One of the most important questions in experimental economics is whether individual behavior in certain games is predictive for behavior in other games or contexts. Critics often claim that individual behavior in the lab is limited to that situation and does not contain much information about how people behave outside the lab. This question appears to be particularly relevant for 'other-regarding' or social preferences, i.e. preferences which are not captured by the standard assumption of purely payoff-maximizing agents and which have been largely investigated in the lab. The results of the experiments testing the stability of other-regarding preferences across multiple games or multiple contexts are ambiguous. The ambiguity may be partly explained by the ‘constructive-preference’ approach, which suggests that individuals construct preferences spontaneously when they are confronted with an unfamiliar decision situation. With more experience in a certain situation, preferences consolidate over time and become more stable. This paper sheds light on this issue by investigating and comparing charitable donations and dictator game allocations in an experimental setting for a better understanding of what drives the decision in each task and whether behavior in one task is predictive of behavior in the other. The donation decision is a familiar decision situation where individuals are likely to have existing preferences. In contrast, the dictator game, though very simple, represents a rather unfamiliar decision situation where subjects may construct their preferences ad hoc. In the experiment, we pay special attention to the sequence in which the games are played. Our experimental results show a significantly positive correlation between both decision tasks if (and only if) the more familiar donation decision is presented first and the rather unfamiliar dictator game is played thereafter. Moreover, the dictator game allocations depend on the sequence of games while the charitable donations do not. Hence, social preferences elicited in the donation context are predictive of subsequent behavior in the dictator game but not vice versa. Thus, if experimenters try to elicit social preferences to make predictions about behavior in other contexts, it seems reasonable to confront individuals with more familiar decision situations where preferences exist and do not need to be constructed. Furthermore, special attention should be paid to the sequence of games if they present a new decision situation for the experimental subjects.
Borgloh, Sarah, Astrid Dannenberg and Bodo Aretz (2010), On the Construction of Social Preferences in Lab Experiments, ZEW Discussion Paper No. 10-085, Mannheim. Download