Are preferences stable?

Whenever we model economic behavior, we assume preferences are given and stable. If we observe from empirics that preferences change, we attribute this to changing circumstances: with a different choice set, people take different decisions. If the econometrician cannot observe this choice set, he then concludes that preferences change.

This conclusion is generally thought to be wrong, and it is concluded that something else must have changed. But what if preferences change indeed? This is the question that Anderson, Harrison, Lau ad Rutström have tried to answer by conducting a field experiment in Denmark. They performed a standard lab experiment to elicit risk aversion measures by offering several choices of lotteries to a diverse segment of the Danish population. Up to 17 month later, they revisited their subjects, again measuring their risk aversion and obtaining information about changing economic circumstances.

They obtain changes, nicely normally distributed around zero, from raw data, that is, not taking into account changes in circumstances. Doing the latter, they find that a more positive outlook on personal finances leads to lower risk aversion. But the changes in the coefficient of relative risk aversion are statistically significant but modest, in the order of 0.20. So we are still not too wrong to assume constant preferences.