V. C. Kharadia avd Robert A. Collins (1980-81), “Forecasting
credit union failures," Journal of Economics and Business,
This study used stepwise regression in internal financial rations to predict which credit unions will fail. Data were
selected from 162 credit unions that failed and from 162 that did not fail. These were split into estimation and validation sub-samples. Stepwise regression was used on the estimation sample, with some intervention by the authors to exclude a variable whose sign was clearly inappropriate. Overall, correct classification of the failed credit unions exceeded 90 per cent of the validation sample (versus 50 percent by chance). It was not clear how much lead time the model would provide. The currently used method, judgmental analysis of the same data by an examiner, appeared to be the poorer method. Surprisingly, however, the advantage of the regression method over judgmental analysis was not large and may not have been statistically significant. (The study did not provide a clear comparison of the two methods.) Finally, this paper concluded that credit union failures did not depend on the economy. Some readers may find that difficult to believe.