Rules versus Discretion

In an excellent piece on the response to the financial crisis, John Taylor highlights the important concept of rules versus discretion in government policy.

Taylor shows that monetary policy during the start of the housing boom deviated significantly from the "rule" it had been following for the previous 20+ years.  Taylor and his coauthors find evidence that this amplified the boom and bust in the housing market.   While not everyone agrees with this conclusion, Taylor presents some convincing counter factual experiments.
More importantly, Taylor explains just how the government has failed with its management of the crisis.  First, he points out the misdiagnosis of the problem.  Many of the early actions of the Fed, like the opening of the TAF, were aimed at solving a liquidity problem, not the solvency problem, which was the real reason there were liquidity issues.  Second, Taylor points to the uncertainty caused by the actions of the Fed and Treasury.  Again, this uncertainty is another benefit of "rules"- and Taylor points to the positive impact of the transparency the IMF uses in determining how they will aid countries experiencing financial crises.

Contrary to Taylor's advice is this quote from Bernanke presented in today's Wall Street Journal; "Roosevelt's specific actions were, I think, less important than his willingness to be aggressive and to experiment..."

I'd like to run an experiment- for 30 years, lets tie the hands of those making monetary and fiscal policy decisions and see what happens.


I just started reading through Taylor's analysis. But one problem that may be important is that his counterfactuals are based on an interest rate rule (the Taylor rule) that is only based on a loosely optimizing argument. To be more precise, the Taylor interest rate rule can be derived from a monetary authority whose objective is to minimize deviations of inflation and output from some target levels. Some argue that these quadratic loss functions are second order approximations to utility maximization, but that is not true for any commonly used utility function. In addition, this quadratic loss function assumes the form of the Taylor rule as a model primitive. John Taylor said himself in an interview with Russ Roberts on EconTalk in August 2008 ( that his rule was just a lucky good approximation for how monetary authorities act.

The point here is that the Taylor rule is a rule of thumb that seems to approximate what monetary authorities do in practice (although no one that I have seen has shown in a convincing way why it comes so close). So showing a counterfactual based on the Taylor rule is just showing another potential interest rate path that the Fed could have followed. It is not clear that this path would have been better.

That said, I think the Taylor rule does seem to have some properties that are attractive, and I think that the arguments in the paper you refer to hold some weight. I just think we can do better in terms of defining what the benchmark optimal response should be (should have been).

"Well, it's pretty hard to argue with someone who won three super bowls," said the health economist.