Overview

An empirical researcher measuring one corporate event starts by naming it. In this pathway the event is a quarterly earnings announcement: thirty Dow firms, 600 announcements, and a sort of the I/B/E/S surprise into good news, no news, and bad news at a 2.5 percent band. The design then separates the days that train the normal-return model from the days that are allowed to contain the announcement, and it chooses the conditional mean that will be subtracted. The abnormal return itself is the next topic.

By the end of this note you should be able to:

• place an earnings date inside the seven-step outline;
• mark an estimation window and an event window that do not share days;
• choose between a constant-mean return and a market model using the $R^2$ variance reduction.

1. Contents
2. The big picture