CEO overconfidence and the linguistic representation of corporate culture in 10-K filings: evidence from upper echelons theory

Purpose This study examines whether chief executive officer (CEO) overconfidence is associated with the linguistic representation of corporate culture in mandatory 10-K filings. Drawing on upper echelons theory, we argue that executives' psychological traits surface not only in strategic and financial outcomes but in the language through which corporate culture is communicated to external stakeholders, across the dimensions of control, collaboration, competition, and creation. Design/methodology/approach Using 27,742 firm-year observations from 2000 to 2022 drawn from Compustat, ExecuComp, and BoardEx, we measure CEO overconfidence through established option-exercise, stockholding, and investment proxies, and culture through textual analysis of 10-K filings. We estimate ordinary least squares models with year and industry fixed effects and assess robustness using firm fixed-effects models, lagged specifications, propensity score matching, entropy balancing, Heckman selection models, and instrumental variables. Findings CEO overconfidence is negatively associated with all four dimensions of culture-related language, with particularly pronounced associations for collective and procedural themes. The findings are robust across alternative specifications and identification strategies. The association is attenuated under female CEOs and more gender-diverse boards, concentrated in the discretionary Management Discussion and Analysis section, and accompanied by more optimistic and self-referential language, consistent with a managerial communication channel. Culture-related language is also associated with greater audit effort and market uncertainty and lower future performance. Supplementary pathway analysis reveals significant indirect associations for two overconfidence measures, linking the documented disclosure channel to economically relevant firm outcomes. Originality/value We extend upper echelons theory to the linguistic domain of regulated disclosure, showing that culture-specific language carries traces of managerial cognition even in collectively produced filings. By distinguishing control, collaboration, competition, and creation, the study also provides a practical framework through which auditors, boards, and reporting professionals can interpret culture-related language as a supplementary signal of managerial orientation, governance tone, and narrative-reporting risk.