Identifying P2P Lending Frauds Based on Ownership Structure

摘要

The once-booming but scandal-ridden peer-to-peer lending industry robbed hundreds of thousands of retail investors of their life savings in the past several years. This paper examines the possibility and feasibility of identifying P2P lending firm non-normal business operation and the exits with the nature of frauds by using company's ownership structure characteristics. This paper adopts the penetrated shareholder compositions of P2P lending firms to construct their shareholding Ego-networks, and proposes an Ego-network clustering approach to describe and compare ownership structures of P2Ps. We find that there are two kinds of groups: the majority is that with a small amount of shareholders spread in shallow equity levels, while the minority is that with more shareholders, deeper equity structures and more complex cross-holding relationships. This paper mainly studies whether equity structural characteristics can be used to detect the frauds of P2P firms. Hence, classification algorithms are used, and logistic regression models with different P2P shareholding Ego-network characteristics are also employed to complete and explain the identification results. We confirm that firm's ownership structure can provide useful information for fraud detection. For example, P2P lending companies which are held by fewer direct and indirect shareholders, have fewer connections among related equity systems, and are held by shareholders with lower centralities in ownership Ego-networks are more likely to non-normally exit because of fraud risk events.