Can linguistic vagueness in annual reports predict supply chain risk?

Annual reports serve as a critical information channel for corporate stakeholders, especially for supply chain partners. We argue that linguistic vagueness in these reports can function as an early-warning signal for predicting supply chain risk. To test our proposition, we develop a theory-grounded measure of linguistic vagueness by integrating communication vagueness theory with the Word2Vec method. Our analysis of publicly listed companies in China reveals a robust positive association between linguistic vagueness and subsequent supply chain risk. Heterogeneity analyses indicate that this predictive capacity varies under different reputational pressures. Furthermore, by employing interpretable machine learning approaches within an out-of-sample predictive framework, we validate the consistent directional predictive signal via SHAP analysis and solidify the indicator’s predictive robustness. Ultimately, this study establishes linguistic vagueness as an incremental signal that carries significant out-of-sample predictive value for supply chain risk, providing stakeholders with a data-driven approach for proactive risk monitoring.