This study proposes a text-based approach to classifying firms that engage in impression management (IM) in corporate narrative disclosures. Listed firms are required to communicate their operating environment and performance to stakeholders; however, prior accounting research documents that managers may suppress unfavorable information or strategically adjust linguistic features (e.g., tone and readability) to present the firm in a favorable light. As large-scale empirical analyses have proliferated, firms engaging in IM—heterogeneous in form and intensity—are pooled with non-IM firms, limiting the ability of linear models that relate performance to textual attributes to detect IM. We address this limitation by defining several performance indicators and applying text mining to firms’ Management’s Discussion and Analysis (MD&A) to identify whether each indicator is explicitly mentioned. We then classify firms as engaging in IM based on the consistency between underlying performance and disclosure. For instance, poor performance that is not discussed in the MD&A is indicative of IM through omission. In addition, by examining references to alternative indicators, we detect cases where managers acknowledge weak results but shift emphasis toward more favorable outcomes, indicative of strategic emphasis. The proposed labeling scheme supports more nuanced empirical tests of IM in narrative disclosures.

Keywords

Impression Management, Text Mining, MD&A, Narrative Disclosure, Japanese Listed Firms