day t, defined as ln(Pi,t) - ln(Pi,t-1), and where RMt is the return on the market index, which is
measured in a similar way as Ri,t. The market index chosen is the Datastream index for the Dutch
market, since this is the only index for which data are available for the whole sample period. The
parameters β0 and β1 are estimated over the estimation period by running an OLS regression of
the stock returns on a constant and the return of the market index.
Denoting the announcement date with t=0, this estimation period ranges from t=-110 to t=-
10. The event window ranges from t=-1 to t=+11. The test statistic is calculated following the
methodology as outlined by Brown and Warner (1985, page 7):
A∕s(At>, (2)
where At is the average abnormal return over the N different firms on day t and s(At) is the
standard deviation of the average abnormal return obtained from the estimation period. The null-
hypothesis is that the test statistic is zero. If the null-hypothesis holds and if the abnormal returns
are independently identically distributed with finite variance, the test-statistic is asymptotically
normally distributed. Besides calculating excess returns for each day in the event period, we also
calculate cumulative average abnormal returns2. Finally, to take into account the fact that the
standard error of the returns will be different for different firms, implying that the abnormal returns
of different firms will not be identically distributed, standardized abnormal returns are calculated as
well.
Event study results
The results of the event study are presented in table 2. This table contains the average abnormal
returns for day -1 to day +1 around the announcement date, as well as the cumulative average
abnormal returns for this event period.
[Insert Table 2]
It can be seen from table 2 that the average abnormal stock returns from day -1 to 1 are
positive but insignificant for both CBs and WBs. Cumulative abnormal returns are not significant
for CBs but they are significantly positive for WBs. The latter effect becomes stronger if
1 Several other estimation and event periods are tried as well. However, the results appear to
be robust with respect to the length of the estimation period as well as the event period, except that
there is no further announcement effect after day +1.
2 In calculating the test statistic for the cumulative abnormal returns it is assumed that excess
returns are not autocorrelated, so that the variance of a two-day excess return is just the sum of the
variances of the corresponding one day returns.