The problem with Taylor Rules

2/08/2014 01:58:00 PM
There was never a time when you could look at John Taylor's version of the Taylor Rule and think it was a good description of Fed behavior.
Yesterday I noted that for the most part, Taylor Rule estimates from all over the ideological spectrum were now in aggrement that a hike in the effective federal funds rate is immanent. Yet, some have rightfully commented that that is not the same as suggesting that the Federal Reserve should increase the nominal interest rate--the Taylor Rule gives a prediction of what the Fed will do, not what would be optimal for the Fed to do. It's something I've written about before.

I think that is a huge oversight for economists to make, both on the policy analysis level and on the forecasting level. As far as policy analysis goes, it makes no sense for people like John Taylor to argue that the Fed should follow his estimated Taylor Rule--to obtain his estimate, all he did was estimate what the Fed has done in the past! What Taylor and others should be doing is estimating instead what the optimal Taylor Rule would be--at what nominal interest rate would actual output equal long-run potential output? There's no reason to suppose that following the historical Taylor Rule will result in full employment.

So clearly, we shouldn't be using Taylor Rules to make monetary policy recommendations. Now, at some level, we could argue that if we are trying to make fiscal policy recommendations, it makes sense to include in the model a Taylor Rule that is predictive rather than optimal--fiscal policy needs to be based on what the Fed actually will do, not what they should do. But even here, the acceptance of the Taylor Rule methodology in macro is curious. For everything else, economists rely on optimization to produce accurate forcasts of entities' choices. In DSGE models, all household and firm choices are predicted by assuming they maximize their objective functions, with perfectly rational expectations no less, while the Fed is still modeled using a 1950s era ad-hoc adaptive expectations process. That's nuts.

My bottom line is this: updating the DSGE methodology to assume that the Fed is rational would probably both improve the monetary policy advice we give as well as improve our forecasts of what the Fed will actually do. When you compare John Taylor's estimated Taylor Rule to what the Fed has actually done, you see that there are pretty huge forecast errors. I suspect the Fed operates much closer to optimal than Taylor gives them credit for.

Nevertheless, while it's important to keep these limitations in mind, I think the broad aggrement among disparate versions of the Taylor Rule that the "predicted" interest rate is now positive is suggestive evidence that the end of the liquidity trap is near. The "natural rate of interest" at which we have full employment is rising, and though maybe not already higher than the actual rate, the gap is closing fast. From here on out, expect monetary policy to start gaining traction.