The Wiley Finance Series : Handbook of News Analytics in Finance

does reflect to some degree the intensity of private information flows not accounted for
byNt. The results are obtained based on maximum likelihood estimation with the left-
censoring point of 0 imposed on the dependent variable.
After censored regression, we apply the EGARCH model of Nelson (1991) in our
investigation of the impact of news arrivals on volatility. According to Bollerslev, Chou,
and Kroner (1992), GARCH-type models are widely used in studies which deal with
volatility modeling. EGARCH has several advantages over the pure GARCH specifica-
tion. First, there is no need to impose non-negativity constraints on model parameters.
Second, the EGARCH model allows for asymmetric response of volatility to positive
and negative shocks, where a negative shock on the financial time-series is more likely to
cause a bigger increase in volatility than a positive shock of the same magnitude. The
following autoregressive of order 1, EGARCH(1,1) model (AR(1)–EGARCH(1,1)), is
estimated for the S&P/ASX 200 Index and the SPI 200 Futures:

Mean equation: rt¼rt 1 þþ"t; ð 12 : 3 Þ

Variance equation: logð^2 tÞ¼!þ logð^2 t 1 Þþ

"t 1

t 1

þ^

"t 1

t 1

þNtþVt 1 ; ð 12 : 4 Þ

in whichrtis either the seasonally adjusted return of the S&P/ASX 200 index or the SPI
200 Futures at thetth interval, and^2 tis the conditional variance of the error process
("t). This model is first estimated without any exogenous variables in the variance
equation, then includes the number of all company announcements (Nt), and finally
the number of all company announcements with lagged de-trended trading volume
(Vt 1 ). The degree of volatility persistence is reflected by the coefficient while
indicates the asymmetric response of volatility to positive shocks and negative shocks.
We estimate the EGARCH(1,1) model where the error term follows Student’st-dis-
tribution in order to incorporate the potential leptokurtic distribution of the error term.
We expect thatis positive and significant in equations (12.2) and (12.4) (i.e., higher
information arrivals lead to higher volatility). In addition, following Lamoureux and
Lastrapes (1990), ifNtis serially correlated and the presence of volatility persistence is
largely induced by serial correlation in information flows, the persistence of volatility ( )
in equation (12.4) should be substantially reduced in comparison with the estimates of
the EGARCH(1,1) model without any exogenous variable.
We also examine the conditional volatility of S&P/ASX 200 Index returns and SPI
200 Futures returns simultaneously by estimating the Diagonal VECH and Diagonal
BEKK models. These two models are simpler forms of the VECH model (Bollerslev,
Engle, and Wooldridge, 1988) and the BEKK model (Engle and Kroner, 1995), in which
conditional variance and covariance depend only on their own lags and the cross-
products of the error term. The Diagonal VECH and Diagonal BEKK models simplify
estimation of the VECH and BEKK models and still have the advantage of controlling
for contemporaneous correlation of residuals across equations. We estimate the follow-
ing models in our analysis:

Firm-specific news arrival and the volatility of intraday stock index and futures returns 285