The Federal Reserve Board is responsible for formulating macro stabilization policy. More specifically, the Federal Reserve Board seeks tradeoffs between inflation and unemployment rates. Fed officials need meaningful data to formulate useful policies. Data on the unemployment rate that coincides with zero inflation provides a starting point for policy formulation. Fed officials also need data on the rate at which inflation reduces unemployment rates. Finally, data on any correlation between inflation and unemployment rates enables Fed officials to estimate the probability of policy success. What do economists know about all of this?
Irving Fisher discovered a tradeoff between unemployment and inflation in the 1920’s.1 Economists did not notice this tradeoff until Alban Phillips published on this topic in 1958. Economists initially assumed that the “Phillips Curve” tradeoff was stable. Hence, central banks officials could choose an unemployment rate by setting the right inflation rate. Most modern economists admit that the Phillips Curve shifts now and then. Milton Friedman asserted that the Phillips Curve shifts whenever we update our inflation expectations.2 Some modern economists say that Phillips Curves don’t exist. Robert Lucas argued that inflation-unemployment tradeoffs change according to how we interpret and react to Fed policy moves, in which case there is no stable-reliable Phillips Curve.
How well does CPI data on inflation correlate with unemployment data in the USA? Unemployment and the year over year CPI have an 85% correlation- from 1948 to 1952 (see figure 2 below). The correlation between monthly inflation and unemployment from 1948 to 1952 is only 5% (see figure1). Both estimates for the 1948-52 Phillips Curve show a correlation for concurrent data, no lead or lag for unemployment.
The correlation between unemployment and year over year inflation shrinks to a paltry 6% from 1954 to 1972 (see figure 4). However, one can conjure up a 21% Phillips Curve for the 1954 to 1972 time period by switching to the monthly CPI (see figure 3). Why should monthly inflation work from 1954-1972, and year over year inflation work better from 1948-1952?
Monthly and year over year inflation rates each correlate with unemployment better during the 1954-1972 period if unemployment leads inflation by three months. This lead for unemployment fits with Franco Modigliani’s Phillips Curve theory- according to which there is a “non-accelerating” level of unemployment at which inflation is zero. Unemployment below the non-accelerating level is inflationary, unemployment above the non-accelerating level is deflationary.3 If it is the case that low unemployment- tight labor markets- cause all prices and wages to rise, then it is likely that the effects of unemployment on inflation would take some time, perhaps two or three months. This makes sense, but why was the Phillips Curve relationship concurrent from 1948 to 1952? Also, Fed officials who believe in the Phillips Curve use interest rate cuts to “stimulate the economy”, to reduce unemployment- at the cost of higher inflation. But, if current unemployment drives inflation rates three months later, then Phillips Curves are useless as a guide for Fed officials who aim to choose an inflation rate so as to achieve an optimal inflation-unemployment tradeoff.
Both the monthly and year over year inflation rates have a near 40% correlation with unemployment from 1973 to 1985, with 3 month leads for unemployment rates (see figures 5 and 6). Since both measures of inflation fit well for 73-85 , both seem legit. However, the slope coefficients for these two 1973-1985 Phillips Curves are different. Which slope coefficient was a better guide for policy for this period, or for right now? This issue matters because none of the slope coefficients for modern Phillips Curve data match, and may differ by large amounts. The idea of a short run Phillips Curve predicts parallel shifts of Phillips Curves. Actual data suggests that Phillips Curves tend to rotate as much as shift.
Federal Reserve officials often rely on the PCE inflation rate, rather than the CPI rate. Phillips Curve estimates using either monthly and year to year PCE data entail the same type of inconsistencies as seen with CPI data, for roughly the same time periods. In fact, PCE data yields even more possibilities for Phillips curve slope and correlation coefficients- more possible Phillips curves to choose from as guides to Fed policy.
Data on unemployment and inflation rates from 1948 to 1985 doesn’t reveal stable estimates of Phillips Curve tradeoffs. Can data since 1985 serve as a better guide for Federal Reserve policy?
[This article first appeared at "On the Other Hand..."]
- 1The Federal Reserve Board has had a legal dual mandate to achieve a reasonable tradeoff between inflation and unemployment rates since 1977.
- 2Ironically, Friedman’s modern opponents have developed his idea of a short run shifting Phillips Curve in an effort to justify activist inflationary policies that Friedman opposed. The best of these models was developed by Guillermo Calvo.
- 3McGraw-Hill. pp. 209–219.) and Hayek F.A., Full Employment, Planning and Inflation (1950) Institute of Public Affairs Review.
