THE ECONOMETRICS OF MACROECONOMIC MODELLING
Lineages of the Phillips curve
Following Phillips’ (1958) discovery of an empirical regularity between the rate of unemployment and money wage inflation in the United Kingdom, the Phillips curve was integrated in macroeconomics through a series of papers in the 1960s. Samuelson and Solow (1960) interpreted it as a tradeoff facing policy makers, and Lipsey (1960) was the first to estimate Phillips curves with multivariate regression techniques. Lipsey interpreted the relationship from the perspective of classical price dynamics, with the rate of unemployment acting as a proxy for excess demand and friction in the labour market. Importantly, Lipsey included consumer price growth as an explanatory variable in his regressions, and thus formulated what has become known as the expectations augmented Phillips curve. Subsequent developments include the distinction between the short-run Phillips curve, where inflation deviates from expected inflation, and the long - run Phillips curve, where inflation expectations are fulfilled. Finally, the concept of a natural rate of unemployment was defined as the steady-state rate of unemployment corresponding to a vertical long-run curve (see Phelps 1968 and Friedman 1968).
The relationship between money wage growth and economic activity also figures prominently in new classical macroeconomics; see, for example, Lucas and Rapping (1969, 1970), Lucas (1972). However, in new classical economics the causality in Phillips’ original model was reversed: if a correlation between inflation and unemployment exists at all, the causality runs from inflation to the level of activity and unemployment. Lucas’s and Rapping’s inversion is based on the thesis that the level of prices is anchored in a quantity theory relationship and an autonomous money stock. Price and wage growth is then determined from outside the Phillips curve, so the correct formulation would be to have the rate of unemployment on the left-hand side and the rate of wage growth (and/or inflation) on the right-hand side.
Lucas’s 1972 paper provides another famous derivation based on rational expectations about uncertain relative product prices. If expectations are fulfilled (on average), aggregate supply is unchanged from last period. However, if there are price surprises, there is a departure from the long-term mean level of output. Thus, we have the ‘surprise only’ supply relationship.
The Lucas supply function is the counterpart to the vertical long-run curve in Lipsey’s expectations augmented Phillips curve, but derived with the aid of microeconomic theory and the rational expectations hypothesis. Moreover, for conventional specifications of aggregate demand (see, for example, Romer 1996: ch. 6.4), the model implies a positive association between output and inflation, or a negative relationship between the rate of unemployment and inflation. Thus, there is also a new classical correspondence to the short-run Phillips curve. However, the Lucas supply curve when applied to data and estimated by ordinary least squares (OLS), does not represent a causal relationship that can be exploited by economic policy makers. On the contrary, it will change when, for example, the money supply is increased in order to stimulate output, in a way that leaves the policy without an effect on real output or unemployment. This is the Lucas critique (Lucas 1976), which was formulated as a critique of the Phillips curve inflation-unemployment tradeoff, which figured in the academic literature, as well as in the macroeconometric models of the 1970s (see Wallis 1995). The force of the critique, however, stems from its generality: it is potentially damaging for all conditional econometric models; see Section 4.5.
The causality issue also crops up in connection with the latest versions of the Phillips curve—the forward looking New Keynesian variety—which we return to in Chapter 7.[18] In the United States, an empirical Phillips curve version, dubbed ‘the triangle model of inflation’, has thrived in spite of the Lucas critique—see Gordon (1983, 1997) and Staiger et al. (2001) for recent contributions. As we will argue below, one explanation of the viability of the US Phillips curve is that the shocks to the rate of unemployment have been of an altogether smaller order of magnitude than in European countries.
