RDP 9216: The Evolution of Corporate Financial Structure: 1973–1990 Appendix 2: Sample Selection Bias

The graphs presented in the paper are based on a balanced data set of 110 firms. A balanced or constant sample of firms was chosen so that changes in the ratios were not induced by changes in the basket of firms being measured. However, using a balanced sample introduces a potential selection bias. Firms that fail are typically highly geared and have low interest cover. Thus, if firms leave the sample, due to failure, the average debt-asset ratio of the balanced sample of firms is likely to underestimate the actual debt-asset ratio of all firms that were operating at a particular point in time. Alternatively, firms may leave the sample by being taken-over by another company. Firms that are taken-over often have relatively low gearing. Thus, if take-overs represent the principal form of attrition, the constant sample of firms may overestimate the actual debt-asset ratio of all firms operating at a particular point in time.

This appendix examines the difference between the ratios obtained from the balanced sample of 110 firms and the ratios obtained from the larger but unbalanced sample. For this purpose the weighted averages from both the balanced and unbalanced samples are presented in Graphs A1 through A4. The unbalanced sample consists of up to 224 firms.

The qualitative conclusions from the unbalanced sample match those reported in this paper. The differences between the ratios generated by the two samples are insignificant relative to the changes in leverage that were occurring over the sample period. Given that the graphs generated from either sample yield the same qualitative conclusions, we are satisfied with reporting the results for the balanced sample only.