Has the Phillhs Curve Shifted? Some Additional Evidence

IN AN EARLIER ISSUE OF Brookings Papers on Economic Activity, George Perry argued that the Phillips curve, measured in the conventional way, had shifted to the right in recent years, and estimated that a 4 percent overall unemployment rate would produce about 1 /2 percentage points more inflation per year than was the case in the mid-1950s.1 Perry's conclusion was based on a wage equation that used, instead of the overall unemployment rate, two other measures of labor market conditions: (1) a weighted unemployment rate, in which the unemployment rate for each age-sex group is weighted by the relative hours of work and wage levels of that group; and (2) an unemployment dispersion index, which measures the variance of unemployment rates among different age-sex groups. The use of weighted unemployment in the wage equation appears to be required simply to take account of the arithmetic of aggregation: A given percentage increase in wages of low-paid and part-time workers has less impact on an aggregate wage index than does the same increase won by higher-paid full-time workers. Perry's use of the dispersion index reflected two hypotheses that appear to be reasonable on a priori grounds: Substitution among different age-sex groups in the labor market is imperfect; and the elasticity of wage response to unemployment increases as the unemployment rate decreases. As a consequence, the overall rate of wage in-