Estimation and Event Windows

Once the earnings date is fixed, the researcher still has to decide which days train the normal-return model and which days are allowed to contain the announcement. Those two sets of days are the estimation window and the event window.

Sharing days lets the announcement revise the benchmark that is then used to measure the announcement. This note sets the timeline, states the non-overlap rule, and records how a longer event window spends the power MacKinlay calculates for a one-day test.

The choice of benchmark is the next lesson; the residual itself is the abnormal return.

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

label $T_0$ , $T_1$ , $T_2$ , $T_3$ and the lengths $L_1$ , $L_2$ , $L_3$ on an earnings timeline;
explain why the estimation window and the event window are kept disjoint;
place an after-the-close announcement on day 0 and day $+1$ ;
read the power comparison between a one-day variance of $0.0004$ and a wider-window variance of $0.0016$ .

In this lesson

Event time
Why the windows stay apart
A numerical earnings timeline
Power and the length of $L_2$

Before you start

Event Definition

The full lesson, with worked examples and exercise solutions, is for members.

Read the full lesson

New to SelfMadeQuant.com? See membership options or browse the Event Studies and Causal Measurement in Finance syllabus.