A. Introduction and Summary

1. The recent behavior of the Australian exchange rate provides the main motivation for this study. In particular, the key development of the past years that has preoccupied observers of the Australian economy has been the sharp weakening of the Australian dollar (SA) against the U.S. dollar. Although, this decline can be seen as a continuation of a much longer term trend depreciation, the fall in 2000 has been particularly accentuated and resembles the pace of decline at the height of the Asian crisis.

2. Set against a background of strong growth and low inflation, generally sound fundamentals, and relatively favorable economic prospects, the recent fall in the exchange rate has been puzzling. In particular, the variables believed to be the main driving forces of the SA in the past—commodity prices, the current account deficit and associated external imbalances, and interest rate differentials—appear not to be relevant in the recent movement of the currency. Indeed, prices of Australian commodities have been strengthening, the current account deficit has narrowed markedly and the trend looks set to continue, and interest rate differentials versus the U.S. have not moved significantly for this reason, many analysts have focused on a number of ad hoc explanations that have not been supported by a rigorous macroeconomic framework.

3. This paper addresses these issues by examining the relative importance of various kinds of shocks in accounting for movements in the real exchange rate since the end of the Bretton Woods period. Although the main focus of this study is Australia, the inclusion of two other “commodity currencies”—the Canadian dollar and the New Zealand dollar—which the $A has generally tended to track closely may provide additional insights into the behavior of the Australian dollar. In particular, the paper focuses on the long—term relationship between the exchange rate and other macroeconomic variables and makes explicit use of the properties of a standard open economy macro model. A four variable structural vector autoregression (VAR) model is estimated to analyze more systematically how different shocks feed into exchange rate movements and to assess their relative importance. Another important part of this exercise will be to analyze how various shocks have contributed to real exchange rate movements over time.

4. Methodologically, this paper follows the approach proposed by Blanchard and Quah (1989) for the identification of “fundamental” shocks through restrictions on their long—run effects and builds on the literature that has extended that empirical strategy to an open economy setting.² The advantage of this methodology is that it identifies the macroeconomic shocks that could simultaneously affect the variables in the system (e.g., exchange rates, output, relative prices, etc.), without making assumptions on the short—run dynamics or on the contemporaneous exclusion of any shock from any of the equations in the system. An early contribution to the literature was made by Clarida and Gali (1994), who estimated a model with relative output, relative prices, and the real exchange rate for a group of G-7 countries. However, the application of similar models to small open economies has been limited by the additional complication introduced by the need to control for the terms of trade.

5. To take account of the fact that real exchange rate variations are associated with changes in the relative levels of domestic versus foreign variables, this paper estimates the structural vector autoregression (VAR) model with the real effective exchange rate, relative output, the relative price level, and the terms of trade all measured relative to a trade weighted basket of countries (Table III.1). Using long—run identifying restrictions, this methodology allows the identification of four different types of shock: shocks that directly affect the terms of trade; shocks to the long—run level of relative output; demand shocks; and pure monetary, or nominal shocks.³

Table III.1.

Trade Weights

Source: IFS. Sample period 1966-1999.

Table III.1.

Trade Weights

	Australia	New Zealand	Canada
United States	21.9	16.6	56.2
Japan	20.2	15.3	11.8
Germany	8.9	8.6	5.6
United Kingdom	7.8	8.5	3.9
France	4.9	5.4	3.4
Italy	4.6	5.0	2.8
Taiwan POC	3.8	2.7	2.7
Korea	3.4	2.3	2.8
New Zealand	3.3
Canada	3.2	2.9
China	2.9	1.8	1.3
Netherlands	2.9	4.8	1.8
Belgium	2.5	2.8	1.5
Singapore	2.5	1.4
Hong Kong SAR	2.3	1.6	1.4
Sweden	1.8	1.7	1.4
Switzerland	1.6
Brazil	1.3		1.2
Australia		13.9
Denmark		1.6
Spain		1.4
Mexico			2.0

Source: IFS. Sample period 1966-1999.

View Table

Table III.1.

Trade Weights

	Australia	New Zealand	Canada
United States	21.9	16.6	56.2
Japan	20.2	15.3	11.8
Germany	8.9	8.6	5.6
United Kingdom	7.8	8.5	3.9
France	4.9	5.4	3.4
Italy	4.6	5.0	2.8
Taiwan POC	3.8	2.7	2.7
Korea	3.4	2.3	2.8
New Zealand	3.3
Canada	3.2	2.9
China	2.9	1.8	1.3
Netherlands	2.9	4.8	1.8
Belgium	2.5	2.8	1.5
Singapore	2.5	1.4
Hong Kong SAR	2.3	1.6	1.4
Sweden	1.8	1.7	1.4
Switzerland	1.6
Brazil	1.3		1.2
Australia		13.9
Denmark		1.6
Spain		1.4
Mexico			2.0

Source: IFS. Sample period 1966-1999.

6. The estimated model can then be used to describe the reaction of the different variables to the four identified macro shocks and to decompose the historical variation of the forecast error into four fundamental components. The estimated impulse response functions for Australia confirm the positive relationship between the terms of trade shocks and the real exchange rate, but do not support the idea of a positive impact of relative supply innovations. Nominal and demand shocks have the expected response: demand shocks cause an appreciation of the real exchange rate through an increase in the interest rate differential, while nominal shocks are associated with a depreciation. The historical decomposition exercise delivers more interesting results. In particular, it suggest that a negative relative demand component, possibly associated with the compression in interest rate spreads following the fiscal consolidation of the 1990s, accounts for most of the recent downward pressure on Australia’s real exchange rate.

7. The paper is organized as follows; the next section briefly reviews the related empirical literature; Section C reports some stylized facts about the Australian dollar and reviews some of the explanations proposed for the recent exchange rate developments; Section D briefly describes the properties of the general open economy macro model underlying the empirical estimations; Section E describes the empirical methodology; Section F describes the data and reports some time—series characteristics of the variables; Section G reports the main results; and Section H concludes.

B. A Brief Survey of the Literature

8. There is a relatively long tradition of empirical papers focusing on the identification of the sources of exchange rate fluctuations. Overall, these studies have had mixed success in explaining movements in the exchange rate. In an early paper on the subject, Lastrapes (1992) uses restrictions on the long—run behavior of economic variables to distinguish between real and nominal shocks to the exchange rates for the G-7 countries over the 1973-1989 period. His finding suggest that real shocks tended to dominate nominal shocks for both the real and the nominal exchange rates. One shortcoming of this study is that by using a bivariate VAR, it can only identify two kinds of shocks. Indeed, the author acknowledges, that in the presence of multiple kinds of real shocks (e.g. real demand and supply shocks), his identification strategy would be potentially compromised. Clarida and Gali (1994) solve this problem by introducing a three equation open macro model and estimating it using Blanchard—Quah type restrictions to identify supply, demand, and monetary shocks. They find that demand and monetary shocks play an important role in explaining real exchange rate fluctuations, while supply shocks contribute only marginally.

9. Using cointegration analysis, Gruen and Wilkinson (1994) examine the relationship between Australia’s real exchange rate, terms of trade, and interest rate differential over the period 1969-1990. They cannot find a stable relationship between the terms of trade or interest rate differentials and the real exchange rate over the full period. However, they find that since the floating of the Australian dollar at the end of 1983, both interest rate differentials and the terms of trade have stable relationship with the exchange rate. Henry and Summers (1999) estimate a GARCH model of the real $A/US$ bilateral exchange rate for the 1971-1998 period. They conclude that changes in the $A/US$ exchange rate are influenced by changes in commodity prices and in the real US$/Yen exchange rate. Interest rate differentials do not appear to play a significant role.

10. The present study closely follows the methodology used by Fisher (1996) and Prasad (1999) and related to Clarida and Gali (1994). Fisher (1996) estimates a structural var model for Australia and New Zealand with the nominal exchange rate, the price level, and the terms of trade, all measured in relative terms vis—à—vis the U.S. He finds that real terms of trade shocks are far more important than nominal shocks in accounting for nominal exchange rate fluctuations in both countries. In addition, positive real terms of trade shocks have a significantly positive effect on the price level in New Zealand, but the effect is negligible for Australia. Prasad (1999) estimates a similar model with relative output, real effective exchange rate, and the relative price level, all measured with respect to a trade—weighted basket of countries. His main findings are that for Australia real demand shocks have been the main determinant of the real exchange rate movements; while for New Zealand nominal shocks have been relatively more important The present study adds to these results by integrating the approaches in Fisher (1996) and Prasad (1999) and considering a four—variable VAR, which permits the distinction between terms of trade shocks and relative supply shocks that are not directly related to the terms of trade.

C. Stylized Facts

11. This section summarizes some stylized facts about the behavior of the Australian dollar with a view to examining some of the explanations that have emerged for its recent depreciation. In this context, some similarities and differences with the Canadian dollar and the New Zealand dollar will be also highlighted.

12. One stylized fact is the relationship between the real effective exchange rate and the terms of trade—which in these countries is closely related to commodity prices (see Figure III.1). The relationship is particularly strong in Australia, where the terms of trade have traditionally been used to track exchange rate movements.⁴ For this reason, the significant divergence of the Australian dollar from the value predicted by the terms of trade has been especially puzzling.⁵ One possible explanation for such a divergence is a break in the structural parameters of the equation describing the relationship between the exchange rate and its determinants. In particular, some analysts have argued that over the years Australia has become more diversified, gradually losing the characteristics of a commodity based country. Thus, terms of trade shocks may have become progressively less important, relative to other influences on the exchange rate such as interest rate differentials and relative income growth.

View Full Size

Terms of Trade and REER (1970 = 100)

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS and RBA

• Download Figure
• Download figure as PowerPoint slide

13. In a simple exercise, a real exchange rate “tracking equation” based on interest rate differentials and the terms of trade is estimated. To examine the evolution of the relative importance of these two factors, the equation is estimated for two separate periods: 1985:1-1991:4 and 1992:1-2000:2. More specifically, we regress the real effective exchange rate on Australia’s terms of trade and real interest rate differentials. Note that this regression involves variables that should be treated as endogenous. Consequently, the results should be considered as correlations and no causal interpretation should be given to the coefficients. Real rates were calculated subtracting actual inflation for the year ahead from long—term rates.⁶ The interest rate differential was calculated as the difference between Australia’s real rate and the arithmetic average of U.S.’s, Japan’s, and Germany’s real rates.

14. The equation was estimated in an error correction form as in Beechey et al. (2000). The examination of the results of the two regressions (see Table III.2) provides evidence to suggest that the relationship between interest rate differentials and the real exchange rate has become increasingly more important.⁷ In particular, all the coefficients of the error correction specification have the expected sign in both sample periods. However, the terms of trade coefficient is significant only in the first period, while the interest rate differential coefficient is significant only in the second period. This evidence appears consistent prima facie with the idea that the interest rate spread compression associated with Australia’s fiscal consolidation of recent years is one important factor in explaining the recent depreciation of the Australian dollar.

Table III.2.

Australia: Real Exchange Rate Tracking Equation

Source: IFS, * and ** represent significance at the 1 percent, 5 percent, respectively. Constant not reported.

Table III.2.

Australia: Real Exchange Rate Tracking Equation

	1985Q1:1991Q4		1992Q1:2000Q2
	Coefficients	T-Statistics	Coefficients	T-Statistics
REER(1 lag)	-0.31*	-2.9	-0.41*	-3.8
Terms of Trade (1 lag)	0.29**	2.1	-0.06	-0.42
Interest Rate Differential (1 lag)	0.21	1.1	0.95*	4.5
Change in the Terms of Trade	1.1*	4.9	1.18*	4.4

Source: IFS, * and ** represent significance at the 1 percent, 5 percent, respectively. Constant not reported.

View Table

Table III.2.

Australia: Real Exchange Rate Tracking Equation

	1985Q1:1991Q4		1992Q1:2000Q2
	Coefficients	T-Statistics	Coefficients	T-Statistics
REER(1 lag)	-0.31*	-2.9	-0.41*	-3.8
Terms of Trade (1 lag)	0.29**	2.1	-0.06	-0.42
Interest Rate Differential (1 lag)	0.21	1.1	0.95*	4.5
Change in the Terms of Trade	1.1*	4.9	1.18*	4.4

Source: IFS, * and ** represent significance at the 1 percent, 5 percent, respectively. Constant not reported.

15. A second stylized fact is the relationship between the country’s net foreign asset position and the real effective exchange rate (see Figure III.2). It has been argued that in the long run the increasing service burden of a growing foreign liabilities requires larger trade surpluses, and hence a more depreciated real exchange rate. Although theoretically compelling, that argument does not seem to be supported by the data. First, the explanation is at odds with the evidence from Canada, whose currency followed long—run trends similar to Australia, while its NFA position remained roughly stable and even improved relative to GDP toward the end of the sample. Likewise, no clear pattern is discernible with respect to the REER and the NFA position in New Zealand. Second, the argument would require that, over time, the income balance worsens and that the REER depreciation is matched by a progressive improvement in the trade balance. However, in Australia both variables have remained roughly stable relative to GDP.

View Full Size

Net Foreign Asset Position and REER

(NFA as percentage of GDP - secondary axis)

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: Lane and Milesi-Ferretti (2000).

• Download Figure
• Download figure as PowerPoint slide

16. In view of the inability of these long—run regularities to account for the recent real depreciation, a number of alternative less conventional explanations have also emerged in the attempt to explain the behavior of the $A. However, only anecdotal evidence exists to support or refute these hypotheses. The first contends that the weakness of the Australian dollar is closely linked to the technology divide: Australia, with a smaller share of the economy involved in the development and production of ICT products and services relative to the U.S., has failed to attract foreign investment flows related to the technology sectors. While it is true the Australian stock market did not soar to levels of the NASDAQ or other “high—tech heavy” stock markets elsewhere in the world, this explanation sits sharply at odds with the fact that in the second half of 2000 the depredation of the $A continued even while a sharp correction was underway in U.S. technology stock prices.⁸

17. A related story suggests that it is the gap in relative prospects between the U.S. and Australia that is the chief driving force behind recent exchange rate movements. According to this thesis downward revisions of expectations on Australia’s output growth relative to U.S.’s output growth were at the root of the sharp exchange rate depreciation until the end of the year. The recent appreciation of the $A appears to reflect a downward revision of expectations about the near—term prospects for the U.S. Although there appear to be little good reason for markets to have been skeptical about Australia’s growth prospects, this explanation at least appears to be consistent with actual exchange rate movements observed in 2000.⁹

D. Theoretical Considerations

18. This section reviews briefly the main characteristics of a standard open economy macro model whose properties are exploited in this paper’s empirical framework. In particular, the differential short—run and long—run properties of a model with sticky prices can be used to identify the different “fundamental” shocks to the real exchange rate in the empirical estimation. In what follows variables will be expressed in relative terms, i.e. domestic relative to foreign levels.

19. The theoretical framework begins with the familiar hypothesis of long—run monetary neutrality. According to this assumption, real variables are, in the long—run, invariant to innovations in money supply and in money demand. Hence, any model incorporating such an assumption will have the property that the long—run coefficients of nominal (or monetary) shocks on real variables are zero. In the context of this paper, this means that “nominal shocks” can be identified as innovations that have no long—run effect on the terms of trade, relative output, and the real exchange rate. Exogenous monetary policy innovations are an example of this kind of shock. In the short run, unless prices are fully flexible, an exogenous increase in the money supply lowers real interest rates leading to higher output growth and to capital outflows that in turn cause an exchange rate depreciation. However, in the long run, prices increase, causing a real exchange rate appreciation and reducing real money balances that in turn lead to higher interest rates and lower growth. In equilibrium, all real variables return to their pre—shock levels.

20. A second building block is the assumption of sticky prices. With sticky prices, exogenous shocks to aggregate demand have only short—run consequences for output, with such consequences being reversed as prices adjust in the long run. However, exogenous demand shocks would have long—run effects on the real exchange rate and the price level. Consider fiscal tightening as an example of this kind of shock: in the short run, a fiscal tightening compresses output through a reduction in aggregate demand, while interest rates drop responding to the reduction in the demand for money. This in turn leads to capital outflows and an exchange rate depreciation necessary to maintain interest rate parity. In the long—run, prices fall leading to a recovery in aggregate demand in turn resulting in an increase in output. The new long—run equilibrium will be characterized by the same output level as in the initial equilibrium, lower prices, and a lower real exchange rate as, for a given output, an improved current account balance partly compensates for the reduced public component of aggregate demand. In the context of the empirical model in this paper, these properties permit the identification of “demand” shocks as innovations that have long—run effects on the real exchange rate, and potentially the price level, but not on output and the terms of trade.

21. Finally, the assumption of a small open economy permits the assumption that the terms of trade are fully exogenous, so that “supply shocks” can be identified as innovations that have long—run effects on output, the real exchange rate, and the price level, but not the terms of trade; while autonomous “terms of trade shocks” are the only innovations that have long—run consequences for all variables.

22. One caveat is worth noting about the assumption that innovations to relative output or “supply shocks” have no long—run effects on the level of the terms of trade. Because all the countries considered in this paper are small open economies, it is reasonable to assume that changes in domestic output do not affect the terms of trade. However, changes in relative output may also reflect changes in the output of partner countries which may, through a change in the demand for commodities, have an effect on the terms of trade for this reason the robustness of the results is tested by estimating the model under the alternative identifying assumption that the terms of trade do not affect relative output, while supply shocks may affect the terms of trade in the long run,

E. Empirical Methodology

23. The theoretical considerations in the previous section imply that relative output, the real exchange rate, and the price level are non—stationary. Hence, the first step in implementing the empirical methodology is to test and confirm the non—stationarity of these series. The model is then estimated in first differences that are confirmed to be stationary. A reduced—form VAR is set up with the real effective exchange rate, relative output, the relative price level, and the terms of trade. The model builds on the framework used by Clarida and Gali (1994) by including relative terms of trade in the analysis. The inclusion of this additional variable is necessary since the terms of trade have been identified as an important determinant of the real exchange rate for small resource—based economies like Australia and New Zealand.¹⁰

24. In the spirit of Blanchard and Quah (1989), and as discussed in the previous section, the following identifying restrictions are used. First, only terms of trade shocks are expected to affect the relative terms of trade in the long run (this assumption implies three restrictions on the four—variable VAR). Second, the paper follows the standard macro literature and assumes that demand shocks (a shift in the IS curve) and nominal shocks (a shift in the LM curve) have no long—run effect on the level of output (two more restrictions). Finally, a sixth restriction is that nominal shocks do not affect the long—run level of the real exchange rate. These six restrictions give the model a lower triangular structure in the long—run. The restrictions are sufficient to make the model exactly identified and can be used to transform the errors from the reduced—form VAR model into a set of structural disturbances that can be interpreted as terms of trade, supply, demand, and nominal shocks (see Annex III.1 for details). An important advantage of this approach is that short—run dynamics are left unconstrained, as all shocks are allowed to affect any of the variable in the short—run.

F. Data and Preliminary Analysis

25. This section describes the data used in the VAR and presents some descriptive statistics from the preliminary analysis of the dataset.

26. For Australia, the levels of these four variables from 1963:3 to 1999:4 are reported in Figure III.3.¹¹ Variables were normalized to 1970:1 and transformed into logarithms. Relative output was relatively volatile around a broadly unchanged level until the 1990s when it witnessed as significant upward trend. The real exchange rate fluctuated substantially since the floating of the Australian dollar, partly reflecting movements in the terms of trade. In addition, a long—run declining trend is discernible in the exchange rate, again partly in reflection a similar evolution of the terms of trade. The increases in exchange rate volatility after 1973 (end of the Bretton Woods period) and 1983 (end of the $A crawling peg) are clearly visible, and so is the sharp devaluation of the $A after the floating of 1983.

View Full Size

Australia: Descriptive Statistics (in logs, 1970=100)

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, OECD

• Download Figure
• Download figure as PowerPoint slide

27. Figure III.4 reports the same variables for New Zealand. Contrary to Australia and Canada, in New Zealand relative output declined substantially over the past four decades. The real exchange rate depreciated sharply in 1974-75, and then again in 1984. The New Zealand dollar appears to be on a long—run declining trend, albeit a less pronounced trend than for the Australian dollar. The exchange rate appreciation and the relative price stabilization in the early 1990s are both likely to reflect the monetary tightening associated with the introduction of inflation targeting.

View Full Size

New Zealand: Descriptive Statistics

(in logs, 1970=100)

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, OECD.

• Download Figure
• Download figure as PowerPoint slide

28. Figure III.5 shows that in Canada relative output increased sharply in the 1970s, stabilized in the 1980s, and exhibited a decline in the 1990s, probably mainly reflecting the unilateral strength of the U.S. economy. Canada’s real exchange rate and terms of trade both followed a long—run declining trend starting in the 1970s.

View Full Size

Canada: Descriptive Statistics

(in logs, 1970=100)

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, OECD.

• Download Figure
• Download figure as PowerPoint slide

29. Before proceeding to estimate the VAR, the time—series properties of the variables are examined. In particular, the variables are tested for deterministic—trend non—stationarity using an ADF test Results are reported in Table III.3. For all three countries it is not possible to reject the null hypothesis of a unit root for relative output and prices. Similarly, it is not possible to reject the unit root hypothesis for the terms of trade and the real exchange rate for Canada and New Zealand; while in the case of Australia the rejection is very weak and depends crucially upon the order of augmentation.

Table III.3.

Augmented Dickey-Fuller Statistics

Source: IFS. Sample period 1966-1999. Regressions include an intercept and a linear trend for the levels. Regressions include an intercept but no trend for the first differences. Order of augmentation selected using the Akaike Information Criterion.

Table III.3.

Augmented Dickey-Fuller Statistics

	Australia	New Zealand	Canada
Terms of Trade	-3.8	-3.1	-2.1
Relative Output	-1.0	-3.2	-1.4
REER	-3.9	-3.2	-2.6
Relative CPI	-2.5	-1.0	-1.4
Critical value (95 percent)	-3.4	-3.4	-3.4
Terms of Trade (1st Diff.)	-6.3	-6.2	-7.0
Relative Output (1* Diff.)	-9.9	-6.7	-4.6
REER (1st Diff.)	-4.1	-4.6	-5.0
Relative CPI (1st Diff.)	-3.1	-4.2	-5.5
Critical value (95 percent)	-2.9	-2.9	-2.9

Source: IFS. Sample period 1966-1999. Regressions include an intercept and a linear trend for the levels. Regressions include an intercept but no trend for the first differences. Order of augmentation selected using the Akaike Information Criterion.

View Table

Table III.3.

Augmented Dickey-Fuller Statistics

	Australia	New Zealand	Canada
Terms of Trade	-3.8	-3.1	-2.1
Relative Output	-1.0	-3.2	-1.4
REER	-3.9	-3.2	-2.6
Relative CPI	-2.5	-1.0	-1.4
Critical value (95 percent)	-3.4	-3.4	-3.4
Terms of Trade (1st Diff.)	-6.3	-6.2	-7.0
Relative Output (1* Diff.)	-9.9	-6.7	-4.6
REER (1st Diff.)	-4.1	-4.6	-5.0
Relative CPI (1st Diff.)	-3.1	-4.2	-5.5
Critical value (95 percent)	-2.9	-2.9	-2.9

Source: IFS. Sample period 1966-1999. Regressions include an intercept and a linear trend for the levels. Regressions include an intercept but no trend for the first differences. Order of augmentation selected using the Akaike Information Criterion.

30. These results suggest that it should be possible to obtain stationarity by taking the first differences of these variables. Indeed, as Table III.3 reports, the ADF test (without trend term) for the first differences confirms that it is possible to reject the null of non—stationarity of the first differences for all four variables in all three countries.

G. Results

31. This section presents estimates of the structural var described above. Although the system was estimated using first differences, for ease of exposition and interpretation the results are presented in levels. The number of lags or the order of the var (four for all three countries) was selected by the Akaike information criterion. The baseline sample period was 1974:1-1999:4. However, the system was also estimated over different sub—samples in order to test the robustness of the results and examine possible changes in how the real exchange rate reacts to fundamental shocks. One important caveat is that, give the sample period ends with the fourth quarter of 1999, the results in this section cannot rigorously explain the most recent developments in the Australian real exchange rate.

32. The effects of fundamental shocks on the levels of the terms of trade, relative output, the real exchange rate, and the relative prices are examined in two ways. First, impulse responses are reported and analyzed; and second, the historical decomposition of the real exchange rate forecast error is examined.

Impulse Responses

33. Figures III.6-III.8 present the impulse responses for Australia under different specification and for different samples. Figure III.6 reports the full—sample impulse responses for each variable to one—standard deviation positive innovation in each of the fundamental shocks: terms of trade, supply, demand, and nommal (standard error bands were computed using the Monte Carlo method and are not reported for brevity of exposition). These impulse responses were obtained by cumulating those resulting by the first—differences VAR. Key findings are as follows:

View Full Size

Australia: Impulse Responses

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, staff’s calculations.

• Download Figure
• Download figure as PowerPoint slide

• The three identifying restriction that supply, demand and nominal shocks have no long—run effect on the terms of trade clearly show in the top—left panel, where only terms of trade shocks have permanent effects.

• Similarly, in the bottom—left panel, relative output reacts significantly only to supply and terms of trade shocks.

• More interestingly, the top—right panel shows that, as expected, innovations in the terms of trade have a strong and permanent effect on the level of the real exchange rate; while positive relative supply shocks are a source of only temporary real appreciation, and to a much more limited extent (not significant).

• Consistently with Prasad (1999) and Fisher (1996), and the predictions of standard open economy macro models, permanent demand shocks have a positive effect on the real exchange rate; while nominal shocks cause only temporary shifts in the real exchange rate.

• Finally, the bottom—right panel shows that nominal shocks have a permanent positive effect on the relative price level, while supply shocks have a permanent negative effect; demand and terms of trade innovations have no significant effect on the relative price level.

34. In summary, these results confirm the importance of the terms of trade in determining Australia’s real exchange rate, while offering only very weak support to the idea that the depredation of the $A in the late 1990s was due to a negative shock to the expectations on the relative output performance of Australia and the U.S.¹² Rather, they are consistent with the thesis that a relative contraction of aggregate demand may have been responsible for the weakness of the Australian dollar. Indeed, the estimates suggest that, over the sample period, shocks to relative output have had only small and temporary positive effects on the real exchange rate.

35. Figure III.7 reports the impulse responses for Australia’s real effective exchange rate over different periods. The top panel reports the results for an extended sample, which includes some of the Bretton Woods years and the first oil shock: 1963:3 – 1994:4. The center panel graphs the impulse responses estimated over the baseline period 1974:1 – 1999:4. Finally, the bottom panel reports the results for the period since the floating of the Australian dollar, 1984:1 –1999:4. There are no striking differences among the three panels. However, not surprisingly, the relatively larger impact of a terms of trade shock over the floating period is noticeable.

View Full Size

Australia: REER Impulse Responses over Different Sample Periods

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, staff’s calculations.

• Download Figure
• Download figure as PowerPoint slide

36. Finally, as a test of the robustness of the results to the identifying assumptions, Figure III.8 reports the impulse responses of Australia’s real effective exchange rate under some alternative identifying assumptions (bottom panel). As noted above, the main concern related to the assumption of exogeneity of the terms of trade movements. It is possible that large changes in demand may affect the terms of trade calling into question the validity of the identifying assumption that the terms of trade are independent from relative output in the long run. One way to test the robustness of the results from the baseline specification is to reverse this identifying assumption. Under this alternative specification, terms of trade shocks have no long—run effect on relative output, while supply shocks may have long—run consequences for the terms of trade. These impulse responses are very similar to those for the baseline specification and confirm the robustness of the results from the corresponding identification assumptions.

View Full Size

Australia: REER Impulse Responses Under Alternative Identifying Assumptions

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, staff’s calculations.

• Download Figure
• Download figure as PowerPoint slide

37. Turning to the comparison with other “commodity currencies,” Figure III.9 reports the full sample results for New Zealand.

View Full Size

New Zealand: Impulse Responses

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS, staff’s calculations.

• Download Figure
• Download figure as PowerPoint slide

• Again, the three identifying restriction that supply, demand and nominal shocks have no long—run effect on the terms of trade can be seen clearly in the top—left panel, where only terms of trade shocks have permanent effects; and a similar argument applies to the bottom—left panel.

• A striking difference with Australia is the negative response of relative output to a positive terms of trade shock. One interpretation of this results is that the strict use of MCI as a trigger in the inflation targeting framework may have caused the New Zealand authorities to react in a contractionary manner to positive terms of trade innovations. However, this result could be also due to the strong correlation between Australia’s and New Zealand’s terms of trade (0.64 in the 1960-2000 period) and the large weight of Australia in New Zealand’s relative output.

• Turning to the top—right panel, in New Zealand as in Australia, both a terms of trade shock and demand shock have a positive permanent effect on the real exchange rate. However, the relative importance of the terms of trade in the exchange rate determination is much less pronounced than in Australia.

• Another difference relative to Australia is the significant positive effect of a terms of trade shock on the relative price level (bottom—left panel). This result is consistent with Fisher (1996) who also finds the terms of trade to be important for relative prices in New Zealand, but not in Australia. This difference may be partly explained by the different reaction of Australia’s and New Zealand relative output to terms of trade shocks.

38. Finally, Figure III. 10 reports the impulse responses for Canada.

View Full Size

Canada: Impulse Responses

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

Source: IFS. staff’s calculations.

• Download Figure
• Download figure as PowerPoint slide

• The top—right panel shows that, unlike for Australia and New Zealand, a terms of trade shock does not have a positive effect on the real exchange rate; while it has a positive effect on relative output (bottom—left panel).

• The bottom—right panel shows that the relative prices impulse responses are similar to those for Australia and New Zealand with the exception that a positive terms of trade shock has a negative effect on the relative price level.

Historical Decomposition of the Forecast Error

39. Next, this section examines the historical decomposition of the unconditional forecast error for the real effective exchange rate. This error represents the unforecastable innovation in the level of the exchange rate and can be decomposed into components attributable to each of the four fundamental shocks identified in the structural VAR.

40. Figure III.11 depicts the exchange rate forecast error and its decomposition for Australia. The top panel reports the actual forecast error and each of the subsequent panels depicts what the forecast error would have been if only that particular kind of shock had hit the system in the sample period. Overall the terms of trade and demand components account for most of the forecast error, while the supply and nominal components seem to have a more limited role.

View Full Size

Australia: Historical Decomposition of the REER Forecast Error

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

• Download Figure
• Download figure as PowerPoint slide

• Turning to the evolution of the various components, it appears that in the late 1970s the terms of trade component accounted for most of the downward pressure on the real exchange rate; the demand component was first marginally positive and turned negative at the end of the decade; while the supply and nominal components were broadly neutral.

• In the first part of the 1980s (characterized by a wide U.S. dollar cycle) the terms of trade were mostly a depreciating factor, while real demand sustained the exchange rate. Both factors and the nominal components account for the 1986 depreciation and the turnaround of the terms of trade is mainly responsible for the recovery after 1987.

• A similar, although less pronounced, pattern can be described for the terms of trade component in the 1990s. In addition, the nominal component which played a broadly neutral role in the first half of the sample became an overall positive factor in the late 1980s and the early 1990s, reflecting a relative tightening of monetary policy associated with the adoption of inflation targeting. Finally, in the late 1990s, the negative demand component possibly associated with the fiscal consolidation and the consequent compression of interest rate spreads was the main factor putting downward pressure on the exchange rate, while the supply and terms of trade components played a mitigating role.

41. Figure III.12 show the decomposition of New Zealand’s exchange rate forecast error. As for Australia, the terms of trade and real demand components play an important role, and the supply component represents only a marginal contribution. However, unlike for Australia, in New Zealand the nominal component seems to have played an important role as well.¹³

View Full Size

New Zealand: Historical Decomposition of the REER Forecast Error

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

• Download Figure
• Download figure as PowerPoint slide

42. In the late 1980s, the terms of trade and the demand component accounted for most of the appreciation of the New Zealand dollar. In the early 1990s, the demand (likely associated with the marked fiscal consolidation) and the supply components contributed to the depreciation, while the nominal component represented a sustaining factor reflecting (as for Australia) the new monetary policy discipline. Indeed, since the late 1980s the transition to an inflation targeting regime is clearly visible in the sharp increase of the nominal component of the forecast error. Recent developments seem to be accounted by a downward movement in demand and nominal components.

43. Finally, Figure III.13 presents the error decomposition for the Canadian exchange rate. Overall the terms of trade component, although important, accounts for a smaller percentage of the forecast error than for Australia and New Zealand. Instead, the real demand component is the main contributor, accounting for an overwhelming part of the unforecastable real exchange rate movements.

View Full Size

Canada: Historical Decomposition of the REER Forecast Error

Citation: IMF Staff Country Reports 2001, 055; 10.5089/9781451802030.002.A003

• Download Figure
• Download figure as PowerPoint slide

H. Conclusions

44. This paper set out to examine in a systematic manner the sources of real exchange rate fluctuations and possibly shed some light on some recent developments in the Australian dollar exchange rate. A structural VAR model of the terms of trade, relative output, the real effective exchange rate, and relative prices was estimated for Australia, New Zealand, and Canada.

45. The estimated model was then used to describe the reaction of the real exchange rate to four different macro shocks and to decompose the historical variation of the forecast error into four fundamental components. The results for Australia confirm the positive relationship between the terms of trade shocks and the real exchange rate. In addition, the historical decomposition exercise shows that terms of trade and real demand innovations account for most of the unforecastable real exchange rate variation. The empirical evidence does not support the idea—popular amongst market analysts—of a positive impact of relative supply innovations on the real exchange rate, and suggests that a contraction in Australia’s relative aggregate demand, possibly associated with the fiscal consolidation and the Asian crisis, has accounted for most of the downward pressure on the Australian dollar in the late 1990s. The terms of trade and relative output seem to have represented mitigating factors. It is well—known that models of exchange rate dynamics typically have very poor predictive power and there is no reason to presume that the framework employed in this paper should be any different. That said, it appears reasonable to state that the evidence in this paper is consistent with the idea that the recent fall in the SA has been due to demand shocks, but does not rule out the exchange rate having “overshot,” especially during 2000.

46. For New Zealand the estimation results suggest that, in addition to terms of trade and real demand innovations, nominal shocks also play an important role. Finally, the results confirm the relatively lower importance of terms of trade shocks for Canada, where the real demand component seems to have an overwhelming importance.