RDP 1999-01: The Phillips Curve in Australia Appendix B: Exogeneity Assumptions Relevant to Estimating Phillips Curves

As the Phillips curve is a single equation in a system, estimating Equations (6) and (8) in Section 2 means that some exogeneity assumptions must be invoked. In many recursive vector autoregression (VAR) studies, prices are ordered after output, implying that the latter is weakly exogenous in the price equation. For examples, see the US studies by Sims (1980) and Leeper and Gordon (1992). Australian VAR research has also continued this tradition, as seen in the work by Smith and Murphy (1994) and Dungey and Pagan (1997). The logic behind this ordering comes from the perceived short-run rigidity of prices and the role of inventories in facilitating a de-coupling of prices and output in the short run. Moreover, when unemployment appears as the variable affecting prices, the argument for weak exogeneity is even stronger because movements in the unemployment rate are generally regarded as lagging output. In any case, some identification assumption needs to be invoked and Phillips curve research in Australia has invariably chosen the identification assumption we have just described.

A further consideration is that the forward-looking component of inflation expectations may be correlated with the error term in the price equation if the information set which agents use to forecast inflation is much broader than that formed from the past history of inflation and the contemporaneous and lagged gaps between the unemployment rate and the NAIRU. For example, this may be the case in Equation (6) since it is well known that import prices are useful for forecasting inflation in Australia. This is, however, only an argument for expanding the specification of Equation (6), which we do in the paper. The crucial assumption in Equation (6) is that expectations of inflation are formed before prices are determined.