RDP 9601: Why Does the Australian Dollar Move so Closely with the Terms of Trade? 1. Introduction

It has long been observed that there is a close relationship between commodity prices and the Australian dollar. When the world price of commodities rises, the Australian dollar tends to appreciate. When world commodity prices fall, the Australian dollar tends to fall. As a consequence, the Australian dollar is sometimes described as a ‘commodity currency’.

Since commodities account for a large proportion of Australia's exports, this close link between commodity prices and the Australian dollar is also reflected in a close relationship between Australia's terms of trade and the real exchange rate, as shown in Figure 1 below. As the figure shows, the relationship has, if anything, become closer since the float of the Australian dollar in December 1983.

Figure 1: Terms of Trade and Real Exchange Rate (Real TWI)
June 1970 = 100
Figure 1: Terms of Trade and Real Exchange Rate (Real TWI)

Note: The figure shows the terms of trade for goods and services and the 22 country trade-weighted Australian real exchange rate derived using CPIs to deflate nominal exchange rates. Underlying CPI is used where available.

This relationship has been widely interpreted as evidence of a close link between the exchange rate and fundamentals. McKenzie (1986), Blundell-Wignall and Gregory (1990), Blundell-Wignall, Fahrer and Heath (1993) and Gruen and Wilkinson (1994) all argued that the terms of trade are a fundamental determinant of the real exchange rate for a small commodity-exporting country like Australia. However, while the direction of this link is consistent with economic theory, it is the contention of this paper that the extent to which the Australian real exchange rate responds to movements in the terms of trade provides evidence of short-sighted behaviour in the foreign exchange market. Australia's real exchange rate appears to move too closely with the terms of trade to be consistent with the actions of rational market participants with long investment horizons.

Our argument is easily summarised. As we show, there is substantial predictability of future movements in the terms of trade. This predictability, together with the fairly close co-movement of the Australian real exchange rate and the terms of trade often implies a large expected real exchange rate change. Furthermore, in general, this expected exchange rate change is not offset by an expected Australian/foreign real interest differential, implying the existence of substantial predictable excess returns to holding Australian assets. This observation, however, suggests a relative scarcity of rational investors with long-horizons, because if such investors were instead relatively numerous, they should move the current exchange rate to eliminate most of any predictable excess returns to holding Australian assets.

We develop this argument in four steps. First, we show that Australia's terms of trade are well-described as fluctuating around a slowly declining trend. Deviations from this slow downward trend are quite long-lived, but do not appear to be permanent. Instead, there is strong evidence that the terms of trade revert to their historical trend over time. Thus, at times when the terms of trade are below trend, they can be expected to improve relative to trend, and when they are above trend, they can be expected to fall.

The second step is to estimate simple time-series models describing the terms of trade as well as the relationship between the terms of trade and the real exchange rate since the float of the Australian dollar. These models are used to generate forecasts of the change in the real exchange rate over horizons ranging from one to eight quarters ahead. The generated forecasts are truly ex ante; that is, they are based solely on the past behaviour of the real exchange rate and the terms of trade.

The third step is to compare these forecasts with the actual change in the real exchange rate over the forecast period. As we show, the forecasts provide no statistically-significant information about the change in the real exchange rate over short horizons – one, two or three quarters. By contrast, however, the forecasts provide significant, and apparently unbiased, estimates of the change in the real exchange rate over horizons from four to eight quarters ahead.

The final step is to combine the forecasts of the change in the real exchange rate with estimates of the Australian-foreign expected real interest differential. With these two ingredients, we can estimate the predictable excess return to holding Australian dollar-denominated assets.

The key point of the paper is that the link between the terms of trade and the real exchange rate is strong enough to generate quite large predictable excess returns to holding Australian assets over horizons of a year or more. At different times since the float, the predictable excess return to holding either one or two-year Australian bonds has varied in a range from about −15 to +15 per cent per annum. That is, at some times, an investor using our simple model would have expected a return of about 15 per cent per annum less holding either one or two-year Australian bonds than holding a portfolio of foreign bonds with the same maturity. At other times, the predictable excess return to holding Australian one or two-year bonds was about +15 per cent per annum.

The presence of such large predictable excess returns, sometimes positive, sometimes negative, suggests less than efficient processing of relevant information by the foreign exchange market. There appears to be a relative scarcity of forward-looking participants in the market with an investment horizon of a year or more. It may be that central banks are among the few active portfolio managers in the market with an investment horizon this long.

The next section of the paper, Section 2, examines the dynamic properties of Australia's terms of trade. It reports the results of three statistical tests which suggest that Australia's terms of trade exhibit only temporary fluctuations around a slowly declining trend. A preferred time-series model of the terms of trade is also presented.

Section 3 describes two simple models of Australia's trade-weighted real exchange rate over the post-float period. For both models, the terms of trade is the sole explanator of the real exchange rate. These models are used to generate ex ante forecasts of the change in Australia's real exchange rate. Section 4 explains the forecasting procedure and examines the models' out-of-sample forecasting performance over different time horizons. Estimates of the one and two-year expected excess returns to holding Australian dollar assets are also derived. Section 5 discusses the results and concludes.