When using the event study methodology, the initial task is to
identify the events and the event window, the period over which
stock prices are examined. In our case we define an event as the arrival
of news related to the crisis in Indonesia. This is a difficult and
manual job. As described in the data section below we use a number
of data sources and hand collect the events. Choice of the event
window entails some arbitrariness in a study like this (Kho and
Stulz, 2000). One difficulty with defining the event window is that
it is not always clear when an event takes place. As we describe in
detail in the data section we take into account the time differences;
for example, an event taking place in the US on day t gets incorporated
into prices in Indonesia on day t + 1. Still we do not know the
intraday timing of the announcements. Therefore we use a conservative
approach. We use a 2-day event window (t = 0, 1), covering
the day of the announcement and the day after the announcement
to capture the price effects of announcements, which occur after
the stock market closes on the announcement day.16 A second difficulty
is when event windows overlap in calendar time. Due to the nature of the crisis, sometimes news arrival is fast, and in consecutive
days we have a series of news of a different nature. For example,
the IMF makes an announcement and the government reacts to it
and then the public reacts further and riots erupt. We calculate
ARs for such event clusters by using extended event windows. In
Table 3 we give the details of such extended event windows in historical
sequence of events.