Revisiting methods for assessing parameter estimation error in empirical datasets

File pdf 559KB

In this paper, we propose a robust methodology for assessing estimation error in empirical datasets. In contrast to recent findings, we show that established portfolio strategies outperform the equally weighted portfolio rule even in larger portfolios, despite increasing absolute estimation error. We find that, when short-sale constraints are introduced and in consideration of portfolios that do not rely on expected return estimates the relationship between portfolio size and measurement error, as previously hypothesised in literature, does generally hold. Measurement error in these strategies, however, is significantly reduced and their application to larger asset universes is yet beneficial. Finally, we discuss the usefulness of statistics to assess estimation error and propose an intuitive measure of return-loss due to “unfavourable estimation error” based on the downside deviation of the return distribution that quantifies the share of measurement error investors should ultimately be concerned about.

Author(s):

Benedikt Himbert    
WHU - Otto Beisheim School of Management
Germany

Julia Kapraun    
WHU - Otto Beisheim School of Management
Germany