On the death of the resurrected short-run Phillips curve: a further investigation.

• Author: Moghaddam, Masoud

In a recent issue of this journal, Richard Reichel (2004) takes issue with the resurrected Phillips curve (PC) in William Niskanen's (2002) article. Accordingly, Niskanen's reformulation of the PC provides empirical evidence for a weak, but statistically significant, short-run (in the same year) tradeoff between inflation and unemployment rates in the United States. Furthermore, the unemployment rate is directly and significantly determined by the one-period lagged value of the inflation rate, which implies an upward sloping PC, consistent with the type of PC explicated by Milton Friedman (1987). Reichel's main point of contention is that the variables in the reformulated version of PC are nonstationary, meaning that statistical properties (such as conditional mean and variance) vary with time. Thus, Niskanen's findings are spurious.

However, variables with the same order of integration, though nonstationary in level, can share a linear trend in the long run--that is, they are cointegrated. If cointegration is confirmed, then the implied error correction model (ECM) is capable of depicting long-run equilibrium along with short-term dynamic adjustments. In that spirit, Reichel examines inflation and unemployment data for the 1960-2001 period in the United States as well as in 15 industrialized nations for integration, cointegration, and error corrections. His overall findings suggest that there appears to be no evidence of cointegration between inflation and unemployment rates, although the short-term PC is observed in Italy and the United States. Most notably, the United States is the only nation in which the short-run PC has also been confirmed in the context of the implied ECM utilized by Reichel. Consequently, he concludes that the U.S. case is an exception to the rule, and that for all practical purposes the PC phenomenon appears to have been dead in the rest of the world.

This article addresses a number of issues both theoretically and empirically associated with the above studies that deserve further consideration. First, the proposed autoregressive distributed lag (ARDL) model reformulated by Niskanen is subject to a specification error in that it lacks the demand pressure mechanism that is necessary for a meaningful test of the contemporary PC. (1) Second, the empirical findings of the ARDL model have been heavily affected by the apparent structural changes. Indeed, when the ARDL model is adjusted for these changes (e.g., adverse supply shocks in the 1970s), the significant short-term PC disappears quickly. Third, the estimated value of the so-called nonaccelerating inflation rate of unemployment (NAIRU) is understated. Fourth, a reliable integration test should take into account the structural breakpoints (temporary or permanent) within the sample period. In fact, when the aforementioned breakpoints are incorporated into Niskanen's model, the variables appear to be stationary--that is, integrated of order zero. Hence, Reichel's main argument regarding integration, cointegration, and the ECM does not hold at least for the United States. Finally, assuming that Reichel's...