個股研究中事件窗口（pdf 24）

It is common practice to assess the average effect of some type of
announcement on stock prices by performing event studies on a large
sample of firms and then averaging or otherwise combining the results.
One benefit of this procedure is that the event window length can be
standardized across observations because the errors from having too
long or short an event window should have a small impact on the
average by the Law of Large Numbers. Here, we examine various
potential rules for determining the length of an event window when
looking at a limited number of observations. We find that rules based
on continuing price movements yield window lengths that correlate
with the “size” of the news, as measured by the magnitude of earnings
surprises, while a rule based on abnormally high volume does not have
this property.
..............................