Paper Type
Short
Paper Number
PACIS2025-1846
Description
Financial constraints may result in inefficient investments, hinder corporate growth, and obstruct the financial market’s role in effectively allocating resources. Thus, developing a robust financial constraint forecasting model has become essential. Numerical data, widely utilized due to its clarity and interpretability, can effectively depict a corporate’s past and current operational status. However, numerical information alone does not adequately convey prospective risks or corporate responses to future market dynamics. To address this limitation, textual data, which complements numerical information by capturing forward-looking and contextual insights, is integrated into the forecasting process. This study incorporates both textual and numerical information into a twin support vector machine (TWSVM) to predict financially constrained corporates. Empirical results demonstrate superior forecasting performance when textual information is included. These findings align with recent recommendations from accounting professionals advocating increased narrative disclosure of risk factors to reduce information asymmetry.
Recommended Citation
Hsu, Ming-Fu; Yeh, Dr. Ching-Chiang; Chang, Te-Min; and Lin, Sin-Jin, "Beyond Numerical Data: The Impact of Textual Information on Financial Constraint Prediction" (2025). PACIS 2025 Proceedings. 6.
https://aisel.aisnet.org/pacis2025/aiandml/aiandml/6
Beyond Numerical Data: The Impact of Textual Information on Financial Constraint Prediction
Financial constraints may result in inefficient investments, hinder corporate growth, and obstruct the financial market’s role in effectively allocating resources. Thus, developing a robust financial constraint forecasting model has become essential. Numerical data, widely utilized due to its clarity and interpretability, can effectively depict a corporate’s past and current operational status. However, numerical information alone does not adequately convey prospective risks or corporate responses to future market dynamics. To address this limitation, textual data, which complements numerical information by capturing forward-looking and contextual insights, is integrated into the forecasting process. This study incorporates both textual and numerical information into a twin support vector machine (TWSVM) to predict financially constrained corporates. Empirical results demonstrate superior forecasting performance when textual information is included. These findings align with recent recommendations from accounting professionals advocating increased narrative disclosure of risk factors to reduce information asymmetry.
Comments
AI ML