Are Credit Scoring Models Able to Predict Small Enterprise Default? Statistical Evidence from Italian Small Enterprises

From as early as the 1960s, there have been a large number of studies aimed at assessing the application of statistical models to corporation data with a view to predicting business failure. This issue has become increasingly important in recent years, as the New Basel Capital Accord (Basel II) linked capital requirements to banks' models for company default prediction. Empirical research shows that economic-financial ratios can help to predict company default through the implementation of statistical techniques. The literature focuses mainly on large and medium sized enterprises that systematically produce detailed financial documentation. However, the financial statements of small enterprises (SEs) tend to disclose less (and are therefore more difficult to interpret), and this prevents a widespread use of statistical models. The issue is of vital importance in countries like Italy with large numbers of SEs. This paper applies mainstream statistical techniques (linear discriminant analysis and logistic regression) to a sample of over 6,000 Italian firms in the attempt to develop two distress prediction models, specifically constructed for SEs and taking into account diversity of size, geographical location and business sector. For both models, prediction accuracy increases progressively with larger firms, and is higher in the North and in manufacturing firms. The success rate is lower in smaller firms and for those located in Southern Italy. Our results suggest that the limited information in SE accounts affects a model's prediction success rate, and also that SEs need to be assessed with specifically designed models