On the death of the Phillips curve: further evidence.

• Author: Reichel, Richard
• Position: Unemployment and inflation rate analysis

In a recent contribution to this journal, William Niskanen (2002) convincingly showed that the Phillips curve does not exist in the United States, at least not in the form usually proposed. Though there is a negative short-run relationship between unemployment and the inflation rate, the long-run correlation is positive. During the sample period 1960 through 2001, the nonaccelerating inflation rate of unemployment (NAIRU) is estimated at a low 3.7 percent, corresponding to zero inflation. These striking results provide a starting point for further research. In this article, I apply cointegration techniques to reproduce Niskanen's results and expand the analysis to other industrialized countries. The estimation results show that the concept of the Phillips curve should indeed be buried. As a policy guideline it is totally useless.

Niskanen's Model

Niskanen proposed the following functional form of the relationship between the unemployment rate U and the inflation rate I (t is the time index):

(1) [U.sub.t] = [mu] + [[gamma].sub.1][U.sub.t-1] + [[beta].sub.0][I.sub.t] + [[beta].sub.1][I.sub.t-1] + [[epsilon].sub.t].

This autoregressive distributed lag model (ARDL) relates current unemployment to the current inflation rate, the inflation rate lagged once, and the lagged unemployment rate. An ordinary least squares (OLS) estimation based on the period from 1960 through 2001 (annual U.S. data) yielded the following equation:

(2) [U.sub.t] = 1.487 + 0.594[U.sub.t-1] - 0.229[I.sub.t] + 0.464[I.sub.t-1].

The estimated coefficients are highly significant, and the overall fit ([R.sup.2] = 0.81) is quite impressive. Surprisingly, the alleged Phillips curve is entirely short-run in nature. An increase in this year's inflation rate of 1 percentage point lowers overall unemployment by 0.23 percentage points in the short run. However, in the long run the negative effects of inflation dominate. Calculating the minimum sustainable unemployment rate by dividing the constant term by (1-0.594) gives a "natural" rate of 3.7 percent, corresponding to a zero inflation rate.

Thus, the major lesson of Niskanen's analysis is that there is no tradeoff between unemployment and inflation that could be exploited by policymakers. On the contrary, a stable price level will minimize unemployment in the long run.

Comments on Niskanen

Although Niskanen provides convincing evidence for the death of the Phillips curve, several comments are warranted:

1. It would be interesting to estimate his model for other countries in order to cheek whether the Phillips curve is still alive elsewhere. Theoretically, Niskanen's U.S. results could be an outlier.

2. Preliminary estimates of Niskanen's ARDL model for other countries gave somewhat puzzling results. For example, I obtained a "natural" rate of unemployment of 14.5 percent for Germany and 21.9 percent for the Netherlands, which seems unrealistically high. In most eases, the constant term of the equation on which the computation of the "natural" rate is based, is hardly significant. Thus, something may be wrong with the ARDL specification.

3. Methodologically, the estimation of an ARDL model by OLS is appropriate if the standard regression assumptions are not violated. Classical OLS is an efficient estimator when the variables included are stationary or integrated of order zero (I(0)). This implies that they do not exhibit trends or a changing variance over time (Greene 2003: 608). In our ease, however, this is doubtful. The hypothesis of stationary data...