RDP 9608: Modelling the Australian Exchange Rate, Long Bond Yield and Inflationary Expectations 5. Conclusion

There is no single, simple conclusion to be drawn from this research, but, rather, a series of points can be made.

Interest rates and exchange rates now form part of the transmission mechanism by which policy changes feed through to the broader economy. Expectations play a critical role in this mechanism, affecting both the timing and speed of transmission. Theoretical discussions of interest and exchange rate markets typically characterise expectations as forward looking. However, it has been difficult to model this type of behaviour within an empirical framework.

One approach has been to rely on the relevant components of full-scale, intertemporal macroeconomic models. These models typically embody theoretically consistent, long-run properties and rational, forward-looking expectations in the financial sector. In Australian macroeconomic models, such exchange rate and bond yield equations are not estimated; they reflect orthodox theoretical considerations including uncovered interest parity and the term-structure hypothesis. But the textbook-style results produced by these macro models have limited relevance for practical policymaking.

Alternatively, single-equation, behavioural models can be used to document the observed historical relationships in the data. These have typically assumed that expectations are formed adaptively, that is, are backward looking. The research in this paper concentrates on introducing forward-looking, policy elements into behavioural models of the Australian real exchange rate and long bond yield.

Given that expectations play a central role in determining the responses to various shocks, the macroeconomic and behavioural model approaches are probably best distinguished by a comparison of impulse response functions. In particular, these two methodologies provide different characterisations of the behaviour of the real exchange rate. In the macro-model framework, monetary policy shocks elicit an instantaneous change in the real exchange rate which is subsequently and gradually unwound. In contrast, the behavioural model developed in this paper does not return this instantaneous ‘jump’ response. Instead, the real exchange rate only gradually transmits a change in monetary policy through to the broader economy so that the full impact of the policy change through this channel is felt with a lag. Despite very different adjustment paths, both models produce final responses of a similar order of magnitude.

On the other hand, about half of a sustained terms of trade shock is finally passed through to the real exchange rate in the macro models; this occurs through an initial jump in the exchange rate, followed by gradual adjustment towards the long run. While this result is theoretically appealing, it does not describe the actual behaviour of the Australian real exchange rate. The behavioural model estimates that the real exchange rate moves much more closely with terms of trade shocks, regardless of whether the shocks are temporary or sustained over very long periods. Some overshooting is estimated to occur immediately. This result is puzzling, but it is consistent with the idea that agents in the foreign exchange market look forward over only a relatively short horizon. The inherent difficulty of incorporating inefficient mechanisms into the macro-model framework may be one source of the disparity between the macro-model results and those recorded by the behavioural models.

Incorporating forward-looking behaviour into a bond yield equation is less straightforward. In this paper, it is achieved by explicitly modelling the formation of inflation expectations. Expectations are generated from a series of assessments about the probability of shifting between a high and a low inflation regime. This is particularly apt in Australia, since a discrete shift in inflationary expectations occurred in the early 1990s. The superior performance of the shorter-horizon measures of inflation expectations suggests that some myopia may exist in this market as well. Further work in this area might consider whether there are roles for both forward and backward-looking elements within the long bond rate model.