May 01, 2003

A little knowledge is a dangerous thing

Megan McArdle has a lengthy - and rather silly - post on economics as a fortress built upon the solid foundations of the Real Scientific Mindset, as opposed to political science and sociology, which are shoddily built outposts of the humanities. In so doing, she shows us yet again that Economics 101 and "Atlas Shrugged" is a dangerous combination. Not that I'm opposed to economic methodology - far from it - unlike Chris Lawrence I'm one of those political scientists who uses economic theory quite a lot. But I'm continually surprised at how often people who don't necessarily know all that much about economics tend to overestimate its "scientific" validity, especially if their politics are libertarian.

Daniel Davies, as usual, gets it right in the comments section for McArdle's post - the kinds of economic theory that McArdle etc are interested in are partial equilibrium models, whose general scientific validity is precisely diddly-squat. Unfortunately, McArdle doesn't seem to know what the difference between partial and general equilibrium is. Donald (now Deirdre - long story) McCloskey has a lovely little book called "The Rhetoric of Economics" which shows how dependent partial equilibrium models are on their initial assumptions - by jiggering these a little, you can get whatever result you want, more or less. Models of this sort are little more than "Just So" stories with nice differentials. While there is such a thing as general equilibrium analysis (the Arrow-Debreu model), that provides more general results, it relies on hopelessly unrealistic assumptions (as its co-creator, Ken Arrow cheerfully admits), and thus is perhaps more interesting as an abstract result in social choice theory, than as a specific contribution to our scientific knowledge of how markets work. Game theory - don't start me on game theory. A notorious little result called "the folk theorem" means that pretty well everything goes in the infinitely iterated games that are needed to model moderately complex social interactions - the best that economists and game theoretic social scientists can do is to show that whatever particular constellation of strategies that they're interested in is an equilibrium, happily ignoring the fact that there are umpteen billion other possible equilibria out there which are equally plausible from a game theoretical point of view. Game theorists have been engaged in the search for convincing equilibrium refinements that would get rid of this problem for decades; Ariel Rubinstein (a very distinguished game theorist) rather rudely dismisses this as analogous to the quest for the Holy Grail.

Not only does McArdle overestimate the scientific power of economics, but she underestimates the contribution of political science and sociology. James Joyner is right on the nail - the problem is that the subject matter of political science is vastly more complicated than that of economics. And it's not only Joyner who says this. The late Mancur Olson, who straddled the division between economics and political science rather nicely, talks of how politics involves goods that are "indivisible", and thus vastly more difficult to measure - politics is more complicated than economics because it doesn't involve goods that can be divided up into neat and discrete little bundles. Doug North, who has a Nobel Prize in economics says that an unthinking extension of economic theory to political science is a Very Bad Idea, precisely because politics is so much more complex. Instead, he advocates a transaction costs approach based on the assumptions of costly information, of subjective models on the part of actors to explain their environment and of imperfect enforcement of agreements.

As I said, there's something a little strange about the marriage of economics (particularly public choice theory) and libertarianism - Julian Sanchez described it quite well in a post that doesn't seem to be available anymore, where he talks about how public choice offers a kind of legitimating scientific ideology for libertarian ideas. And it is an ideology - and not all that much more than that. Which isn't to say that economics can't be used quite well to illustrate libertarian concepts - but it can be used to illustrate other views of society very nearly as easily, and with precisely as much "scientific" validity. Which is to say, none at all. Where economic and rational choice approaches to explaining human behaviour have their value is not in scientific validity in any general sense of the word, but rather in logical consistency. Constructing a game theoretic model forces you to be precise about what microfoundations you are employing, what preferences actors have over outcomes etc, and if it's done right, it allows you to be sure that the outcomes of the model are consistent with the premises. And that's it. Lousy arguments will still remain unconvincing, even if they're tarted up as sub-game perfect equilibria. Good arguments will still be good - for reasons that can't be reduced to a constellation of actors, strategies and outcomes. Partial equilibrium models and game theory provide us with a good language for describing certain kinds of social and economic relationships, and for testing the internal consistency of the arguments that we make about these relationships. They don't provide us with much of a scientific foundation beyond that.

Update: Kieran Healy fillets McArdle with savageness and style.

And Dan Drezner (temporarily compering with the Volokhs) physically winces at McArdle's post.

Posted by Henry at May 1, 2003 09:28 PM
Post a comment

Remember personal info?