Default or profit scoring credit systems? Evidence from European and US peer-to-peer lending markets

Financial Innovation

Table 2 Out-of-sample performance of credit and profit scoring models—Bondora

Model	% invested loans	Performance across invested loans		Performance across all loans		Total profit in mil. EUR
		Average return	SD	Average return	SD
Panel A: Credit scoring models
LR	59.8	19.57	33.06	11.70	27.30	0.53
\(LR^{{\lambda_{\min } ,\alpha = 1}}\)	59.9	19.72	32.98	11.82	27.30	0.54
\(LR^{{\lambda_{\min } ,\alpha = 0}}\)	59.3	19.69	32.96	11.69	27.17	0.53
\(LR^{{\lambda_{\min } ,\alpha_{\min } }}\)	59.7	19.58	33.17	11.70	27.37	0.53
RFC	75.8	19.49	38.48	14.78	34.53^†	0.69
NNC	60.6	17.61	36.85	10.66	29.93	0.47
Panel B: Profit scoring models
LM	75.7	19.34	39.76	14.65	35.58	0.66
\(LM^{{\lambda_{\min } ,\alpha = 1}}\)	76.3	19.66	39.54	15.00	35.53^†	0.70
\(LM^{{\lambda_{\min } ,\alpha = 0}}\)	76.2	19.69	39.29	15.02	35.32^†	0.70
\(LM^{{\lambda_{\min } ,\alpha_{\min } }}\)	76.2	19.67	39.55	15.00	35.53^†	0.70
RFR	75.4	20.55	38.80	15.49	34.83^†	0.72
NNR	71.8	20.35	38.82	14.62	34.15^†	0.68

\(LR^{{\lambda_{\min } ,\alpha = 1}}\), \(LR^{{\lambda_{\min } ,\alpha = 0}}\), \(LR^{{\lambda_{\min } ,\alpha_{\min } }}\) are lasso, ridge and elastic net versions of logistic regression, \(LM^{{\lambda_{\min } ,\alpha = 1}}\), \(LM^{{\lambda_{\min } ,\alpha = 0}}\), \(LM^{{\lambda_{\min } ,\alpha_{\min } }}\) are lasso, ridge and elastic net versions of linear regression
SD standard deviation, LR logistic regression-based models, LM linear regression-based models, RFC random forest classification, NNC neural network classification, RRR denotes random forest regression, NNR neural network regression
^†Denotes a model that belongs to the set of superior models