encompasses. For SPI 200 Futures, trading volume is the total number of contracts
traded in each 30-minute interval.
In order to construct a continuous time-series for SPI 200 Futures, we use the method
of rolling over futures contracts. In particular, we use the data of the closest-to-maturity
contract and ‘‘roll over’’ to the second closest-to-maturity contract on the day before the
expiration date of the closest-to-maturity contract.^3 Day trading on SPI 200 Futures
commences at 9:50 am and ends at 4:30 pm. In other words, the day-trading session of
SPI 200 Futures starts 10 minutes before the open and finishes 30 minutes after the close
of the equity market. Following Chan, Chan, and Karolyi (1991), we deal with this issue
by removing SPI 200 Futures data before the open and after the close of the equity
market. Consistent with Kalev et al. (2004), we collapse the period from the previous
day closing (4:00 pm) until the next day 10:30 am into one single period. Although our
approach may not be a perfect solution, it delivers the benefit of having a continuous
return series for both the index and futures markets.
Information flow in our study is proxied by the total number of news announcements
made by companies listed on the ASX and stored in the Signal G database. This
database is collected from SIRCA and provides headlines of announcements made
by companies on the ASX. Signal G gives details on the code of the announcing
company, the announcement date and time, and the categories of announcements, such
as takeover announcements, shareholder details, periodic reports, dividend announce-
ments, company administration, issued capital, and asset acquisition and disposal.^4
Based on the announcement date and time, we compute the total number of announce-
ments made by all companies in the S&P/ASX 200 Index in each 30-minute interval.
Similar to the case of returns, all overnight announcements are accumulated into the
first period (i.e., from the close of the previous day to 10:30 am of today).
In order to deal with the intraday pattern of returns (see, e.g., Wood, McInish, and
Ord, 1985), we regress our raw returns on 12 different dummy variables for the 12
intraday intervals. We use the residuals of this regression as our measure of seasonally
adjusted returns. In addition, similar to the observation of Kalev et al. (2004), we find an
upward trend for trading volume over the sample period. Therefore, to avoid spurious
results we de-trend the trading volume of both the S&P/ASX 200 Index and the SPI 200
Futures by estimating the following OLS regression:
Vt¼ 1 þ 2 tþ 3 t^2 þet; ð 12 : 1 ÞwhereVtis the aggregate trading volume in the intervalt. We useet, the residual from
the regression, as the de-trended trading volume.
In this study, similar to Kalev et al. (2004), we adopt the total number of news
announcements of all constituent stocks in the S&P/ASX 200 Index as the main proxy
for information arrivals. We test the relation between news arrivals and the volatility of
S&P/ASX 200 Index returns and SPI 200 Futures returns by estimating the censored
regression model with the absolute value of returns as our proxy for volatility. This
relation is also examined using univariate and multivariate conditional volatility models
Firm-specific news arrival and the volatility of intraday stock index and futures returns 275(^3) The results are qualitatively similar when we roll over futures contracts three or five days before the expiration date of
closest-to-maturity contracts. This is consistent with Carchano and Pardo (2009), who find that, for stock index futures,
different ways of rolling over futures contracts do not result in differences between resultant futures time-series. 4
Signal G also provides additional information such as the industry subgroup the announcement belongs to or the location at
which the documents were lodged.