RDP 2001-03: The Response of Financial Markets in Australia and New Zealand to News About the Asian Crisis 2. Motivation and Data

2.1 Why Australian and New Zealand Financial Markets Might be Affected

There are a number of reasons why negative events relating to the Asian financial crisis might be expected to have a negative effect on financial markets in countries such as Australia and New Zealand. Firstly, to the extent that financial crises in some countries result in a generalised increase in uncertainty in world financial markets, we should expect increased volatility in financial markets in non-crisis countries, which usually results in lower (risk-adjusted) returns.

Secondly, the Asian crisis countries are important markets for Australian and New Zealand exports. As such, a pronounced recession in the crisis countries might be expected to have a negative effect on activity in Australia and New Zealand via the current account; these expectations would then flow through to financial market returns.^[1]

Thirdly, some market participants might have factored in some possibility – however remote – that contagion of the crisis could have spread as far as Australia and New Zealand, perhaps due to financial institutions' debt exposures to the crisis countries.

Finally, even if financial market participants do not expect that countries such as Australia and New Zealand will experience financial crises, they may expect that portfolio rebalancing behaviour could result in sharp declines in asset prices in countries with unrelated fundamentals. Kaminsky and Schmukler (1999) describe how market participants, in responding to a crisis in country A by selling country-A assets and buying country-B assets, may rebalance their portfolios by selling country-C assets, where country C is similar to country B. This ensures that the share of B and C assets in the portfolio remains at the desired level. This results in an apparent contagion of the crisis from country A to the unrelated country C. The effect is also consistent with the portfolio adjustment model of contagion in Lowell et al (1998). In addition, the effect might be compounded if there is a significant number of uninformed traders in the market, as they may also sell country-C assets if they interpret the sell-off as reflecting a change in fundamentals.

The factors listed above could explain some co-movement between Asian financial markets and those in Australia and New Zealand. On the other hand, there may be reasons for Australian and New Zealand markets to move in the opposite direction to their Asian counterparts. If a financial crisis in one region caused overseas investors to repatriate or otherwise reallocate their funds, it is possible that markets such as Australia and New Zealand could have received them, putting upward pressure on asset prices in those countries. That is, Australia and New Zealand could have been country B, not country C, in the portfolio-rebalancing scenario of Kaminsky and Schmukler (1999).

Further reactions to crisis events may occur, related to the actual or expected response by monetary policy-makers. For example, if the authorities raise short-term interest rates in response to an exchange-rate depreciation – or market participants expect that they will do so – this may result in a fall in stock prices and movements in long-term bond rates.

2.2 The Impact of News on Financial Markets

A large literature exists on the impact of macroeconomic news on financial market prices sampled at high frequencies (Campbell and Lewis (1998); Fleming and Remolona (1997); Almeida et al (1998); Kim and Sheen (1998) and Kim (1999) are some recent examples).

One distinction between most of this ‘event study’ literature and the present paper is that the former generally examines the effects of news events on financial markets in the country in which the news originated. We focus on the effects of news on third-country markets. In addition, most of the previous literature examines the effect of official macroeconomic data releases, which generally have pre-scheduled release dates and times. Exceptions to this are releases of German macroeconomic data, which do not follow a pre-determined schedule. In this case, market participants are less likely to be able to plan reaction strategies upon the release of the data. Almeida et al (1998) find that the response of the USD/DEM bilateral exchange rate to German releases is somewhat more drawn out than the response to US releases, which are pre-scheduled, although the difference can be measured in hours.

The set of news events we consider go even further than this, however. Although the precise timing of German macroeconomic releases is not known in advance, they are approximately regular. So although market participants may not know the exact timing of the German CPI release, they know that a release will occur each month. By contrast, news events during the Asian financial crisis were not always predictable. This would tend to increase the ‘surprise’ value of news about the Asian crisis, relative to the surprise value embodied in regular releases of macroeconomic data.

Limitations of the available data, described in the next section, prevent us from examining the response of Australian and New Zealand financial markets to news at ultra-high frequencies of hours or minutes. Also, since we do not have information on the times that most of the news events occurred, we are restricted to examining news effects on a daily frequency.

Asian time-zones largely overlap the Australian and New Zealand domestic trading zones. We would, therefore, expect that in most cases the reaction of Australian and New Zealand markets would begin on the same day that the Asian news events occurred. There will be some instances, however, in which the news events in Asia occurred after the market closes in Australia and New Zealand, and so the reaction will have occurred on the following day.

2.3 Identifying the Timing of News Events

The first step in assessing how news about the Asian financial crisis affected other countries' financial markets is to identify the events that constitute news. We use a combination of two pre-existing chronologies, one from the BIS and the other from the IMF (BIS 1998 Table VII.6, p 131; IMF 1998 Box 2.12, p 49), as well as the RBA's daily market reports. A table listing the events from these sources is shown in Appendix A. It should be noted that in some cases the dates cited in the IMF chronology differ from other IMF papers (e.g. Lane et al 1999). Where possible, we have verified the dates using newswire stories and other sources. The IMF and BIS chronologies ended in June and March 1998; we extended the chronology in this paper to end-August 1998 using the RBA's daily market reports.

Positive news will have the opposite effect on markets to negative news, suggesting that we should distinguish between events that are considered ‘good’ or ‘bad’ news. We classify events relating to agreements between international agencies and crisis countries, announcements of rollovers of debt and certain reforms as ‘good news’; all other news events listed in Appendix A are considered to be ‘bad news’. The classification of events as positive or negative is shown in the right-most column in the table in Appendix A. Our listing is similar to the classification used by Kaminsky and Schmukler (1999), based on the chronology compiled by Nouriel Roubini (Roubini 1999), and to that of Baig and Goldfajn (1998), compiled from newswire stories.^[2]

Kaminsky and Schmukler (1999) report that days on which some of the most volatile movements in Asian financial markets occurred were not necessarily associated with specific news events relating to the crisis. There are a number of possible explanations for this. Firstly, markets might react to cumulations of news, so that a seemingly ‘small’ or unimportant news event can engender a greater response if it follows a series of news events (the ‘straw that broke the camel's back’ effect). Secondly, there may be some herding behaviour by traders, so that sudden changes in financial prices can occur even in the absence of significant news. Thirdly, the news events considered may be less relevant to asset markets than the trading strategies used by market participants. To maximise returns from these trading strategies, it may be necessary to take advantage of particular market conditions, such as thin volume, which may not occur on news-event days.

2.4 The Financial Market Data, Episodes and Volatility

The data used to measure financial market returns and volatility for Australia and New Zealand in this study are: the broad indices of stock prices – the All Ordinaries Index (AOI) for Australia and the NZSE40 for New Zealand; bilateral exchange rates for the AUD and NZD against the USD; and the prices on futures contracts for Australian and New Zealand 10-year bonds, which trade on the Sydney Futures Exchange (SFE) and the New Zealand Futures and Options Exchange (NZFOE).^[3]

We use daily market-close data for stock prices and bond-futures prices, and 4 pm (AEST) readings for the bilateral exchange rates. Given these data series, we need to derive an appropriate measure of volatility: for daily data, the usual approach is to take the absolute value of daily percentage changes in prices (returns), or squared percentage changes. To avoid introducing spurious autocorrelation into our measure of financial-market volatility, we do not use measures such as rolling standard deviations of daily returns. Although the daily series will be considerably noisier than series that incorporate information from a run of days, their time-series properties will be more informative.

An alternative approach would be to use the diffusion-theoretic measure of daily realised volatility, which can be calculated (to a close approximation) as the daily summation of squared intra-day returns (Anderson et al 1999). It is not clear, however, that volatility within the day is the appropriate measure of interest to policy-makers. In any case, one of the principal attractions of this alternative measure of realised volatility is that some transformations of it may be normally distributed; this did not seem to be the case for the intra-day data available to us. This could, at least in part, reflect that this intra-day data set had a large number of missing observations.^[4]

We examine financial-market behaviour in Australia and New Zealand from the beginning of 1994 to the end of August 1999. We compare times of crisis with other times by dividing our sample into four sub-periods or episodes: ‘Pre-crisis’ – from 1 January 1994 to 30 April 1997; ‘Asian crisis’ – from 1 May 1997 to 31 August 1998; ‘World crisis’ – from 1 September 1998 to 31 December 1998; and ‘Post-crisis’ – the first eight months of 1999.^[5] The Asian crisis period spans sixteen calendar months, starting at the beginning of the month in which the first major news event occurred (Appendix A). We defined the end of the Asian crisis as being the onset of financial crises outside the Asian region; accordingly, we separately identify a ‘world crisis’ period, which we take as ending at the end of 1998 when most markets had calmed down considerably. The post-crisis period is therefore limited to the first eight months of 1999.

We were constrained from beginning the pre-crisis period any earlier than January 1994 by the availability of the composite Asian financial indices described and used in Section 4. We also wanted to avoid selecting a sample for the pre-crisis period that was too short, as the exact beginning of the Asian crisis is not necessarily clear. As early as July 1996, there was notable pressure on the Thai baht, following the collapse of the Bangkok Bank of Commerce. There was also pressure in January 1997, following the release of poor export and fiscal data (IMF 1998). Therefore, we chose to start the sample long before there was any indication of trouble in the region.

Another advantage of the 1994 start-date is that it captures the onset of the global bond bear market in February 1994. This period was characterised by falling bond prices and more volatile financial markets in general. It was followed by a substantial recovery in financial markets, which continued through to the beginning of the Asian crisis period. Capturing both market phases seemed a balanced approach, rather than constructing a sample period characterised by a bull or bear market alone. Moreover, differences between the pre-crisis and Asian crisis periods might then be reasonably attributed to the Asian crisis, rather than simply being due to the comparison between a turbulent period and a relatively calm period in financial markets.

2.4.1 Stock market volatility

Figure 1 plots the absolute daily percentage change in Australian and New Zealand stocks during the four periods described above. The standard pattern of financial-market volatility is apparent: in both countries, stock-market volatility fluctuates over time and tends to ‘cluster’, i.e. particularly turbulent days tend to be followed by turbulent days and relatively calm days tend to be bunched together. Volatility of Australian stocks appears, on average, to be slightly lower than for New Zealand, although overall, the patterns of fluctuations look very similar. This is evident throughout most of the sample, but most clearly during late October 1997 – where the large spikes represent the large stock market sell-off at that time – and subsequently, in the world crisis period.^[6] There does not appear to be much difference in volatility between the pre-crisis, Asian crisis and post-crisis periods (with the exception of the large spike in October 1997), whereas the world crisis period clearly exhibits a higher level of volatility for both countries.

2.4.2 Bond market volatility

It is clear that volatility in bond market returns – the absolute percentage change in the price on the futures contract – is much smaller than stock-price volatility (Figure 2). This partly reflects the pricing conventions on the Sydney Futures Exchange. However, there appears to be more evidence of volatility clustering in the bond market, with the 1994 period characterised by very volatile returns, followed by a period of relative calm in the second half of 1995. Again, these patterns are evident in both Australia and New Zealand, although, unlike the case for stock-price volatility, bond-price volatility is much higher for Australia and appears to be more persistent. Overall, however, volatility in the Australian and New Zealand bond markets seems highly correlated, with volatility in the pre-crisis period much higher for both countries than in the other periods. This is consistent with the global sell-off in bond markets throughout 1994 and early 1995.

2.4.3 Foreign exchange market volatility

Volatility of both the AUD/USD and the NZD/USD exchange rates increased markedly during the Asian crisis, building towards the end of the period, and remained high into the world crisis period (Figure 3). This result suggests that the Asian and world crises had their largest impacts on the exchange rates of the two countries. The increased daily volatility during the later part of the Asian crisis period and in the world crisis period was associated with large depreciations in the AUD/USD and NZD/USD exchange rates. By contrast, the bond and stock markets rallied during most of this period. In part, this may reflect a ‘flight to quality’ by investors.

Although the volatility in the exchange rates of these currencies against the USD varied considerably in the crisis periods, the volatility in the AUD/NZD cross-rate was relatively stable (Figure 4), despite the differences in the operational regimes and stances of monetary policy between the two countries. During the Asian and world crises, the monetary policy instrument was the cash rate in Australia, whereas in New Zealand, it was a monetary conditions index (MCI), based on the trade-weighted index for the NZD and the 3-month bank bill interest rate. The relatively constant volatility of the AUD/NZD cross-rate reflects that these two currencies are generally traded as a bloc.

Footnotes

This vector of contagion is essentially the economic linkages model of Lowell, Neu and Tong (1998). [1]

Although this classification is somewhat arbitrary, it did not seem to be crucial to our results. [2]

The bonds data are for the ‘next’ contract to be delivered, which is a very close substitute for the underlying spot instrument, i.e. physical 10-year bonds. The markets in these instruments on the futures exchanges are deep and liquid and provide reliable price readings. These markets are generally considered to be more liquid than those for the corresponding physical securities. [3]

We calculated a measure of daily realised volatility (the logarithm of the summation of log-intra-day returns – see Anderson et al (1999) for a derivation) using ten-minute observations of Australian stocks, AUD/USD bilateral exchange rates and NZD/USD bilateral exchange rates. We then estimated the density of these series using a standard kernel density estimation procedure, with an Epanechnikov kernel and Silverman (1986) bandwidth selection. We found a considerable degree of excess kurtosis relative to the corresponding normal (Gaussian) distribution. These results are available from the authors. [4]

This rather arbitrary dating is not the only way to define periods of crisis. Eichengreen, Rose and Wyplosz (1995), and Eichengreen et al (1996) define a crisis period by the occurrence of extreme values of an index of ‘exchange market pressure’, defined as a weighted average of movements in exchange rates, interest rates and international reserves, relative to interest rate and reserves changes in a numeraire country. [5]

Over the whole period, the average absolute daily per cent change in Australian stocks was 0.6 per cent, compared to 0.7 per cent for New Zealand. However, in the period since October 1997, average volatility has increased to 0.7 per cent and 0.9 per cent. [6]