Determinants of Default Risk in Online Peer-to-Peer Lending – Evidence from Finland

Abstract

Purpose of the study

The development and widespread adoption of information technology has led to the emergence of electronic marketplaces, where the traditional intermediaries are being bypassed by an online platform directly connecting the market participants. Within the financial services industry, the consumer lending business of traditional intermediaries such as banks is being disrupted by online peer-to-peer lending platforms, that allow the borrowers to borrow money from the depositors directly without any intermediary. Since its global inception in 2005, and especially after the credit crunch of 2008, peer-to-peer lending has gained significant momentum in various markets, with some US platforms reaching both multi-billion-dollar loan circulation volumes and market capitalizations. The development of peer-to-peer lending lags far behind in Europe, but exponential growth is expected to follow there and across the globe. However, the sustainability and continuing widespread adoption of peer-to-peer lending is highly dependent on the availability of reliable means to assess the default risk of the borrowers. For this purpose, I study the factors that determine the default risk of peer-to-peer loans using real loan data and provide the first empirical observations of the Finnish peer-to-peer lending market.

Data and methodology

I obtain a novel data set of successfully funded loans from the leading peer-to-peer lending platform in Finland. The final sample consist of 6 936 funded loans that had closed as either repaid or defaulted at the time of data extraction. Various loan-, borrower- and application specific characteristics are included in the data that I use to analyze the borrower default behavior. After the descriptive analysis of the data, I use logistic regression to study the factors affecting the default probability. I construct a model using only ex-ante information and test the predictive power of the model in classifying between ex-post good and bad loans. In addition, by means of survival analysis, I employ Cox Proportional Hazards model to examine the different factors that predict not only if, but also when, will the loans default.

Key findings

I find significant evidence of various factors affecting the default risk of peer-to-peer loans. Among these factors are various hard/traditional loan- and borrower characteristics. In addition, I find interesting and new evidence of the importance of soft/non-traditional information in form of orthographic characteristics found in the loan applications. More specifically, I find that th lack of capitalization (i.e. use of lower-case only even when a word should start with a capital letter) in the submitted text content of the loan application is associated with significantly higher risk of subsequent default by the borrower.