RDP 2001-03: The Response of Financial Markets in Australia and New Zealand to News About the Asian Crisis 3. The Response to News

In this section, we use some simple summary statistics and econometric techniques to measure the impact of news on financial-market volatility and returns during the Asian crisis.

Within the Asian crisis period, we distinguish between ‘news’ days and ‘no-news’ days, defined as days on which a news event did not occur, and which neither immediately preceded nor immediately followed a news day. Days on which a news event did not occur, but which were adjacent to a news day, are identified separately as ‘pre-news’ and ‘post-news’ days.

3.1 Summary Statistics

3.1.1 Stock prices

The top panel of Table 1 summarises volatility in the Australian and New Zealand stock markets – as measured by the average absolute percentage change in Australian and New Zealand stocks – for all news-event days (pre-news, news and post-news days) and no-news days during the Asian crisis period. The table also shows the corresponding measures for the world crisis, pre-crisis and post-crisis periods, as well as the Asian crisis period taken as a whole. Table 2 and Table 3 present mean-difference tests of the significance of the differences between these measures.

Several facts stand out. Firstly, during the Asian crisis, all news-event days were noticeably more volatile for both Australian and New Zealand stock indices than were days when news events did not occur. Secondly, volatility in both stock indices in the pre-crisis period was significantly lower (in a statistical sense, using a one-tailed test with a significance level of 5 per cent) than during the Asian crisis, but similar to no-news days during the crisis. It was also lower than in both subsequent periods (world crisis and post-crisis). Thirdly, volatility in the world crisis period was similar to the Asian crisis for Australian stocks, but for New Zealand stocks, the world crisis period exhibited significantly higher volatility.

3.1.2 Bond futures prices

The variation in bond-market volatility was much smaller than for the other financial markets considered. For both Australia and New Zealand, there was seldom more than 0.01 percentage points difference between the mean absolute movements in the bond futures prices across the sub-periods (Table 1). The mean difference tests shown in Table 2 and Table 3 do not indicate any significant news effects during the Asian crisis period for Australia or New Zealand. Pre-news days, news days and post-news days did not engender any greater volatility in Australian and New Zealand bond markets, on average, than days when news events did not occur. Reflecting the severe sell-off in bond markets in 1994, mean volatility in the pre-crisis period was significantly greater than for the Asian and post-crisis periods for both the Australian and New Zealand markets, but not greater than in the world crisis period. Although these are statistically significant differences, they are very small from an economic perspective.

3.1.3 Exchange rates

The effect of the Asian crisis on Australian and New Zealand financial markets is particularly evident for exchange rates. There was an apparent news effect: the mean absolute returns on all news-event days were significantly greater than for no-news days for both exchange rates. In the Asian crisis, world crisis and post-crisis periods, both exchange rates were significantly more volatile, on average, than in the pre-crisis period. This suggests that these differences reflected a generalised increase in volatility stemming from heightened uncertainty triggered by the crises. Moreover, the world crisis period exhibited greater volatility than the Asian crisis period in both countries, although not significantly so for Australia.

3.1.4 Comparing Australia and New Zealand

In Section 2.1 above, we discussed a number of reasons why financial markets in Australia and New Zealand might react to news events in Asia. The degree of the responses, however, may not be the same. For example, there may be differing degrees of macroeconomic integration with the crisis countries. There could be different expectations about the likelihood of the crisis spreading to these economies. The reactions could also reflect differences in markets' expectations of the potential responses by the monetary authorities in each country, or market reactions to different monetary policy actions that actually occurred. (Australia and New Zealand were conducting monetary policy using different operational regimes at the time of the crisis.) Finally, there is a possibility that financial markets in different countries react differently to policy actions that appear identical.

In Table 4, we compare the average volatility of financial markets in Australia and New Zealand, using the same mean-difference test statistic as in the previous subsections.^[7] For the stock market, the results are unambiguous: in the crisis periods and the post-crisis period, the mean volatility is larger in New Zealand. However, this difference between countries is significant only during the world crisis and post-crisis periods. There could be a number of reasons for this, not least that the New Zealand stock price index, being relatively small, was more susceptible to being moved by large liquidity flows during the second half of 1998. In any case, this difference is unrelated to the Asian crisis period and, therefore, cannot be attributed to differences in the authorities' responses to the Asian crisis, or to different market expectations about the implications of the crisis. A similar pattern can be seen in the results for bonds and exchange rates: where differences between Australia and New Zealand exist, they occur in the pre-crisis or post-crisis periods. The crisis periods seem to have resulted in greater similarity between markets. A possible explanation for this is that both markets were driven by overseas events during the crises, and to about the same extent, while at other times they were driven by country-specific shocks.

While the volatility in the two countries' financial markets were very similar during the Asian crisis, the levels of the financial-market variables suggest that conditions in Australian and New Zealand stock and bond markets were rather different during this period (Figure 5).

3.2 Econometric Evidence

In this section, we seek to further quantify the effect of news on financial markets using econometric methods. Based on our chronology, we constructed a news event ‘dummy’ series which took the value +1 for good news, −1 for bad news, and zero otherwise. We then estimated vector autoregressions (VARs) of the daily returns on Australian and New Zealand assets and on a benchmark US financial asset (the S&P500 stock price index for the VAR explaining stock returns and the futures contract on the 30-year benchmark Treasury bond for the bond price VAR), for the pre-crisis, world crisis and post-crisis periods. For the Asian crisis period, we augmented the VAR with the current and lagged values of the news event dummy series. This is similar to the methodology used by Baig and Goldfajn (1999).

Since bilateral exchange rates are relative prices – in this case to the US dollar – it is not possible to use this exact approach for the exchange rates. Instead, we estimated VARs of the AUD/USD and NZD/USD with the CRB Commodity Price Index, which is intended to proxy for the effects of global shocks on commodity-exporting countries.^[8] For each of the VAR systems, we used two lags of the endogenous variables, which was the preferred number of lags according to the Schwartz Information Criterion. We included the current-dated and first lag of the news variable for the Asian crisis period.

The results from these models should be taken as indicative rather than decisive, not least because linear VARs are hardly the best available model of financial asset returns. In particular, the residuals from most of these models are non-normal; specifically, they have marked ARCH properties. However, when we estimated single-equation models incorporating the same variables and lag structure to these VARs, allowing for GARCH residuals, the qualitative results on the importance of the news events in Asia and US developments were unchanged. It is also not feasible to estimate multivariate GARCH models using our data set. Because non-trading days are not identical across markets, there are missing values, which can distort estimation of the process for the error variance.

The VAR results for the stock market are shown in Table 5. The estimated coefficients on the news dummy series are positive but insignificant for Australian and New Zealand stocks. The coefficients on the lagged S&P500, however, are large and highly significant for both countries in all periods. This suggests that the news dummies do not appear to have much independent effect on Australian and New Zealand stock markets, once overnight events in US markets are controlled for; these markets are dominated by overnight developments in the US.^[9] However, there is some evidence that Australian and New Zealand market participants react to events in Asia indirectly via the US. The contemporaneous news dummies are just significant in the equation for the S&P500, and they are of the expected sign. This might explain why the post-news days exhibited greater average volatility in both countries' stock markets than did news days (Table 1). It also suggests possible inefficient information processing. If Asian news had systematically moved the S&P500, which then systematically moved Australian and New Zealand stock markets, it begs the question why the Australian and New Zealand markets did not react on the day of the news event. One answer may be that timing issues prevented these markets from reacting contemporaneously, for example, if the event occurred after the markets closed.

The results for bonds indicate an even smaller response to the news events, once the overnight movements in the US Treasury market are controlled for (Table 6).^[10] The estimated coefficients are broadly similar across the four sub-periods, with the inclusion of the news-event dummies making little difference to the estimation results for the Asian crisis. Again, overnight movements in the US long bond mattered more for Australian and New Zealand bond returns than did the Asian crisis news events.

The picture for the exchange rates is somewhat different in that the contemporaneous news dummies are of the right sign but are insignificant, while the lags of the dummies are significant in both the AUD and NZD equations. The significance of the lagged dummies and not the contemporaneous dummies could possibly be attributed to the timing of the news announcements or to foreign exchange markets waiting for the US stock market reaction. The estimated coefficients on the news dummies are positive, implying that bad news in Asia resulted in a depreciation of the AUD/USD and NZD/USD.

Interestingly, the CRB index became more significant in later periods. This suggests that market participants looked more closely at commodity price series, such as the CRB index, when assessing the fundamentals underlying these exchange rates.

Footnotes

Using a two-tailed test, not a one-tailed test as in the previous section. [7]

Westpac Banking Corporation produces a real-time commodity price index that better reflects the composition of Australia's exports. Although back-data are available, this index was not available to traders until 1999. In any case, estimation of the exchange rate VAR using the WBC index instead of the CRB index gives similar results. [8]

The US market's day t occurs after the close of Asian, Australian and New Zealand day t, but before their day t+1. This also applies to the timing of the CRB series; we therefore only include lags of the CRB index, not its contemporaneous value, in the equations for explaining returns on Australasian exchange rates. [9]

The estimated coefficients on the lagged US bond futures are substantially less than one, despite the yields on the underlying securities moving closely together, because of differences in the quoting conventions used in the markets trading the futures contracts. This does not affect those coefficients' significance or the values of the other coefficients. [10]