Discussion Paper

No. 2007-14 | March 26, 2007
Evaluating Inflation Targeting Using a Macroeconometric Model


This paper uses a structurally estimated macroeconometric model, denoted the MC model, to evaluate inflation targeting in the United States. Various interest rate rules are tried with differing weights on inflation and output, and various optimal control problems are solved using differing weights on inflation and output targets. Price-level targeting is also considered. The results show that 1) there are output costs to inflation targeting, especially for price shocks, 2) price-level targeting is dominated by inflation targeting, 3) the estimated interest rate rule of the Fed (in Table 4) is consistent with the Fed placing equal weights on inflation and unemployment in a loss function, 4) the estimated interest rate rule does a fairly good job at lowering variability, and 5) considerable economic variability is left after the Fed has done its best. Overall, the results suggest that the Fed should continue to behave as it has in the past.

JEL Classification:




Cite As

Ray C. Fair (2007). Evaluating Inflation Targeting Using a Macroeconometric Model. Economics Discussion Papers, No 2007-14, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2007-14

Comments and Questions

anonymous - Referee Report
June 21, 2007 - 16:56

see attached file

Ray Fair - Response to Referee
June 28, 2007 - 15:53

This referee's report does not have any detailed suggestions for revising my paper, and I generally agree with the referee's main points. I am assuming that people do not have rational expectations, and, as the referee says, I am clear on what I am doing given this assumption. ...[more]

... Models Models like the New Keynesian model, which assume rational expectations, do not fit the data well (as the referee agrees), and so it would seem that other approaches are worth pursuing at this point. The following are a few minor differences of emphasis between the referee and me.

Third paragraph: In my case disequilibrium behavior does not depend on the policy regime except as agents respond to policy-variable changes over time. This is where the assumption that expectations are not rational is so important.

Fourth paragraph: I wouldn't say that my agents are "self-destructive." They simply don't have rational expectations. Whether it is closer to the truth to model expectation formation as being rational or not is an empirical question, and so far I would argue that the evidence favors the assumption that expectations are not rational.

I can add a reference to the Nessen and Vestin paper. They say in their last paragraph that more empirical work is need on the issues raised in their paper, such as the degree of persistence in the Phillips curve, and, of course, I agree. I would argue that my paper is an example of this kind of empirical work.

Finally, my macroeconometric model is a structural model, not simply a forecasting model. A forecasting model might imply that there is no theory behind it, which is not true of my model. As I discuss in the paper, there is plenty of theory, just not theory that assumes rational expectations.

anonymous - Referee Report
July 03, 2007 - 09:30

see attached file

Ray Fair - Response to Referee 2
July 05, 2007 - 19:22

Response to Referee 2. Ray C. Fair.

I agree with the bottom line of the referee (second to last paragraph) regarding the contribution of my paper. Regarding my discussion of the NK model, I get mixed advice on how much I should compare my MC ...[more]

... model to the the NK model in a paper like this. Some want more; some want less. The tone of my discussion in the present paper is to use the similarities and
differences between the NK and MC models to help explain the MC model. Since most people are more familiar with the NK model, this discussion provides a useful way for giving one a sense of the MC model without having to have them read my book on the model. I don't agree with the referee that the MC model is ad hoc. Both it and the NK model have utility maximizing households, for example. The main difference is that the MC model does not assume that expectations are rational, but this is not really ad hoc. I can spell out in the paper, as the referee suggests, more about how expectations are assumed to be formed in the MC model and how this leads to disequilibrium. Expectations are sophisticated, but they are never based on knowledge of the complete model. (Again, this is all spelled out in my book, and there is always the question of how much I should get into it in any paper.) Finally, on the fit of NK models, I am not aware of outside-sample RMSEs being generated for any version of the NK model other than the one I compare in Table 1. I think Table 1 in my paper is useful for giving a quantatitive sense of how a typical NK model currently fits the data relative to other models.