Distortion effects and extreme observations in empirical research: An analysis of the incremental information content of cash flows

Abstract

AbstractLinear regression models are used in a number of studies examining the presence or absence of incremental information content in cash flows. The results of these studies are not consistent. This paper provides evidence of the impact that extreme observations can exert on parameter estimates in a regression model. Two techniques commonly used to address the problem of extreme observations are considered. These techniques, winsorising the data and trimming the data, are compared to a regression diagnostic technique, Cook's distance. The comparison of these techniques provides evidence that the choice of technique can determine the significance or otherwise of regression results. This paper concludes that the inconsistency in reported results examining the incremental information content in cash flows may be attributed to the techniques adopted to address the issue of extreme observations.