Marketplace Lending

Winning @ Lending Club

By including information concerning the economic vitality of a borrower’s locality as a decisioning criteria, investors can improve the performance of their Lending Club loan portfolio.  Overlaying local economic vitality data provides additional insights about borrower creditworthiness that is not captured by traditional credit scoring models.  Our framework blends this additional information into an expected loss methodology to create a loan selection scheme that provides transparent choice between portfolio returns and the underlying asset quality.

We analyzed all loans originated by Lending Club in 2013—over 130 thousand loans worth nearly $2 billion.  These loans are seasoned and recent enough so that the impact of evolving platform characteristics is minimal.  Based on the borrower’s geographic information, we appended to each loan information derived from a sub-index of the WAIN Street Business Default Index concerning sole proprietors (BDX-Solo).  We used the sub-index values from the month prior to the origination month.  Previous analysis has demonstrated the association between the BDX-Solo and consumers.  We used two related local economic vitality metrics—forecasted business default rate and default gradient.  Previous analysis has shown these two metrics as valuable segmentation criteria.

Using the two economic vitality metrics and the grade assigned to each loan by Lending Club, we identified loan pools with similar historic returns and expected losses.  Expected losses are an assessment of credit risk.   Returns are the rewards to the investor for taking that risk.  With additional performance analysis, we distilled loan selection criteria that balance returns and the asset quality generating those returns.  Finally, we tested these selection criteria against loans originated each month between January 2014 and June 2014 by appending local economic vitality metrics based on the WAIN Street data from the month prior to the origination month.  The results are reported below.

Lending Club Performance Improvement

Each selection criteria represents a set of rules that identify a collection of loans characterized by the Lending Club assigned grade and WAIN Street assigned forecasted default rate and default gradient for the borrower’s locality.  The “Aggressive” criteria increase portfolio returns at the cost of selecting riskier loans.  The “Conservative” criteria sacrifice returns for better asset quality.

To facilitate an objective assessment of the riskiness of selection criteria, we not only report the improvement in returns, as is commonly done, but also the improvement in charge-offs.  Higher charge-offs correspond to poorer underlying asset quality.  And in that spirit, two portfolios with identical returns can be compared based on their respective charge-off rates.  We feel such an approach provides a clearer picture of the asset quality generating the returns.  It is also more transparent than only reviewing “alpha”.  Ex. The “Aggressive I” criteria improves returns by 4.5%– a nearly 50% improvement in returns.  However, by examining the charge-offs for the same portfolio, one can clearly gauge that the excess return is on the back of lower quality assets.  Similarly, the “Conservative II” criteria sacrifices 3.6% in returns but the 4% improvement in charge-offs clearly shows that the selected loans are of much better quality.

Things to note

  • The returns reported above are annualized without compounding and do not account for the roughly 1% in fees that investors are charged.
  • The returns include an adjustment for potential future losses based on the current under-performing status of the loan. Ex. Loans “In Grace Period” are assumed to have lost 28% of the principal.  This “charges” portfolio returns for deterioration in asset quality.
  • The selected loans are likely more diversified than a pool created by using criteria already incorporated into the platform’s credit scoring model.
  • The results clearly show what practitioners already know – there is no free lunch.  If you seek higher returns, you must accept lower asset quality.  True, loans might be mispriced.  But platforms are continuously improving their underwriting.  A more viable approach to improving performance is to use new information and make an efficient choice between portfolio returns and asset quality.

Going forward we will update the loan selection criteria with each monthly refresh of the WAIN Street data and quarterly as Lending Club releases its new origination data.  The strong results described here notwithstanding, the model is basic and allows for further enhancements such as segmenting by the loan’s term, purpose, and perhaps, reconsidering credit factors that are already incorporated into the assigned grade but might interact differently with locality.

Using the selection criteria
For loan investors, this framework provides a starting point that can be tuned to their unique needs. The rules described here are largely coarse-grained as reflected in the percent of loans that are selected and finer grained rules can be identified.  We can append the local economic vitality metrics to loans as a service and also license the metrics for model development.

For platforms, this framework offers an additional means for benchmarking and tuning underwriting.  We can append local economic vitality metrics for benchmarking and also license the data for incorporating into a credit scoring model.

For individual investors, we are looking into collaboration for developing an online service.