We explore the quality of risk assessment for entrepreneurs/small business borrowers as compared to consumers, when the same information on previous credit history is used for both segments in marketplace lending. By building several cross-sectional logistic regression and machine-learning models and applying them separately to small business loans (SBL) and consumers we can measure models’ predictive accuracy for different segments, and thus, make observations about the value of the information used for screening. We find the differences in profiles between SBL and consumers, hence they should be assessed by separate models. Yet separate SBL models do not perform well when applied to a future time period. We attribute this to the relatively low predictive value of personal credit history for entrepreneurs as compared to the consumers. We advocate the use of additional information for risk assessment of entrepreneurs, in order to improve the quality of credit screening. This should lead to improved access of small business borrowers to credit in situations when they have to compete with consumers for funding.
|Number of pages||25|
|Journal||Journal of Financial Management, Markets and Institutions|
|Early online date||3 Jul 2021|
|Publication status||E-pub ahead of print - 3 Jul 2021|
- small business finance
- marketplace lending
- risk of default