Thank you, Mr. Chairman. Our presentation this afternoon,
entitled “Explaining Low Inflation since the Mid-1990s” reflects joint work with Flint
Brayton and David Lebow and will be presented by myself and David. As shown in
the upper left panel of your first exhibit, the rate of unemployment fell through the
late 1990s, eventually reaching a level not seen in thirty years. Nevertheless, core
consumer price inflation—shown at the upper right—remained in the 1½ to 2 percent
range throughout this period. This outcome struck many observers as being at odds
with the view that low unemployment necessarily pushes up inflation. The objective
of our briefing today is to provide an explanation for these developments. To do this,
we find it helpful, as summarized in the middle panel, to step back and ask whether
changes may have been occurring in the economy over a longer span of time that may
have altered the inflation process. We then focus on the recent past and ask why
inflation remained so low in the late 1990s, when unemployment was also low.
Finally, in light of the changes that we identify as being important, we evaluate the
likelihood that the recent conjunction of low inflation and low unemployment can be
repeated in coming years.
By our reckoning, three factors account for much of the recent experience of
simultaneously low inflation and low unemployment, as listed in the lower panel of
the exhibit. First, a change in the way monetary policy has been conducted over the
past two decades, compared with the 1960s and 1970s, may have reduced the short-
run responsiveness of inflation to resource utilization. This factor is a relatively small
part of our overall explanation for why inflation has remained so low. The second
factor is the pickup in productivity growth that started in the mid-1990s. As output
per hour advanced more rapidly, hourly compensation responded only gradually, and
the pressure on prices from unit labor costs temporarily waned. This element is the
largest part of our explanation for the late-1990s experience. Looking to the future, the influence of the pickup in productivity growth on inflation should diminish over time as hourly compensation adjusts more fully to the faster trend growth in output per hour. Lastly, structural developments in labor markets may have caused the natural rate of unemployment today to be about ½ percentage point lower than in the mid-1980s. These effects are likely to persist.
I should note that our presentation does not discuss the influence on recent inflation of movements in the prices of imports, energy, and food. Although these standard “supply-shock” variables have at times had significant effects on the year-to year pattern of inflation, we see their net contributions to the good inflation performance of the past several years as being relatively small.
The framework for our analysis, the FRB/US model of inflation dynamics, is sketched out in exhibit 2. This model and others like it help inform the staff projection. In this framework, inflation (π) is a function of lagged inflation, expected future inflation (πe), the difference between the unemployment rate (U) and its natural rate (Un), relative price shocks (such as movements in food, energy, and import prices), and unit labor costs. In FRB/US, households and businesses form expectations of inflation based on their knowledge of the structure of the economy, including the manner in which monetary policy is conducted.
The factors we examine fit into this model as follows. First, for reasons we will expand on shortly, changes in the conduct of monetary policy alter the influence of the unemployment rate on expected inflation. Second, changes in labor productivity growth affect inflation through unit labor costs. Finally, the labor market developments we highlight influence the natural rate of unemployment directly and thereby alter the amount of inflation impetus associated with any given level of the actual unemployment rate.
Although economic slack, as measured by the gap between the unemployment rate and the natural rate of unemployment, clearly affects inflation in this framework, the equation also makes clear that unemployment is by no means the only determinant of inflation. According to the FRB/US model, since the mid-1960s, movements in this unemployment gap account for only about 20 percent of the variation in consumer price inflation, when measured in terms of its year-to-year change. This estimate includes the direct influence of unemployment on inflation shown in the equation as well as the indirect influence that is transmitted through the behavior of expected inflation.
The top panels of exhibit 3 show the change in inflation from one four-quarter period to the next plotted against the unemployment rate over the past forty years, with a split at the end of 1983. In each panel, we present the simple regression line. As you can see, the lines are downward sloping, indicating that periods of high unemployment tend to be associated with declining inflation. But it is also clear that the points in the post-1983 period are clustered in a narrower range than in the 1960 83 period. And as Art Rolnick noted earlier, the average slope of the relationship is considerably shallower in the post-1983 period than in the earlier period—indeed, the slope falls by about half.
The change in the relationship between inflation and unemployment could have occurred for many reasons, but we find it plausible to attribute at least some of it to a change in the way that monetary policy has been conducted. As highlighted in the first bullet point of the lower panel, many studies suggest that since the early 1980s monetary policy has moved more aggressively to stabilize the economy than it did in the 1960s and 1970s. In the staff’s FRB/US macroeconomic model, such a change in monetary policy reduces the sensitivity of inflation to unemployment by altering the way households and businesses form their expectations of inflation. In particular, under a more aggressive policy, low unemployment is no longer as strong a signal of higher future inflation, and so expectations are better “anchored.”
Based on simulations of the FRB/US model, we find that a conventional representation of this change in policy would reduce the sensitivity of the change in inflation to unemployment by about a third and thus would account for more than half of the reduction in the sensitivity seen in the top panels. Other models imply similar reductions, although the precise results are sensitive to the model used and the exact specification of the change in policy. While the policy shift predates the low-inflation episode on which our attention is focused, this timing means that by the late 1990s the change in policy had been in place long enough that it was likely reflected in the expectations-formation process. As a consequence, the low unemployment rate of the late 1990s induced less deterioration in inflation expectations, and thus in actual inflation as well, than would have been expected from the average historical relationship between inflation and unemployment.
It is important to keep in mind that policymakers cannot “exploit” the new, lower sensitivity of inflation to the unemployment rate. If policymakers were to respond to this lower sensitivity by moving less aggressively when unemployment fell below the natural rate, this reversion to a less aggressive policy would, over time, alter the way in which expectations are formed, leading to an increase in the sensitivity of inflation to economic slack. David Lebow will now continue our presentation.