This policy brief introduces a novel statistical framework designed to capture latent behavioral dynamics in European small and medium-sized enterprises (SMEs), accounting for data contamination and transition uncertainty.
The model integrates a bias-adjusted three-step latent Markov estimation procedure with the robust Optimally Tuned Robust Improper Maximum Likelihood Estimator (OTRIMLE). The model is implemented entirely in-house using custom-developed R code, tailored to analyze SME balance sheet data under real-world imperfections.
While empirical application is forthcoming, the model has been extensively tested on simulated data, demonstrating strong performance in accurately recovering latent states and transitions in the presence of outliers.
Once estimated on the EU SMEs sample, the latent state variable is mapped to ESG (Environmental, Social, and Governance) scores, enabling a forward-looking interpretation of firm sustainability. This methodology opens a pathway toward robust, longitudinal risk segmentation for SMEs, with potential for broader deployment across economic and financial domains.