A Firm’s Operational Risk: Data Set and Empirical Evidence

Problem definition: A firm’s stakeholders would benefit from a quantitative measure of the managerial disclosure of its operational risk and the implications of this disclosure. However, a quantitative and scalable measure of a firm’s disclosed operational risk is unavailable. Thus, the firm’s stakeholders remain unaware of its disclosed operational risk and the risk’s implications. Methodology/results: U.S. law requires a public firm to textually disclose its operational and nonoperational risks in Item 1A of its annual report (i.e., Form 10-K). We train 64 transformer models on U.S. public firms’ Item 1A text to score 131,920 firm-years (16,959 firms, 2005 to 2024) on eight risk factors: (1) accounting, (2) finance, (3) international, (4) legal, (5) management, (6) marketing, (7) operations, and (8) technology. We measure each transformer’s performance on eight metrics. Next, our Python code retains each risk factor’s best-performing transformer (among the eight). Subsequently, our regression estimates report that a firm’s disclosed operational risk is positively associated with its operational cost and that its disclosed nonoperational risk strengthens this positive association, thus supporting our two hypotheses. Our OSF repository includes an Excel file that contains a data dictionary and count and probability scores of the eight risk factors for 131,920 firm-years (16,959 firms, 2005 to 2024). The repository also includes our Python code file and trained models’ files. Managerial implications: First, our empirical evidence informs managers and corporate stakeholders that a firm’s disclosed operational and nonoperational risks are associated with its operational cost, thus showcasing disclosure’s relevance. Second, our Excel data file provides stakeholders with eight risk factors for 131,920 firm-years (16,959 firms, 2005 to 2024). Third, one can also use our Python code files and trained transformers to measure risk reflected in other sources of firm-generated text (e.g., managers’ answers in earnings calls, CEO interviews, and press releases). Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2025.0134 .