A holistic model for earnings quality measurement using partial least squares

Abstract

Earnings quality is an important topic in empirical research of financial accounting. This makes necessary developing appropriated empirical measures for this concept that allow the extraction of valid conclusions about its causes or consequences. Despite multidimensional nature of earnings quality, the majority of studies adopt unidimensional scopes for its measurement. These models suffer from misspecification and estimation bias for not considering the rest of dimensions. Furthermore, the scant number of multidimensional measures are not including all dimensions, with equal weights for all of them and without controlling for their correlations. The main methods traditionally used to estimate earnings quality suffer as well from several problems. For that reason, we propose the use of PLS as a systematic, double-step process that first analyses the validity and internal consistency of the concepts that are represented by the proxies and, if this is met, the representativeness of the relationships between the different variables.