RDP 9706: Is the Phillips Curve A Curve? Some Evidence and Implications for Australia 5. Policy Implications of Non-linearities

The basic policy prescription of the linear natural-rate framework is to tighten (loosen) policy whenever unemployment exceeds (is less than) the natural rate/NAIRU (recall that these are equal in the linear model).^[17] In the simple linear model, while symmetric policy mistakes will affect the variability of unemployment, they do not affect its average level (De Long and Summers 1988). If the model is extended to allow for asymmetric adjustment of inflation expectations there may be a linkage between the mean of unemployment and its variability. Nevertheless, in general, the concerns raised by Staiger, Stock and Watson (1997) about the uncertainty in the linear natural-rate framework are of second-order importance, unless there is some additional cost of output/unemployment variability, for example, if output variability is directly in the policy-maker's objective function.

In contrast, in the non-linear natural-rate framework, the variability of the unemployment rate has a direct effect on the average level of unemployment.^[18] This means that policy mistakes that increase the variability of unemployment will increase the average level of unemployment (which we refer to as the natural rate of unemployment). For example, if monetary policy is slow to respond to a rise in inflationary pressures caused by unemployment falling below the NAIRU, the convexity of the curve requires that to return inflation to its targeted level, unemployment will have to be above the NAIRU for a longer period (or alternatively will need to be higher) than it was below it. Thus, the average unemployment rate – the Friedman natural rate – will be higher. Good stabilisation policy, on the other hand, may reduce the gap between the natural rate and the NAIRU by reducing the variability of unemployment.

Consequently, convexity of the short-run Phillips curve also points to the need for pre-emptive monetary policy. Monetary policy needs to be forward-looking so that inflationary pressures are quickly doused (to prevent inflation expectations from rising), in order to avoid costly recessions down the track. Clark, Laxton and Rose (1995) present estimates of the gains from a forward-looking monetary policy approach in a non-linear world, in terms of the ability to avoid larger recessions. These results also provide support for a conservative approach to policy-making, as the inflationary consequences of a slight overheating of the economy are greater than the consequences of a slight underheating.

While a non-linear Phillips curve provides support for pre-emptive monetary policy, it also implies that deep recessions should be avoided. The convexity means that the extra disinflationary impact from a slightly deeper recession is likely to be marginal, so that deep recessions are unnecessarily costly. Consequently, non-linearity implies that a gradualist approach to disinflation is preferable. In contrast, Ball (1994) exposits a model where a ‘cold turkey’ approach is preferable. However, his results depend on the effects of monetary policy actions on the public's expectations through credibility channels which we do not model here.

In the recent Journal of Economic Perspectives volume, a number of authors argue that the Federal Reserve Board could test the limits of the economy without any adverse long-run consequences (Gordon 1997 and Stiglitz 1997). Such a conclusion may be justified in a linear world. However, in a non-linear world such a prescription could be dangerous. If the central bank, in testing out the limits of the economy, induces overheating, again, the resultant period of depressed activity necessary to bring down inflation would more than offset the gains from the period of higher activity. Given the large uncertainty about the true level of the NAIRU, there may be some justification in attempting to test the limits, but the convexity implies that such an approach must proceed cautiously.

Moreover, an assumption of concavity in the short-run Phillips curve (as suggested by Eisner (1996)) would be dangerous if, in fact, there is convexity. In the limit, concavity implies that increasing the variance of unemployment (that is, larger amplitudes of business cycles) will reduce the average level of unemployment. This implication does not square well with the developments in the Australian economy over the past twenty years.

Footnotes

This prescription has recently been challenged by Chang (1997) and Espinosa and Russell (1997) who argue that the uncertainty about the value of the NAIRU renders the framework inoperative from a policy perspective. [17]

This basic implication of convexity was noted by Mankiw (1988) in commenting on De Long and Summers (1988). [18]