The key words in the title of this conference, as far as I am concerned, are “uncertain times.” They are key because there is almost always considerable uncertainty when it comes to economic policymaking. I recognize this statement may sound typically like a Federal Reserve official trying to build sympathy largely for imaginary difficulties in policymaking, but I will try to convince you that uncertainty is real and has significant implications for policy. I also recognize that, as the last speaker at a daylong conference, I am in an unenviable position, especially because earlier participants were both distinguished and knowledgeable. Thus, I take as my charge: to be mercifully brief and to be provocative, at least by Federal Reserve standards. Let me note, in this regard, the now obligatory disclaimer that I am speaking only for myself and not for others in the Federal Reserve.
As a monetary policymaker, I cannot avoid making forecasts of future economic performance, because I would like policy to be anticipatory and to head off prospective economic problems, or simply to contribute to stability if that is all that is required. We are all familiar with the story of long policy lags which make preemptive or anticipatory policy desirable. But the other traditional characteristic of policy lags, their variability, is almost synonymous with uncertainty and suggests caution in pursuing preemptive policy.
In any event, anticipatory policy requires forecasts and requires that I pay some attention—give some weight—to them. Unfortunately, neither my personal forecasting record, nor that of the economics profession, is particularly good. To illustrate this, consider a couple of episodes from recent experience; I pick these because they are at hand, but since I have been at this for more than 30 years now, I could find many other examples if need be.
- 2nd half of 1999
- 1st half of 2000
- 2nd half of 2000
- Q4 2001
- Q1 2002
The facts are, no one has sustained a good record of forecasting the
short-term performance of the U.S. economy. What does this mean for policymaking?
The normal prescription is that, in the face of substantial uncertainty,
policy actions should be modest and infrequent so that, at worst, they
will not be destabilizing. I endorse this conclusion and would add a couple
of observations.
First, with regard to real growth, I think we have seen over the past twenty years that the U.S. economy is terrifically resilient. Recall that the economy grew uninterruptedly from late 1982 till mid 1990, then again from the spring of 1991 till the spring of last year. This was a gratifying performance made even more so by the shocks and storms the economy weathered along the way, including the stock market crash of October 1987 (and several other corrections of note); financial crises in Latin America, parts of Asia, Russia, and so on; significant downsizing of the defense industry in the wake of the collapse of the USSR; virtual demise of the S&L industry and domestic banking problems of note; and so on.
Despite this litany of problems, economic growth proceeded, and I think it is fair to say that on average over the past twenty years the surprises largely have been on the upside. To be fair, some of these shocks and disruptions did in fact provoke a policy response, and thus we do not have a clean test of resilience, but I for one am under no illusions that policy was so appropriate and precise that it deserves the lion's share of the credit for this economic performance.
Upside surprises in economic growth are usually easy to take, especially if they are accompanied by persistently low inflation, as was the case in much of the 1990s. As noted earlier, we appear to have had another bout of this in the last quarter or two, with growth exceeding earlier expectations. Interestingly, while I think most forecasters estimate growth of at least four percent in the first quarter, many in the business community remain quite cautious about the economy and its prospects. This divergence between the views of business leaders and those of economists is not trivial, and yet the numbers are “neutral,” speak for themselves, and are undeniably positive. What accounts for the divergence?
The short, honest answer to this question is “I don't know,” but I am willing to speculate a bit. In a nutshell, I think many businesses are still adjusting to a low inflation environment. In general, profits cannot be improved by price increases because such increases will not stick. In a highly competitive, low inflation environment, cost containment or outright cost reduction is critical to improved profitability, but this will not necessarily be easy to achieve. Thus, I would speculate, business continues to experience bottom line pressure even though the recovery is well underway in at least several sectors.
The preceding commentary was a bit of a digression but it serves the purpose of introducing the subject of inflation, another topic challenging in my view to those responsible for short-term forecasts and to monetary policymakers. There is a voluminous body of evidence for the long run indicating that monetarists are correct about inflation. In the long run, inflation is a monetary phenomenon, where by the long run I mean periods of five or ten years or more.
But this relation doesn't get us very far if accurate short-term forecasts of inflation are the goal. One favorite way to produce short-term, say annual, forecasts of inflation today is with some variation of the Phillips curve, usually involving the concept of NAIRU (nonaccelerating inflation rate of unemployment). The underlying notion, albeit perhaps an oversimplification, is that unemployment is a reasonable proxy for labor market pressure and that labor market pressure ultimately translates into inflation, presumably because of its implications for wages, compensation more broadly, and unit labor costs. NAIRU is key because significant upward pressure on compensation, say, is expected only when actual unemployment drops below it.
Depending on your “tastes and preferences,” this may be a compelling story, but for this policymaker an important question is how well do the NAIRU models work in practice? The answer to this question is, in short, “not well,” and this has been true since 1984 according to our (FRB Mpls.) analysis. In short, our research suggests that unemployment has not contributed significantly to forecasts of inflation for quite some time.
Specifically, two of our economists, Andy Atkeson and Lee Ohanian, both now faculty members at UCLA by the way, reviewed a variety of inflation forecasting models. Their results revealed that, beginning in 1984, Phillips curve type NAIRU models did no better at forecasting inflation than a simple model which assumed that next year's inflation rate would equal last year's. And as a practical matter, many of you may recall that many forecasters generally overpredicted the pace of inflation, from about the mid 1990s on.
In light of this evidence, am I ready to discard the NAIRU concept in its entirety? No, although it's tempting. But I do think, speaking only for myself, that the burden of proof has shifted to those who believe it's a valuable concept. In fact, the whole exercise is beginning to remind me of the gyrations in which we used to engage in the 1980s to try to preserve or to resurrect various short-run demand for money equations as they went increasingly off track. Ultimately, the exercises proved futile and we surrendered.
What are the policy implications of these circumstances? In my view, they exacerbate the degree of uncertainty and, other things equal, ought to contribute to caution in the response of monetary policy to changes in conditions and forecasts. Indeed, one way of interpreting the conduct of policy from about 1996 through 2000 is that a good deal of caution was in fact exercised. To remind us, the economy was into its sixth year of expansion by the spring of 1996 and the unemployment rate dropped below five percent by 1997. It continued to decline, on average, through 2000, yet there was at most minor tightening of monetary policy through this episode. Recall, finally, that virtually all estimates of NAIRU at the time were in excess of five percent, and some were as high as six percent.
Two obvious reservations about a persistent policy prescription of caution are in order. First, such a prescription runs counter to desires to be preemptive and, secondly, it would seem to risk an inadequate response to a serious problem should one arise. I take these reservations seriously but am not yet overly concerned. Effective preemptive policy action in fact requires an ability to forecast short-run fluctuations in the economy accurately, and I have already expressed reservations on this score. Perhaps more importantly, I have also emphasized the fundamental resilience of the U.S. economy and would add that much of the time excesses in the economy, positive or negative, largely self correct. Policy rarely has to try to save the day. Put another way, given the fundamental soundness and strength of the U.S. economy, it is important that we get policy approximately correct, but it is not necessary that we get it precisely correct. And the costs associated with overreacting are probably at least as great as those associated with caution.