This study investigates whether CEO characteristics improve the prediction of firm-level maturity mismatches between investment and financing and develops an improved framework to evaluate their contribution.
We compare the out-of-sample performance of a CEO-augmented model (which includes CEO characteristics) with a baseline model that contains only firm fundamentals in predicting firm-level maturity mismatches. The comparison spans machine learning algorithms (Lasso, Elastic Net, Decision Tree and XGBoost) and traditional linear approaches (standard OLS and OLS with random effects). After identifying XGBoost as the most effective algorithm in capturing the predictive value of CEO characteristics, we use it to explore the economic implications of enhanced prediction.
We construct an XGBoost-based predictive framework and demonstrate that incorporating CEO characteristics enhances the prediction of firm-level maturity mismatches. This indicates that CEO characteristics can also function as predictors with incremental value beyond firm fundamentals and that XGBoost effectively captures their complex and nonlinear associations. These findings are robust across validation checks. Variable importance analysis identifies CEO tenure, annual cash compensation and shareholding as key predictors exerting intricate effects on maturity mismatch patterns.
While this study generates several valuable findings, some limitations remain. First, although incorporating CEO characteristics improves predictive performance, the gain is relatively modest. This likely reflects both the inherent complexity of maturity mismatch and the limited informational contribution of CEO features relative to firm fundamentals in this setting. Second, as a predictive approach, machine learning restricts causal interpretation. Although our feature importance and pattern analyses offer some insights, they cannot fully uncover the underlying mechanisms. Future research could explore double machine learning or hybrid methods to better integrate causal inference with prediction.
The proposed framework yields practical implications. To our knowledge, this is among the first studies to develop a predictive system for firm-level maturity mismatch. By integrating CEO characteristics, our framework enables regulators and investors to flag potential mismatch risks ex ante. Moreover, the use of machine learning allows for adaptation to region-specific variables, enhancing contextual relevance and predictive accuracy, especially in diverse emerging market settings. Our analysis also uncovers economically meaningful patterns. In particular, the observed trade-off between equity- and cash-based compensation may reflect their differing effects on agency costs and maturity decisions. These insights offer firms a new perspective for incorporating financing risk considerations into executive compensation design.
This study uncovers a nonlinear predictive relationship between CEO characteristics and firm-level maturity mismatches and proposes a CEO-augmented predictive framework to improve prediction performance. The framework helps regulators and investors identify hidden risks and encourages firms to consider CEO characteristics in executive selection and incentive design to mitigate financial fragility.
