Measuring Long-Horizon Security Price Performance
Working Paper FR 95-19
Posted: 1 Jul 1998
Date Written: October 1995
This paper studies the specification of tests for long-horizon (i.e., multiyear) abnormal security returns around firm-specific events, using samples of randomly selected securities and simulated event dates. The main result is that typical long-horizon tests are misspecified. The tests have a severe tendency to indicate abnormal performance when none is present, with rejection frequencies sometimes exceeding 5 to 10 times the significance level of the test. The result is not sensitive to the specific model used for estimating abnormal returns. The parametric test statistics do not satisfy their assumed properties. We identify several sources of misspecification.
JEL Classification: G30
Suggested Citation: Suggested Citation