Outlier Impacts in Stock Pricing

Jianlong Wang,Saiful Hafizah Jaaman,Humaida Banu Samsudin

doi:10.5296/jmr.v7i2.7050

Abstract

Outlier is the observation that is not consistent with the rest of observations. It exists not only in stock prices but also in the economic variables. In multifactor asset pricing model, the ordinary least square method (OLS) is commonly used to estimate coefficients. The existence of outliers can lead to inadequate results under the OLS framework. Huber’s robust method (HRM) can be used to avoid the bad impacts of outliers and the abnormal problems. Appling both methods to Shanghai stock market, the outlier observations are analyzed to examine its influence on the results and parameters estimation. The result of this study found that HRM outperforms OLS.

Highlights

As it is known, ordinary least square method (OLS) method is to minimize the square of residuals, in which the weight to each factor is the same
The result of this study found that Huber’s robust method (HRM) outperforms OLS
Since the different result of the OLS and HRM, further study and comparison are done

Summary

Introduction

OLS method is to minimize the square of residuals, in which the weight to each factor is the same. The regression results will be influenced directly. Robust regression model is an alternative to OLS when outliers exist. McDonald, Michelfelder, and Theodossiou (2009) claim that OLS is not the best estimation method of betas and may lead to erroneous estimates. P. Theodossiou (1998) and Bali and Weinbaum (2007) reject the normality assumption for returns as well as some macro-economic factors. This is even severe in terms of the outlier and non-normality distributed variables. By comparing OLS and robust regression method, he concludes that the robust regression method can help detect outliers and provide resistant results in the presence of outliers

Methods

Results

Conclusion