Predicting Enterprise Performance by Network-Based Risk Profiling Approach

Abstract

Deriving the firms’ risk profile based on specific features has important implications in risk controlling and investment. Recently, much research demonstrates that the firms’ ownership networks substantially impact the firms’ risk profile. In this paper, we propose a framework of risk profiling approach built upon information retrieved from the firm's ownership networks. The method considers the non-linear relationships between firm fundamentals with network structures. To test the performance of the proposed method, we construct a new dataset of Chinese listed firms with their financials and network parameters in the period between 2005 and 2020. We show that the proposed method significantly outperforms traditional ones in predicting a firm's market value changes. Specifically, we first use the conventional linear method, like logistic regression and linear discriminant analysis, as the performance benchmark. Then, the more advanced technique based on information theory like Gradient Boosting is adopted and has shown remarkable performance with at least 85% area under the curve (AUC) compared with the 60% AUC of the traditional linear model. The proposed method has implications in risk management, portfolio management, and corporate finance. As a special implication example in risk management, we demonstrate that a network-based approach can effectively detect duplication of individual names in a unique dataset.