Does More Market-Wide Information Originate While an Exchange is Open: Some Anomalous Evidence from the ASX.

1. Introduction

In a seminal paper, French and Roll (1986) categorise price-relevant information into public information, private information and noise. They assert that the flow of all three categories of information is clustered around the trading time of an exchange: public information arrives more frequently when the exchange is open because it is also the period of normal business activity; informed traders can trade on their information only when the exchange is open and therefore private information affects asset prices during trading hours; and the noise that is introduced into asset prices by the trading process would also affect asset prices when an exchange is open. Since volatility of asset prices is intimately linked with the arrival of information (Ross 1989), the implication is that equity prices should be more volatile when the exchange is open than when it is closed.

The existing empirical studies appear to provide unequivocal support to the argument of French and Roll (1986--asset returns are more volatile during exchange trading hours than during non-trading hours.(1) Oldfield and Rogalski (1980) compare the variance of daytime (open-to-close) returns to the variance of the overnight (close-to-open) returns for five large stocks listed on the NYSE, and find that the average daytime variance is 4.26 times larger than the overnight variance. In a more comprehensive study, Stoll and Whaley (1990) use a sample of all stocks listed on NYSE and document an average daytime to overnight variance ratio of 5.40.(2) Lockwood and McInish, (1987) use index returns of the US market and find that daytime returns are more volatile than overnight returns. Further confirmation for higher daytime variance comes from studies that examine the S&P 500 index futures and the Japanese stock market. The daytime to overnight variance ratio is 5.07 for S&P 500 index futures (Becker, Finnerty & Tucker 1993) and 2.11 for individual Japanese stocks (Amihud & Mendelson 1991).(3) Thus, the evidence, that asset returns are more volatile when financial markets are open, appears overwhelming. The evidence is robust to the choice of individual stock or portfolio returns, the selection of cash or futures markets, or the use of different sample periods.

In this paper, however, we provide evidence from the Australian Stock Exchange (ASX) that is in contrast to the evidence accumulated for the US and Japan. We find that the return variance of the All Ordinaries Index (All Ords) during trading period is less than the return variance during non-trading period. This implies that, contrary to the arguments of French and Roll (1986), a larger amount of information relevant to the pricing of the Australian equities is released while the ASX is closed. We present and test the hypothesis that the dominant influence of the US market, which operates during the overnight period in Australia, causes the Australian overnight volatility to be higher than its daytime volatility. Our results provide support for this hypothesis. During public holidays in the US, the overnight volatility in Australia reduces to only one third of its normal level, and the ratio of trading to non-trading time return variance becomes somewhat similar to that found in prior studies for the US and Japanese asset markets.

The rest of the paper is organised as follows. In section 2, we provide evidence on the ratio of trading to non-trading period return variances for the equity indexes of Australia, US and Japan for our total sample and for each of three sub-periods. In section 3, we review prior studies on international information flows, and outline our foreign information hypothesis. We assert that the distinct behaviour of the variance ratio for the Australian equities could be due to the dominant influence of the US information. We develop a simple test of this hypothesis in section 4 where we also report our results. Finally, we outline the implications of our study in section 5.

2. Trading and Non-Trading Return Variances

We obtain daily opening and closing values of the All Ordinaries index (All Ords) from 1 January 1991 to 31 July 1996. All Ords is a market capitalisation weighted index that covers a broad range of equities listed on the ASX.(4) We give particular attention to the choice of the opening value of the index, as it is used for calculating both the overnight (close-to-open) and daytime (open-to-close) returns. As has been frequently mentioned, the non-synchronous trading or stale quotes problem makes the value of an index reported at the very open of the exchange a bad proxy for the underlying opening value of the index (see Stoll & Whaley 1990). This is particularly so for the All Ords, as under the electronic trading system of SEATS at the ASX, it takes about ten minutes before all the stocks are even allowed to be traded.(5)

Previous studies find that the problem of stale quotes can be overcome for the US (Japanese) market if one uses the index values 30 (15) minutes after the opening of the exchange (Stoll & Whaley, 1990; Lin, Engle & Ito 1994). For the ASX, Aitken, Brown and Walter (1994) find that, on average, it takes 8 minutes (45 minutes) for large (medium) stocks to trade after the opening of the exchange. As stated above, the stocks are weighted by their capitalisation in the calculation of All Ords. Larger stocks, thus, dominate the calculated value of the index. An analysis of the 15-minute return volatility of the All Ords (not reported here) shows that the volatility is unusually high in the first 15 minutes and then declines sharply after that. We treat this as evidence that most of the overnight information gets incorporated in the Australian index in the first 15 minutes after the open of the exchange. We, therefore, choose the value of the All Ords 15 minutes after the open of the exchange, that is, at 10:15 AM Australian time, as opening value of the index.(6)

To compare and contrast our results with that for the US and the Japanese stock markets, we also obtain daily opening and closing values of the S&P 500 and Nikkei 225, respectively, for our sample period. As in prior studies, we use the value of the index 30 (15) minutes after the opening of the exchange to represent the opening value of the index for S&P 500 (Nikkei 225). We calculate the percentage returns for the three indexes by taking the first difference of the log of the successive prices; where overnight and daytime returns represent close-to-open and open-to-close returns, respectively.

[R.sub.t] = (1n[P.sub.t] - 1n[P.sub.t-1]) x 100

In table 1, we present the variances of returns over trading and non-trading periods for the All Ords. In contrast to the overwhelming evidence in prior research, the ratio of trading to non-trading return variances is less than one (i.e. 0.88) for the Australian equity index. A simple two-tailed F-test confirms that the ratio is statistically less than one at 5% confidence level.(7) To confirm that our result for the Australian equities is anomalous not merely due to the adoption of delayed opening values of the index, we also report the trading to non-trading period variance ratio for the S&P 500 and Nikkei 225 in table 1. As stated above, both these indexes are also recorded at 15 and 30 minutes after the opening of the respective exchanges. For our total sample period, the average daytime to overnight variance ratios for the US and Japan are 2.57 and 3.57, respectively. These ratios are both statistically and economically greater than one, and are in line with prior studies conducted for these markets.

Table 1 Ratio of Daytime to Overnight Return Variances for Stock Indices

                                     Total       Sub
                                     Sample    Period 1

Australia

  Daytime Variance                   0.268      0.299
  Overnight Variance                 0.303      0.285
  Ratio of Daytime to
   Overnight Variance                0.88(*)    1.05
  (Critical value of F-statistic
   at 5% significance)              (0.90)     (1.19)

US
  Daytime Variance                   0.324      0.411
  Overnight Variance                 0.126      0.139
  Ratio of Daytime to
   Overnight Variance                2.57(**)   2.96(**)
  (Critical value of F-statistic
   at 5% significance)              (1.11)     (1.19)

Japan
  Daytime Variance                   1.386      1.982
  Overnight Variance                 0.389      0.427
  Ratio of Daytime to Overnight
   Variance                          3.56(**)   4.64(**)
  (Critical value of F-statistic
   at 5% significance)              (1.11)     (1.19)

                                      Sub        Sub
                                    Period 2   Period 3

Australia

  Daytime Variance                  0.284       0.208
  Overnight Variance                0.360       0.255
  Ratio of Daytime to
   Overnight Variance               0.79(**)    0.81(*)
  (Critical value of F-statistic
   at 5% significance)             (0.84)      (0.82)

US
  Daytime Variance                  0.234       0.324
  Overnight Variance                0.094       0.149
  Ratio of Daytime to
   Overnight Variance               2.49(**)    2.17(**)
  (Critical value of F-statistic
   at 5% significance)             (1.19)      (1.21)

Japan
  Daytime Variance                  0.991       1.129
  Overnight Variance                0.312       0.440
  Ratio of Daytime to Overnight
   Variance                         3.18(**)    2.57(**)
  (Critical value of F-statistic
   at 5% significance)             (1.19)      (1.21)

Note: Total sample period is from 1 January 1991 to 31 July 1996. The first, second and third sub-periods cover the time periods of 1 January 1991 to 31 December 1992, 1 January 1993 to 31 December 1994, and 1 January 1995 to 31 July 1996, respectively.

(*) and (**) in the table indicate that a ratio is statistically distinguishable from one at 5% and 1% levels of significance, respectively.

As a further check of robustness, we divide the total sample period into three sub-periods,(8) and report the variance ratios for the three indexes for each of the three sub-periods in table 1. In each of the three sub-samples, the daytime return variances of Australian equities have been equal to or less than the overnight variances, as reflected by a ratio of 1 or less than one. The F-test confirms the statistical reliability of this result for two of the three sub-periods. For S&P 500 and Nikkei 225 indexes, however, the variance ratios remain significantly greater than one in each of the three sub-periods.

The evidence that index returns at the ASX are more volatile during the non-trading period of the exchange is challenging for two reasons. First, it is in sharp contrast to accumulated evidence for the US and the Japanese markets. Second, it implies that a greater proportion of the market-wide information originates during the overnight period in which the ASX is closed. This is contrary to the arguments of French and Roll (1986) that both information and noise are primarily generated when financial markets are open.(9) In the following paragraphs, we propose and test one possible explanation for this puzzling finding.

3. The Foreign Information Hypothesis

We propose the dichotomy of domestic and foreign information for understanding the varying rates of information arrival in the Australian market. The businesses and financial markets in Europe, Middle East and the Americas open and close during the overnight period of the ASX. There is strong anecdotal evidence and market folklore that suggest the importance of foreign, in particular the US, information for the Australian stock market. Consider, for example, the following comments appearing in the Australian financial press, which are similar to the ones frequently echoed in the Australian financial circles:

   The only absolutely concrete conclusion to be drawn from the latest share
   market plunge, is the way we remain hostage to what happens on and around
   Wall Street. We continue to live in a world made in America, and it is made
   while we--at least most of us--sleep. It comes packaged with the dawn
   across the Pacific. (Terry McCrann, The Daily Telegraph, 19 August 1997, p.
   25)

The literature on international stock market linkages identifies two channels by which information in one country affects equity prices in other countries (see, Becker, Finnerty and Friedman 1995; Karoyli and Stulz 1996). First, news originating in one country may affect stock prices in other countries because the news has a global component and directly affects the expected cash flows or required returns of foreign equities. Consistent with this hypothesis, Becker, Finnerty and Friedman (1995) find that macroeconomic announcements in the US contain global content and affect equity prices in the UK stock market.

However, the release of identifiable public information may not be the only channel by which news originating in one country affects other stock markets. King and Wadhwani (1990) present a contagion model in which traders in one country are unable to observe directly all the information released in foreign countries. Consequently, they rely on the observed price movements in foreign markets to extract the global component of whatever information might have been released in those countries. In the contagion model of King and Wadhwani (1990), foreign price movements, per se, affect domestic equity prices. Consistent with their contagion model, King and Wadhwani (1990) find that the volatility of stock prices in the UK reduces markedly when the US stock market is closed, indicating that US price movements per se are a source of information for the UK traders. Chang, Jain and Locke (1995) also provide evidence that price volatility is elevated when a related market is open and lower when it is closed.

Corresponding to the two channels of international news transmission identified in the existing literature, we note two reasons as to why overnight information would be more important relative to daytime information (resulting in a lower trading to non-trading time variance ratio) for the Australian market than for the US and Japanese markets. First, as depicted in figure 1, operating hours of most of the major stock exchanges of the world overlap with the non-trading period of the ASX. The normal business hours for most of the world (the Middle East, Africa, Europe and the Americas) also coincide with the non-trading period of the ASX. It is, thus, plausible that most of the global economic or political news, like shock to world money supply or coal price, that affect the fundamentals of the Australian equities would emerge during the 18-hour overnight period. Second, the Australian economy as well as the stock market is much smaller than those of the US. As a result, the Australian traders are likely to pay greater attention to price movements in the dominant US equity market in order to draw inferences about the sentiment for equities of international investors. The special significance accorded to the US market may be due to the recognition of the US market as a producer of information or it may result form the systematic tendency of international traders to overreact to the US price movements (Becker, Finnerty & Friedman 1995).(10)

[Figure 1 ILLUSTRATION OMITTED]

Therefore, although the domestic information is more likely to be generated when the ASX is open as implied by the arguments of French and Roll (1986), most of the foreign, in particular the US, information is likely to emerge during the non-trading hours of the ASX. The special significance of the US information for Australian traders may help explain why more of the market-wide information that is relevant to the Australian equities originates during the non-trading period as opposed to trading period of the ASX.

4. A Test of the Foreign Information Hypothesis

In order to test the hypothesis that much of the information that arrives during the non-trading period of the ASX is not domestic and, instead, originates in the US, we develop a simple test. We separate the days during our sample period of 1 January 1991 to 31 July 1996 on which the businesses and exchange in the US were closed due to public holidays (other than weekends) while the ASX and the Australian businesses remained open during those days. In total, there were 29 such days.(11) These are listed in table 2. We calculate the return variance for the All Ords for these 29 overnight periods when the US stock exchanges and businesses were closed.

Our hypothesis states that the special significance of the foreign, in particular the US, information is the primary source of a less than one trading to non-trading period variance ratio for the Australian equities. The flow of information that originates in the US would be minimal on days when the US exchange and businesses are closed. Consequently, our hypothesis predicts that on such days, the trading to non-trading period variance ratio calculated for the Australian index returns would be greater than one as observed for the equity markets of the US and Japan. More specifically, our hypothesis predicts that the overnight return variance of All Ords calculated on days when the US has holidays should be significantly lower than the overnight variance on other days, as the flow of information from the US is minimal on these days. However, the daytime return variance of the ASX should remain unaffected as there is no interruption in the normal business or exchange operating hours in Australia and, thus, the flow of domestic information remains unchanged. Since the denominator should reduce while the numerator should remain constant, our hypothesis predicts that the ratio of daytime to overnight return variances should be significantly higher on days when the US businesses are closed compared to the ratio during the rest of the sample period.

Finally, since the US market may not be the only influencing market for Australia, we also extend the above hypothesis to other large European and North American stock markets that open and close during the overnight period of the ASX. Table 2 lists the days on which the stock markets in UK, France, Germany, Switzerland and Canada were closed while both the Australian and the US markets were open. If one or more of these markets are also important source of information for the Australian traders, one would expect the Australian overnight volatility to be less on days when these markets are closed. As table 2 shows, however, the European markets share many of the holidays. This leaves us with very few unique country-specific holidays, making any individual-country empirical tests difficult to conduct. For example, for the stock markets of Germany, UK, Switzerland and Canada, table 2 lists only 8 unique country-specific holidays. Because we have so few unique holidays for countries other than the US, we base our variance ratio tests for these countries for all the holidays reported in table 2.

Table 2 List of Market Closures for Australia and Other Major European and North American Stock Markets

This table lists the days on which various markets were closed over the period 1 January 1991 to 31 July 1996 as follows: For Australia, the table contains all holidays. For the US, the table lists all market closures except those that coincide with the Australian holidays. For all other countries, the table contains market closures except those that coincide with the Australian and the US market closures.

                     Australia          US           France

1991                 01-Jan-91      18-Feb-91     01-May-91(*)
                     28-Jan-91      27-May-91     08-May-91
                     29-Mar-91      04-Jul-91     09-May-91(*)
                     01-Apr-91      02-Sep-91     15-Aug-91
                     25-Apr-91      28-Nov-91     16-Aug-91
                     10-Jun-91                    01-Nov-91
                     25-Dec-91                    11-Nov-91
                     26-Dec-91

1992                 01-Jan-92      17-Feb-92     01-May-92(*)
                     27-Jan-92      25-May-92     08-May-92
                     17-Apr-92      03-Jul-92     28-May-92(*)
                     20-Apr-92      07-Sep-92     02-Nov-92
                     08-Jun-92      26-Nov-92     11-Nov-92
                     25-Dec-92
                     28-Dec-92

1993                 01-Jan-93      15-Feb-93     20-May-93(*)
                     09-Apr-93      31-May-93     21-May-93
                     12-Apr-93      05-Jul-93     14--Jul-93
                     14-Jun-93      06-Sep-93     16-Aug-93
                     27-Dec-93      25-Nov-93     01-Nov-93
                     28-Dec-93                    11-Nov-93

1994                 03-Jan-94      21-Feb-94     12-May-94(*)
                     26-Jan-94      27-Apr-94     14-Jul-94
                     01-Apr-94      30-May-94     15-Jul-94
                     04-Apr-94      04-Jul-94     15-Aug-94
                     25-Apr-94      05-Sep-94     31-Oct-94
                     13-Jun-94      24-Nov-94     01-Nov-94
                     26-Dec-94                    11-Nov-94
                     27-Dec-94

1995                 02-Jan-95      20-Feb-95     01-May-95(*)
                     26-Jan-95      29-May-95     08-May-95(*)
                     14-Apr-95      04-Jul-95     25-May-95(*)
                     17-Apr-95      04-Sep-95     05-Jun-95(*)
                     25-Apr-95      23-Nov-95     14-Jul-95
                     12-Jun-95                    14-Aug-95
                     25-Dec-95                    15-Aug-95
                     26-Dec-95                    0l-Nov-95

1996                 01-Jan-96      19-Feb-96     01-May-96(*)
                     26-Jan-96      27-May-96     08-May-96
                     05-Apr-96      04-Jul-96     16-May-96(*)
                     08-Apr-96
                     25-Apr-96
                     10-Jun-96

Total Closures          41              29             36
Unique Closures                                        24

                      Germany           UK           Swiss

1991               01-May-91(*)    06-May-91      09-May-91(*)
                   09-May-91(*)                   01-Aug-91
                   20-May-91(*)                   31-Dec-91
                   03-Oct-91
                   20-Nov-91
                   24-Dec-91
                   31-Dec-91

1992               01-May-92(*)    04-May-92      02-Jan-92
                   28-May-92(*)    31-Aug-92      01-May-92(*)
                   18-Nov-92                      24-Dec-92(*)
                   24-Dec-92(*)                   31-Dec-92

1993               20-May-93(*)    03-May-93      20-May-93(*)
                   24-Dec-93(*)    30-Aug-93      24-Dec-93(*)
                   31-Dec-93(*)                   31-Dec-93(*)

1994               12-May-94(*)    02-May-94      12-May-94(*)
                   23-May-94(*)                   23-May-94(*)
                   02-Jun-94                      01-Aug-94
                   03-Oct-94                      30-Dec-94

1995               01-May-95(*)    08-May-95(*)   01-May-95(*)
                   25-May-95(*)    28-Aug-95      25-May-95(*)
                   05-Jun-95(*)                   05-Jun-95(*)
                   03-Oct-95                      01-Aug-95

1996               01-May-96(*)    06-May-96      02-Jan-96
                   16-May-96(*)                   01-May-96(*)
                                                  16-May-96(*)

Total Closures          24              9              21
Unique Closures          8              8              8

                      Canada

1991               20-May-91(*)

1992               18-May-92
                   12-Oct-92

1993               01-Jul-93
                   11-Oct-93
                   24-Dec-93(*)

1994               10-Oct-94

1995               22-May-95
                   09-Oct-95

1996               20-May-96

Total Closures          10
Unique Closures          8

Note: (*) The common holidays among the five smaller foreign markets.

The results of our tests are presented in table 3. We report the overnight return variance and variance ratios of All Ordinaries index for days on which the US and other foreign markets were closed during our sample period. The results in this table should be compared with those in table 1 for the overnight variance and variance ratios for the ASX. The variance of the Australian overnight returns coinciding with the US holidays (2nd row in table 3) is only about one third of the overnight variance for the total sample period. The standard F-test confirms that this difference in variance is statistically significant at 1% level of confidence. This suggests that the US is the primary source of information that affects the Australian equities during the non-trading period of the ASX: on days when the US has public holiday, an important source of information is missing, leading to a marked reduction in the flow of information to the ASX which is reflected in a much lower volatility of the Australian index returns.

If we use the variance of overnight returns when the US had holidays and compute the daytime to overnight variance ratio for the All Ords, it significantly exceeds one, a result similar to the results of all the previous studies on the larger stock markets. The ratio is 2.3 (and statistically greater than one at 1% level of significance) if, in numerator, we use daytime variance calculated for the total sample period. Alternatively, the ratio is 1.8 if, in the numerator, we use daytime return variance calculated for days immediately subsequent to USA holidays. The arrival of the US information, thus, appears to be the primary source of a less than one ratio of trading to non-trading period return variances for All Ords. When the flow of the foreign information slows down due to public holiday in the US, this ratio is somewhat similar for Australia as it is for other stock markets. Incorporating the dominant impact of the US information, thus, appear to reconcile the results of the trading versus non-trading period volatility of the ASX with the prior evidence for larger markets of the US and Japan.(12)

The results in table 3, however, also indicate some influence of the major European markets on the volatility of the Australian stocks. As table 2 indicate, many of the holidays are common among these European markets. So, any evidence of reduced Australian volatility when one or more of these markets is closed cannot be seen as the influence of that one foreign market. Second row of table 3 shows that the Australian market volatility does go down when the major European markets are closed, but this reduced volatility is not statistically significant except for the UK stock market at 5% level of confidence. However, given that the number of UK holidays in our sample is only 9, it is hard to treat the result for the UK market as fully reliable. Overall, by looking at the results in table 3, it appears as if the closure (or joint closure) of the major European markets reduces the overnight volatility of the Australian markets but the results are not as strong as for the US market.

The Canadian market, however, dos not seem to have much impact, as the Australian market volatility is, in fact higher on the days of the Canadian market closures. The results in table 3 are consistent with prior studies of international market linkages (see Eun & Shim 1989) that find that the US market is the most dominant in influencing the stock markets in other countries.

5. Summary and Implications

In this paper, we report the relative return variances of the Australian equity index for trading and non-trading periods. The results seem inconsistent with the argument that most of the information affecting equity prices originates while financial markets are open. We propose and test the hypothesis that this contrasting evidence is due to the special significance of the US information for the Australian equities. Our results support this hypothesis. In particular we find that proportionately a larger amount of the information that affects the Australian stock market in a 24-hour day is generated when the ASX is closed, and that the US is the primary source of this non-trading-period information.

We believe our study contributes towards a deeper understanding of the arrival and processing of information in financial markets, and contributes to the existing literature several ways. First, by reporting that the Australian overnight volatility is only one third of its usual level during US holidays, our paper provide new and direct evidence of informational linkages between the US and the Australian stock markets. Our evidence is based on volatility tests and is unlike most prior work on international stock-market linkages that uses correlation or regression coefficients. As von Furstenberg (1994) argues, correlation coefficients are unable to establish or refute the existence of information linkages across markets and thus are unable to help us determine whether information in one market affects equity values in other markets.(13)

Second, our findings suggest a predictable pattern in the Australian overnight volatility, that is, the equity returns at the ASX are less volatile on days when the US has a public holiday. This could be useful for active portfolio management at short horizons. Finally, our results suggest caution in generalising the empirical findings based on one or two largest equity market data. While for equity and future markets in the US, prior studies have repeatedly shown that more information is released when the exchange is open, this result cannot automatically be extended to other equity markets. In fact, we conjecture that other smaller markets whose non-trading hours coincide with the US business hours may also have trading to non-trading period variance ratios similar to that of the ASX rather than those of the US and Japan.

It would, thus, be useful to examine the data from other smaller markets. Moreover, while in this paper we focus on the index return volatility at the ASX, an analysis of individual stock return variances over trading and non-trading periods may yield further refinement of the inferences we have drawn in this paper. We look forward to engaging in these research endeavours in future.

(Date of receipt of final transcript: September, 2000. Accepted by Stephen Gray, Area Editor)

(1.) In addition to information-related argument of French and Roll (1986), Miller (1989) presents another explanation for higher volatility of the NYSE equity returns during the daytime period. He argues that the specialists at the NYSE attempt to maintain the continuity of the prices and, thus, reduce the variability of equity prices at the open of the exchange. This results in an artificially lower volatility of overnight returns compared to that for daytime returns. However, Amihud and Mendelson (1990) cast doubt on the empirical relevance of Miller's proposition. They provide evidence that markets that do not have specialists and operate under different institutional arrangements also display a lower overnight volatility.

(2.) Other studies that also use individual stock returns for the US market and reach similar conclusions include French and Roll (1986), Barclay, Litzenberger and Warner (1990), and Forster and George (1994).

(3.) Hertzel, Kendall and Kretzmer (1990) also report higher return variance for trading period than for non-trading period for five currency futures contracts listed on the Chicago Mercantile Exchange.

(4.) The number of stocks in the index is not fixed and stocks are constantly added and deleted based on the twin criteria of size and turnover. At the end of September 1996, the index included 321 companies that accounted for about 93% of ASX capitalisation and 95% of ASX equity turnover (ASX Investor Handbook, 1997).

(5.) The daily opening of the ASX involves orders carried over from the previous trading day and new orders entered between 7:30 AM to 10:00 AM when SEATS is in its Pre-Opening mode. In the first ten minutes after 10:00 AM, the trading opens randomly in alphabetical order for different groups of stocks. The algorithm for executing orders at the open involves sequentially taking the best overlapping bid, calculating a volume weighted average price using the best overlapping ask, and then executing the two orders at this price. This matching procedure is repeated for the next best overlapping orders, until all such outstanding orders are executed. Thus, the SEATS system creates a series of prices at the open to match and execute the accumulated buy and sell orders. The officially quoted opening price is the first price generated by the opening algorithm of SEATS. It, therefore, does not represent the equilibrium price that would result after executing the entire buy and sell orders outstanding at the opening of the market. As a result, the quoted opening value of the index is inappropriate as a representative opening value of the listed equities.

(6.) The use of a delayed opening value also allows the observed stock prices to capture and reflect the influence of strategic trading by informed investors. The strategic trading occurs when the trading behaviour of informed traders causes elevation of price volatility and trading volume at the opening and close of daytime trading sessions (Admati & Pfleiderer 1988; Foster & Viswanathan 1994). Since in this paper, we attempt to draw inferences about varying rates of information flow during overnight and daytime periods, we must allow all the overnight information, including the private information of informed traders, to be reflected in the opening value of the index.

(7.) The standard approach for testing the hypothesis of equality of the two population variances is to calculate the variance ratio or F-statistic as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] are sample variances. To reject the null hypothesis of equal variances, one needs to compare the computed variance ratio (or F-statistic) with the critical value of F for appropriate degrees of freedom. For details of the variance ratio test, see Daniel and Terrell (1995).

(8.) The first, second and third sub-periods cover the time periods of 1 January 1991 to 31 December 1992, 1 January 1993 to 31 December 1994, and 1 January 1995 to 31 July 1996, respectively. The break points that define these sub-periods have been chosen arbitrarily around calendar years, as, a priori, we were unable to identify any significant economic or institutional event that could potentially create structural changes in the behaviour of the ASX intraday return volatility.

(9.) Although on a per-hour basis, their argument is still valid. That is, per-hour volatility of overnight return is still less than per-hour daytime volatility.

(10.) Copeland and Copeland (1998) also conclude that `the Americas lead Europe and the Pacific by one day, presumably because the drawing of information is generated in the Americas' (p. 76).

(11.) These comprise five public holidays that occur each year: Presidents' Birthday, Memorial Day, Independence Day, Labour Day and Thanksgiving Day, and an Exchange Holiday on 27 April 1994, that was a one off instance.

(12.) It is worth pointing out that the foreign information hypothesis that we present here is the sufficient but not the necessary condition for the pattern of reduced Australian overnight return variance during US holidays that we document here. If the information flow to the ASX from some other important source also slows down due to one reason or another on the days of US holidays, we may observe lower variances on these days. However, we are unable to think of any possible reason to expect so.

(13.) Von Furstenberg and Jeon (1989) assert that theory is not unequivocal as to what observed coefficients should be if markets are informationally linked. The markets that are informationally linked may exhibit positive correlation, negative correlation or no correlation at all. Von Furstenberg and Jeon point out that three possibilities can generate correlations between the stock indices of two countries ranging from negative to positive. If the news released in a country is redistributive (e.g. major patentable innovations), it would have opposite effect on the respective industry (stock) indices of the two countries. If the information is about the unconnected special situations (e.g. merger announcements) in one country, it is not going to cause any effect on the other country's equity prices. But if the news affects industries as a whole more or less irrespective of country (e.g. a world-wide glut or shortage of steel), it leads to positive correlation between equity prices across countries.

References

Admati, A. & Pfleiderer, P. 1988, `A theory of intraday patterns: Volume and price variability', Review of Financial Studies, vol. 1, pp. 3-40.

Aitken, M., Brown, P. & Walter, T. 1994, `Intraday patterns in returns, trading volume, Volatility and trading frequency on SEATS', working paper, Department of Finance, University of Sydney.

Amihud, Y. & Mendelson, H. 1990, `Explaining intraday and overnight price behaviour: Comment', Journal of Portfolio Management, winter, pp. 85-6.

Amihud, Y. & Mendelson, H. 1991, `Volatility, efficiency, and trading: Evidence from the Japanese stock market', Journal of Finance, vol. 46, no. 5, pp. 1765-89.

Australian Stock Exchange Limited (ASX), 1991-96, Weekly Index Analysis.

Australian Stock Exchange Limited (ASX), 1997, The Investor Handbook, Australian Stock Exchange Limited.

Barclay, M.J., Litzenberger, R.H. & Warner, J.B. 1990, `Private information, trading volume, and stock-return variances', Review of Financial Studies, vol. 3, pp. 233-53.

Becker, K.G., Finnerty, J.E. & Friedman, J. 1995, `Economic news and equity market linkages between the US and U.K.', Journal of Banking and Finance, vol. 19, pp. 1191-210.

Becker, K.G., Finnerty, J.E. & Tucker, A.L. 1993, `The overnight and daily transmission of stock index futures prices between major international markets', Journal of Business Finance and Accounting, vol. 20, no. 5, pp. 699-710.

Chang, E., Jain, P. & Locke, P. 1995, `Standard and poor's 500 index futures volatility and price changes around the New York stock exchange', Journal of Business, vol. 68, pp. 61-84.

Copeland, M. & Copeland, T. 1998, `Lead, lags and trading in global markets', Financial Analysts Journal, vol. 54, pp. 70-80.

Daniel, W.W. & Terrell, J.C. 1995, Business Statistics for Management and Economics, 7th edition, Houghton Mifflin Company.

Eun, C.S. & Shin, S. 1989, `International transmission of stock market movements', Journal of Financial and Quantitative Analysis, vol. 24, pp. 241-56.

Fama, E.F. 1976, Foundations of Finance, Basic Books Inc. Publishers, New York.

Forster, M.M. & George, T.J. 1994, `Trading hours, information flow and international cross-listing', Working Paper No. 191, Kellog School of Management, Northwestern University.

Foster, F. & Viswanathan, S. 1994, `Strategic trading with asymmetrically informed traders and long-lived information', Journal of Financial and Quantitative Analysis, vol. 29, pp. 199-518.

French, K.R. & Roll, R. 1986, `Stock return variances: The arrival of information and the reaction of traders', Journal of Financial Economics, vol. 17, pp. 5-26.

Hertzel, M.G., Kendall, C.S. & Kretzmer, P.E. 1990, `The volatility of asset returns during trading and non-trading hours: Some evidence from the foreign exchange markets', Journal of International Money and Finance, vol. 9, pp. 335-43.

Karolyi, G.A. & Stulz, R.M. 1996, `Why do markets move together? An investigation of US-Japan stock return comovements', Journal of Finance, vol. 51, no. 3, pp. 951-56.

King, M. & Wadhwani, S. 1990, `Transmission of volatility between stock markets', Review of Financial Studies, vol. 3, pp. 5-33.

Lin, W.L., Engle, R. & Ito, T. 1994, `Do bulls and bears move across borders? International transmission of stock returns and volatility', Review of Financial Studies, vol. 7, no. 3, pp. 507-38.

Lockwood, L.J. & McInish, T.H. 1987, `Stability tests for variances and means of intraday/overnight returns during bull and bear markets', Working Paper No. 87.114, Department of Finance and Real Estate, University of Texas at Arlington.

Miller, E.M. 1989, `Explaining intra-day and overnight price behaviour', Journal of Portfolio Management, vol. 15, no. 4, pp. 10-16.

Oldfield, G.S. Jr. & Rogalski, R.J. 1980, `A theory of common stock returns over trading and non-trading periods', Journal of Finance, vol. 35, no. 3, pp. 729-51.

Ross, S.A. 1989, `information and volatility: The no-arbitrage martingale approach to timing and resolution irrelevancy', Journal of Finance, vol. 45, pp. 1-17.

Stoll, H.R. & Whaley, R.E. 1990, `Stock market structure and volatility', Review of Financial Studies, vol. 3, no. 1, pp. 37-71.

von Furstenberg, G.M. & Jeon, B.N. 1989, `International stock price movements: Links and messages', Brookings Papers on Economic Activity, no. 1, pp. 125-79.

von Furstenberg, G.M. 1994, `Comment', in The Internationalization of Equity Markets, ed. J. Frankel, University of Chicago Press, pp. 338-43.

G. Mujtaba Mian([dagger])

Christopher M. Adam([sections])

([dagger]) Department of Finance and Accounting, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260. Email: [e-mail]

([sections]) Australian Graduate School of Management, UNSW Sydney, NSW 2052. Email: [e-mail]

We thank the area editor, Stephen Gray, and an anonymous referee for their helpful comments in improving the paper. Roger Loh provided useful research assistance for the completion of this paper. Any remaining errors are our responsibility.