Credit Variables Explained: Public Records

LendingClub borrowers are asked for their employment and income information during the loan application process.  It is seemingly obvious that this information would be valuable in making investment decisions, and our Does Income Matter post from earlier this week examines this in detail.   However, there is another aspect of borrower income that that bears examination: verified income.

According to this recent blog post, LendingClub verifies income for 60% of borrowers before the loans are issued.  There is a true/false flag available for investor use both in the “browse notes” filters as well as in the downloadable CSVs.  Let’s see if this flag tells us anything about risk and return on a historical population of LendingClub Notes:

This graph leads us to make a few curious observations:

  • The default rate is nearly the same for both groups, where one might have expected the “not verified” population to be riskier.
  • The benefit in net return for the “verified” group is entirely due to a higher interest rate.  This is something we might not have predicted before making this graph, so it makes sense to do some more research.

First, let’s take a look at why the interest rate might be higher for verified-income loans.  As we see in the table below, LendingClub does not make their verification decisions randomly.  It appears that they verify higher-risk borrowers at a much higher rate than lower risk borrowers.  68% of G-grade loans get verified vs. 23% of A-grades.  Higher-risk loans carry a greater interest rate, which explains why the verified population skews towards higher-interest.


We know what you might be thinking.  If LendingClub concentrates its verification efforts on the higher-risk, higher-interest loans, shouldn’t we expect verified-income loans to have a higher default rate?  This brings us to an important concept in modeling – variable interactions.  Put simply, a predictive variable (such as whether income is verified or not) may perform differently given different values of another variable. Let’s take a look at the verified flag when plotted against LC’s credit grade:


For loans in grades A-C, verification status does not appear to have a significant impact on default rate.  However, for loans in grades D-G, verified-income loans carry far lower risk than their non-verified counterparts.  For G-rated loans, the difference is an astonishing 633 basis points!  Put simply: verification status does not effectively separate risk when used as an across-the-board filter or model input, but it has a major impact when just applied to loans within the higher-interest credit grades.

Conclusion – Challenge of Deployment

As we can see from the data presented, verified income can serve as a predictive component of a loan-selection strategy, as long as it is applied in a nuanced way.  However, as described in the above-referenced blog post, lenders are having a hard time effectively utilizing this filter, given that the massive surge in investor demand is causing many loans to reach 100% funding before income verification can be completed.  We hope that this will eventually cease to be a constraint as LendingClub scales up operations to keep pace with its surging growth.