Applying Variable Deletion Strategies in Bankruptcy Studies to Capture Common Information and Increase Their Reality

Abstract

In financial distress studies selection of variable is commonly based<br />on the success of variables in variable sets employed in earlier<br />bankruptcy studies, suggestions in the literature or an accompanying<br />data reduction in a large set of variables. If seemingly different variable<br />sets exhibit a strong relationship then heterogeneous variable sets<br />capture common information. Canonical correlation analysis appropriately<br />examines the relationship between two sets of measured variables.<br />The main purpose of the present study was to illustrate the value of<br />variable deletion strategies in canonical correlation analysis for more<br />parsimonious to capture common information. In research contents, the<br />law of parsimony states that the fewer variables used to explain a<br />situation, the more probable that the explanation will be closer to reality.<br />Therefore, as variable sets become more parsimonious there are greater<br />probabilities that the results of the analysis will be replicable. To<br />determine the common information between variable sets in financial<br />distress studies, the study selected two specific bankruptcy models:<br />Altman, the most famous model, and Deakin, the biggest model. The<br />results indicated that as the number of variables increase, the probable<br />effect of these sources of error variation on the canonical correlation<br />increases. Therefore, the goal of a variable deletion strategy is toestimate as much variance with the smallest variable set possible. In this<br />study the goal was achieved by removing the three variables in variable<br />sets employed in selected bankruptcy studies.