In dynamic strategic interactions, a player who spies the opponent's actions might have incentives to feign ignorance and forgo immediate payoffs so that he can earn higher future payoffs by manipulating the opponent's suspicion. I model and experimentally implement the situation as a two-stage hide-and-seek game. A substantial share of the spying players fails to feign ignorance, despite the empirical suboptimality of the behavior and their largely correct predictions about opponents' suspicion. Subjects are highly heterogeneous in their tendency to feign ignorance and show only moderate learning. The players who are spied on hold empirically correct beliefs and mostly best-respond.

In legislative bargaining, the proposer is often able to extract a greater proportion of the surplus. However, a higher likelihood of being selected as the proposer can backfire, as it may reduce the probability that the agent is included in a winning coalition. We experimentally test the theoretical prediction of potentially negative returns to recognition probability in two-period legislative bargaining noted in Baron & Ferejohn (1989). We find that higher recognition probability benefits subjects in all treatments, except one in which we automate the second period. It is because proposers often favor the member with the greater recognition probability as a coalition partner, and such tendency varies depending on the proposer’s recognition probability, counter to the theoretical prediction. In all treatments, a vast majority of subjects exhibit a strict preference for higher recognition probability.

Observing others' actions can resolve strategic risk but may also result in second-mover disadvantage and inefficiencies. For example, a high level of transparency in workplaces may enhance operational control but stifle innovation. I theoretically analyze 2-by-2 cyclical games in which a player(manager) selects a probability of monitoring the other(employee)'s action. Employees know the probability but do not know whether their actions are observed. The model predicts the conditions for Transparency Paradox (Bernstein, 2012) in which an intermediate probability of monitoring is optimal for the manager and for the firm's efficiency. If innovation exposes employees to greater potential exploitation by the manager, managers must monitor and exploit with smaller probabilities. However, the laboratory experiment finds that the subjects in the role of manager persistently over-monitor under such a condition, and even those who do not over-monitor tend to over-exploit. In best response, employees refrain from innovation irrespective of the monitoring probability, which incentivizes over-monitoring by managers. Under over-monitoring and over-exploitation, employees bear most of the efficiency loss. The findings highlight the incentive conditions and dynamic processes that make firms and organizations particularly vulnerable to invasive surveillance/monitoring practices that lead to inefficiency and inequality.

In economic and social relationships, such as employment and marriages, participants often have the option to separate from their partner. This study experimentally investigates how the option to separate with or without a cost affects cooperation in indefinitely repeated prisoner’s dilemma game. For costly separation, I consider the cost of establishing a new match as well as the opportunity cost of missed interactions due to delays in matching. The theoretical model assumes the coexistence of non-strategic defectors and strategic conditional cooperators and predicts the long-run cooperation rate to which the population converges as players learn the initially unknown type distribution through experience. The experimental data support the theoretical prediction that compared to no separation, costless separation lowers the cooperation rate in the long run. The low cooperation rate and positive assortment through separation from defectors result in greater equity within matches. Costly separation increases the cooperation rate compared to no separation, yet the higher cooperation rate improves efficiency only under the high continuation probability. 

Previous experiments on delegated decision making find seemingly contradictory results: some experiments find that people take greater risks when they decide for others than for themselves, while other experiments find the opposite. I hypothesize that these mixed conclusions are the result of the type and the timing of feedback provided to subjects and conduct an experiment to identify these causes. In a choice between two binary lotteries, subjects either learn the outcome of only the chosen lottery or the outcome for both the chosen and the unchosen lottery. Feedback is provided immediately after each decision or after a sequence of ten decisions. I find that when subjects receive an immediate feedback on the outcome of both the chosen and the unchosen options, they make a risky shift. That is, subjects take greater risks for others than for themselves. If I alter either the timing or the content of feedback, the risky shift disappears. If I alter both the timing and the content of feedback so that the feedback is given at the end, only on the outcome of the chosen option, I find a risky shift again. These findings reconcile the contradictory results in the previous studies. I present a theoretical model and analyze how subjects' risk-taking behavior evolves as they make more decisions.

Recent studies have shown that consumers’ general lack of knowledge and misperception of the climate impact of their food choices prevent them from realizing the great potential for mitigating greenhouse gas (GHG) emissions in our food system. We propose a randomized controlled trial experiment in the field that measures the effect of climate impact information on restaurant consumer food choices. In partnership with local restaurants and technology startup that provides carbon footprint information for restaurant and grocery food items, we design an information intervention that is automated, scalable, and broadly applicable to various food environments. In addition to estimating the causal effect of the intervention, our consumer survey allows us to link their behavior to their demographic characteristics, food-related knowledge, beliefs, and attitudes to better understand the heterogeneity and identify the channels through which the intervention affects behavior.

We investigate the determinants of public support for a meat tax. Using survey data, we first estimate the distributional effects of a meat tax under several revenue-recycling scenarios. Then we conduct a survey experiment: In addition to eliciting their demographic characteristics, political position, and beliefs regarding the health, environmental, financial, and distributive effect of this policy, we also use a set of randomized information treatments to obtain exogenous variations in these beliefs and examine their causal effect on the support for a meat tax.