US auto finance: Managing credit losses in a challenging environment

The US auto finance industry is showing signs of distress. Despite reaching an all-time peak of over $1.2 trillion in outstanding balances, there are multiple indicators of potential challenges. The industry is now sitting on its largest share of seriously delinquent loans ever–50 percent higher than at the peak of the recession in 2010,[1] with serious delinquencies representing 4.69 percent of the total outstanding in the first quarter of 2019. All of this is happening while underlying auto sales have been flat over the past few years, due to a variety of factors, including the exhaustion of post-recession pent-up demand, and rising gas prices.

No single reason fully explains the rising delinquencies in the auto lending space. At a macro level, the recent strong economy and low unemployment rate have supported the steady growth of US household consumer debt to over $13 trillion[2] –the highest ever recorded. At the same time, many household incomes are becoming less predictable: today, over one-third of the workforce relies on temporary work in the “gig economy” as a primary or secondary source of income.[3]

Looking specifically at the auto lending industry, several issues seem responsible. Firstly, until late 2018, auto lenders have underwritten about 20 percent of loans to subprime customers, with this percentage rising to roughly half for auto lending companies.[4] Secondly, auto lenders’ underwriting practices appear less advanced than those used by other originators. For example, their limited use of advanced predictive modelling and marginal reliance on external data sources and features engineering translate into limited power to predict a customer’s financial reliability.

Thirdly, auto lending has experienced increased levels of fraud and misrepresentation at origination since the recession (e.g., income inflation and misrepresentation of second jobs and associated income). According to Transunion, auto loans suspected of synthetic fraud have grown by approximately 5 percent year over year.

To complicate matters, antiquated collections practices and structural under-investment have increased the risk of these high delinquencies turning into actual losses. Over the past few years, the benign credit environment has led loss-management operations (especially collections departments) to prioritize greater efficiency over effectiveness (for example, cutting costs to reduce operating costs) without corresponding reinvestments in effectiveness.

Further, borrowers are less responsive to traditional contact methods like phone calls and letters. For example, a recent McKinsey study demonstrated that lenders contact early delinquent customers primarily through traditional contact channels, even though consumers respond better to emails, texts, online banking prompts, and mobile pushes.[5] This issue is becoming even more relevant after the most recent CFPB proposal to set a limit on phone contacts of delinquent customers.

Another potential factor to consider regarding delinquencies is the rise of ride-sharing services. In previous recessions, delinquent customers who used their vehicles for work had a strong incentive to retain them; a percentage of these customers may now turn to ride-sharing as an alternative.

The net result is that delinquencies in auto lending are rising and collections operations are increasingly ill-equipped to deal with this rise (Exhibit).

Implications for auto lenders

To remain competitive by maintaining profitability while not increasing the cost of lending, auto lenders can learn from the journeys being taken by other originators. These include redesigning collections and underwriting operating models to rebalance investments towards areas with the highest return on investments, making more effective contacts through digital channel orchestration and strategy, and improving underwriting practices using more extensive data and better prediction techniques.

Building next-generation collections and underwriting operations

In an environment of constrained capacity and investments, retooling operations can free up capacity from processes that companies can eliminate, automate, and simplify - or move to a more efficient party.

Auto lenders have historically relied upon a high amount of manual work across the whole loss management cycle. This can range from very manual (and often opinion-based) underwriting processes to the document gathering and validation needed to begin legal or repo processes. These tasks require capacity that companies could redeploy to activities that add more value, such as developing a task force to target first payment default accounts or building predictive models to identify the best way to contact customer microsegments.

Beyond pure capacity redeployment, auto lenders have an opportunity to increase the effectiveness of collections by introducing an auto-specific Value-at-Risk (VaR) segmentation that compliments an early and regular VaR approach with behavioral, auto-specific segments, such as separate queues and strategies for accounts with a high likelihood to skip with no profitable repo option. Best-in-class auto lenders are also developing account-level treatment strategies and best-time-to-repo models to maximize reinstatements and take advantage of the enhanced ability to track vehicle usage and position.

Financial institutions that have gone through similar journeys have seen increases in straight-through-processing rates of 30 to 40 percentage points and reductions of cost/income ratio of 15 to 20 percentage points (before reinvestments). In addition, we have seen a focus on collections, repossession, and recoveries lead to reductions in net charge-offs of 10 to 15 percent.

Developing effective omnichannel orchestration

Effective omnichannel orchestration involves coordinating all of the digital and physical ways to reach delinquent customers. It enables each customer to engage with the lender in ways they prefer, yielding higher response rates. The most advanced lenders have set up agile test-and-learn environments to optimize their return on digital investments on a reoccurring basis.

We have seen auto lenders increase payments from delinquent customers by 10 to 20 percent, by contacting those customers through their preferred channels.

Moving beyond basic variables and logic in loan underwriting

Improving credit performance begins with better underwriting–moving away from simple credit box models to more complex machine-learning tools that take advantage of a variety of data available to the captive lender. Captives can take advantage of the richness of the information available on the OEM–creating a strong data partnership that benefits both the sales and the captive side. The use of external data (such as employment history and financial product ownership with other lenders) has also increased thanks to the emergence of data aggregators.

With better data, auto lenders are beginning to combine analytics with triggers from behavioral/live events, to build personalized actions (e.g., account-level best-time-to-repossess indicators, or individualized loan modification structuring).

By using more sophisticated models and adding new data sources, lenders can increase the predictive power of models by 10 to 20 GINI points, vastly increasing the success rate of identifying and engaging with the delinquent customers that are hardest to reach.

Given the challenges of the current lending environment, US auto lenders must take a fresh look at their credit management strategy and operations.

Running really transformative cross-functional operational and digital efforts require significant coordination and executive commitment. While no one knows exactly when the next recession will arrive, the time to start building renewed and more effective operations to ensure resilience against a potential economic downturn is now.

The authors would like to thank Amit Garg, Vijay D’Silva, Akshay Kapoor, Jon Steitz, and Diana Goldshtein for their contributions to this article.

[1] Quarterly report on household debt and credit 2019 Q1, Federal Reserve Bank of New York

[2] Quarterly report on household debt and credit 2019 Q1, Federal Reserve Bank of New York

[3] Forbes

[4] Quarterly report on household debt and credit 2019 Q1, Federal Reserve Bank of New York

[5] McKinsey survey of delinquent borrowers at North American Financial institutions.