Lending in Turbulent Times

Dear Colleague,

In the midst of the last Recession, dubbed, “The Great Recession,” lenders of all shapes and sizes “tightened” their credit standards. I recall much confusion about this term and why it was happening. And we spent many hours with regulators, consumer advocates and others illustrating what tight credit is and why it was occurring.

In this newsletter we’ve included a relevant and timely synopsis of a white paper illustrating the relationship between a credit score and the risk it represents. This relationship is “dynamic,” in that risk associated with credit scores change over time. The paper discusses why it’s doubly important for lenders to validate and test their models during volatile times such as these.

The way this relates to credit “tightening” and “loosening” is this: depending on what level of risk a credit score represents, lenders will adjust things like their “credit score cut-off.” This refers to the lowest credit score required for a product or particular interest rate or loan terms.

Let’s take, for example a credit score cut-off of 700. In a good economic climate, this credit score represents a fairly low likelihood of default. But, in a highly volatile economic climate with high levels of unemployment, a credit score of 700 can represent a much higher likelihood of default.

A lender usually targets an acceptable percentage of new accounts that are more likely to default, and for which they may take a loss. The lender must keep this percentage steady or risk its own financial stability. So, in order to keep the percentage or number of accounts that are expected to default steady, the lender must move its credit score cut-off higher in order to achieve the same risk level. Thus, the lender will tighten credit availability and there would potentially be fewer loan approvals but a more stable likelihood that those loans may go into default status. On the other hand, as risk is driven out of the economy, say by greater employment, the cycle reverses because as risk associated with a score diminishes, lenders will lower their score cut-offs to maintain a level risk of default.

In the current environment and in the foreseeable future, lenders will continue to monitor their portfolio of loans and move their minimum score cut-offs higher in order to maintain safe and sound credit policies. It is a balancing act for sure because lenders want to lend. It is how they remain in business!

Another area we might see “credit tightening” is in credit card limits. Lenders will also monitor their already issued portfolio of credit cards by pulling the holders’ credit scores, usually monthly. This process is known as portfolio management, which constitutes a FCRA- (Fair Credit Reporting Act) permissioned purpose for pulling a credit score. This is considered a “soft pull” and doesn’t harm your credit score.

When they do these soft pulls, they may see increased risk exposure because of a degradation of credit quality of the overall portfolio. In order to mitigate their exposure to sizable losses, credit card issuers may reduce the maximum limits for some card holders in order to limit the lender’s potential losses should those consumers not be able to pay back the amount owed. I would expect these credit limit reductions to occur over the next few months.

For card holders, a reduction in one’s credit limit can cause their credit scores to decline because it reduces the amount of credit a person has at her/his disposal, thus causing their balance-to-credit limit or “revolving utilization” ratios to increase; unless, of course, cardholders are paying down their balances in excess of the minimum payment due and using less available credit.

And I get it. You’re thinking: “Wait, my score went down and now it goes down again because my credit card issuers reduced my limits?”

There are two factors at play here. First, by reducing the amount a consumer is allowed to charge, this prevents that consumer from going into a potential debt spiral. Secondly, as economic conditions improve, consumers can ask for credit card limits to return to their previous levels, or even higher. Indeed, we did see credit loosen as the Great Recession got further and further in the rearview mirror as risk was driven out of the system.

Please read on for the aforementioned whitepaper synopsis and other educational content. And please stay safe and healthy.



CEO and President, VantageScore Solutions

Did You Know:
How to Correctly Interpret Credit Risk in an Economic Downturn

Examination uncovers how credit scores can be used as continued economic dislocation looms

VantageScore Solutions, LLC, developer of the VantageScore credit score model, released today a whitepaper on the VantageScore credit score’s role in understanding consumer credit risk and the need to closely monitor the risk represented by a given credit score in an increasingly volatile economic environment.

VantageScore’s latest whitepaper “The Dynamic Relationship Between a Credit Score and Risk: How to Correctly Interpret a Credit Score During an Economic Downturn”, provides lenders and other users of credit scores transparent details about how the score-to-default risk relationship changes over time, and notes the importance of timely and active credit risk management in order to make proper portfolio adjustments and credit score cut-off recalibrations in response to shifts in the economy.

Some insights from the study are below:

  • Default levels fluctuate over time because they reflect macroeconomic conditions, such as the unemployment rate, interest rates and home values, all of which can either increase or decrease risk levels.
  • Data from the 2008-2010 financial crisis provides a useful indication for how risk can shift, where default rates for new originations as well as exiting accounts were 200-250% higher for a given credit score band when compared with a more stable economic timeframe (2017).
  • Close monitoring of risk results and building in forward-looking expectations and scenarios around how those results may change, are necessary in setting (or calibrating) the score cut-offs.
  • Given the unprecedented impact of the COVID-19 pandemic and the high degree of uncertainty about the shape of the economic recovery and longer-term consumer impacts, the need for closely monitoring and managing risks is even greater.

“Credit scores are often improperly thought of as absolute predictors of whether a borrower will default on a loan,” said Barrett Burns, president and CEO of VantageScore Solutions. “In reality, a credit score is a representation of risk accomplished by rank ordering the scoreable consumer population based on who is least and most likely to default. Inherently, as the overall economy slows, the risk associated with scores will shift in accordance. By understanding this relationship, lenders can make better, safe and sound decisions while protecting consumers from becoming overleveraged.”

For more details from the “The Dynamic Relationship Between a Credit Score and Risk – How to Correctly Interpret a Credit Score During an Economic Downturn,” study, visit:

VantageScore Podcast w/ the “Personal Profitability” Blog

The latest podcast from the VantageScore Podcast series has posted. This month, we visited with Personal Profitability founder and popular finance blogger Eric Rosenberg. 

Rosenberg 4 has written for TD Ameritrade, SoFi, Citizens Bank, Business Insider, Investopedia, Huffington Post, Credit Karma, Forbes, MSN, Nasdaq, Mint by Intuit, and many other outlets. Before writing about finance and investing, Rosenberg was a former bank manager and corporate finance and accounting professional for 10 years.

In this episode, Eric shares his journey on how he became one of the most in-demand financial bloggers. Take advantage of this down time in quarantine to “spring clean” your finances. Learn how to organize your financial goals, budget, save an emergency fund, tips to spring clean your finances, reduce debt, monitor credit, and how COVID-19 bankruptcy can impact credit.




5 Questions with AICurio, Inc

Lester Firstenberger is CEO of AiCurio, Inc which has developed the first deep learning, artificial neural net that predicts at unprecedented  accuracies all cashflows, defaults, and prepayments for residential mortgage whole loans, MSRs.  Lester is recognized nationally as a regulatory attorney and expert in consumer finance, securitization, mortgage, and banking law.  In a variety of capacities over the past 30 years, Lester has represented the interests of numerous financial institutions in securitization and other transactions valued in excess of $1Trillion USD.

He was appointed to and served a three-year term as a member of the Congressionally created Consumer Advisory Council of the Board of Governors of the Federal Reserve System in Washington, DC. He has extensive governmental relations experience in the US and Canada at the federal, state and provincial levels. Lester has co-founded several FinTechs that relate to mortgage banking, digital identity and data privacy. 

1) First off, when was AiCurio launched and what business needs are you serving?

So that’s an easy one. We launched AiCurio in December 2018. That’s when we put a stake in the ground to become the first commercially available machine learning model for residential mortgage loan performance and cashflows.  Currently our Model is only for residential mortgage loans.  Building a deep learning artificial neural net model that is good enough to take to market is no small feat and takes some time, but, frankly, we were a bit stunned by how quickly we were able to come up with our predictive Model.  After 11 months of work, AiCurio’s Model was trained and validated to a 96.5% accuracy on a confusion matrix, which is a standard academic evaluation method for model predictions of all True Positives, but just as importantly, all True Negatives. 

As for the business needs, well, where to start?  In a nutshell, AiCurio gathers, normalizes, and processes data, and then we convert the data to usable and actionable information so that our clients can make better profitable decisions.  Specifically, at the individual loan level, we predict the likelihood of the occurrence of each monthly transition and of the cash flows associated with each such monthly transition at that 96.5% accuracy. (current to current, current to 30, 30 to 60, 60 to current etc.) . Our 96.5% accuracy is worth repeating.  That level of predictability in default and prepayment scenarios is truly unique and unparalleled in the industry.  It is what we and frankly others believe is a game changer.  You can imagine just some of the many applications of a Model this accurate: loan level valuations for entire portfolios and pools, marks to model, trader pricing on whole loans or MSRs bids and sales, CECL at the loan level, refinance portfolio retention strategies, the list really goes on and on.  The net-net of Model recommended actions yields gains and improvements to cashflows on any given portfolio that can exceed 100 basis points per annum, and even more on NPL and RPL pools.

I can say without hesitation that AiCurio provides the best guidance available in the market for month-to-month, life of loan decisioning by all owners and servicers of the mortgage cash waterfall.  We call these abilities of the Model the Mortgage Risk Decision Engine.

2) How has the COVID-19 pandemic affected your marketing and sales strategy?

I think of this current shock, and the initial wave of unilateral forbearances as a replay of many of the issues we experienced in 2008, 2009, and 2010, but this time, all crammed into a three-month period.  Like everyone else, we have shifted the direction of our sales strategy to focus more immediately on asset owners and their servicers’ need for assistance with defaults/loss mitigation and loan modifications. Efforts to contain the pandemic have increased unemployment, impacted the consistency and amounts of income and employment, begun to reduce housing prices, and consequently, for a not-insignificant portion of borrowers, limited their ability to refi or purchase.  It’s our belief that these circumstances and resulting behavior will continue to trend for through the medium future at least.  Unfortunately, so also will the pain initially experienced by servicers as the increase to servicer workload will be directly attributable to the increased demand and need for simultaneous modification and default processing along with regular customer service.  These additional required efforts will tax servicers’ abilities to meet borrowers’ and investors’ needs, not to mention the requirements associated with compliance, regulatory oversight and the Cares Act. 

The good news is our model provides consistent, statistically-validated recommendations for each individual borrower in any portfolio.  To that end, we are in discussions with a number of entities about how our Servicing Optimization and Performance Improvement Model (SOAPi) operates to ameliorate the risk of each loan predicted to default on any payment.  SOAPi does this by running over 200 synthetic loan modification simulations to make the best recommendations that will yield the greatest NPV per loan as it also considers regulatory concerns of service levels and customer attention.

3) Historically the mortgage industry has lagged in terms of adopting new technologies and models. Do you think the industry is ripe for a paradigm shift and will therefore aggressively adopt new technologies like AiCurio and why?

 “Lagged” is probably a polite understatement. In 2010, our chief data scientist, Mike Biddle, attempted to use machine learning (ML) to support ratings of legacy RMBS but was met with stiff resistance.  As stated by a rating agency executive at the time: “…if you cannot give us the calculations in Excel [the rating agency] will not use it”.  Since ML cannot be delivered in Excel, a lower-level methodology with far-inferior accuracy was, and continues to be, employed. This Excel-level of analytics, which has long-served the trade as well as it can, has been the near-universally used standard despite it being orders of magnitude inferior to the predictive and reporting capabilities of AiCurio.

In 2018, the mortgage industry began to show signs that the it was ready to start the process of implementing ML. The FNMA Q3 topic analysis survey of 184 lending institutions found that 13% had some level of AI implementation and another 14% were in trials.  The survey also found that 58% of all mortgage lending institutions planned to use ML within 2 years, with only 2% stating that they had no need for ML within their operations.

That’s a lot of background, and I could give more. However, as everyone in the trade well knows, we as an industry have been slow to adopt new technologies, but not without good reason related to change and workflow management, among others.  And while there are lots of challenges, we set out by design and have put in the work to create something unique and very useful for originators, servicers, traders, and owners of mortgage loans; and just as importantly, AiCurio is designed to be easy to use by people used to Excel Spreadsheets. The staff of AiCurio has been creating forecasting models for the residential mortgage industry for over two decades, with the AiCurio Model being the 11th major model build.  AiCurio is the first artificial neural net ever created for residential portfolio analysis at the loan level.  AiCurio’s Model calculates over 21,000 interrelated actual data dependencies to arrive at the 96.5% predictive accuracy at the loan level.

Probably one of the single-most important indications of the mortgage markets’ readiness to adopt technology like AiCurio came with the June 2018 announcement that Freddie Mac intended to integrate ML into all aspects of their operation.  In her June 15, 2018 announcement, Loretta Ibanez, Mortgage Innovation Director Single-Family Strategic Delivery, described the reimagining of the mortgage experience by taking advantage of “an enormous opportunity to embed artificial (AI) and machine learning into the company’s internal operational processes.”  Following Freddie’s announcement,  a number of articles were published in support of this vision of an AI-driven residential mortgage industry:  American Banker, Can Freddie Mac’s embrace of AI pull the industry along?; National Mortgage News, Freddie Mac said to test artificial intelligence underwriting software; Forbes, AI is coming to Take Your Mortgage Woes Away. 

4) What segments of the mortgage sector do you think will be early adopters of your AiCurio model?

We are initially targeting investors in mortgage assets, due to the direct and immediate financial benefit of having better information. This includes investors who are buying, selling, holding, or securitizing residential whole loans as well as MSRs.  AiCurio’s level of accuracy in predicting cash flows significantly helps these and their associated secondary and capital market teams make better decisions that directly increase their bottom line and makes them more profitable.  It has not gone unnoticed by many early users that they have a significant competitive advantage in knowing, at a very high level of accuracy, the probable NPV of all cashflows when trading with those that do not have the benefit of the AiCurio Model.

From our many ongoing discussions, we believe that servicers will soon also begin to adopt the AiCurio Model, either directly as they see the benefits realized by the investors that use the output to guide their servicing efforts, or indirectly as investors and asset owners begin to utilize AiCurio’s SOAPi Model as a servicing directive and oversight tool. While the servicers’ benefits are less direct, the Model supports operational efficiency to reduce costs and makes the loan modification and loss mitigations efforts much more effective.  Successful implementations of AiCurio’s Model enhance and enable servicing staffs to have better and more efficient and meaningful interactions with borrowers.  Our Model enables “real people” to utilize SOAPi’s precision loan modification and loss mitigation recommendations that better meet each borrower’s needs and are more likely to have a mutually beneficial outcome for all involved.

5) How would you explain the benefit of AI and your model in particular to the average homeowner with a mortgage being serviced by one of your clients?

Our long-term goal is to introduce the AiCurio Model into the lender’s origination process in order to extend mortgage financing to a greater portion of mortgage-seeking consumers, especially to those who would have otherwise been unapproved for mortgage credit.  The AiCurio technology allows lenders to identify otherwise good credit risk borrowers that currently fall outside of today’s QM “underwriting boxes.”  This analysis helps lenders better serve borrowers from a business and human point-of-view.  The Model reveals to underwriters and servicing managers alike actionable data driven insights at hyper-levels of accuracy that have not existed to date.  This permits better origination and servicing decisions to be made based on data and science.  It will ultimately lead to a much more transparent and vibrant marketplace for both borrowers and lenders. 

The last point I would make is that adopting the AiCurio Model does not reduce the power of relationships — it actually moves them to a higher level.  When implemented correctly, which, is  not burdensome since we do all the work, AiCurio replaces routine tasks fraught with human or other error with accurate information that ultimately bolsters the creditor-borrower process and relationship with transparency, certainty, and speed.   Moreover, the AiCurio Model doesn’t desensitize your organization to the customer, rather, it increases your ability to know and focus on each customer’s individual need in seconds instead of days.  The knowledge gleaned from the Model enhances a lender or servicer’s ability to anticipate borrower and organizational needs while optimizing your interactions with all borrowers, performing or otherwise.




Valued partners:
VantageScore Licensees:
Equifax Experian TransUnion