Estimation Period

The estimation period in an Event Study is the pre-event window used to fit the baseline ("business as usual") model. Coefficients estimated here are projected forward into the Observation Period to compute predicted values and Abnormal Returns. The estimation window should be long enough for reliable estimation but free of anticipation effects that contaminate the baseline.

When to use

Pick the estimation period before fitting anything. PS_3 chose Jan 1 1986 – Nov 19 1989 (~3.9 years pre-HNC) — long enough for four full annual cycles to identify seasonal coefficients, and ending exactly at the policy date to avoid post-policy contamination. The trade-off: longer estimation periods give more precise baseline coefficients but risk the data-generating process changing slowly over time; shorter periods are more current but noisier.