We analyze the data sets of all majoritarian Baron and Ferejohn (Am Political Sci Rev 83(4):1181–1206, 1989) experiments through 2018. By exploiting the variation of the experimental parameters, we are able to identify how group size, discount factor (cost of agreement delay), voting weights, and communication affect bargaining outcomes and dynamics. The outcomes are qualitatively in line with the stationary subgame perfect equilibrium, i.e., minimum winning coalitions are modal; proposers demand larger shares than non-proposers; and most agreements are reached without delay. Experience and communication between players move outcomes closer to the equilibrium. However, bargaining dynamics are not stationary. Behavior following a disagreement is history-dependent in the form of retaliation towards failed proposers and their supporters, which, if rationally expected, may deter proposers from demanding high shares.

Many situations involve trading-off what is safe for the individual and what is beneficial for the group. This tension is extensively studied as the “Stag Hunt” coordination game. By combining a new theoretical approach with two experiments, this paper shows a disconnection between behavior in Stag Hunt games and the predictions of many models. Any Stag Hunt game can be uniquely decomposed into three payoff components: strategic, behavioral, and kernel. The behavioral component exists in every Stag Hunt game despite being largely ignored by previous models. Arbitrarily many Stag Hunt games exist where these models predict the same behavior. Despite the constant predictions, behavior in repeated and one-shot games systematically vary. The mechanism underlying these results is that all models of Stag Hunt behavior fail to properly account for the influence of the behavioral component.

In this paper, we use experimental data to study players’ stability in normal-form games where subjects have to report beliefs and choose actions. Subjects saw each of 12 games four times in a regular or isomorphic form spread over two days without feedback. We document a high degree of stability within the same (strategically equivalent) game, although time and changes in the presentation of the game do lead to less stability. To look at stability across different games, we adopt the level-k theory, and show that stability of both beliefs and actions is significantly lower. Finally, we estimate a structural model in which players either apply a consistent level of reasoning across strategically different games, or reasoning levels change from game to game. Our results show that approximately 23% of subjects apply a consistent level of reasoning across the 12 games, but that they assign a low level of sophistication to their opponent. The remaining 77% apply different levels of reasoning to different games. We show that this may be due to subjects being attracted to the action with the highest possible payoff.

The paper examines whether an institution has a differing impact on cooperation if it is introduced by a representative of the affected subjects rather than exogenously imposed. The experimental design controls for selection effects arising from the endogenous policy choice. The treatment varies whether the decision-maker is elected or randomly appointed. There is evidence of a large democracy premium in the sense that endogenously chosen institutions lead to more cooperation than identical exogenous institutions, but only if the group leader is democratically chosen. Especially the subjects who initially did not prefer the policy are more likely to cooperate if it was brought about by an elected representative. There is no democracy premium for randomly appointed group leaders.

Economists model self-control problems through time-inconsistent preferences. Empirical tests of these preferences largely rely on experimental elicitation using monetary rewards, with several recent studies failing to find present bias for money. In this paper, we compare estimates of present bias for money with estimates for healthy and unhealthy foods. In a within-subjects longitudinal experiment with 697 low-income Chinese high school students, we find strong present bias for both money and food, and that individual measures of present bias are moderately correlated across reward types. Our experimental measures of time preferences over both money and foods predict field behaviors including alcohol consumption and academic performance.

We design a novel experiment to study how subjects update their beliefs about the beliefs of others. Three players receive sequential signals about an unknown state of the world. Player 1 reports her beliefs about the state; Player 2 simultaneously reports her beliefs about the beliefs of Player 1; Player 3 simultaneously reports her beliefs about the beliefs of Player 2. We say that beliefs exhibit higher-order learning if the beliefs of Player k about the beliefs of Player $k-1$ become more accurate as more signals are observed. We find that some of the predicted dynamics of higher-order beliefs are reflected in the data; in particular, higher-order beliefs are updated more slowly with private than public information. However, higher-order learning fails even after a large number of signals is observed. We argue that this result is driven by base-rate neglect, heterogeneity in updating processes, and subjects’ failure to correctly take learning rules of others into account.

We exploit testing data to gain better understanding on framing effects on decision-making and performance under risk. In a randomized field experiment, we modified the framing of scoring rules for penalized multiple-choice tests. In penalized multiple-choice tests, right answers are typically framed as gains while wrong answers are framed as losses (Mixed-framing). In the Loss-framing proposed, both non-responses and wrong answers are presented in a loss domain. According to our theoretical model, we expect the change in the framing to decrease students’ non-response and to increase students’ performance. Under the Loss-framing, students’ non-response reduces by a 18%-20%. However, it fails to increase students’ scores. Indeed, our results support the possibility of impaired performance in the Loss-framing.