Thursday, September 24, 2009

Half Full or Empty ?

I don't think much has been accomplished by the effort to base macroeconomics on microeconomic models of rational agents. On a couple of occasions, I have claimed that something major was accomplished in this way long ago by Milton Friedman. That was the example of how the approach could be useful and a major inspiration for the whole research program. Right now I am struck by painful doubt about that one key example.

The example of how previously mysterious facts were actually explained by an economist once is the permanent income hypothesis. Basically this is a model of rational consumption in which Friedman noted that optimal consumption depended not only on current income and wealth but predicted future labor income -- in particular it should depend on the expected present discounted value of future income. The model explained some puzzles.

1. in the cross section consumption increases less than one for one with income -- people with higher incomes save a larger fraction of that income. Based on this, many economist really predicted that with economic growth the share of consumption in GNP would decline. It didn't.

2. There are odd patterns in the cross section of income and consumption. One is that African Americans consume less than White Americans with the same income. There are many other interesting patterns which I can only remember by remembering that they are explained by the PIH.

The explanation is that permanent income increases less than one for one in current income, people with high income now might have just gotten lucky this year. Similarly with race, average African American income is lower so if an African American has the same income as a white American it is likely that the African American had an unusually good year or that the white American had an unusually bad year.

Now, to my dismay, I notice that this evidence just shows that Friedman's model is better than Hick's model of the consumption function in the IS-LM model which makes consumption a function only of current income. In itself, it doesn't show much more.

Let me make a very very different model. Consumption depends only on wealth. One way of imagining this is that people are trying to maximize the discounted value of the logarithm of consumption (so interest rates don't effect consumption as a function of wealth) and irratinally predict that their future labor earnings will definitely be zero. That's pretty pessimistic. Can Friedman's facts show that, even if people aren't fully rational, at least they aren't that irrational ?

Not at all. In the time series national income and wealth grow proportionally. This is the strange fact that growth theorists are constantly trying to explain. The constant times wealth model works fine.

Income and wealth are, of course, correlated. However they aren't perfectly correlated. The regression slow of wealth on income could easily be less than one. If so the pattern in the cross section is explained.

It is plain true that African Americans have lower wealth than white Americans with the same income.

There is proof that consumption depends on some sort of smoothed of income. I haven't mentioned any evidence that this is smoothed future income not smoothed past income.

Now there is such evidence, but it is relatively subtle compared to the facts which I mentioned (a punch in the nose is relatively subtle compared to the fact which I mentioned). It is of the form of my hard to understand estimates reject the total myopia model and not "yes of course. Why didn't I see that ?"

Now at the moment I really should begin to worry about Friedman's critique of the Phillips curve, but if there is as much less to that as there is to the triumph of the hypothesis that permanent income matters (not the PIH which says only permanent income matters) then I will be upset.

I will just note that Keynes very firmly and clearly warned people not to look for a Phillips curve in The General Theory.

No comments: