Lenders, be prepared. Headwinds may be ahead. After enjoying a strong consumer credit market for a decade following the Great Recession, the lending industry faces a growing set of challenges. With a much anticipated credit downturn looming on the horizon, it’s time for lenders to sharpen their pencils.
Consider the changing marketplace. With a growing U.S. economy, a 50-year low in unemployment, and interest rates remaining near historic lows, Americans are carrying more consumer debt than ever—with particular growth in riskier areas such as credit card balances and unsecured personal loans.
Total non-mortgage household debt exceeded four trillion dollars at the end of 2018. While new originations have been healthy, there are signs of an uptick in delinquencies, for instance in credit card and auto loans.
The competitive landscape is changing, too. Nontraditional players, including a new class of fintech platforms, are gaining strong momentum and taking market share away from banks. According to TransUnion, in 2018 fintech lenders originated 38% of all personal loans, up from just 5% in 2013.
The rapid rise of fintechs comes in part from their speed and flexibility. They tend to offer more favorable terms to near-prime and subprime borrowers, with quicker approvals and funding of loans. Millennials are an obvious target for these lenders. This new competition has put more pressure on traditional lenders’ bottom lines.
In this environment, lenders need to continue to innovate and adapt to changing consumer behaviors to remain competitive and grow their portfolios, while maintaining a laser sharp focus on credit risks. As always, the fundamental task is to target well-qualified prospects from a competitive universe and match them with the right products, safely and profitably.
That means lenders need to understand and accommodate changes in credit use patterns and attitudes when designing products. They need to leverage the best available data and consumer insights in underwriting, pricing, and portfolio management strategies. And in doing so, they must fully satisfy risk and compliance expectations.
Having the right analytical tools, including an effective credit scoring model that can accurately assess risk for a broad population of consumers, is as critical as ever.
The VantageScore 4.0 credit scoring model offers significant performance benefits over other credit scoring models by utilizing better data, including trended credit data, and by taking advantage of more advanced modeling techniques. The model also allows lenders the ability to assess a significantly broader population of credit-eligible consumers, creating opportunities for borrowers and lenders alike.
An ever-growing number of lending institutions offer VantageScore credit scores to consumers.
VantageScore 4.0 offers distinct advantages to lenders in addressing the challenges and opportunities the current marketplace presents. First, the model can accurately assess previously “unscoreable” consumers, making the credit markets more accessible to creditworthy consumers while creating a corresponding opportunity for lenders. Secondly, the model provides superior predictive performance across all credit segments, driven by deeper insights on credit behavior.
According to the latest census, there are 252 million people over the age of 18 in the U.S. Conventional credit scoring models can score only 201 million of them, leaving 51 million deemed “unscoreable” because they do not meet the traditional criteria of having an update on their credit file in the past six months and having an account that is at least six months old.
When using VantageScore 4.0, however, 241 million consumers are now scoreable—including 40 million who could not be scored by other conventional models.
Who are these previously “unscoreables” hiding in plain sight? Which consumers aren’t getting scored by conventional credit scoring models?
Many are young to credit, with accounts less than six months old: young adults starting their careers or newly arrived immigrants. Then there are dormant customers, with no credit activity in the past six months, but previous credit updates; and the no-trade segment, who’ve lost access to credit and have external collections, public records, and inquiries on their credit files.
The newly scoreable represent a full 16% of the adult population in the U.S. overall, and as much as 20% in the South.
Conventionally unscoreable consumers represent a nationwide opportunity.
The opportunity is particularly significant in two key demographic groups:
While not all of the newly scoreable consumers will be ready to take on the responsibilities of managing credit, many are well qualified for certain loan products. In fact, about 10 million of the newly scorable consumers, or a quarter of them, will have credit scores greater than 620.
Leveraging modern modeling techniques, such as machine learning, VantageScore 4.0 can effectively separate consumers with a clear track record of unfavorable credit behaviors from those who are simply starting to develop, or re-develop, credit histories. The result is a model that accurately scores more people, opening up pockets of opportunity for consumers and lenders alike.
According to Pew Research, Millennials will soon become the largest segment of the adult U.S. population, with their numbers expected to reach 73 million this year. They use credit far less than prior generations; fewer than 60 percent use credit regularly, and 21 percent have “thin” files, meaning they have fewer than three credit accounts, even though they have similar or greater income and assets as compared with their “thick file” contemporaries.
Remember, this generation now shoulders 46% of all $1.5 trillion of U.S. student debt outstanding. No wonder they’re shy about taking on more obligations. That’s also why Millennials have a disproportionately low share of America’s credit card (14%) and personal installment loan (12%) balances.
As a result, conventional models which tend to look at depth, breadth, and tenure of credit put millennials at a disadvantage.
VantageScore 4.0 reduces the emphasis on these conventional factors while bringing in other, more predictive attributes related to trends in the consumer’s credit management activity. These refinements make it possible to deliver highly accurate assessments of risk.
VantageScore 4.0 brings two cutting-edge innovations to the table: Trended credit data, and attributes driven by machine learning techniques.
Among the many new data sources becoming available, perhaps none holds greater promise than trended credit data. Trended credit data captures the trajectory of borrower behaviors over a period of time, as opposed to the typical snapshot at a point in time, revealing additional insights about the consumer’s credit behavior.
As an illustration, consider two consumers, Connor and Jim:
Who’s the bigger default risk?
The conventional “snapshot” view shows that their incomes are similar. But Jim has a thicker credit file with three times as many accounts as Connor, and a longer credit history. Connor’s bankcard utilization rate is 27%, compared to Jim’s 46%.
You might conclude that Jim, who’s never missed a loan payment, is the better risk.
However, a review of the trends in credit behaviors over time tells a different story. Connor has been focused on paying down debts, including his student loans, and lowering his utilization of credit, while Jim has been ramping up his balances and utilization. In fact, the data suggests that Jim may have higher volatility in his need for credit, with utilization reaching 75% at one point within the last year.
As it turns out, Jim is more likely to default over the next 24 months. But without trended credit data, you might never know.
With VantageScore 4.0, the trended data related to a consumer’s credit management activities become a more significant contributor to the credit score, driving superior predictive performance.
While there seems to be a lot of attention on machine learning recently, the truth is that machine learning (ML) has been in use in the financial services industry for many years.
Recent advances in computing power and ML algorithms provide significant efficiencies in data science, making it possible to look for deeper insights in data by exploring multi-dimensional relationships across many data elements. Rather than simply looking at individual attributes such as available credit, age of collections, and so on, the model can identify relationships between disparate pieces of data that strongly correlate with a borrower’s likelihood of default.
Identifying those meaningful relationships is only a first step. Building credit scoring models is as much as art as science; it requires significant domain knowledge. In building VantageScore 4.0, the company’s data scientists applied their expert judgment to convert those newly discovered data relationships into attributes that can be understood and interpreted. This development process ensures that the resulting model is intuitive, explainable, compliant with all regulatory requirements, and generates the necessary adverse action logic.
The insights provided by machine learning algorithms have particular value in cases where the available data on a borrower is limited, such as consumers with sparse credit files, or whose recent credit activity has gone dormant.
Innovation is one thing, performance as well as safety and soundness is critical. A common way to assess the value provided by a more predictive credit scoring model is through what’s called a “swap set analysis.” For a given risk level cutoff, the swap set analysis compares the Accept/Reject decisions provided by two competing credit scoring models.
Such an analysis was performed for VantageScore 4.0 versus another commercially available generic credit scoring model, to demonstrate how the VantageScore model’s superior ability in order to assess risks translates to better decisions in credit card acquisitions.
For a similar risk cutoff set at 1.5% cumulative defaults, VantageScore 4.0 allows an incremental 5% of the prospects to be accepted. A comparison of the swap sets highlights that the population “swapped in” by VantageScore 4.0—those borrowers accepted by VantageScore 4.0 who would have been rejected by the other model—had 45% lower risk compared to the population “swapped out” (i.e., rejected by VantageScore 4.0 but accepted by the other model), based on actual performance in the subsequent 24-month period.
Further analysis of the credit attributes of the swap-in and swap-out populations reveals a significantly more favorable pattern in utilization and balances for swap-ins despite a slightly shorter credit history and more blemishes in payments historically. The swap-outs exhibit the opposite behaviors, with balances and utilizations ramping up to nearly twice those of the swap-in consumers.
VantageScore Solutions is the only credit score model developer to publish detailed results of its annual model assessments, along with whitepapers, user guides, and other communications aimed at model users and consumers.
VantageScore firmly believes in the value of transparency and the importance of knowledge sharing to enable model users to gain comfort with how the model works and performs and to complete model governance and other approval processes.
In light of the changing consumer credit marketplace and the fierce competition among lenders, this level of transparency is vital. It enables the users of credit scores to find a scoring model that offers the greatest performance and leads to the most successful business decisions. Such a model has the potential to open up new lending opportunities among previously unscoreable populations. It also allows lenders to take smarter risks across an entire portfolio.