Bankruptcy studies andad hocvariable selection: a canonical correlation analysis

Purpose - To determine the relationship among covariates used in financial distress studies. Design/methodology/approach - The study selects four specific bankruptcy studies and employs canonical correlation analysis to determine the relationship among the different variable sets that these studies used as predictors of financial distress. Canonical correlation analysis identifies the relationship and provides an indication of the amount of redundancy that exists between two variable sets. The four studies are representative of the genre, similar as to choice of statistical technique, and frequently cited by researchers. Findings - The research findings indicate that the relationships between the alternative variable sets are very weak and alternative variable sets do not represent similar financial relationships. Redundancy coefficients suggest that, if one variable set is redundant to another variable set, it is because the redundant variable set, is much smaller than the predictor variable set. Research limitations/implications - The results suggest that there is not much similarity among the variable sets used in financial distress studies; to the extent that there is any similarity, it is due to variables common to each set or one variable set being larger than the other variable set.Ad hocvariable selection in financial distress studies results in the use of alternative variable sets containing heterogeneous variables unrelated to one another. Originality/value - A common criticism of financial distress research is that a theory of corporate failure does not exist. Variable selection is not prompted by economic theory but is based upon suggestions in the literature, the success of variables in earlier studies, or the selection of a large set of variables with an accompanying data reduction procedure. Despite nearly 30 years of research in the area, the absence of an inter‐correlational structure among alternative variable sets highlights the atheoretical nature of financial distress research.