RDP 9808: What Moves Yields in Australia? 3. Data, Episodes and News

3.1 Data

The data used for interest rates in this study are the yields on futures contracts for 90-day bank bills and 10-year bonds, which trade on the Sydney Futures Exchange (SFE). These data are for the ‘next’ contract to be delivered, which is a very close substitute for the underlying spot instrument, i.e. physical 90-day bank bills and physical 10-year Treasury bonds. The markets in these instruments on the SFE are deep and liquid, and provide reliable readings for relevant yields.

The SFE has made available timed tick-by-tick readings of bill and bond yields, which are the basis of the current study.^[5] The floor of the SFE opens for trading in these contracts at 8.30 am each day, and closes at 4.30 pm.^[6] Accordingly, almost-continuous readings on yields are available between these times for the period of the study. We measure observations at intervals of five minutes throughout the day, although, since announcements typically occur on the stroke of the half-hour, each observation is taken as the first ‘tick’ in the minute immediately preceding the regular five-minute interval, to avoid missing any of the reaction of yields to news.^[7]

3.2 Episodes

The starting point for the study is determined primarily by data availability: the tick-by-tick readings are available from 1993 but the coverage of the data-set on market expectations of economic statistics (discussed below) deteriorates noticeably as the period is extended much before January 1994. We conclude the study at end September 1997. For purposes of looking for ‘flavour-of-the-month’ effects, we divide the period between January 1994 and September 1997 into three episodes, each of 15 months duration:

• the first dates from the start of January 1994 to end March 1995;
• the second is from the beginning of April 1995 to end June 1996; and
• the third is from the beginning of July 1996 to end September 1997.

As shown in Figures 1 and 2, a casual observer might characterise these periods in terms of some simple stylised facts: the first episode was dominated by a savage bear market; the second was initially a correction to earlier overshooting and subsequently a return to a measure of stability; while the third saw a rally in fixed-interest markets based on growing confidence that low inflation had returned globally.

The global bear market was triggered by the Fed's move to tighten monetary policy in early February 1994 in its pre-emptive strike against inflation (Figure 1). In Australia, the Reserve Bank tightened monetary policy in the second half of 1994, as reflected in the rise in the target cash rate; the sequence of moves was widely anticipated by the market, with the bill yield moving above the policy rate some time before the first (and each subsequent) tightening (Figure 2). After the third tightening in Australia in 1994, the market overshot, pricing further large tightenings into bills.

In the second episode, monetary policy was unchanged in Australia (Figure 2). But bill yields remained above the level of the target cash rate for some considerable period after the last move in the tightening phase. Once the bill yield eventually returned to the target cash rate, however, it oscillated around this level for the next year or so. At various times in this period, markets moved to price in small easings or tightenings according to the run of news. Bond yields in Australia moved down for most of the period before they rose from March 1996, initially on uncertainty about the outcome of the federal election, and subsequently on an upward revision to forecasts of the budget deficit. These trends in yields in Australia again were not dissimilar to those in the United States.

The third episode was characterised by falling bond yields globally as fears of inflation receded around the world. Monetary policy was eased in Australia (in five steps each of half a percentage point) as inflation declined and confidence rose that low inflation could be sustained. The extent to which markets moved ahead of these monetary policy adjustments was less marked than in 1994. In the instances in which a change in policy had not been fully anticipated, bill yields adjusted to reflect the new stance of monetary policy virtually as soon as the policy change was announced. Short-term interest rates generally hewed much closer to the policy target during this easing phase, rather than overshooting to the extent evident in the first episode.

3.3 News

What information is likely to be relevant for movements in interest rates? Hardouvelis (1988) argues that newly arriving information can affect interest rates through two channels – either through revisions to expectations about the setting of monetary policy, which he shows can dominate movements in short-term interest rates, or through revisions to expectations about inflation, which might dominate long-term interest rates. Since announcements about monetary policy are now explicit in a number of countries, including Australia, such announcements might also be important. Edison (1997) argues that information about economic activity as well as inflation is likely to be important because it can affect interest rates either directly, by influencing inflationary expectations, or indirectly, by encouraging expectations that such news might prompt the monetary authorities to adjust interest rates.

Bill yields are likely to be heavily influenced by expectations about the near-term setting of monetary policy. These expectations might be revised in response to news^[8] about monetary policy itself or, more often, about other economic announcements that might influence the policy setting. On the other hand, bond yields might be expected to reflect longer-term influences. While temporary changes in monetary policy might be expected to have a smaller effect on bond yields than on bill yields, markets can never be certain about how temporary a change in monetary policy will turn out to be. Simple correlations also suggest that there is a tendency for bill and bond yields to move broadly in parallel. In practice, therefore, short-term and long-term factors are difficult to disentangle, especially since portfolio performance might be measured (and published) with a frequency as high as monthly. For these reasons, as discussed below, the list of news items tested for bond yields is the same as for bill yields.

As noted, in the spirit of Fleming and Remolona (1997), the current study identifies – in advance – the news items that might be expected to affect both yields, rather than adopting the approach of identifying individual large movements in these yields and then looking for ‘causal’ items of news.

The announcements which are the focus of this paper fall into three categories:

1. announcements about, and official commentary on, monetary policy;
2. statistical releases on inflation, wages and economic activity likely to be perceived by investors as affecting the policy, inflation or economic outlook; and
3. some statistical releases, such as for the balance of payments and financial aggregates, which have not recently figured prominently in discussion about monetary policy but which are perceived as having been important in the past.

We also investigate, to a limited extent, whether the Commonwealth Budget speech has an announcement effect.

As to the first category, the following announcements are investigated:

• the Bank's monetary policy announcements; and
• the Bank's regular report on The Economy and Financial Markets (including, since May 1997, the Governor's Semi-Annual Statement on Monetary Policy to the House of Representatives Standing Committee on Financial Institutions and Public Administration) and speeches by Governors.
• In addition, the Bank's Board meets once a month, on a timetable well known to the market, to consider monetary policy. Given normal routines, the first opportunity for the Bank to announce a change in policy after a Board meeting is at 9.30 the next morning, when it announces its dealing intentions for that day. Even though fewer than half of the policy announcements since January 1990 have occurred on this schedule, the market sometimes looks for such announcements at this time, especially during periods of heightened uncertainty. A routine statement of the Bank's dealing intentions, with no accompanying policy announcement, can itself be ‘news’, since it would flag that policy is on hold, if only for that day. In the empirical tests discussed in Section 5, we therefore examine whether the announcement of the Bank's dealing intentions on the day after Board meetings contains market-moving information. As to timing of the Bank's commentary and Governors' speeches, we assume the main messages are transmitted from the time of their official embargo.

On the second group of announcements, the items of particular interest are those related to inflation and, more importantly, the formation of inflationary expectations. The Consumer Price Index (CPI) and measures of wages are vital. Currently, the most widely observed measure of wages in Australia is average earnings of full-time employees working a standard week (AWOTE).

As discussed, indicators of economic activity can also have implications for interest rates. Since no single measure of activity is adequately comprehensive or timely, we look for effects from:

• employment as the most timely major indicator of the state of the economy each month, and of intrinsic interest for policy;
• retail trade which provides relatively timely information on the largest single component of spending each month;
• the monthly release on local government building approvals, a leading indicator of housing investment, which can have a major influence on the economic cycle;
• the monthly release on merchandise imports which can provide an early reading on the tempo of domestic spending or on possible developments in the balance of payments; and
• the quarterly GDP release which, although the least timely of these indicators, is the most comprehensive measure of economic activity.

On the third set of announcements, a view held by some is that, from time to time, the stance of monetary policy has been determined by pressures on Australia's balance of payments, specifically the deficit on the current account.^[9] While the current account is not an objective of monetary policy, the belief that it is might still linger in some quarters. Moreover, it could be argued that the financing of the current account deficit could affect the risk premium on bonds, or that unexpected deterioration in it would lead to expectations of a depreciation of the Australian dollar and perhaps to higher interest rates. Also, between 1976 and 1985, monetary targeting was a feature of the monetary policy framework in Australia. While this regime has long since been abandoned, it might still be the case that financial markets regard financial aggregates as containing useful information about the outlook for interest rates. For these reasons, we investigate the announcement effects on bill and bond yields of monthly news on the balance of payments and financial aggregates.

‘Unanticipated’ information is calculated as the difference between the information as published and the median market forecast. For the first two episodes, median market forecasts were obtained from a survey of about 30 market economists conducted by Dow Jones the week before the release of each statistic.^[10] For the third episode, the median forecasts were obtained from a similar survey carried out by Reuters. For monetary policy announcements, the unanticipated element is measured as the difference between the 30-day bill yield at noon the day before a policy change was announced and the new cash rate target after the policy announcement.

The news releases examined for market-moving influences are summarised in Table 1, with their frequency, normal time of release, and the approximate publication lag after the end of the reference period. We use the first-published (i.e. unrevised) estimates since these are the estimates to which markets initially respond.^[11] All of the statistical news and monetary policy announcements are released while the floor of the SFE is open for trading. Fiscal policy statements are usually an evening affair but the floor of the SFE opens for the Budget speech from around 7.00 pm to 8.15 pm.

The items of news in Table 1 are carried by all of the major screen services virtually instantly at the time of their planned publication. The Australian Bureau of Statistics (ABS) and the Reserve Bank of Australia – the authorities responsible for their publication – are highly conscious of their market sensitivity and ensure that their public release occurs precisely on schedule. Monetary policy announcements aside, the timing of the release of each of these items is publicised in advance in regularly issued publication schedules. Monetary policy announcements and the Bank's daily dealing intentions always appear at 9.30 am; given their nature, no schedule of monetary policy announcements is possible.

A summary of the frequency of news is contained in Table 2.

Taking the period as a whole, there were 370 items of news, on 289 days, leaving 646 days when no announcements were made.

Footnotes

Tick-by-tick means that a reading of yield is available for every trade carried out in these contracts over the period of the study. [5]

The trading floor is closed between 12.30 pm and 2.00 pm each day. [6]

For example, we measure the movement in yields, say, for the period from 11.30 am to 11.35 am as the movement from the first ‘tick’ at 11.29 am to the first ‘tick’ at 11.34 am. [7]

‘News’ in the economic literature is generally defined as the unanticipated information contained in any announcement. This convention is followed in the current study. [8]

The Australian Bureau of Statistics ceased publishing monthly estimates of the current account deficit in December 1996, and began publishing monthly figures on the balance on goods and services. Before January 1997, we use the figures for the current account deficit; thereafter, we use those for the balance on goods and services. [9]

A gap in these forecasts for AWOTE was kindly filled by BT Economics providing consensus market forecasts from their records. [10]

Markets can also respond to revisions to significant items of news, especially if these are released separately from preliminary estimates for subsequent time periods. In Australia's case, this mainly affects AWOTE. We have not investigated the effect of announced revisions. [11]