Sunday, May 31, 2009

self reported vs revealed preferences

So often I read in economics papers skepticism towards self-reported preferences, or willingness to pay. I admit, it's pretty unreliable, but a good starting point. But so often, revealed preferences are just as fishy sounding to me, and I wonder why that is not focused on more.

For example, there is an assumed monetary value of time depending on your income. That is, if you spend a certain amount of time doing something, the value of that time is your revealed willingness to pay. Does this not sound ridiculous to you? Most people work for some fixed amount of money. The marginal hour of work that I do will not affect my paycheck in the least. Even IF I worked for an hourly wage of $15, I can't just decide willy nilly to work an extra hour and get an extra $15. The number of hours along with the wage is part of the agreement with my employer. So saying that if I go out of my way by half an hour to avoid some consequence, that consequence is worth at least $7.50 to me to avoid, is nonsense. If I weren't avoiding a consequence with that half hour, I'd probably be sitting at home reading or playing with my cats or something else equally non-money-earning.

Then there's just the fact that people misestimate the consequences of their actions. For example, if the probability of dying in a motorcycle accident is .5 if I don't wear a helmet, and .25 if I do, and a helmet is available for $100, and I don't wear one, does not mean that I value my life by at most $400, my "revealed" preference. I'm just underestimating the probability of dying, or overestimating my abilities as compared to the general population of drivers from which those two statistics were calculated. In this case you're much better off asking me how much I'd be willing to pay to guarantee that I won't die in a motorcycle accident, even as ridiculous and hypothetical that question is.

I think good research should attempt to use both measures, and discrepancies in answers does not automatically mean that one is less reliable than the other. We should be trying to figure out what biases our statements and actions in general, and use that to determine which measure is more accurate.


shanedicks said...

Kinda seems like you're challenging the rational actor assumption.

Self-reported preferences are biased by a sense of a rational self, one that is continuous and whose actions must be justified in a coherent narrative. Revealed preferences seem to be biased by the real life irrationality and incoherence of that same self.

Not sure what the implications of this, other than that I agree that its silly to argue over which preferences are more useful.

Vera said...

Yeah I think the gut trust of revealed over stated preferences is due to habitual belief in the rational actor assumption, even among behavioral economists who spend their lives refining that assumption. Most of the time that may be true but glossing over the debate ignores some really interesting questions.

somebody said...