U.K. cross-sectional equity data: do not trust the dataset! The case for robust investability filters

We propose a novel approach to cross-sectional equities sample selection, derived from best market practice in index construction and focused on investability. Using the U.K. market as a template, we first demonstrate how the popular Datastream dataset is plagued by data deficiencies that would surely invalidate statistical inferences, and that are not addressed by commonly used filters. We show the benefits and need for a supplementary data source. We then develop robust investability filters to ensure statistical results from cross-sectional analysis are economically meaningful, an issue overlooked by most studies on cross-sectional risk pricing