Assessing the Spatial Dimension in Financial Analysis: Evidence from Italian SMEs

Abstract

Effective risk assessment is a prerequisite for informed financial decision-making and sustainable economic growth, as it supports efficient capital allocation and helps mitigate systemic vulnerabilities. This requires advanced predictive frameworks that account for the broader economic and financial context in which firms operate. This study addresses the research gap related to the limited adoption of spatial models in financial analyses by assessing how the integration of the spatial dimension into machine learning classification models enhances their overall performance. Focusing on Italian small and medium enterprises (SMEs) as a case study and using firm-level data from the AIDA database for the year 2022, the analysis compares results from the non-spatial probit model with the probit spatial autoregressive and spatial Durbin models. The findings demonstrate that spatial models consistently outperform the non-spatial specification in terms of accuracy, sensitivity and specificity, particularly when the spatial proximity parameter is optimally defined. Moreover, the results show that financial distress tends to cluster geographically, and that firms’ default risk is influenced not only by intra-firm indicators but also by the characteristics of nearby firms.