For many years, the so-called Phillips curve has provided the inflation forecast framework underlying much of the FOMC’s policy discussions. Yet both before and especially after the financial crisis, the Fed and FOMC members have consistently under-forecasted inflation. A new working paper from the Federal Reserve Bank of Philadelphia adds to what is now a large and growing literature spanning the past 20 years, documenting the poor performance of the framework as a basis for policy.
Briefly, as Chair Yellen discussed in a speech at UMass-Amherst, the model posits that inflation is a function of real side variables, and most specifically that movements of inflation away from its trend value are a function of the deviation of the rate of unemployment from its natural rate (the rate of unemployment that would exist if the real economy were in a state of long-run equilibrium). In other words, the model suggests that the current rate of unemployment, 4.3% ,which is arguably below the natural rate, should be causing a surge in inflation as employers bid up wages and pass them on in the form of higher prices. But that isn’t happening and hasn’t happened in the US economy since the Great Recession. In fact, as unemployment has dropped, inflation has also declined somewhat.
The problem of a bankrupt theory was highlighted in the most recent FOMC minutes but has gone largely unnoticed by economic commentators. What did the minutes suggest was the nature of the discussion at the table? First, participants argued that the unemployment rate was likely to decline further and potentially significantly overshoot the full employment level. Despite this potential – and despite the Phillips curve theory – there has been a notable absence of wage pressure. One rationale offered was that the lack of wage pressure could be due to compositional changes in the labor market, as most of the hiring has been for lower-wage workers. On the inflation front, the minutes were vague in explaining the recent softness in inflation, referring to “idiosyncratic factors” that were unnamed but hypothesized to keep inflation low for the second half of the year.
But then the minutes get interesting, because they reflected doubt about the framework’s suggestion that for a given rate of “expected inflation, the degree of upward pressures on prices and wages rose as aggregate demand for goods and services and employment of resources increased above the long-run sustainable levels.” The minutes note that a “few” participants questioned the usefulness of the framework for inflation forecasting. Nonetheless, “most” of the participants reaffirmed the “validity” of the model but then went on to attempt to rationalize why it might not be working. The reasons given included a reduction in the responsiveness of inflation to low unemployment (a reduction in the coefficient on unemployment in the model), the problem of using the unemployment rate, which may be an imperfect measure of labor market slack in the model, lags in the model, the influence of international market conditions reducing the impacts of tight labor markets on pricing power, and the influence of technology. Finally, it was noted that at least two participants argued that there may actually be a nonlinear relationship between unemployment and inflation, with the impact becoming increasingly strong as labor markets become increasingly tight. However, other participants argued there was little or no empirical support for such a nonlinear relationship.