Econometric modeling for proactive risk management of financial failure in Moroccan SMEs: a stepwise logistic regression approach in python

Abstract This study develops a predictive model of financial failure specifically tailored to Moroccan small and medium-sized enterprises, based on financial data collected from a matched sample of 30 companies over the period 2019–2021. The methodology incorporates classical statistical techniques, including principal component analysis for dimensionality reduction, followed by stepwise logistic regression to construct the econometric model. The objective is to design a parsimonious, interpretable, and operational tool for the early detection of financial difficulties. Three financial ratios emerge as significant predictors: inventory turnover, economic profitability, and commercial profitability. The model demonstrates consistent predictive performance one, two, and three years before bankruptcy, with respective accuracy rates of 87%, 87%, and 83%, corroborated by tenfold cross-validation, thus confirming its empirical robustness. Unlike approaches based on complex artificial intelligence algorithms, this study adopts a transparent and interpretable methodological framework that is well suited to environments where data is limited, such as those frequently encountered in emerging economies. While the limited sample size is a constraint, the results underscore the continued relevance of traditional financial indicators in early warning systems. Future research could improve this model by incorporating macroeconomic and qualitative variables, thereby expanding its analytical depth and practical applicability.