RDP 2003-04: Identifying the Efficacy of Central Bank Interventions: Evidence from Australia 3. Reserve Bank of Australia Intervention Data

The estimation in this study uses daily interventions in the foreign exchange market by the RBA over the period from July 1986 to November 1993. This period is chosen because it contains a single distinct change in intervention policy, in October 1991, that allows us to identify the parameters in our exchange rate and intervention system. Additionally, the characteristics of intervention in this sample facilitate accurate estimation. Over the seven years of data (1,930 daily observations) intervention is frequent (48 per cent of days) and often large (up to A$1.3 billion). Another advantage of our sample is that other central banks do not intervene in the AUD/USD market and so we need only focus on the actions of one central bank.

The data are the daily net purchases of Australian dollars by the RBA and include all transactions made by the RBA, including those on behalf of the Government. Since the RBA can, and does, meet the Government's demand for foreign currency transactions from its own reserves, it has discretion as to the timing of transactions on behalf of the Government. For this reason the inability to exclude transactions on behalf of the Government does not appear to be problematic. Similarly, Neely (1998) argues that not excluding client transactions for the Fed has little influence. Over the sample used, almost all RBA intervention was conducted in the spot market verses the US dollar (Andrew and Broadbent 1994).[7] Rankin (1998) reports that the RBA has always sterilised its interventions.[8]

The exchange rate is measured as US dollars per Australian dollar (AUD/USD). The intervention and exchange rate returns data are aligned to cover exactly the same 24-hour period, commencing at 9am Sydney time. This avoids inaccurate results from using misaligned data, which have potentially hampered previous studies on central bank intervention. Aligning the intervention and exchange rate returns data is important since central banks can, and do, intervene outside their business hours, as shown in Dominguez (2003), and stated for the RBA by Rankin (1998).[9] The 24-hour period over which the data are measured will be referred to as a day. The intervention data for any given day will include transactions conducted in any of the major markets around the world on that calendar day.[10]

Rankin (1998) outlines five distinct periods of intervention policy used by the RBA since the float of the Australian dollar in December 1983. Table 1 summarises the pattern of intervention over these five episodes. The two periods used in this study are the second and third. During the period immediately following the float of the Australian dollar – December 1983 to June 1986 – the RBA allowed the dollar to float freely with only very small interventions. Our study commences using data from July 1986 when the RBA took on a distinctly proactive policy to foreign exchange market intervention. Interventions became more frequent (the RBA was active on 69 per cent of days) and substantially larger (the average absolute intervention was A$63 million). The change in policy in our sample occurs in October 1991 when the RBA all but ceased to make very small interventions. Large interventions continued to be used as in the preceding era as seen in Figure 1. As Rankin (1998) puts it, the RBA ‘sought to maximise the impact of its intervention through careful management of their size and timing’. This strategy of not making small interventions drastically reduced the frequency of interventions (to 24 per cent of days), but substantially increased the average size of interventions (to A$144 million). We do not consider the last two episodes when the RBA did not intervene from December 1993 to June 1995, or when interventions resumed in July 1995.

Table 1: Summary Statistics of Interventions
Regime Full I II III IV V
  Dec–1983 Dec–1983 Jul–1986 Oct–1991 Dec–1993 Jul–1995
  Jul–2000 Jun–1986 Sep–1991 Nov–1993 Jun–1995 Jul–2000
Number of interventions
Sales 1,283 99 780 15 0 389
Purchases 504 230 143 116 0 15
Total number of days 4219 640 1,338 554 397 1,290
Probability of intervention
Unconditional 0.42 0.51 0.69 0.24 0 0.31
Conditional on INTt −1 ≠ 0 0.73 0.62 0.80 0.50   0.75
Conditional on INTt −1 = 0 0.20 0.40 0.44 0.15   0.12
Average intervention (A$ million)
Sales 46 9 56 35   36
Purchases 81 16 100 159   285
Absolute value 56 14 63 144   46
Maximum intervention (A$ million)
Sale 661 44 661 150 0 286
Purchase 1,585 90 1,026 1,305 0 1,585
Memo items (A$ million)
Money base 20,404 11,421 16,206 19,853 22,608 28,995
M1 60,976 22,831 37,261 57,435 74,242 102,921

Notes: Purchases are of Australian dollars; sample starts 13 December 1983.
Source: Authors' calculations

Figure 1: Proportion of Days with Intervention
Figure 1: Proportion of Days with Intervention

Table 2, drawn largely from Neely (2001), compares the frequency and size of RBA interventions with those of the Fed, the Bundesbank and the Swiss National Bank. Over the period covered in the table, from the early 1980s to the end of the 1990s, the RBA intervened on 42 per cent of days, compared to between 4 and 12.5 per cent for the other central banks. Typical central bank interventions are tiny relative to the huge daily turnover in foreign exchange markets. The daily turnover in the AUD/USD market was estimated to be US$17.9 billion in 1992 by BIS (1993), although only US$4.8 billion of this involved a non-financial counterparty.[11] While the Australian economy is substantially smaller than the US and German economies, and the Australian dollar less heavily traded than all three other currencies in Table 2, the RBA interventions are of a comparable size to those of the larger central banks.[12] The large magnitude of some RBA interventions is also confirmed by comparing the maximum interventions to the money base and M1 in Table 1. In the two episodes considered in this study, the largest intervention is seen to be over 5 per cent of the money base, or 2 per cent of M1. Note that while the turnover in the foreign exchange market has increased substantially, and certainly did so through our sample, neither of the hypothesised transmission channels are affected by the volume of turnover.

Table 2: Comparison of Interventions by Central Banks
US dollar purchases
(US$ million)
Proportion of
days with
  Minimum maximum (Per cent)
US DEM/USD 01/07/1983 31/12/1998 −797 950 5.6
  JPY/USD 01/07/1983 30/12/1998 −951 722 4.9
Germany DEM/USD 01/07/1983 31/12/1998 −833 800 12.5
Switzerland CHF/USD 03/01/1986 29/12/1999 −545 150 3.9
Australia AUD/USD 13/12/1983 30/07/1999 −932 436 42.3


Over the past decade the RBA has moved to using a combination of spot market transactions and currency swaps, which together replicate a forward, to intervene and sterilise its intervention (see Rankin (1998)). [7]

Dominguez and Frankel (1993b) and the references contained therein suggest that the Fed, Bundesbank and Bank of Japan have typically only partially sterilised their interventions. [8]

Given Australian trading hours do not overlap with either London or New York trading hours, and BIS (1993) shows one-quarter of Australian dollar trading occurs in markets outside of Australian trading hours, this is potentially important. [9]

For a few 24-hour periods that end on an Australian public holiday we do not have an observation in the Sydney 9am exchange rate series. Instead, the Australian dollar exchange rate measured at noon EST (2am or 4am Sydney time, depending on daylight saving) by the New York Fed is used. [10]

The average (absolute) intervention for the regime including 1992 was just 0.6 per cent of the total average daily turnover. [11]

The BIS survey lists the AUD/USD currency pair as the ninth most traded at that time. The average daily turnover for the USD/DEM, USD/JPY and USD/CHF was US$192.2 billion, US$154.8 billion, and US$48.8 billion, respectively. Only considering trades with non-financial counterparties the AUD/USD was the fifth most traded currency pair with the three pairs mentioned before respectively US$26.9 billion, US$32.3 billion, US$7.0 billion. [12]