http://www.stat.columbia.edu/~cook/movabletype/archives/2008/12/risk_aversion_a.html
tl;dr: There is something more to risk aversion than concave utility functions, yet many economists pretend not to see the issue.
I agree with him. I think this concavity + expected utility theory thing is a clever trick, but it only *happens* to generate some risk aversion. It neither makes it a necessary nor a sufficient explanation for the risk aversion we observe in the real world.
What do you think?
It's pretty clear that the EU hypothesis does have some flaws as a positive theory of behavior... but economics as a discipline is certainly already aware of that. That said, many of the experiments that "prove" hyperbolic discounting don't strike me as particularly credible. You have to build in the fact that the people running the experiment are not omnipotent, and that subjects believe there is a nonnegative probability they will not get the money they were promised (at least not without some work), a probability that changes with the date at which the money is supposed to come. It's easy to imagine that this risk dominates pure time preference at the horizons being discussed.
Besides the Rabin stuff, there is the Gilboa-Schmeidler stuff on probabilities not adding to one, plus there is a broader array of behavioral alternatives than just prospect theory (which arguably dates to the late 40s with Friedman and Savage) and hyperbolic discounting, which does seem to be a real phenomenon, one of those elucidated helpfully by neuroeconomics. The very short term is driven by a different, and more primitive, part of the brain than the longer term. It is all about "fight or flight" and "**** her/him now to reproduce," hard wired in evolutionarily, even if "irrational" in a modern world with functioning intertemporal financial markets that did not exist a million years ago plus when the wires were getting hardened in.
Unfortunately the calibration critique applies to all major models of decision making under risk. But Rabins calibration result relies on an empirically invalid hypothesis, namely, people are CARA over a relatively large wealth range.
Btw, if Rabin gets the Swedish bank prize it will be for his work on fairness not this calibration thing - an idea someone else had investigated before ( see footnote on Rabins paper).
Calibration theorem is what Rabin is going to get his Nobel prize for. Not all the subsequent nonsense.
----------
Subsequent nonsense?!? Its ****ing arrogant bastards like you who have produced squat and never will be 1% of the researcher Rabin is that makes me sometimes hate being an academic economist.
Rabin does NOT assume cara utility. The theorem holds for any globally concave utility function. Results before him relied on specific functional forms such as CARA.
-------
It does. Read the paper you idiot. Which u(.) is consistent with people rejecting a given risk at all wealth levels between [w1,w2] (0<w1<<w2]?
CARA sarisfies it but other utility functions satisfy it as well.
And by the way, the previous person who had proved a similar result was Samuelson, although he interpreted it differently.
On top of that, Safra and Segal have recently generalized Rabin's calibration theorem to most theories of decision making under undertainty as long as they are defined over final wealth. This includes virtually all ambiguity averse models but excludes prospect theory.
It does. Read the paper you idiot. Which u(.) is consistent with people rejecting a given risk at all wealth levels between [w1,w2] (0<w1<<w2]?
b00e is an idiot. Not only did he not understand Rabin's result but he insults people who did. His claim is obviously wrong.
Take for example a log utility (which displays decreasing risk aversion), and consider the lotteries that Rabin proposed: winning 110 or losing 100 with equal probabilities. Then, a simple calculation shows that the decision maker will reject these lotteries for all wealth levels between w1 = 100 and w2=1100.
^ he is not. You cannot generate implausible levels of risk aversion - the ones in which one rejects a 50/50 chance of losing 100 or gaining an infinite amount - with a u(.) that is not strictly concave over this range of wealth. Rabins most striking results require small stakes risk aversion at wealth levels up to 300.000
^ Of course! But this is exactly Rabin's point. You cannot have at the same time small stakes risk aversion (for relatively low wealth levels) and moderate degrees of risk aversion for larger lotteries. Unlike you have claimed, his results aren't based on assumptions about the utility function being CARA.
The point is that either people accept these small gambles and we shouldn't use expected utility in the context of small lotteries, or expected utility theory is not a good descriptive theory (at least for lotteries involving small stakes).