Today's post is the second in the series of three on articles by John List, specifically List (2007) 'On The Interpretation of Giving in Dictator Games' from the Journal of Political Economy.

Results from the Dictator Game - a typical experiment used in behavioural economics - are thought to justify the existence of (pro)social preferences. John List's contribution in this paper is to propose (in line with Bardsley, 2005) that the dictator game should not only involve giving some amount to one's partner in the game, but also taking away. Why? Actions in the real world involve both positive and negative, or give-take interactions. Adding this variation to the game could give us greater insight into the existence (or lack thereof) of social preferences and the extent to which individuals act in the mode of homo economicus.

Models that support prosocial preferences often fall into three classes:

1) reciprocity models (Rabin, 1993)

2) inequity aversion (Fehr & Schmidt, 1999)

3) altruism and spite (Levine, 1998)

In most dictator game experiments it is customary for 60%+ of individuals to pass some positive amount of money to their partner. Mean transfers tend to be around 20% of the total. Social preferences are typically used to explain this.

List adds two different treatments to the original game, along with a third to test differences between perceptions of deservingess vs. windfall cash.

1) Give up to $5, but also able to take away up to $1.

2) Give and take up to $5.

3) Have the subjects do basic administrative work for 30 minutes and then play treatment 2.

The inclusion of any possibility to take money from the partner results in drastic changes: "many fewer agents are willing to give money when the action set includes taking" (484). Moreover, "by allowing choices that are not entirely selfish in the nonpositive doman, the social norms of the game change, providing the dictator with the "moral authority" to givenothing" (484). But, the dictator "now has many choices that signal she is not entirely selfish" (485). What is meant by this is that, although they might no longer give positive amounts, the outcome predicted by economics (taking the maximum amount) does not occur.

To report: adding the 'take $1' option shifts the median offer to $0 dollars, and a mean offer of $0.33. For the symmetrical (give-take up to $5) the median offer is -4.50 dollars (i.e. the median individual takes away $4.50 from their opponent), the mean is -$2.48. Hence, when one "makes the action set symmetric, nearly all giving vanishes" (487). But still, the 'most selfish' behaviour is avoided. 'Taking' is much lower with the 'earnings' treatment indicating a difference in beliefs about taking 'windfall' vs. 'earned' money. This indicates an underlying moral structure with respect to interpreting gains and losses (see Levitt and List, 2007).

Importanly, List argues that "the data suggest that current interpretations of dictator game data likely need revision. Rather than represeting social preferences as currently modeled in the oft-cited literature, the data are consistent with the power of changing the giver and recipient expectation" (490). This, furthermore, highlights the importance of generalizability, or the fact that it is very difficult to generalize laboratory results to the 'real world'. List argues that we should recognize the specific institutions structures in which the activities occur and how these gain salience in different situations. We must manipulate experiments to emphasise the different and contrasting contexts to which individuals adhere, or which frame individual action. Using such manipulations will, hopefully, give us greater insight into the generalizability of laboratory results.

My comments

Something I would like to know is whether adding the new actions would alter individual behaviour in the original game, i.e. play 10 rounds adjusted game, then play 10 rounds 'original' dictator game. Here we would be able to interpret whether it is a state-based difference, or whether it would bring about a shift in preferences (endogenous preference formation). If we see endogenous preference formation, then we could assume that individuals are learning about the game and their own responses to different actions as a consequence of playing different versions. This would be an intriguing possibility and would be an interesting path for future research. The reason I think this would be good is because you could get interesting within subject variation, whereas List only looks at between subject variation.

In addition, it would be fascinating to observe subject interactions (as Levitt and List, 2007 report) of the experimental subjects after the experiments, in a field experiment to test, again in order to detect the state-based or endogenous nature of the preferences. Ultimately, the pairing up of field and laboratory experiments is crucial in order for us to gain greater insight into the relevance of social preferences in human decision-making.

References