Thursday, October 15, 2009

Experimental economics

I always thought that results from experimental economics did not generalize and now I find that this is true. From Tim Harford:

There is the “ultimatum” game, in which player A (Anna) is given $10 and asked how much, if any, she proposes to offer to player B (Bernard). Bernard can accept the offer, but if he rejects it, neither Anna nor Bernard get anything. If Anna and Bernard were rational income-maximisers, Anna would offer one cent and Bernard would accept it as better than nothing. This never happens, so Anna and Bernard are not rational income-maximisers.

Then there is the “dictator” game, introduced by Jack Knetsch, the Nobel laureate Daniel Kahneman and Richard Thaler, co-author of Nudge and perhaps the world’s leading behavioural economist. In the “dictator” game, Anna divides the $10 as before, but Bernard cannot reject her offer, so Anna can’t lose. Nevertheless, Anna will often throw Bernard two or three dollars. A third game, “gift exchange”, begins with Bernard offering Anna a payment. Anna then decides how to respond – effectively, an initial peace offering followed by “dictator”.

The results are astonishingly consistent: these games seem to demonstrate a taste for fairness. People offer more than they have to, reject unequal offers and reciprocate generosity. This has been a thorn in the side of conventional economics for more than 20 years.

List’s contribution ... has been to show that these results stem from the experimental set-up. In one set of experiments, he gently varied the rules of “dictator”. Anna, in addition to dividing up the $10 between herself and Bernard, was given the option to take a further dollar from Bernard. This option should be irrelevant. Because most Annas offer money to Bernard, they should hardly be tempted to pick his pocket. But in fact, when offered the chance to take money, far fewer Annas decide to give Bernard anything and one in five actually took Bernard’s dollar. Another experiment showed that Anna’s willingness to take from Bernard was dramatically less if she thought Bernard had earned his money. As the experimenter, List found he could nudge his subjects into being generous or mean with small variants in the set-up.

No comments: