Variable selection for international bankruptcy forecasts

Corporate financial distress not only incurs serious financial loss to its creditors but also has a high cost to the society and the country´s economy. Consequently, financial distress prediction studies are important to all those involved: owners, shareholders, lenders, suppliers, and government. In this paper, we focus on the corporate bankruptcy prediction for international market using CompuStat Global database. First, we introduce a robust variable selection technique, called adaptive lasso (least absolute shrinkage and selection operator), on the global corporate bankruptcy data in search for a parsimonious default forecasting model. Second, we demonstrate a comprehensive case study on Japan bankruptcy prediction. Comparing to US and UK, significant variables selected are generally different across countries. Only firm´s activity indicator Sales/Total Assets displays uniform significance across three countries.