Exchange Market Pressure in Australia

Introduction

Since the collapse of the Bretton Woods system in the early 1970s, most industrialized economies have moved toward a managed floating exchange rate regime-allowing some exchange rate flexibility, but often intervening in the foreign exchange market to influence the path of the exchange rate. Recent studies have argued that under a managed floating regime, monetary authorities should focus on the pressure on the exchange rate rather than on changes in exchange rates and foreign exchange reserves alone (Tanner, 2001). This has led to the development of a large number of different measures of the magnitude of exchange market pressure (EMP), some based on structural models of the economy (e.g., Weymark, 1995, 1997, 1998; and Spolander, 1999) and some model-independent (e.g., Eichengreen, Rose, and Wyplosz, 1996; and Pentecost, Van-Hooydonk, and Van-Poeck, 2001).

Many of the measures of EMP based on structural models have been generated using a methodology proposed in Weymark (1995, 1998) that stands independently of the structural model employed. This paper provides an application of Weymark's methodology to the Australian case since the float of the Australian dollar using the small open-economy model outlined in Spolander (1999). It also provides an application of the model-independent approach developed by Eichengreen et al. (1996). The resulting EMP indices are then examined to determine if they provide plausible descriptions of the pressure on the Australian dollar during the post-float period. As such, this paper contributes to the EMP literature by examining how well two existing methodologies perform when applied to another situation. The role of foreign exchange intervention by the Reserve Bank of Australia (RBA) is then evaluated by calculating degree of intervention (DI) indices, as described in Weymark (1995, 1998).

Background

The term exchange market pressure has been widely used in the intervention literature to describe movements in two key external sector variables: holdings of international reserves and the nominal exchange rate. One of the earliest studies to examine EMP by Girton and Roper (GR) (1977) used a monetary model to explain exchange rate movements and defined EMP as the "volume of intervention that is required to achieve any desired exchange rate." Since the development of the Girton and Roper model, many researchers have applied the model to a number of both developed and developing countries.1 Modified versions of the Girton and Roper model also have been applied to various countries. For example, Wohar and Lee (1992) applied a less restrictive version of the Girton and Roper model allowing for deviations from purchasing power parity (PPP) and incorporating foreign disturbances, to the Japanese case. Their results indicated that the less restrictive model performed better. This finding more recently has been supported by Pollard (1999) for Barbados, Guyana, Jamaica, and Trinidad and Tobago. Mah (1998) also highlighted the importance of incorporating dynamic specifications of some of the independent variables in the Girton and Roper model.

Roper and Turnovsky (1980) and Turnovsky (1985) introduced the idea of using a small open-economy model in constructing an EMP formula. In a seminal contribution, Weymark (1995, 1998) provided a general methodology for calculating EMP that was model independent, under which the approaches of Girton and Roper (1977) and Roper and Turnovsky (1980) could be viewed as special cases. Weymark also proposed a model-independent definition of EMP as:

The exchange rate change that would have been required to remove the excess demand for the currency in the absence of exchange market intervention, given the expectations generated by the exchange rate policy actually implemented.

Weymark (1995, 1997) has applied her methodology to various simple open-economy models and used summary statistics from actual changes in the exchange rate and foreign exchange reserves held by the central bank to calculate measures of EMP. Spolander (1999) extended the simple model in Weymark (1995) to incorporate a monetary policy reaction function and the sterilization of foreign exchange intervention, thereby constructing a more realistic model of the economy.

Eichengreen et al. (1996) argued that dependency on a particular model was an undesirable feature for an EMP index. As an alternative, they proposed a measure of speculative pressure that is a linear combination of a relevant interest rate differential, the percentage change in the bilateral exchange rate and the percentage change in foreign exchange reserves. The weighting allocated to each of the three components is chosen in order to equalize their conditional volatilities. In a similar paper, Pentecost, Van-Hooydonk, and Van-Poeck, (2001) determined the weights using principle components analysis. The EMP indices resulting from these approaches are model independent 'because neither the components of the index nor the weighting scheme is derived from a structural model of the economy' (Weymark, 1998).

Notationally, m^sub t^ is the natural logarithm of the stock of money (with superscripts d and s denoting demand and supply, respectively), p^sub t^ is the natural logarithm of the domestic price level, c^sub t^ is the natural logarithm of real domestic income, i^sub t^ is the domestic short term interest rate, p*^sub j^ is the natural logarithm of the foreign price level, e^sub t^ is the natural logarithm of the exchange rate expressed as units of domestic currency per unit of foreign currency, i*^sub t^ is the foreign short term interest rate, d^sup a^^sub t^ is autonomous domestic lending by the central bank divided by the one period lagged value of the money base (B^sub t-1^), r^sub t^ is the stock of foreign exchange reserves divided by the one period lagged value of the money base (B^sub t-1^), y^sup trend^^sub t^ is the long-run trend component of real domestic output (y^sub t^), and y^sup gap^^sub t^ is the difference between y^sub t^ and y^sup trend^^sub t^.

In this model, it is assumed that agents form expectations rationally and also that a constant proportion (?) of intervention is sterilized. Furthermore, under this model, the sterilized portion of intervention is ineffective. As empirical evidence [see Edison (1993) for a survey] is still mixed with regards to the efficacy of sterilized intervention, this is not an unreasonable assumption. Uncovered interest parity is assumed to hold, which rules out the existence of a portfolio balance effect, while future expectations about the exchange rate are held constant when imputing EMP, which eliminates the possibility of a signaling effect.

Equation (2) describes changes in money demand as a positive function of domestic inflation and changes in real domestic income and a negative function of changes in the domestic interest rate. Equation (3) describes the purchasing power parity condition where exchange rate changes and foreign inflation determine domestic inflation. Equation (4) describes uncovered interest rate parity. Equation (5) explains changes in the money supply as a positive function of autonomous changes in domestic lending and unsterilized changes in the stock of foreign exchange reserves. Equation (6) states that changes in foreign exchange reserves are a function of the exchange rate and a time-varying response coefficient, p^sub t^. Equation (7) describes the evolution of the central bank's domestic lending. It suggests that changes in domestic lending are a positive function of domestic inflation and changes in trend real output, and a negative function of the gap between real output and its trend. Equation (8) is a money market clearing condition that states that money demand is continuously equal to money supply.

The EMP formulas given for the model-dependent and model-independent approaches in equations (1) and (13), respectively, are essentially identical. The difference between the two methods is the way in which ? is calculated, as specified in equations (11) and (14). The EMP indices calculated from these formulas represent changes in the exchange rate that would have occurred if the RBA had unexpectedly refrained from intervening in the foreign exchange market. Negative values indicate pressure for the Australian dollar to appreciate during that period, while positive values indicate pressure for the Australian dollar to depreciate.

When DI = 1, the central bank intervenes to keep the exchange rate fixed, and when DI = 0, the central bank does not intervene, thus allowing the exchange rate to float freely. Negative values of DI indicate that intervention magnifies changes in the exchange rate. That is, when the exchange rate is under pressure to depreciate, intervention magnifies the depreciation. Values of DI between 0 and 1 indicate that intervention has acted to reduce the pressure on the exchange rate. When DI > 1, intervention reverses the exchange rate movement. That is, the exchange rate is induced to move in the opposite direction to the movement that would have occurred in the absence of foreign exchange intervention. Due to the discontinuity of the specification of equations used in defining the degree of intervention, DI can take extremely large absolute values.2 Hence, extremely large values of DI will be replaced by 2 when DI > 2 and -1 when DI < -1.

Data Description

Quarterly measures of the data series were obtained for the period from 1984:1 to 2003:4. see Appendix A for a detailed description of the data series and their sources. Time series properties of each of the data series are reported in Appendix B. Unit root tests indicate that almost all of the variables are first difference stationary. The one exception is also found to be first difference stationary if the number of lags in the augmented Dickey-Fuller test is increased. The first difference stationarity of the variables is what prompted Spolander (1999) to specify equations (2) and (3) in first differences.

The model is estimated for the bilateral Australian dollar against the U.S. dollar exchange rate (e^sub t^). The Australian 90-day bank-accepted bill rate is used as the domestic interest rate (i^sub t^) and the 3-month U.S. certificate of deposits rate represents the foreign interest rate (i*^sub t^). The M1 monetary aggregate is used as the domestic money stock (m^sub t^). The Australian and U.S. consumer price indices proxy for the domestic price level (p^sub t^) and the foreign price level (p*^sub t^), respectively. Australian and U.S. real gross domestic product represent the domestic level of output (y^sub t^) and the foreign level of output (y*^sub t^), respectively. Australian real gross national income is used as the domestic national income (c^sub t^). The Australian money base (B^sub t^) and a measure of net spot foreign exchange transactions (?R^sub t^) are also used in estimation. The measure of foreign exchange transactions includes transactions between the RBA and market participants as well as transactions between the RBA and the Australian government.

Empirical Analysis and Results

As the equations have endogenous variables on the right side, they were estimated using a two-stage least squares approach. Instrumental variables for the two-stage regressions were chosen by considering all of the exogenous variables and the one period lag of all of the endogenous and exogenous variables as possible instruments, and selecting significant variables from initial regressions of each endogenous variable on the possible instruments. Two-stage least squares was used in preference to three-stage least squares or full information maximum likelihood estimation due to its greater robustness in the presence of misspecification and the relatively small sample size.

The two-stage least squares estimation results are provided in Table 1. All of the parameters are correctly signed and most are significant. Some of the parameters in equation (18) are statistically insignificant. This is in line with Spolander's findings, and he attributed the problem to misspecification of the relationship governing the evolution of domestic lending. The a^sub 2^ parameter in equation (17) is also statistically insignificant.

Diagnostic tests for each equation are also included in Table 1. The tests suggest that the residuals from equation (17) are autocorrelated; therefore, the standard errors for this equation are corrected using the Newey-West procedure. Furthermore, the residuals from all three equations are non-normally distributed. This means that the t-tests of statistical significance in Tables 1 may be misleading. This does not reduce the validity of the parameter estimates.

The two EMP indices are presented in the first two columns of Appendix C, while the two DI indices constructed from the EMP indices are presented in the last two columns. A talk by Ian Macfarlane (who was then Deputy Governor of the RBA) that was published in the RBA Bulletin in 1993 gives some insight into a number of intervention episodes in the period of 1985-1991. This provides some basis against which to compare the estimated indices. The two EMP indices are reasonably plausible and follow a fairly similar pattern, although the model-dependent measure displays more volatility.

Both EMP indices suggest that between 1984:1 and 1986:3 the Australian dollar was generally under pressure to depreciate. The DI indices indicate that at least at the beginning of this period the RBA seemed content to let the Australian dollar depreciate, as significant depreciation pressure in 1984:2, 1985:1, and 1985:2 was offset by intervention only to a small degree. Depreciation pressure in 1984:3 was slightly exacerbated by intervention. By mid 1986 the Australian dollar was down to around the U.S.$0.60/A$ mark. Macfarlane (1993) suggests that at this point the RBA started to feel that the Australian dollar was becoming undervalued, and, consequently, the RBA intervened heavily during 1986:3 in support of the Australian dollar. Under the model-dependent EMP index the Australian dollar was under pressure to depreciate by 16.7 percent during this quarter, which was offset by RBA intervention.

The EMP indices indicate that the Australian dollar rebounded from late 1986 to late 1988; the DI indices suggest that the RBA intervened substantially to smooth the upward progress of the Australian dollar during this time. Two important exceptions during this period of pressure on the Australian dollar to appreciate were 1987:1 and 1987:4. The second of these anomalies is due to the global stock market crash. Macfarlane (1993) suggests that in both of these periods the RBA intervened heavily in support of the dollar, and this is supported out by the DI indices.

By late 1988, Macfarlane (1993) suggests that the RBA was starting to feel that the Australian dollar was becoming overvalued as it neared the U.S.$0.90/A$ level. From 1989:1 to 1989:3 the RBA was content to let the Australian dollar depreciate, as intervention only slightly offset depreciation pressure in the first two quarters and reversed appreciation pressure in the third quarter. Macfarlane (1993) also suggests that appreciation pressure was offset to a significant degree in the December quarter of 1990 and the June quarter of 1991 and this is supported out by the DI indices. The EMP values for these quarters, however, suggest that the pressure to appreciate was not substantial in these quarters.

The model-dependent DI index tells an interesting story about the recent history of the Australian dollar. From 1997:1 until 1998:3 the Australian dollar faced pressure to depreciate, which was exacerbated in four of the seven quarters by RBA intervention. In 1999:4, 2000:4, and 2001:2 the RBA intervened to reverse sizable pressure on the Australian dollar to appreciate. This provides some evidence that the RBA may be at least partly responsible for the fall of the Australian dollar from about U.S.$0.80 per A$ at the end of 1996 to around U.S.$0.50 per A$ by mid 2001. The model-independent DI index displays the same overall pattern, but much less strongly.

A number of limitations must be considered when interpreting the results of this analysis. First, two of the parameter estimates needed for the construction of the model-dependent EMP index are insignificantly different from zero. This suggests that there is a problem either with the model specification or with the estimation procedure and casts doubt on the accuracy of the model-dependent EMP index. Second, the parameter estimates are sensitive to the choice of instrumental variables, and small changes in the parameter estimates have a large impact on the EMP values. Third, the Spolander model does not allow any effect from sterilized intervention. As the sterilization parameter (?) is insignificantly different from one (fully sterilized intervention) it would be worthwhile in future to extend the model to allow sterilized intervention to have an impact on the exchange rate. Finally, both of the EMP indices used in this paper were derived under the assumption that changes in domestic lending (or the domestic interest rate) are not used to affect the exchange rate. There have been times, however, where the RBA has acknowledged that the state of the exchange rate has played some part in determining its monetary policy stance. Another possible future extension of this work is to allow for indirect intervention operating through changes in domestic lending or the domestic interest rate.

Conclusion

This paper estimates exchange market pressure (EMP) indices for the Australian dollar against the U.S. dollar exchange rate over the Australian post-float period using a model-dependent approach proposed by Weymark (1995, 1998) and a model-independent approach developed by Eichengreen et al. (1996). Although there are some concerns in the estimation of the model-dependent index, both indices appear to provide plausible descriptions of the pressure on the Australian dollar. Degree of intervention (DI) indices are also constructed. The results suggest that the Reserve Bank of Australia (RBA) contributed to the large depreciation of the Australian dollar between 1997 and 2001. In general, there is some evidence to suggest that over this period RBA intervention magnified pressure for the Australian dollar to depreciate and reversed pressure for the dollar to appreciate.