Theil's nonparametric estimation in event study methodology

Date of Completion

January 1997


Statistics|Economics, Finance




Event studies have become a frequently employed tool for researchers in financial economics. The identification of the direction, magnitude, and speed of security price changes in response to new information provides immediate feedback to investors, management, and government regulatory bodies.^ Crucial to the correct determination of whether or not an event actually has had an impact is the measurement of excess security returns. Previous research about event study techniques has supported the use of nonparametric test statistics because of the non-normality observed in security returns. These studies have used a parametric procedure in the estimation of parameters for the market model. This dissertation, in contrast, applies Theil's nonparametric regression in the estimation of abnormal returns; an approach that is distribution free and provides a more internally consistent nonparametric approach for the detection of abnormal performance.^ Simulation results indicate that Theil's estimation offers improved power in the detection of abnormal performance over the traditional methodology. The results are robust as to choice of stock market examined (NYSE or NASDAQ), size of the portfolio, and event induced variance increases. The nonparametric combination of the Theil estimation procedure and the rank statistic is found to provide higher power in the symmetrical detection of both negative and positive abnormal performance. ^