THE STABILITY OF BANKRUPTCY PREDICTORS IN THE CONSTRUCTION AND MANUFACTURING INDUSTRIES AT VARIOUS TIMES BEFORE BANKRUPTCY

This article focuses on the design of bankruptcy models, specifically the selection of suitable predictors. Previous research has drawn mainly on data concerning manufacturing companies one year before bankruptcy. Our research examines financial ratios that are suitable bankruptcy indicators in two different industries (the construction and manufacturing industries) over a period of five years prior to bankruptcy. Our main objective is to verify whether bankruptcy predictors are industry-specific. Another objective was to determine which indicators can detect signs of bankruptcy earlier than one period before bankruptcy. We presume that the application of industryspecific indicators can help increase the predictive accuracy of bankruptcy models when applied to a particular industry. Per analogiam, we assume that the inclusion of indicators capable of detecting signs of bankruptcy more than a year before its occurrence will increase their predictive capacity. Significant predictors were first identified on a linear basis using the parametric t-test or F-test; for the sake of comparison, a non-linear non-parametric Boosted Trees method was also applied. Data for a total of 34,229 active companies and 304 companies that went bankrupt during the relevant period was analyzed. The research confirmed our presumption that bankruptcy predictors are both industry and time specific. Four years before bankruptcy, the indicators return on assets, inventory turnover and asset structure are important predictors in both the manufacturing and construction industries. The net working capital to total assets ratio is a specific predictor for manufacturing companies in the third year before bankruptcy, as is the short-term indebtedness indicator. In the construction industry, specific predictors are the net working capital to sales ratio in the third and first years before bankruptcy, and the interest coverage indicator in all four years preceding bankruptcy. Were these indicators to be included in a model for an alternative industry, they would be likely to reduce its accuracy.