Friday, May 24, 2013

experimental results are artifacts of experiments...

A lot of economists hate dictator games. There are two main classes of reasons for this: first, people allegedly behave differently in the dictator game than they do in the real world. Second, (the more academic critique), it's not useful scientifically because too many factors go into the dictator game decision so it's not clear what the game measures.

The second critique can be rephrased positively: the dictator game is incredibly useful for measuring a wide array of effects, by comparing the results between two closely related variants of the game. It's wonderfully simple and flexible.

The same response also addresses the first complaint. Of course dictator games aren't straightforward to compare to the real world. Many factors go into the decision, and changing the setting to something like this changes a large number of them. The two settings are so different it's not even interesting to compare them anymore.

So yes, all dictator game results are artifacts of the game setting. ALL experimental results are "artifacts" of the experimental setting. That doesn't mean it isn't useful as a measurement device (which I give the author of the link above a lot of credit for bringing up), nor does it mean it isn't useful for understanding the real world, you just have to be more nuanced about your analogizing.

(In particular, in this case, expectations about behavior, framing of the question, and differing claims to the property, I suspect are the main things that differ between a standard laboratory dictator game and giving someone some casino chips and suggesting he might like to share with another guy. The defining attribute of a dictator game is not just that the money in question didn't come from the dictator's previously earned income. In the lab, both people show up expecting to earn some money, so if you find yourself in the position of allocating that money, of course you'll think it's fair to share, and of course you'll think that your partner will expect you to share and to resent you for making his trouble of showing up not worthwhile if you don't, and of course you think the experimenter is watching you, and of course you interpret the situation as a sharing game rather than a lucky random windfall that only affects you. Does anyone really think we needed an experiment to verify that people don't share randomly acquired money like they share money that they're instructed to allocate with another person who has come looking for it?? These results are not surprising nor newsworthy nor do they prove anything about the uselessness of laboratory dictator games...)

No comments: