Monthly Debt and Disposable Income

Underwriting consumer loans can be a complicated exercise, in part because of how much data is available to assist in making the approve / decline decisions.  Most credit models include data from a potential borrower’s application, credit bureau and possibly other data sources if available.

From the application, a lender attempts to understand the intentions of the borrower and some personal information that is not available on the credit bureau.  This includes the use of proceeds and income of the borrower.  For Small Business lending, the application usually includes business metrics such as revenues, expenses, inventory levels and receivables.  The combination of application and credit bureau data gives insight into the previous behavior of a borrower and the current financial situation.  Assuming that the borrower has a decent amount of history on the credit bureau, the data is very powerful.  Some important factors that can be understood from this data are:

  1. Credit seeking: Has the borrower been searching for credit?  This can be assessed by counting the number of inquiries that the borrower has in recent months.
  2. Historical derogatory activity: Does the borrower have any credit facilities that are currently or recently past due?  Has the borrower had any major issues (e.g. tax liens, bankruptcies, foreclosures) recently?  These issues may arise with out an actual loan going past due.
  3. Ability to Pay: can a potential borrower take on more debt?  This is generally assessed by comparing the amount of income and the amount of current debt obligations.

Today’s blog post will focus on the existing debt of Prosper borrowers.  We will look at the trends over time, whether debt is correlated to loan performance, and ways to calculate ability to pay.

Since 2011, the average monthly debt of borrowers (at the time of application) has risen from ~$800 / month to ~$1,100 / month.

By credit grade, we see the increase has been more prominent on the AA, A and B credit grades.

The next obvious question is about performance.  Are borrowers with higher monthly debt more likely to default?  The performance of loans booked from 2011 and 2012 show that there is not much correlation.  The group with lowest monthly debt has the lowest default rate, but the difference in performance with other groups is not correlated.

From this graph, it is clear that in order to truly understand the implications of the debt that a borrower has, their income must also be taken into account.  Simply looking at the monthly debt obligations is not very informative.  This is because a potential borrower with $100,000 in annual income and $2000 / month in debt obligations is very different from someone with $25,000 in annual income.

In general, Prosper borrowers have annual income over $25,000, but there is a large range above that and it is widely distributed , as can be seen below.

The distribution of income by credit grade is as would be expected, the higher credit grades have higher income groups.

Now, considering that the objective is to assess a borrower’s ability to make additional debt payments given existing income and commitments, income and debt together is a good start.  Below is monthly income minus monthly debt, also called disposable income.  Not surprisingly, most Prosper borrowers have a decent amount of disposable income.

Disposable income by Credit Grade is distributed as expected given the independent distributions of income and monthly debt by Credit Grade.

The below chart shows that, as disposable income increases, the historical default rates go down significantly.

This is also the case within the different credit grades (the only exception being AA with very low disposable income, which includes a total of 20 loans).

Based on this analysis, it is clear that monthly debt on its own is not sufficient to determine the relative riskiness of a consumer’s current debt profile.  However, comparing the debt to the income with a metric such as disposable income, results in a powerful variable that predicts risk well.