At the Monetary Economics Workshop of the National Bureau of Economic Research Summer Institute, Cambridge, Mass.
July 10, 2007
Inflation Expectations and Inflation Forecasting
I would like to thank Christina Romer and David Romer for giving me the chance to address participants in the Summer Institute, sponsored by the National Bureau of Economic Research (NBER). As an academic, I regularly attended the Summer Institute and presented or commented on research here. I also served for a time as the director of the Monetary Economics group, the position now shared by David and Christina. The informal nature of the institute, the large number of talented people in attendance, and the opportunity to hear about the very latest work in the field—often while in early draft form—made these few weeks each summer one of the most stimulating times of the year for me. In my current position, I am keenly aware of the long history of fruitful interaction between economists inside and outside of central banks, and I am eager to see this interaction continue. This ongoing intellectual exchange, by improving our understanding of the economy and the workings of monetary policy, has had and will continue to have sizable benefits.
Today I will offer a few remarks on the relationships among monetary policy, inflation, and the public's expectations of inflation, focusing—as seems appropriate for this audience—on some important open questions. I will also give a short overview of the way the Federal Reserve Board staff forecasts inflation, including some discussion of how the staff incorporates information about expected inflation into its forecasting process.
As you know, the control of inflation is central to good monetary policy. Price stability, which is one leg of the Federal Reserve's dual mandate from the Congress, is a good thing in itself, for reasons that economists understand much better today than they did a few decades ago. Inflation injects noise into the price system, makes long-term financial planning more complex, and interacts in perverse ways with imperfectly indexed tax and accounting rules. In the short-to-medium term, the maintenance of price stability helps avoid the pattern of stop-go monetary policies that were the source of much instability in output and employment in the past. More fundamentally, experience suggests that high and persistent inflation undermines public confidence in the economy and in the management of economic policy generally, with potentially adverse effects on risk-taking, investment, and other productive activities that are sensitive to the public's assessments of the prospects for future economic stability. In the long term, low inflation promotes growth, efficiency, and stability—which, all else being equal, support maximum sustainable employment, the other leg of the mandate given to the Federal Reserve by the Congress.
Admittedly, measuring the long-term relationship between growth or productivity and inflation is difficult. For example, it may be that low inflation has accompanied good economic performance in part because countries that maintain low inflation tend to pursue other sound economic policies as well. Still, I think we can agree that, at a minimum, the opposite proposition—that inflationary policies promote employment growth in the long run—has been entirely discredited and, indeed, that policies based on this proposition have led to very bad outcomes whenever they have been applied.
Inflation Expectations: Conceptual Frameworks
Undoubtedly, the state of inflation expectations greatly influences actual inflation and thus the central bank's ability to achieve price stability. But what do we mean, precisely, by "the state of inflation expectations"? How should we measure inflation expectations, and how should we use that information for forecasting and controlling inflation? I certainly do not have complete answers to those questions, but I believe that they are of great practical importance. I hope my remarks here will stimulate some of you to work on these issues.
What is the right conceptual framework for thinking about inflation expectations in the current context? The traditional rational-expectations model of inflation and inflation expectations has been a useful workhorse for thinking about issues of credibility and institutional design, but, to my mind, it is less helpful for thinking about economies in which (1) the structure of the economy is constantly evolving in ways that are imperfectly understood by both the public and policymakers and (2) the policymakers' objective function is not fully known by private agents. In particular, together with the assumption that the central bank's objective function is fixed and known to the public, the traditional rational-expectations approach implies that the public has firm knowledge of the long-run equilibrium inflation rate; consequently, their long-run inflation expectations do not vary over time in response to new information.
But in fact, as I will discuss in more detail later, long-run inflation expectations do vary over time. That is, they are not perfectly anchored in real economies; moreover, the extent to which they are anchored can change, depending on economic developments and (most important) the current and past conduct of monetary policy. In this context, I use the term "anchored" to mean relatively insensitive to incoming data. So, for example, if the public experiences a spell of inflation higher than their long-run expectation, but their long-run expectation of inflation changes little as a result, then inflation expectations are well anchored. If, on the other hand, the public reacts to a short period of higher-than-expected inflation by marking up their long-run expectation considerably, then expectations are poorly anchored.
Although variations in the extent to which inflation expectations are anchored are not easily handled in a traditional rational-expectations framework, they seem to fit quite naturally into the burgeoning literature on learning in macroeconomics. The premise of this literature is that people do not have full information about the economy or about the objectives of the central bank, but they instead must make statistical inferences about the unknown parameters governing the evolution of the economy. In a learning context, the concept of anchored expectations is easily formalized in a variety of ways; in general, if the public is modeled as being confident in its current estimate of the long-run inflation rate, so that new information has relatively little effect on that estimate, then the essential idea of well-anchored expectations has been captured.
Allowing for learning has important implications for how we think about the economy and policy. For example, some work has shown that the process of learning can affect the dynamics and even the potential stability of the economy (see, of many possible examples, Bullard and Mitra, 2002). Considerations of how the public learns about the economy affect the form of optimal monetary policy (Gaspar, Smets, and Vestin, 2006). Notably, in a world with rational expectations and in which private agents are assumed already to understand all aspects of the economic environment, talking about the effects of central bank communication would not be sensible, whereas models with learning accommodate the analysis of communication-related issues quite well (Orphanides and Williams, 2005; Bernanke, 2004). Macroeconomic models with learning also give content to the idea of an economy moving gradually from one regime to another, particularly if the central bank as well as the public is assumed to be updating its beliefs. For example, if the central bank and the public learn from experience that high inflation imposes greater costs and fewer benefits than previously thought, then the equilibrium will adjust toward one with lower inflation and lower inflation expectations. This line of explanation of how economies move between monetary regimes, which has been explored by Sargent and others, strikes me as quite plausible as a historical description (Sargent, 1999). In sum, many of the most interesting issues in contemporary monetary theory require an analytical framework that involves learning by private agents and possibly the central bank as well.
Implications of Anchored Inflation Expectations
Why do we care about the variability of inflation expectations? As my colleague Rick Mishkin recently discussed, the extent to which inflation expectations are anchored has first-order implications for the performance of inflation and of the economy more generally (Mishkin, 2007). Mishkin illustrated this point by considering the implications of the fact that inflation expectations have become much better anchored over the past thirty years for the estimated coefficients of the conventional Phillips curve, which I define here to encompass specifications that use lagged values of inflation to proxy for expectations or other sources of inflation inertia. As he noted, many studies of the conventional Phillips curve find that the sensitivity of inflation to activity indicators is lower today than in the past (that is, the Phillips curve appears to have become flatter);1 and that the long-run effect on inflation of "supply shocks," such as changes in the price of oil, also appears to be lower than in the past (Hooker, 2002). These findings are of much more than academic interest. To the extent that the Phillips curve may have flattened, inflation will now tend to be more stable than in the past in the face of variations in aggregate demand. (Of course, this can be a good thing or a bad thing, depending on whether inflation expectations are anchored in the vicinity of price stability.) Likewise, a lower sensitivity of long-run inflation to supply shocks would imply that such shocks are much less likely to generate economic instability today than they would have been several decades ago. Notably, the sharp increases in energy prices over the past few years have not led either to persistent inflation or to a recession, in contrast (for example) to the U.S. experience of the 1970s.
Various factors might account for these changes in the Phillips curve, but, as Mishkin pointed out, better-anchored inflation expectations—themselves, of course, the product of monetary policies that brought inflation down and have kept it relatively stable—certainly play some role. If people set prices and wages with reference to the rate of inflation they expect in the long run and if inflation expectations respond less than previously to variations in economic activity, then inflation itself will become relatively more insensitive to the level of activity—that is, the conventional Phillips curve will be flatter.
Similar logic explains the finding that inflation is less responsive than it used to be to changes in oil prices and other supply shocks. Certainly, increases in energy prices affect overall inflation in the short run because energy products such as gasoline are part of the consumer's basket and because energy costs loom large in the production of some goods and services. However, a one-off change in energy prices can translate into persistent inflation only if it leads to higher expected inflation and a consequent "wage-price spiral." With inflation expectations well anchored, a one-time increase in energy prices should not lead to a permanent increase in inflation but only to a change in relative prices. A related implication is that, if inflation expectations are well anchored, changes in energy (and food) prices should have relatively little influence on "core" inflation, that is, inflation excluding the prices of food and energy.
Although inflation expectations seem much better anchored today than they were a few decades ago, they appear to remain imperfectly anchored. A number of studies confirm that observation. For example, Gürkaynak, Sack, and Swanson (2005) found that long-run inflation expectations, as measured by the difference in yields between nominal and inflation-indexed bonds, move in response to news about the economy, rather than remaining unaffected. Levin, Natalucci, and Piger (2004) have shown that some survey measures of inflation expectations in the United States respond to recent changes in the actual rate of inflation, which would not be the case if expectations were perfectly anchored. Models of the term structure of interest rates better fit the data under the assumption that both inflation expectations and beliefs about the central bank's reaction function are evolving (Kozicki and Tinsley, 2001; Rudebusch and Wu, 2003; Cogley, 2005).
An indirect but elegant way to make the point that inflation expectations remain imperfectly anchored comes from a statistical analysis of inflation by Stock and Watson (2007). Stock and Watson model inflation as having two components, which may be interpreted as the trend and the cycle. Changes in the trend component are highly persistent whereas shocks to the cyclical component are temporary.2 The key finding of this research is that the variability of the trend component of inflation (and thus the share of the overall variability of inflation that it can explain) appears to have fallen significantly after about 1983. That is, unexpected changes in inflation are today much more likely to be transitory than they were before the early 1980s. Because it seems quite unlikely that changes in inflation could persist indefinitely unless long-run expectations of inflation also changed, I interpret the Stock-Watson finding as consistent with the view that inflation expectations have become much more anchored since the early 1980s. At the same time, that the variability of the trend component of inflation, though modest, remains positive, implies that long-run expectations of inflation are not perfectly anchored today.
The policy implications of the much-improved but still imperfect anchoring of inflation expectations are not at all straightforward. To evaluate these implications, we must understand better the historical variation in inflation expectations, the effect of this variation on actual inflation and economic activity, and the relationship between policy actions and the formation of inflation expectations. With the hope of promoting progress on these broad topics, I pose three questions to researchers, the answer to any of which would be quite useful for practical policymaking.
First, how should the central bank best monitor the public's inflation expectations? Theoretical treatments tend to neglect the fact that in practice many measures of inflation expectations exist, including the forecasts of professional economists, results from surveys of consumers, information extracted from financial markets such as the market for inflation-indexed debt, and limited information on firms' pricing plans. In a very interesting paper, Mankiw, Reis, and Wolfers (2003) compared the available measures, emphasizing in particular that median measures of inflation expectations often obscure substantial cross-sectional dispersion of expectations.3 On which measure or combination of measures should central bankers focus to assess inflation developments and the degree to which expectations are anchored? Do we need new measures of expectations or new surveys? Information on the price expectations of businesses—who are, after all, the price setters in the first instance—as well as information on nominal wage expectations is particularly scarce.
Second, how do changes in various measures of inflation expectations feed through to actual pricing behavior? Promising recent research has looked at price changes at very disaggregated levels for insight into the pricing decision (Bils and Klenow, 2004; Nakamura and Steinsson, 2007). But this research has not yet linked pricing decisions at the microeconomic level to inflation expectations; undertaking that next step would no doubt be difficult but also very valuable.
Third, what factors affect the level of inflation expectations and the degree to which they are anchored? Answering this question essentially involves estimating the learning rule followed by the public or various components of the public, although one could consider alternative frameworks like Carroll's (2003) epidemiological model of the propagation of information among private agents. A fuller understanding of the public's learning rules would improve the central bank's capacity to assess its own credibility, to evaluate the implications of its policy decisions and communications strategy, and perhaps to forecast inflation. Realistically calibrated models with learning would also inform our thinking about policy and the economy.
Inflation Forecasting at the Federal Reserve
I would like to shift gears at this point to tell you a bit about how the Federal Reserve Board staff goes about forecasting inflation. Obviously, this activity provides critical inputs into the making of monetary policy, and as I will discuss, the staff's long-term track record in forecasting inflation is quite good by any reasonable benchmark. I hope that my brief description will stimulate your interest in the complex and challenging problems of real-time macroeconomic forecasting. But, as you will see, the discussion of practical inflation forecasting will bring us back to one theme of my remarks—that our ability to forecast inflation and predict how inflation will respond to policy actions depends very much on our capacity to measure and to understand what determines the public's expectations of inflation.
The Board staff employs a variety of formal models, both structural and purely statistical, in its forecasting efforts. However, the forecasts of inflation (and of other key macroeconomic variables) that are provided to the Federal Open Market Committee are developed through an eclectic process that combines model-based projections, anecdotal and other "extra-model" information, and professional judgment. In short, for all the advances that have been made in modeling and statistical analysis, practical forecasting continues to involve art as well as science.
The forecasting procedures used depend importantly on the forecast horizon. For near-term inflation forecasting—say, for the current quarter and the next—the staff relies most heavily on a disaggregated, bottom-up approach that focuses on estimating and forecasting price behavior for the various categories of goods and services that make up the aggregate price index in question. For example, we know from historical experience that the prices of some types of goods and services tend to be quite volatile, including not only (as is well known) the prices of energy and some types of food but also some "core" prices such as airfares, apparel prices, and hotel rates. The monthly autocorrelations of price changes in these categories tend to be low or even negative. In contrast, changes in inflation rates in some services categories, such as shelter costs, tend to be more persistent. In assessing what price changes in a particular category imply for future price changes in that category, the staff uses not only various forms of time-series analysis but also specialized knowledge about how the various indexes are constructed—for example, whether certain categories are sampled every month in all localities and how seasonal adjustments are performed. In making very near-term price forecasts,