Daily Speculations

Forum: The Promiscuous Hypothesis (September 2003)

Victor Niederhoffer: Kahneman and Thaler are known quite justly, IMHO, at the Univiersity of Chicago, as men of promiscuous hypotheses. They always got an ad hoc explanation for any findings. I find their contrived experiments with college students quite unconvincing of irrationality in the real world, as there is a big difference between how you solve a decision-making problem under uncertainty in the real world with a backdrop of numerous other considerations and how you solve it for credit in a psych class for a buck or two.

Adi Schnytzer: Experimental economics does have followers and not only one Nobel Prize winner in Economics, but two. Furthermore, experimental papers are published in all the top journals. Maybe it has a point or two to make.

Ross Miller: All of the original Kahneman and Tversky experiments had neither monetary nor (course) grade payoffs. Absolutely nothing was at stake. Furthermore, K&T anomaly probabilities only consider subjects who make choices in the early stages of the experiment that lend themselves to anomalies. For example, the first choice in the experiment might be between Payoff A and Payoff B. Half the subjects may choose Payoff A and the other half Payoff B, but those choosing A are uninteresting because no anomaly is possible in this situation. Then, the half choosing B are given a choice between Payoffs C and D, where the B/D combination is the anomaly. 60% of the B choosers may choose D, and so K&T report the anomaly percentage as 60%; however, only 30% of _all_ subjects actually fell into the anomaly category (B/D) and that percentage is not reported. While some proportion of the Payoff A choosers may well be prone to that anomaly, the experimenters may lead the unwary reader to the conclusion that this percentage is also 60% by only reporting this number. It would have been nice if they had offered Payoff A individuals a different set of alternatives that might elicit the anomaly rather than eliminate them from the sample. The bottom line is that most Kahneman-Tversky type experiments can only concretely demonstrate that a minority of their subjects exhibit anomalous behavior.