Impact of Outliers on Stock Return Models: Implications for Event Studies and the Pricing of Risk

Abstract

This paper investigates the impact and implications of outlier returns for event studies and the pricing of risk. A mixed regression process consisting of a regular and an outlier component is used to model returns for individual stocks. The regular component of stock returns is estimated using Huber’s Robust M estimation method. Estimates of betas from the regular return regression models are shown to be appropriate measures for systematic risk. Moreover, regular return models are appropriate for the computation of cumulative abnormal return (CAR) statistics employed in event studies to test for significant impact events. This paper shows that when outliers are present in the data, models based on OLS estimation method lead to erroneous estimates of systematic risk and CAR test statistics.

Keywords: CAPM, CAR tests, Fama-French model, M-estimator, portfolio management, systematic risk

JEL Classification: G11, G12