RDP 9601: Why Does the Australian Dollar Move so Closely with the Terms of Trade? 5. Discussion and Conclusions

The paper has documented the predictability of the Australian real exchange rate over horizons of between one and two years. It has also presented results suggesting the presence of quite large and variable expected excess returns to holding Australian dollar assets. How can we explain these results?

One potential explanation is the presence of a time-varying risk premium. That is, the excess returns we document may not represent unexploited profit opportunities but, instead, the compensation demanded by risk-averse investors for bearing risk. Under such an interpretation, the positive (negative) excess returns displayed in Figure 4, would represent the risk-premium (discount) to holding Australian dollar assets relative to the trade-weighted portfolio of foreign bonds – ie, the additional compensation demanded by investors for holding the riskier Australian (trade-weighted foreign) bond. If this explanation is valid, then Australian bonds must be viewed by investors as much riskier than foreign bonds when Australia's terms of trade are low but expected to improve, and much less risky when the terms of trade are high but anticipated to fall. But the idea that the risk premium on Australian bonds could be as large and volatile as suggested by the profile of expected excess returns shown in Figure 4, seems implausible to us.

A second possible explanation is that rational foreign exchange market participants only gradually learn of the predictability of exchange rate changes. It may take time for the market to learn of the existence of considerable longer-run predictability in Australia's terms of trade, and that this in turn renders medium and longer-run changes in Australia's real exchange rate forecastable. In which case, over time, one would expect to see the current level of the exchange rate increasingly reflect the market's implicit forecast of its level into the future. If so, the strong association between fluctuations in Australia's terms of trade and real exchange rate should decline over time.

A third possible explanation for our results is ‘short-termism’ of foreign exchange market participants. Recall that forecasting model B does not provide any statistically-significant information on exchange rate changes over short horizons, but does provide significant and apparently unbiased forecasts over horizons of one to two years. This indicates that in order to profitably exploit the real exchange rate – terms of trade link, it is necessary for investors to have a trading horizon of at least a year. However, it may be difficult for key market participants, such as fund-managers and institutional investors, to take long-term open positions in a foreign currency. For institutional reasons – related, for example, to the time-period over which their performance is assessed – their relevant trading horizon may necessarily be a few months, rather than the one or two years required to exploit the inherent predictability of the Australian currency. With short-term investment horizons apparently widespread in the foreign exchange market, it may not be so surprising to observe a relationship between the real exchange rate and the terms of trade that can only be exploited by (relatively scarce) long-horizon investors.

Finally, our paper has implications for empirical modelling of the Australian exchange rate. Most existing Australian empirical macroeconomic models, such as the Murphy model, the MSG2 model, and the Treasury's TRYM model, use the uncovered interest parity condition, sometimes allowing for a constant risk-premium, combined with the assumption of rational (or quasi-rational) expectations, as the central relationship determining exchange rate outcomes. Our results suggest that this is inconsistent with a key feature of the data. Predictable changes in Australia's terms of trade provide significant information about movements in the Australian real exchange rate over horizons of one to two years. Over these horizons, there appear to be quite large, and variable, expected excess returns on Australian dollar denominated assets – contrary to the predictions of uncovered interest parity.