RDP 1999-11: A Structural Vector Autoregression Model of Monetary Policy in Australia 4. Conclusion

Beginning with Sims (1980), small structural vector autoregression models have become an increasingly popular means of modelling monetary policy. Specifying a structural vector autoregression model suitable for analysing monetary policy in Australia, however, has proved to be a difficult task, as it has for other small open economies. Typically, these types of models predict behaviour that is inconsistent with economic intuition. Most notably, they predict a rise in the rate of inflation and a depreciation of the nominal exchange rate in response to a rise in the policy interest rate. The accepted explanation for these results is that the systematic component of monetary policy has been inadequately modelled. As yet, however, there is no consensus on the best way to resolve these problems.

Faced with these difficulties, we estimate a small dimensioned structural vector autoregression model that has been successfully applied to a number of economies by Kim and Roubini (1999). We find that the Kim and Roubini model, amended slightly, provides a reasonable empirical description of important features of the Australian macroeconomy. Most importantly, the qualitative responses of the price level and the exchange rate to a change in monetary policy are consistent with intuition. Further, in response to changes to monetary policy, we observe changes to output and prices, the magnitude and timing of which are consistent with other empirical work, both for Australia and for other countries. Finally, the model also provides sensible predictions for the effects on Australian output and the price level of external interest rate and exchange rate shocks.

We also use the model to consider the response of monetary policy to either external business cycle shocks or nominal exchange rate shocks. We do so by comparing the response of the macroeconomy with and without the estimated endogenous response of monetary policy to either of these shocks. The model suggests that monetary policy has served to dampen the effects of external business cycle shocks. For the exchange rate shocks, the results indicate that monetary policy has dampened the subsequent effects on the nominal exchange rate of the initial shock.

In terms of specifying a model that provides reasonable predictions, we find two features of the model to be important. First, the monetary policy reaction function, the systematic component of monetary policy, needs to be specified so that policy responds contemporaneously to external business cycle effects. Second, we require a structure that models the strong interdependence of the domestic interest rate and nominal exchange rate. Both of these features seem critical to obtaining price and exchange rate responses to monetary policy that accord with economic intuition. The latter feature, however, poses some difficult estimation problems that we do not fully resolve; it is likely that our estimates of this relationship are not very robust and further work in this direction is still required. A further weakness of the model that we document is parameter instability. A natural extension of the work here is to determine whether the instability we observe can be adequately modelled.