence, being able to measure beliefs is important in many economic environments. In this article, we review the approaches that have been used to measure beliefs and make comparisons of their effectiveness. We also discuss belief elicitation when the truth is not verifiable. We find that introspection (non-incentivized responses) seems to do as well as rather complex incentivized methods. We conjecture that simple and easy-to-comprehend incentivized methods are superior to introspection, in fact there are very few studies making such comparisons; this is an avenue for future research. We also discuss ways in which relatively complex methods could be made easier for usage in experimental work.

How do people reason in complex economic environments involving direct, indirect and feedback effects? We study mental shortcuts using an experiment, in which participants predict the outcome of a complex game where players take turns transferring money over the course of several periods. Participants look up individual transfers using a custom calculator. This allows us to precisely recover their mental model of the game by measuring which effects they consider and which they ignore. We find that there is only one mental shortcut that participants use consistently - considering direct effects only. The share of participants who use this shortcut is approximately 50% and is largely invariant to the complexity of the environment, consideration costs, and prediction incentives. Remaining participants either consider all effects or guess without thinking. Participants' search patterns strongly correlate with their self-reported use of mental shortcuts outside the lab.

How should managers address political and racial conflict in the workplace? We address this question by combining a field experiment embedded into a job training program and a survey experiment with firm managers in Brazil. We establish the causal effect of political and racial conflict on productivity and turnover by randomizing job training participants into either homogeneous or diverse teams. We then evaluate the impact of a cross-randomized deliberation intervention that is currently implemented in a multinational company. In the manager experiment, we survey firms’ policies to navigate political and racial conflict, and elicit incentivized beliefs about the impact of the deliberation intervention. We then randomize an information treatment that provides information on the productivity impact of the intervention. We measure the effect of correcting managers’ beliefs on their demand for the intervention.