Dollars and Jens

Wednesday, January 28, 2009

Taylor rule

Merrill Lynch economist David Rosenbergâ€™s version of the rule suggests that the right target rate for the Fed would be a negative one percent in the coming year. Former Fed governor Laurence Meyer of Macroeconomic Advisers reckons the rule calls for a rate of negative 4% or lower. Goldman Sachs, with its estimate of 9.5% unemployment and a 0.25% core inflation rate by the end of 2010 suggests that by next year the appropriate target rate for the Fed will be a negative 6%.The Taylor rule suggests that the appropriate* interest rate is a linear function of the "output gap" and the inflation rate; the "output gap", as noted at the link, is hard to measure, but essentially means "the extent to which the real economy is underperforming". (I happen to like a time derivative of the unemployment rate, but I'm not an expert. Yet.) The coefficient on inflation should exceed 1†; this is called the Taylor principle, and means that the

*real*interest rate should be higher if inflation is higher, so as to attempt to reduce aggregate demand (this is the way it's usually formulated) to rein in inflationary pressures.

It's usually written that way, in terms of nominal interest rates, in part because nominal interest rates are easier to measure and define and are, at least in the past couple decades, the main policy instrument of the FOMC. The formula can be rewritten, though, subtracting inflation off of each side, so that the real rate is a linear function of inflation and output gap (with the Taylor principle dictating that the coefficient on inflation be positive).‡ At that point an indicated negative real rate suggests a positive inflation target; you can target negative real rates, so long as you have some means other than rates to generate inflation. This is probably what the Fed is going to be attempting for the next year or so; it will be trying to create enough inflation through essentially quantitative means so that a zero nominal interest rate creates a real rate that is as negative as the Taylor rule prescribes.

* I believe the rule was initially stated, at least in part, as a

*description*of what the Fed

*does*as much as a prescription for what it should do. Taylor's arguments in favor of a policy rule do sometimes include claims that the Fed has erred when it has deviated from a Taylor rule, but one of his strongest arguments is that being explicit about what you're doing aids both in passing along institutional knowledge and in helping to refine the practice of policy, giving precision to what has been done in the past and therefore what is likely, perhaps, to be too easy or too tight in the future; another argument he makes is that a policy rule is a good way to frame debate, giving a starting point from which discretion can then take the form of a discussion as to why policy should be easier or tighter than the rule. In each of these more normative arguments, he is making clear that he doesn't believe his simple formula is the endpoint of research in monetary policy.

† Econometric evidence suggests that this coefficient was a bit less than 1 in the sixties and seventies, which may have been part of the problem; since 1982, and particularly since Greenspan took over in 1987, it has been 1.5 or even a bit higher. (I obviously in this context mean it as a measure of what the Fed

*does*, not what it should do.)

‡ When I say that nominal interest rates are easier to measure and define than real rates, I'm implicitly saying that inflation is hard to measure and define. I kind of think, in fact, that the real rate in this latter version of the Taylor rule should be calculated using a full inflation measure, while the inflation on the other side of the equation should be a core inflation measure. I have no real conviction in this belief, though; the problem with full inflation is that it moves around a lot, and if you're trying to set a policy rate for the next six weeks, what you care about is the inflation rate in the next six weeks, not the recent past. Perhaps what I should say is that the inflation that gets subtracted out of the nominal interest rate should be a shorter-term prediction of inflation than the inflation that goes into the linear function.