RDP 9201: The Impact of Real and Nominal Shocks on Australian Real Exchange Rates 5. Conclusions and Summary

Unit root tests of real exchange rates examine the issue of whether or not there are shocks which have permanent effects. In this paper the focus is on two Australian dollar real exchange rates and it is shown that they are characterised by unit roots. In light of models of real exchange rate determination any other result would have been surprising. The more interesting question addressed in this paper is how important are the shocks which have permanent effects relative to those which have just temporary effects. To answer this question a technique developed by Blanchard and Quah (1990) is used, together with restrictions on exchange rate and unemployment dynamics suggested by some standard models of exchange rate determination. Important roles for both types of shocks are found.

The paper begins with an examination of the links between productivity growth and long-run real exchange rate changes. It is argued that productivity growth in Australia's traded goods sector has been very slow compared to that of Japan and roughly comparable to that in the United States. These differences have been reflected in differential rates of change in the relative price of non-tradeables to tradeables in the three countries. It is argued that Australia's productivity growth relative to Japan on the one hand, and the United States on the other, can help explain the difference in the behaviour of the two real exchange rates over the last two decades.

In the second part of the paper the focus turns to an explicit consideration of two types of shocks: those with temporary effects on the real exchange rate and those with permanent effects. Using a model of the real exchange rate these shocks are given a economic interpretation. The “permanent” shock is considered a real shock and the “temporary” shock a nominal shock.

In general, the dynamic responses are similar to those suggested by the model. Positive nominal shocks lead to a substantial temporary real depreciation and a fall in Australian unemployment. Positive real shocks show a similar unemployment response and real appreciation with some short-run undershooting of the real exchange rate. The variance decompositions show that for the AUD/YEN rate, real shocks account for the bulk of the forecast error variance at all horizons. In contrast, Australian nominal shocks account for over half of the short-run forecast error variance for the AUD/USD rate. While the importance of this shock falls as the forecast horizon lengthens, it remains relatively important for some time: at the 2 year horizon it is still accounting for a quarter of the variance.

The results support the view that both real and monetary factors are important in understanding real exchange rate behaviour, especially in the short run. While real exchange rates have a unit root, shocks which have a temporary effect also play an important role. These results, however, must be interpreted with some caution. First, the exchange rate regime during the period of study has not been a completely clean float. Prior to December 1983 the exchange rate was set by a daily adjustable peg against a trade weighted basket of currencies. Since December 1983 it has been floating, but with periods of sizeable foreign exchange market intervention. No account has been made for these deviations from a clean float. Additionally, the difference in some of the results achieved using the AUD/USD and AUD/YEN rates suggest that estimation technique may not be capturing the complete dynamics of the AUD/USD rate. Second, the pseudo standard errors are relatively large making it difficult to quantify the observed effects with any great degree of precision. The results might thus be best interpreted as suggestive rather than definitive. Thirdly, the same caveats that Blanchard and Quah make concerning the low dimensionality of the system and the possibility that unemployment does in fact have a unit root apply here. Finally, the bilateral rates have been examined individually. It may be more appropriate to examine them in one system. Unfortunately, doing so increases the number of identifying restrictions needed beyond that supplied by the model. Notwithstanding these caveats, the results do suggest that moving beyond the standard unit root tests offers additional insight into the behaviour of real exchange rates.