Framing inuences choice. However, little is known about the underlying mechanisms behind framing eects. We study gain-loss framing in binary modied dictator games. Subjects choose the selfish option more often in the loss frame compared to the gain frame. Recording visual xations with eye-tracking, we nd that dictators focus more on their own outcomes when facing losses. This suggests that losses to the own outcome are weighted more than losses to another player. * Financial support by the Max Planck Society is greatly acknowledged.

There is ample evidence that the number of players can have an important impact on the cooperation and coordination behavior of people facing social dilemmas. With extremely few exceptions, the literature on cooperation assumes common knowledge about who is a player and how many players are involved in a certain situation. In this paper, we argue that this assumption is overly restrictive, and not even very common in realworld cooperation problems. We show theoretically and experimentally that uncertainty about the number of players in a Volunteer's Dilemma increases cooperation compared to a situation with a certain number of players. We identify additional behavioral mechanisms amplifying and impairing the eect.

Leadership mechanisms provide a potential means to mitigate social dilemmas, but empirical evidence on the success of such mechanisms is mixed. In this paper, we explore the institutional frame as a relevant factor for the effectiveness of leadership. We compare subjects' behavior in public-goods experiments that are either framed positively (give-some game) or negatively (take-some game). We observe that leader and follower decisions are sensitive to the institutional frame. Leaders contribute less in the take-some game, and the correlation between leaders' and followers' contribution is weaker in the take-some game. Additionally, using a strategy method to elicit followers' reactions at the individual level, we find evidence for the malleability of followers' revealed cooperation types. Taken together, the leadership institution is found to be less efficient in the take-than in the give-frame, both in games that are played only once and repeatedly.