Changes in Firms’ Business Models and Accrual Estimation

Standard accrual models estimate discretionary accruals under the assumption that the relation between firms’ operating activities and accruals is stable. When firms change their business models, this assumption may be violated: non-discretionary accruals arising from structural shifts in operations can be misclassified as discretionary, even absent changes in reporting behavior. This study constructs a firm-specific, time-varying measure of business model change (BMC) by using transformer-based embeddings of consecutive annual Item 1 disclosures. The measure is strongly associated with observable operational shocks but exhibits no significant association with indicators of deliberate earnings management. Incorporating the measure into the first-stage estimation increases explanatory power, and permutation tests indicate that this gain is fully attributable to firms’ own business model changes rather than to mechanical overfitting. The improvement in the separation between normal and abnormal accruals in the Xie (2001) earnings persistence framework is concentrated among firms experiencing substantial business model change: in the top BMC quintile, formal model comparison tests favor the augmented model, and the persistence of standard-model abnormal accruals declines significantly as business model change increases, while the decomposition remains unaffected for firms with stable business models. Overall, the findings indicate that business model change is a systematic determinant of normal accruals omitted from standard models.