Stock return outliers and beta estimation: The case of U.S. pharmaceutical companies

• Data outliers can dramatically influence OLS estimates.
• Pharmaceuticals are characterized by large outlier returns.
• The Hubert Robust M Estimators are employed as an alternative to OLS.
• Hubert Robust M Estimators relate mathematically to OLS.
• OLS bias cannot be eliminated even when observations include 18 years.

Efficient estimation of the equity cost of operating public corporations is essential for a rational investment policy. Traditional OLS beta estimates of a single stock are known to suffer from violations of normality due to outliers – extreme returns caused by large, unpredictable company-specific events. We confirm the presence of an outliers-driven, often significant bias in OLS beta estimates by undertaking parallel estimates with a related method based on a mixed-return model that follows Huber's Robust M (HRM) estimator. We demonstrate that the OLS bias can be substantial even in a sample spanning 18 years of monthly observations.