Whose voice speaks? A multilevel model of CEO sense giving through communication

Abstract

This study reconceptualizes CEO sense-giving communication targeted at investors and stakeholders as a layered, multilevel phenomenon where CEOs' individual voice is nested within firm-level communication norms and, in turn, industry-level norms. We quantify the relative importance of CEO, firm and industry determinants in key conference-call communication attributes that are central to stakeholders' evaluation and to the informational environment surrounding audited financial reporting. Using transcripts of quarterly earnings conference calls for S&P 1500 firms (2010–2018), we construct four validated communication measures – forward-looking information, external focus, tone, and concreteness – using established dictionaries and natural language processing (NLP). We then perform variance partitioning via multilevel modeling (years nested within CEOs, within firms, within industries) to estimate the proportion of variance attributable to CEO, firm and industry factors. We further test whether CEO effects intensify under high-stakes conditions (negative earnings surprises and low analyst following). Across communication attributes, CEO effects are sizable (approximately 19–31% of variance) and coexist with meaningful firm (about 11–21%) and industry effects (about 5–16%), supporting a nested structure of CEO communication. CEO influence strengthens when external scrutiny and information needs are heightened: CEO-attributable variability increases following negative earnings surprises, and when analyst following is low. CEO effects are also more pronounced in the Q&A portion of calls than in the scripted presentation. The study reframes CEO sense-giving from an individual-level account to a layered agency-in-context model of disclosure style and introduces variance partitioning as a direct way to separate CEO voice from firm- and industry-level communication norms. The results inform strategic management, accounting and governance research by clarifying when “CEO style” metrics capture managerial agency versus contextual norms, with implications for interpreting narrative disclosures and for oversight by boards, audit committees and assurance providers.