RDP 8812: International Interest Rate Linkages and Monetary Policy: The Case of Australia 3. The Main Questions

This section addresses directly the main questions raised at the outset of the paper.

3a. Have interest rates become more volatile?

Figure 1 shows monthly changes in the yields on 10-year bonds. There are three distinct periods:

• up to the middle of 1982, there was little volatility in the bond rate. This is to be expected, given that the yield was an administered one in this period;
• from mid 1982 until late 1983, monthly fluctuations in the bond yields were much more marked, as yields became market determined under the tender arrangements;
• from the end of 1983, volatility in yields declined, though it remained higher than in the period of administered rates.

Why is this the case? If, as is commonly believed, the volatility of world interest rates has generally increased, why has the volatility of Australian rates declined since 1983?

The important institutional explanation is that the Australian dollar was floated in December 1983. A major reason behind the floating of the exchange rate was, as in the textbooks, to gain greater monetary policy independence. Whereas a fixed or quasi-fixed exchange rate makes it very difficult to insulate domestic financial conditions from those abroad, a flexible exchange rate allows disturbances to be taken on the exchange rate rather than on interest rates (or asset prices).

Whether this is actually the case depends, naturally, on whether the authorities have any particular objectives for the exchange rate itself. No particular exchange rate has been sought in the floating period in Australia, though on occasions the Reserve Bank has been very active in the market in the interest of maintaining orderly conditions. Given this, after the floating of the exchange rate we might expect to see an increase in volatility of the exchange rate, a decline in volatility of interest rates, and Australian interest rates becoming more independent of those overseas.

Figure 2 shows the remainder of the story. The increased volatility seen in long rates in 1982 and 1983 (i.e. immediately prior to the float) was also observed in short rates. From end 1983, both long and short rates became less volatile, and the exchange rate more volatile, than had previously been the case.

Table 1 gives one simple statistical measure of variability, which shows the same story. Bond rates became more volatile after the introduction of the tender, but then became less volatile after the float. The volatility of short-term rates also declined noticeably after the float. The volatility of the exchange rate increased progressively throughout the period.

3b. Are Australian interest rates more synchronised with foreign rates?

Figure 3 shows interest rates for Australia and the United States over the 1980s. The upper panel shows yields on three-month government securities. Up to 1982, there were on occasion substantial differentials, but these did not persist for long. It is clear, however, that since early 1985, there has been a marked difference in experience. Differentials have been large, and persistent. Trends have frequently been in opposite directions.

The same is true for bond rates. The Australian rate was an administered one up until the middle of 1982, and this explains why it moved in a stepwise fashion. Even so, there was a reasonably close relationship to the U.S. bond yield for much of this period. From 1982, the two rates diverged. The differences became more marked from early 1985, with U.S. rates falling sharply, and those in Australia remaining at historically high levels. Then U.S. rates moved up over 1987, while Australia's rates trended down.

It is apparent from these charts that Australian interest rates have not become more synchronised with U.S. rates. While greater integration of financial markets across national borders would have been working towards greater synchronisation, the economic shocks affecting Australia in recent years, and the policy reactions to them, have caused interest rates to move differently from those in other countries.

Over 1985 and 1986, Australia suffered a major fall in its terms of trade and a widening in its current account deficit. Economic developments since then have been dominated by the need to bring about a medium-term correction of the current-account imbalance. The adjustment was on three fronts:

• the exchange rate depreciated. From the beginning of 1985 to mid 1986, the Australian dollar depreciated by nearly 40 per cent in real terms;
• fiscal policy was tightened. The central government budget moved from substantial deficit (4 per cent of GDP) in 1983/84 to an expected surplus of about 2 per cent of GDP in 1988/89. Over the same period, the total public sector budget position changed from a deficit of 7 per cent of GDP to balance; and
• monetary policy was also tightened substantially. Short-term interest rates at times reached nearly 20 per cent during 1985 and 1986.

The tightening of monetary policy meant that Australian short-term interest rates were much higher than U.S. short-term interest rates. Given that short-term rates affect long-term rates, this would help to explain why Australian long-term rates were higher than those in the other countries. Figure 4 shows the movements in yields on 10-year bonds and three-month Treasury notes over this period. The bottom panel shows a single line representing the difference between long-term and short-term interest rates (above the line is an upward sloping yield curve and below the line is an inverse yield curve). The period of monetary stringency from 1985 to 1987 shows up clearly as a predominantly inverse yield curve.

Another factor explaining the divergence between Australian and U.S. long-term interest rates over recent years was the fall in the Australian exchange rate and the increase in inflation over 1985 and 1986. Under some theories, nominal interest rates can only move out of line with those overseas when there is the expectation of an exchange rate change. If it were possible to measure exchange rate expectations, we could test to see whether the apparent divergence of long-term interest rates would disappear. Unfortunately, we cannot measure exchange rate expectations directly. A crude approximation would be to use past actual inflation rates to measure expected exchange rate changes, thereby assuming that expectations are adaptive and that exchange rates adjust to preserve purchasing power parity. This is the sort of assumption which would underlie, for example, comparisons of real interest rates in different countries. Rather than attempting to do this graphically, we examine this subject in the next section, where the methodology allows for the influences of short-term interest rates, inflation rates and foreign bond rates to be measured separately.

A third factor has been at work over the past year or two, when long-term rates in Australia have been falling whereas those in the U.S. have been tending to rise. The sharp tightening in fiscal policy in Australia in recent years, has produced Commonwealth Government budget surpluses and thereby reduced the supply of Commonwealth government bonds. The effect of this fall in the supply of bonds has been magnified by growing demand for such securities by banks, which are required to hold government securities for ratio purposes, and has led to strong competition for the available securities and a fall in their yields.

The effects of these various shocks affecting the Australian economy in recent years make it difficult to assess the extent to which there has been an underlying tendency for long-term interest rates in Australia to become more closely synchronised with those overseas. While at this stage there does not appear to have been greater synchronisation, a firm answer to the question cannot be given until the effects of the shocks have passed.

3c. The relative importance of domestic and international influences on long-term rates

This section adopts more formal empirical procedures, based on notions of interest parity and the term structure of interest rates, to look at the Australian long-term bond market. The two approaches are used to answer different questions. The interest parity approach examines the determinants of Australian bond yields and how these vary over time. In contrast, the term structure approach examines the behaviour of the bond market to test for market efficiency. Both approaches are compatible; it is possible that domestic long-term bond yields are determined in a way that is consistent with both interest parity and the term structure approach.

(i) The interest-parity approach

The interest parity framework is based on the proposition that the yield on a domestic bond is equal to the yield on an equivalent overseas bond plus or minus the expected change in capital value of holding that bond due to exchange rate variations.

The strength of this relationship depends on the degree of capital mobility and the exchange rate regime. The greater the degree of capital mobility, and the more fixed the exchange rate, the greater is the dependence of domestic bond rates on those in other countries. At one extreme, in the case of a credible fixed exchange rate system with capital mobility, the domestic bond rate is determined entirely by foreign rates. At the other extreme, if there is zero capital mobility, foreign interest rates have no impact on the domestic rate. Most countries, however, are in an intermediate position with some capital mobility and some exchange rate flexibility. There is thus a mixture of domestic and foreign influences that will impinge on the determination of long-term bond rates.

The important implications from this condition for the small open economy are:

• growing capital mobility reduces monetary policy's influence over the long-term interest rate; and
• increased flexibility of the exchange rate provides the authorities with improved control over the nominal long-term interest rate. However, if the authorities adopt exchange rate objectives, then domestic long-term interest rates may still be largely influenced by foreign rates.

The empirical work in this section uses an equation derived from the interest parity condition which separates the variation in domestic bond rates into the effects of changes in foreign bond rates, the differential movements in inflation and the short-term effects of the stance of monetary policy (measured by the short-term real interest rate differential).^[1] The equation is:

where R_t is the long-term bond rate, are rates of change of prices, rr are real short-term interest rates, E represents expectations and an asterisk signifies an overseas variable.

Table 2 presents estimates over the 1980s, using monthly data and a weighted combination of U.S., German and Japanese interest rates as the foreign bond rate. The full-sample results show positive (as expected) and significant effects on the bond rate from a world average bond rate, from the inflation differential and from the short-term real interest rate differential. The results in the table have been corrected for serial correlation in the error term. The large size of the rho value suggests the possibility of missing variables and/or some dynamic mis-specification in the model. The model is extended in Appendix 1 to account for these possibilities: the results are generally consistent with the simpler version discussed here.

Comparing the two sub-sample equations provides some interesting insights:

• First, the size of the coefficient on the bond rate does not rise in the 84:1–88:6 sample; in fact it declines slightly, though this is not statistically significant. This suggests that the synchronisation of domestic and foreign nominal interest rates has not increased in the recent period, when inflation differentials and the stance of monetary policy are taken into account. In fact, decomposing the fitted values from the equation into contributions from the various components (see Appendix 1) suggests that the foreign bond rate exerted a smaller influence on the Australian rate in the second sub-period.
• Second, there is a large and significant constant term in all equations, suggesting that the pure interest-parity condition is not met in Australia. This constant could perhaps be interpreted as a risk premium.
• Third, the real short-term interest differential has a stable, significant coefficient in all equations. This suggests that monetary policy's influence over the long-term rate through short-term real interest rates has not diminished in the post-float period.

(ii) The term structure approach

The conventional way to think about the relationship between domestic short-term and long-term rates is in the context of some sort of term-structure framework. In this approach, the long-term interest rate is thought to be equivalent to the average of expected future short-term rates over the relevant period. This is an equilibrium result which has a good deal of intuition. If it were not so, then there would be an incentive to either sell the long-term security and purchase the stream of short-term securities, or the reverse.

Put algebraically, the term structure relationship can be approximated by the following equation:

where H_t is the holding yield on a long-term security in period t, r_t is the yield on a short-term security, and X_t is any relevant piece of information. ε_t is a zero-mean residual. In the pure “expectations hypothesis”, α=β=0.

This approach can be used to test the expectations hypothesis, and has been so used in many instances. But its particular usefulness in the present context is that it allows a test of whether foreign interest rates have any effect on domestic excess holding yields by including foreign bond rates in X.^[2] It should be noted, however, that the model does not test for any contemporaneous linkages between domestic and foreign yields as in the previous framework. Even if long rates were determined by expected future short rates in Australia, it is possible that expected domestic short-rates move in a way that is consistent with the interest party condition outlined above.

Applying this technique to weekly Australian and U.S. data over the period 1979–1987 yields several interesting findings. Firstly, the pure expectations hypothesis can be rejected. This is a result in keeping with international experience.

Secondly, the long-term rate tends to react more to changes in the short-term rates than would be predicted by the expectations theory: that is, the long-term rate is in some sense overly volatile.

Thirdly, and of more interest in the present context, the U.S. bond rate is statistically significant in the equation when it is estimated prior to December 1983. A rise in the U.S. bond rate was associated with a subsequent rise in the Australian bond rate. This is suggestive of sluggish adjustment of Australian interest rates to U.S. rates. In the second half of the data period, however (i.e. after December 1983), this apparent inefficiency was not statistically significant.

It is now worth trying to summarise and reconcile the results from the two empirical approaches adopted. From the interest parity equations there is no evidence of a growing external influence on domestic bond rates. While rising capital mobility has had the potential to undermine the domestic authorities control over the long rate, the floating of the $A appears to have offset this by giving the Reserve Bank more leverage over domestic inflation and short-term interest rates. However, the term structure model suggests that long rates are not fully explained by expectations of future short-rates; an element of over-reaction to short-term rates was evident, and overseas factors may have had an influence in the pre-float period which was independent of any link through short-term rates. Consistent with the first approach, this influence appeared to have diminished in the post-float period.

The next section looks at the role of long-term rates in the transmission process.

Footnotes

Appendix 1 gives details of the derivation. [1]

Appendix 2 provides much more detail on the technique. [2]