Bryan Caplan Asks an Impossible Question
As usual, one of us econ geek bloggers asks a question, and then the other econ geek bloggers have to respond in ways that show how smart we are. I hope I am not misrepresenting him, but Bryan Caplan is basically saying, "Austrians claim you can't model the future with probability distributions. But I am still waiting for an Austrian to model the future with probability distributions to clarify their claim."
Tyler Cowen and Arnold Kling have the right instinct--namely, Caplan is being goofy--but they get into all sorts of sophisticated examples and analyses. No, the Kirznerian point is quite basic; you don't need to look at derivatives markets to see its implications.
(My own, less wise-alecky answer is in the comments on Bryan's original post. HT2 Pepe for originally alerting me to this debate.)
Tyler Cowen and Arnold Kling have the right instinct--namely, Caplan is being goofy--but they get into all sorts of sophisticated examples and analyses. No, the Kirznerian point is quite basic; you don't need to look at derivatives markets to see its implications.
(My own, less wise-alecky answer is in the comments on Bryan's original post. HT2 Pepe for originally alerting me to this debate.)
I can't post on Bryan_Caplan's blog anymore because I was banned, even though that happened two years ago, and I apologized several times. (Now why would anyone want to ban me? It boggles the mind. Hey, maybe my banning was an example of radical...)
ReplyDeleteOkay, anyway, a lot of the Austrians seem to be saying that radical uncertainty refers to events that were not assigned any probability at all, even implicitly. However, I think Bryan_Caplan adequately showed how mainstream (Bayesian) models incorporated this kind of event, though he didn't mention it in the post, and in my search of the comments I didn't find it.
He explained it in one of his QJAE responses to Walter Block. (mises.org doesn't make it easy to find...) Basically, the model would say that the agent assigns probability to all the knowns (summing to less than 1), and the remainder to "something I'm not aware of as a possibility".
So a better challenge would be, "how is this inaccurate?" or "how does this method fail to capture relevant insights?" Keep in mind that under this model, an agent needn't explicitly label the probabilities this way in his mind.
Silas,
ReplyDeleteHere are two Caplan replies from the QJAE (one and two, both pdfs btw). I'm not sure which one you meant; they both have probability and Block in the title...
I haven't read Caplan's argument--or if I did, I've forgotten it--but based on your summary, it seems goofy. You can't use a rational optimizing framework with open-ended outcomes like that.
Without being able to specify the particular scenarios, you can't assign a realized utility in that outcome. So you can't correctly optimize.
Now I imagine Caplan could come back and say, "Well OK, but then you just pick a best-guess of your expected utility in that vague area."
But why should that utility number be right, if you don't even know what could possibly happen to you in those circumstances? He is just begging the question.
You can't use the neoclassical approach if you don't have the correct probability distribution etc.
BTW, why do you include underscores between people's names? E.g. "Bryan_Caplan"?
OK, trying to mediate: I think what Caplan is protesting is that it is always theoretically possible to assign some (subjective) probability to the actualization of a member from the set {sh*t I've never even imagined occurring}. And I believe Kirzner's more important point is that such a probability estimate has just about nothing to do with any "objective probability" (if that notion really means anything in regards to unique events) that an agent will be entirely surprised by the outcome of his actions.
ReplyDeleteAnyone know how Nassim Taleb might respond to this dispute between Brian Caplan and the Austrians?
ReplyDeleteAraglin
Anyone know how Nassim Taleb might respond to this dispute between Brian Caplan and the Austrians?
ReplyDeleteIf I know Taleb, he's got puts on both sides, and is just waiting for one of them (or both) to really say something stupid. From his latest post, I think the Caplan contracts are almost in the money.
Ba-DUM
Bob_Murphy: I'll admit, I don't know what neoclassicals claim about the optimality of actions, given those probability distributions, so I'm not qualified to comment further about whether assigning x% to "other" affects the line of reasoning.
ReplyDeleteBut you can find Bryan_Caplan's point on pp. 72-73 of the first one:
Hoppe (1997) raises two other noteworthy objections to the
realism of probability theory. The first is that it implies that “it would be possible
to give an exhaustive classification of all possible actions. . . . For without
a complete enumeration of all possible types of actions there can be no
knowledge of their relative frequencies” (p. 56). There is a simple reply: One can always complete an enumeration by adding a catch-all “other” category. In
assigning probabilities to the party affiliation of the next president of the
United States, for example, one need merely break the possibilities down into
Democrat, Republican, and “other.”
BTW, why do you include underscores between people's names? E.g. "Bryan_Caplan"?
Would you be convinced if I said it was a software issue having to do with hot-keys and a convention of space-delimitation of entries, plus my assumption that people aren't actually offended by underscores like that?