RDP 2015-12: Modelling the Australian Dollar 1. Introduction

The decision to float the Australian dollar in 1983 is widely recognised as having been beneficial for the Australian economy (Beaumont and Cui 2007; Stevens 2013). In particular, the floating exchange rate has played a crucial role in buffering the Australian economy from external shocks, in part by allowing the Reserve Bank of Australia (RBA) to better control domestic monetary conditions.

Although the exchange rate has been market-determined for many years, interest in assessing its level relative to some ‘benchmark’ remains strong. This interest has resulted in a broad range of theoretical and statistical models being developed to examine this, both in the academic literature and by practitioners. In part, this variation reflects well-documented difficulties in explaining the behaviour of exchange rates. More fundamentally though, it reflects differences in the questions being addressed: for example, a model that seeks to test theories about the determinants of exchange rates will differ from a model that seeks to forecast future levels of the exchange rate or to estimate the level of the exchange rate that is likely to help achieve particular economic outcomes.

This paper focuses on the level of the real exchange rate. More specifically, it attempts to quantify the extent to which the real exchange rate is consistent with the level that would be expected based on its historical relationship with other variables. While this approach will not directly answer the question of whether a given level of the real exchange rate is likely to help achieve particular economic outcomes, it may nevertheless provide a useful starting point for such discussions.^[1]

Against this background, this paper describes an error correction model (ECM) of the Australian dollar real trade-weighted index (RTWI). The ECM estimates an ‘equilibrium’ level of the exchange rate based on its historical relationships with variables that are likely to have affected the RTWI over the medium term (and which also have a theoretical justification for doing so). Quantifying the extent to which the RTWI is consistent with the ECM's ‘equilibrium’ level is one of the methods used by RBA staff as a starting point when thinking about the level of the exchange rate, and so complements, but is not a substitute for, a broader analysis of economic indicators and other models that have been developed by RBA staff.^[2]

Despite the well-documented difficulties in empirically explaining movements in exchange rates, the RTWI has displayed a strong and consistent relationship with the model's key explanatory variables – most notably the terms of trade – over the medium term. Nevertheless, the RTWI has on occasion displayed large and/or persistent divergences from the model-implied ‘equilibrium’. While such divergences are typically explained largely by the model's short-run dynamics, it is nevertheless still worth examining whether they also reflect other factors that are not adequately captured in the baseline model.

In view of this, this paper examines the deviation between the observed RTWI and the ECM's estimated equilibrium RTWI in recent years in more detail, and considers whether there is evidence that ‘unusual’ factors held up the value of the Australian dollar during this period. This is done in two ways. First, the baseline model is used to examine whether there is any evidence that the estimated relationships between the RTWI and the standard explanatory variables have changed over time. Second, the baseline model is augmented with additional explanatory variables in an attempt to capture two of these potential ‘unusual’ influences more directly: the resources boom; and foreign central banks’ use of unconventional monetary policies. The intent of this latter exercise is to examine whether these recent developments have revealed some omitted variables that have always been theoretically and empirically relevant determinants of the RTWI, but which have previously been more difficult to identify.

The remainder of this paper proceeds as follows. The next section provides a high-level review of the theoretical and empirical exchange rate literatures. Section 3 then provides an overview of the baseline model, while Section 4 examines whether there is any evidence that the relationships between the RTWI and the baseline model's explanatory variables have changed systematically over time using a Markov-switching extension. Sections 5 and 6 then motivate and present augmented versions of the baseline model that attempt to better incorporate Australia's resources boom and foreign central banks’ unconventional monetary policy actions through the addition of different explanatory variables. Section 7 concludes.

All of the analysis in this paper uses data available up until the end of 2014. To preview the results, the paper does find some evidence that ‘unusual’ influences have had some effect on the RTWI in recent years, although it is difficult to quantify these effects. Overall, despite these unusual influences, the baseline ECM continues to display robust explanatory power and none of the extensions presented in this paper is clearly superior over a relatively long time period.

Footnotes

For a discussion of some of the difficulties in using formal models to assess the level of the exchange rate that is likely to help achieve particular economic outcomes, see Stevens (2013). [1]

Other models developed by RBA staff which can help to answer a range of broader questions about the relationships between the exchange rate and other macroeconomic variables include: the DSGE model set out in Rees, Smith and Hall (2015), which models the exchange rate using the uncovered interest parity relationship; the Bayesian vector autoregression model set out in Langcake and Robinson (2013); and the structural vector autoregression model set out in Manalo, Perera and Rees (2014). [2]