I have a new column out in Bloomberg looking at some arguments by conservative economists against Thomas Piketty's work on inequality. I stumbled last week across this post by Tyler Cowen, which Barkley Rosser helpfully put into context. Cowen claimed that we don't really need Piketty because several earlier studies "already give an explanation" for the observed wealth inequality. Really? It turns out, he suggested, that you don't need any stories about returns to investment growing faster than wages. Standard economic models have already shown that inequality may just be the consequence of simple things like differences in personal patience (rich having more, of course, and the poor less), or in the effects of random shocks to peoples' ability to earn over their lifetimes.

Really?

Having looked into it, I now think this is a perfect example of Chameleon Economics, as recently described so brilliantly by Paul Pfleiderer. You tuck some preposterous assumptions A into a model, derive some apparently interesting result X, and then hope that people will soon forget about A so you can go around saying "we've shown that X" holds. The preposterous assumptions A might even include an assumption that is essentially equivalent to X, so you've assumed the result you want to prove. This trick is the real basis of the papers that Cowen pointed to, but jeez -- the authors did such a good job of plastering their arguments over with 50 odd pages of technical mumbo jumbo that it took quite a lot of effort to see what they were up to. In the paper by Krusell and Smith, for example, you can read on and on in utter semi-conscious misery before you begin to find the real secret of what the authors have done, as they finally admit:

When the representative-agent model is altered only by adding idiosyncratic, uninsurable risk, the resulting stationary wealth distribution is quite unrealistic: there are too few very poor agents, and much too little concentration of wealth among the very richest. For this reason, we consider a version of the model with preference heterogeneity: agents have random discount factors, whose values have a symmetric distribution with a small variance and whose transition probabilities are such that the average duration, or life length, of a discount factor equals that of a generation. In this fashion, we incorporate genetic differences in the population that are passed on imperfectly from parents to children. We show that this model does succeed quite well in matching the key features of the wealth distribution.

In other words, they start out seeking an explanation for the unequal distribution of wealth -- why do some people have so much more than others? Ultimately, they find that this result tumbles directly out of their economic model, IF they make the assumption that some people in the model are more patient than others, and are therefore better at saving and accumulating wealth than others. There you go -- the whole result from that one assumption (plus some others)! Science advances!

Anyway, how about the following for a funny coincidence. Courtesy of a kind invitation from Ole Peters, I'm spending May at the London Mathematical Laboratory, a small mathematics center in Central London. Last week we were discussing Piketty's book, which Alex Adamou, one of the researchers here, has been diligently working his way through. Ole boldly suggested that maybe we should try to get Piketty to come here and give a talk on the book, whereupon we all chuckled at the very idea, thinking it preposterous given the outrageous current demands on his time. Piketty seems to be on a worldwide tour of epic proportions.

Yet this afternoon we learned that, at the very moment of our discussion, Piketty was actually in the very same building, one floor above our heads, giving a talk to a public policy think tank! Had we been speaking a bit louder, he might have heard us!

## No comments:

## Post a Comment