Signaling overreporting and underreporting in sustainability: A methodological framework based on ESG rating divergences

This paper proposes a residual-based framework to detect potential ESG-washing and underreporting signals among STOXX600 firms by exploiting the divergence between ESG performance scores from Refinitiv and disclosure scores from Bloomberg. For each ESG pillar and for the aggregate ESG score, we estimate expected disclosure using a linear benchmark and three nonlinear models, and define GAP indices as the unexplained component of disclosure conditional on ESG performance, firm size, sector, and year. Model comparison based on value-based and rank-based metrics shows that nonlinear specifications, especially Random Forest, outperform the linear benchmark and rank firms more accurately along the disclosure distribution. Firms in the upper and lower tails of the GAP distribution are identified as potential washers and underreporters, respectively. The evidence reveals sectoral and pillarspecific asymmetries: washer signals emerge more frequently in Industrial Goods & Services and Financial Services, whereas underreporter profiles are more evident in sectors such as Utilities, Real Estate, and Basic Resources. The proposed framework transforms ESG rating divergence into an informative signal on disclosure credibility.