The silent risk: exploring underreporting bias in occupational accidents through severity-based modelling

Abstract Ensuring safe and healthy working conditions is a fundamental right within the European Union. However, the reliability of occupational accident statistics is often compromised by underreporting, especially for non-fatal injuries. This paper examines the extent to which underreporting bias affects model specification in the analysis of occupational accident rates (OARs), using Italy as a case study. By focusing on a single-country context with a uniform regulatory framework for Occupational Safety and Health (OSH) and Employment Protection Legislation (EPL), we avoid the institutional variability that often complicates cross-country comparisons. Using INAIL and ISTAT data at the provincial (NUTS-3) level for the period 2011–2019, we conduct panel regression analysis, relying primarily on fixed effects estimates. We distinguish among minor, severe, and severe-plus-fatal accidents to assess how potential underreporting influences the consistency and robustness of model results. Our findings indicate that while underreporting—particularly of minor injuries— may attenuate effect sizes, it does not substantially compromise model specification or the relationships between workplace accidents and socio-economic, demographic, and institutional factors. These results suggest that minor injury data should be interpreted with caution, especially during economic downturns when reporting incentives weaken. From a policy perspective, the study highlights the importance of reinforcing both safety measures and reporting mechanisms, particularly in vulnerable sectors and recessionary periods. More broadly, the paper contributes to improving the reliability of OAR indicators and supports more accurate modelling of workplace safety dynamics.