Monday, November 22, 2010

Industrial Policy

I am a huge believer in mixed economies. But it is hard to prove that they are the best option for exactly the reasons that Ms. McArdle is skeptical about industrial planning in China:

It's not that I didn't understand that the government did this; it's that I didn't understand how pervasive it would be, or how popular this would be, at least with the folks we interview. Everyone--including most of the economists and NGOs--seems to think this is swell. No fiddling around with archaic, unplanned systems; just figure out what the country needs and do it!

Perhaps it is just my ideological blindness that makes me believe that this cannot, in the long run, turn out well. But there's a plausible story that the early boom was mostly a matter of removing distortions (and taking advantage of capital, human and otherwise, accumulated in Hong Kong and China). Now the government is much more directly picking winners and losers. They're not trying to manage growth; they're trying to cause it in places where it shows little sign of happening organically.

It's not that I think that no form of industrial policy can ever have good effect. Can government build infrastructure to good effect? Yes, certainly. Can they manage growth? Can it occasionally pick industrial winners? They have in the past--though on average, I'd say it's abundantly clear that governments have more often picked, and sustained, losers. And the more comprehensive the industrial policy, the worse the economic losses have generally been.

The issue here is that successes can be explained by a lot of different factors (as can failures). And it is pretty clear that people have strong "priors" (in the Bayesian sense) for the approach that they favor. This makes it hard to use the (limited in scope) data to decide between approaches. Add in confounding factors and it gets harder. Add in changes in technology and you guarantee issues.

Just consider, for example, stock market returns. What is the relevant period of interest? Some people claim you can consider returns on stock from 1800 to today based on what records are available. But clearly the information available to stock analysts is different today than in 1960. Now consider that you want to look at long term returns (i.e. saving for retirement) and suddenly the noise threatens to overwhelm the data -- as you really have only a very few (tightly correlated) time trends.

How much worse is that for looking at economic growth?

No wonder these arguments are difficult to make . . .

No comments:

Post a Comment