Take care of your business at home, before searching for greener pastures. It’s a key takeaway from Part One of our interview regarding subprime auto: lenders and servicers with modernized collection and recovery systems have a solid foundation for profitability that should be a requisite before embarking on business development plans.
In Part Two of our discussion with David Albers, CEO of Revenue Connections, we asked David to share his advice on ways to positively impact collection efforts for student loans, where, unlike auto, there is no collateral to support recoveries.
Industry Professional’s Advice to Improve Collections for Better Portfolio Performance
In the student loan space, given there is no collateral to mitigate a lender’s loss in case of default, determining borrower attributes, their weightings and correlations and how they affect the probability of repayment are all paramount to a lender’s success. Which borrower attributes are meaningfully and consistently predictive of the risk of non-payment at the time of origination and during the life of the loan?
In my view, the answer lies in creditworthiness (typically indicated by a credit score) and capacity (usually determined by a borrower’s income versus expenses) but not in the way we have historically come to view these two attributes.
Let’s start with creditworthiness. What is the predictive value of a borrower’s credit score in student lending?
Credit scores from one of the big three bureaus is what most lenders, issuers and servicers think of when they talk about creditworthiness. In student lending, borrowers typically have a low credit score due to a lack of borrowing experience which, by definition, indicates that they are more likely to enter collections at some point in the life of the loan. The bureaus do a decent job of highlighting this in their scores, but there are other attributes that may be more predictive of potential collection issues. As such, these traditional credit scores should comprise only a portion of a student lender’s underwriting matrix and eventual servicing or collections scoring and segmentation strategy.
There are a growing number of scores being developed based on “alternative data” sources that do a better job of highlighting collections risk than traditional credit scores. More and more shops are also developing their own internal behavioral scores for segmentation and collections treatment once the loan is booked. I believe that a combined approach of using traditional credit scores as a “baseline” and measuring improvements in predictability and performance driven by the introduction and testing of various sources of data as one builds toward a proprietary internal score is the optimal approach and, with IT advancements, is more feasible than in the past.
How should capacity be evaluated given the unique position of the student borrower?
In student lending, this is a difficult attribute to determine weighting because students historically have insufficient capacity to repay a loan. To the extent a lender can enlist a co-signer this risk can be greatly diminished if not completely extinguished. If not (which is in most cases) the lender has to rely on future expected capacity. That is why it’s an easy decision to make loans to post grads such as doctors, attorneys, etc., but a much more difficult decision to make loans to undergrads or students seeking a trade degree. Some lenders actually limit their lending to certain levels of education and disciplines due to these factors. In banking, some might call this a form of redlining or discriminatory practices, but in private loans and fintech, this is considered smart lending.
The bottom line is that because this income represents the primary source of repayment for a student loan, each lender has to decide how to define capacity and determine how much weight they are willing to give it in their underwriting matrix and in their scoring and segmentation strategy. It’s a tough call.
Given the potentially limited value of traditional attributes such as credit score and capacity, what else should be considered?
The takeaway for lenders or issuers is that they need to clearly define their target customer and risk profiles to determine the attributes that are most important to them and establish a disciplined pilot-and-learn methodology to testing new attributes or weightings not only at the time of origination, but as a regular part of their servicing and collections process. Additional underwriting criteria, scores and attributes and their weightings are for each lender to determine based on their predilection for risk and their sophistication in developing internal scores based on that criteria.
Loans to medical students perform very differently than loans to students working toward their associates or undergraduate degree for obvious economic reasons. What has proven even more predictive of credit quality than the type of degree is whether the borrower graduated or not. There are newer lending companies making the case that schools have different risk profiles that impact credit quality and that there are other attributes that are more important than credit scores and capacity. These are just a few of the many conversations going on regarding student loans as an asset class and what is most predictive of credit quality.
Summing it up, what are some best practices to implement?
Lenders should fully understand the attributes that have contributed to the risk profile their underwriting and servicing practices have created and use this as a baseline in a pilot-and-learn environment. More specifically, create a platform where the lender uses advanced quantitative methods to identify various attributes which appear to have predictive qualities in originations and in servicing and then tests them as part of a well-controlled and disciplined champion-challenger program. If they work, expand their use methodically and gradually until the model is fully validated. Wash, rinse, repeat.
As David points out, “student lending will continue to evolve from an analytical standpoint more than other, more established asset classes.” With this prediction, it seems sure to pass that lenders and servicers who focus their efforts on understanding the results of their underwriting and take an expanded view of borrower attributes will see real benefits relative to those who don’t.