La previsione dell’insolvenza ex art. 13 co. 3 C.c.i.: efficacia del test di classificazione binario e dell’analisi discriminante lineare

We argue that insolvency forecasting for SMEs will soon become a relevant topic for entrepreneurs and consultants alike, in the wake of the upcoming reform of the Italian bankruptcy law (d.lgs. 14/2019 – Codice della crisi d’impresa e dell’insolvenza), and that data scarcity is the main obstacle to the development of predictive tools for SMEs. In order to introduce the topic to a broader audience, we present an analysis of the history of insolvency prediction models under the light of the reformed legislation, outlining a general framework for the construction of insolvency prediction models for and by Italian small and medium enterprises, in-house or with help from their consultants, in compliance with Article 13 sub. 3 of the Business distress and insolvency Code. We employ the framework to build two classes of models. The first class employs an outdated approach, the univariate dichotomous classification test. The second adopts one that is more widely used in SMEs insolvency prediction: linear discriminant analysis (LDA). We then perform a comparison between the predictive abilities of the two. We draw the conclusion that the former is more effective than the latter, within the limited boundaries of the experiment. Such result is mildly inconsistent with the literature on the topic. We underline how the scarcity of data about Italian SME limits, both in the empirical set-up and in real life, the accuracy of the LDA. This leads to the conclusion that, in this context, a simpler statistical approach may yield a more satisfactory output. Finally, we suggest how both models could be improved in future research.