Transcripts of the monetary policymaking body of the Federal Reserve from 2002–2008.

Thank you, Mr. Chairman. Clearly, from the discussion so far— and my prediction is that it will continue this way—there is a tremendous amount of support for the outline that you have presented, and I share that. I want to focus on two points that Richard raised and maybe hit them a little harder. First, on the fed funds path, the fed funds rate is the policy instrument. We don’t want to confuse the public about the difference between the policy instrument and the policy goals. If we want to present something, we ought to present the policy rule by which the fed funds rate is determined. We don’t know how to do that beyond the broad outlines of the Taylor rule, and that is the point that we should be emphasizing. So I am very opposed to presenting a fed funds rate path because I think it is really going to confuse our communication.

Now, on the issue of false precision, I want to use an analogy because I hope that it will help to explain this issue. Let me use a GPS analogy. If I did my calculation correctly, GPS coordinates usually show up on your instrument down to a thousandth of a minute of longitude, let’s say. If I did my calculation correctly, that’s about six feet. For some policy purposes, perhaps locating the piers on a suspension bridge, you would really like another digit. But the GPS instrument can’t produce that other digit. If you were to try to do it with repeated trials, you would just get random garbage. Therefore, if you are locating bridge piers, you have to use a different measuring device. Now, Don said, “What is the difference between 1.9 and 1.8?” Well, that is beyond the measurement accuracy of a lot of the data. It is not just that it doesn’t have very much policy relevance, but it is really beyond the measurement accuracy, and we have multiple measures of inflation. We take the PCE index, and we look at it with and without the nonmarket components. Those relationships change, and therefore we are really making a mistake if we force ourselves to make these judgments to 0.1 percentage point. In the projections, we ought to round these things off to the appropriate number of significant digits. Standard scientific practice is not to present measurements that run beyond the accuracy of the measuring instruments. So what is the appropriate number of significant digits in this business? If you look at the nature of the revisions, the multiple measures, and so forth, I would propose that for GDP and inflation it is probably 0.5 percentage point. For the unemployment rate, it is maybe 0.25 percentage point. But we are really going to tie ourselves in knots over stuff that is like the fourth digit on your GPS instrument. It just doesn’t have any measurement validity to it.

Part of the problem in presenting the histograms that are organized around these very fine intervals is that they will display differences that are totally insignificant and don’t amount to anything. But if we present it, people will start thinking that we must think it means something. So I think we really will aggravate the communication process, and therefore we ought to look at what the appropriate number of significant digits is. I am not trying to obfuscate the differences. I am trying to make sure that we focus on differences that are genuine. If we have a real difference that is 0.5 percentage point apart, well so be it. There is no reason in my view that we shouldn’t present it and explain it.

Keyboard shortcuts

j previous speech k next speech