Smith on Inflation Expectations

One of the risks of typing your homework late at night is that the ease of writing it up lets you slip into editorializing. I did this some last night (after solving the problem of course) complaining about a Barro-Gordon model exercise we did about central bank inflation bias. I think the argument makes very little sense. You basically only get the inflation bias result by assuming the central banker understands all sorts of things about the macroeconomy, but somehow doesn't understand how inflation expectations are formed (I'll give you a hint for inflation expectations in this particular case: they take the expected value of inflation! Real tough, huh? Clearly a central banker can't figure that one out). When you assume a dumbass central banker you get a dumbass policy. That seems dumb to me.

So it was nice this morning to see Karl Smith provide a dose of reality on inflation expectations modeling too - this time, thinking about what agents in the economy know about central bank behavior. He writes:

"My understanding is that the principle fear regarding inflation is that if a sustained period of high inflation were allowed expectations would become unmoored and the Fed would loose it hard won credibility.

The problem with this view is that it flies in the face of the notion that inflation expectations are rational.

A rational agent, however, incorporates knowledge not only of Fed behavior but of the informational and operational constraints facing the Fed. If it is in fact the case that the Fed is only risking higher inflation because it is

- Uncertain about potential output
- Concerned about hysterias
- Has difficulty adjusting policy at the zero lower bound
- Is operating in the wake of a major international financial crisis

Then a rational agent should not conclude that long run inflation targets have meaningfully changed."

Again - if you assume a dumbass agent you get dumbass agent behavior. We really should make sure our claims pass the smell test. Making simplifying abstractions is just fine. I'm hardly a praxeologist, as you all know. But you need to ask yourself if the only reason you're getting the result you're getting is because of an unrealistic assumption. If that's the case, then you've got a problem.

OK... now I need to reread my homework and make sure I didn't put it quite as bluntly as that.