Integration of Accounting-Based and Option-Based Models to Predict Construction Contractor Default

Abstract

This paper aims to predict construction contractor default, which is excluded by most extant studies, due to the distinct characteristics of construction industry. Default predicting models developed in past literatures are mostly built by accounting information, yet accounting sheets have innate flaws. To calculate default probability, several recent studies applied the option pricing theory, which presumes that the stock market is efficient. This presumption isn't always true in real life. In this paper, a hybrid model is proposed. It combines information from both models by inputting the default probability from the option-based model into the accounting-based model. As the measure of models' predicting performance, the Area Under the receiver operating characteristic Curve (AUC) is used. Empirical results show that the hybrid model (AUC: 0.8732) outperforms both the accounting-based model (AUC: 0.7519) and the option-based model (AUC: 0.8581). This result shows that accounting or stock market information alone is not sufficient to explain real-world behavior. It is suggested that the hybrid model be used as an alternative prediction model of construction contractor default.