Skip to main content

Table 1 2016 Unsecured consumer loans–stochastic frontier estimation. Including the volume of unsecured consumer lending. Best-practice (minimum) ratio of nonperforming consumer loans

From: Consumer lending efficiency: commercial banks versus a fintech lender

Parameter

Variable

Coefficient estimate

Pr( >|t|)

β1

Growth rate in consumer lending from 2013 to 2016i

− 0.000511

0.009747

β2

Consumer Loansi (100 billions)

− 0.036515

0.000000

β3

Consumer Loansi (100 billions)]2

− 0.060919

0.000000

β4

Contractual consumer loan ratei

0.059706

0.000000

β5

[Contractual consumer loan ratei]2

0.072303

0.057154

β6

[Contractual consumer loan ratei] × [GDP Growth Ratei]

 − 0.005834

0.167194

β7

[Contractual consumer loan ratei] × [Herfindahl Indexi]

− 0.079739

0.000929

β8

[Consumer Loansi (scaled)] × [Consumer Loan Ratei]

    1.794105

0.000000

β9

[Consumer Loansi (scaled)] × [GDP Growth Ratei]

    0.065616

0.000000

β10

[Consumer Loansi (scaled)] × [Herfindahl Indexi]

− 0.507393

0.000000

σμ = 1/θ

 

    0.026070

0.000000

σν

 

 0.000418

0.008610

  1. The data set includes LendingClub and 387 top-tier bank holding companies at the end of 2016 with plausible values of nonperforming unsecured consumer loans and total loans exceeding 10 percent of assets
  2. Skewness and D’Agostino skewness test of OLS residuals \((\sim N\left(\mathrm{0,1}\right) \,\,\mathrm{asymptotically})\): skewness \(=4.7992 \left(>0\right)\Rightarrow\) positively skewed test statistic \(=16.7911\) with p-value < 2.2E−16 \(\Rightarrow\) positively skewed with statistical significance. Histogram and density of OLS residuals in the relative frequency scale appear in Appendix 1
  3. Likelihood ratio test of \({H}_{0}: {\sigma }_{\mu }^{2}=0\) vs. \({H}_{1}: {\sigma }_{\mu }^{2}>0\) (asymptotically distributed as 50:50 mixture of \({\chi }_{(1)}^{2}\) and \({\chi }_{(0)}^{2}\)): test statistic \(=536.8902\) with p-value < 2.2E−16 \(\Rightarrow\) strongly reject linear regression model (i.e., absence of inefficiency) in favor of stochastic frontier model (i.e., presence of inefficiency)