Are Phillips Curves Useful for Forecasting Inflation?

thumbnail of forecasting-with-phillips-frbm Original by A. Atkeson, L.E. Ohanian, 2001, 10 pages 

This summary note was Posted on

  • Not proven to be useful for forecasting inflation
  • For the last 15 years economist have not produced a version of the Phillips Curve (PC) that makes more accurate predictions than a naive model presuming inflation over the next four quarters is the same as last four quarters
  • Given the weak theoretical and empirical underpinnings of the various incarnations of the PC we conclude that the search for yet another FC based forecasting model should be abandoned.

Types of models

  • Phillips curve is an equation that relates unemployment rate to a measure of the inflation rate
  • When unemployment is below a baseline rate of unemployment inflation rates tend to rise over time (and the otherway round)
  • The baseline inflation rate of unemployment is known as the non-accelerating inflation rate of unemployment (NAIRU)
  • NAIRU Phillips curves are widely used to produce inflation forecasts both in academic literature and policy making institutions such as the Federal Reserve (FR)
  • Modern specifications of NAIRU PC differs from early ones in that unemployment or some other measure of economic activity is used to forecast future changes in the inflation rate rather than the inflation rate itself.
  • Stock and Watson NAIRU models include lagged values of the unemployment rate or the activity index and the inflation rate as measured by the personal consumption expenditures (PCE deflator) and the consumer price index (CPI) rather than just the current unemployment rate to forecast inflation
  • The federal reserve’s Greenbook model use two measures of inflation: the gross national product deflator (GNP) and the GDP deflater
  • Naive model from publication predicts inflation over the next four quarters has equal to the inflation over the previous fours quarters: E_t(P(t+4)-P(t)) = 0 with P the percentage change in inflation
  • The textbook NAIRU model used in the publication specifies that the expected change in the inflation rate over the next four quarters is proportional to the unemployment rate minus the NAIRU: E(P(t+4)-P(t)) = Beta(U(t)-U(nairu)) with U the unemployment rate and Beta the slope of the Phillips Curve.

Short history

  • Unemployment has been suggested as an indicator of future inflation as early as 1926 (Fisher)
  • Early studies (1960s) document a negative correlation between unemployment rate and rate of nominal wage growth or the rate of inflation
  • Equations relating unemployment rates and inflation were first called Philips Curves
  • However there is no presumption that a statistical relationship observed in one economic environment would be stable enough to be useful for forecasting inflation when that economic environment changes
  • Theory predicts than any relationship between current unemployment and future inflation observed in historical data should be expected to change as the economic environment changes
  • There is a break point in the economic environment after 1970
  • Regression line from more recent data (1970-99) shows no relationship between unemployment and inflation
  • Lucas and Sargent (1979) argue that the breakdown of the PC and more complicated derivatives represents an “economic failure on a grand scale”.

More info: