Alternative Data and Modeling Techniques in the Credit Process

Submitted electronically

Monica Jackson
Office of the Executive Secretary
Consumer Financial Protection Bureau
1700 G Street, NW
Washington, DC 20552

Dear Ms. Monica Jackson,

The Milken Institute Center for Financial Markets would like to thank you for the opportunity to respond to the Consumer Financial Protection Bureau’s (CFPB) request for information (RFI) regarding the use of alternative data and modeling techniques in the credit process.

The Milken Institute (the Institute) is a nonprofit, nonpartisan think tank determined to increase global prosperity by advancing collaborative solutions that widen access to capital, create jobs, and improve health. The Center for Financial Markets (CFM) promotes financial-market understanding and works to expand access to capital, strengthen and deepen financial markets, and develop innovative financial solutions to the most pressing global challenges.

Our comments largely reflect on the potential benefits and risks to market participamodelingnts from the use of alternative data and modeling techniques. We are also aware of the Bureau’s recent RFI on the small business lending market, which we intend to respond to in due course. Nonetheless, we have included our observations from the small business lending marketplace in this letter given the importance of alternative data in painting a more holistic picture of a small business beyond the credit score of a small business owner.

The Institute commends the Bureau for taking a closer look at the use of alternative data and models in the credit process. The use of alternative data and incorporation into various credit models, provided the use of such data adheres to various state and federal regulations, has the potential and is already opening up access to capital for a variety of consumers and small businesses.

However, innovations in the credit space are constantly evolving, requiring regulatory frameworks to enable the right kind of flexibility to drive innovation forward without reducing the level of protections to the end user. As such, the comments that follow pertain to the approach of the RFI:

1. The CFPB’s RFI contains a number of detailed questions, which companies may be hesitant to address in a public setting given the proprietary nature of their data and underlying models. As such, and building on CFPB’s efforts under Project Catalyst, we would encourage the CFPB to develop a true regulatory sandbox model where creditors using alternative data for credit purposes can engage with regulators, and where companies can come in under a safe harbor to test innovative credit models. Engagement, particularly in a more private setting, could encourage creditors focused on consumer lending to be more open with regulators, share knowledge and information, and effectively answer the vast majority of questions contained in this RFI in more depth.

2. The CFPB should be open to providing additional guidance related to unfair, deceptive, or abusive acts or practices (UDAAP) and the disparate impact standard. Rather than working backwards to understand how credit decisions were made in a certain area or to a certain segment of the population (which would require painstaking access to evolving and potentially proprietary processes and models), providing additional guidance on what is unfair (e.g. regardless of being deliberate or accidental, average interest rates for a specific gender or race being higher than others purely on that one variable) would go a long way to clarify how the CFPB interprets each provision. The guidance should also incorporate a range of bands that provide the lender with the flexibility to maneuver and tailor their credit models with the understanding that CFPB enforcement actions will not occur provided the creditor does not step outside of said bounds (e.g. regardless of being deliberate or accidental, number of loans to a particular underserved group does not exceed a statistically significant underweighting—more than two standard deviations—relative to their population in that geography or industry). By focusing on clarifying what the destination is, verifiably fair practices, rather than mandating specific steps on a specific road, the CFPB would be able to achieve its objective without needing detailed access to processes that may change faster than can be observed through an RFI.

On the substance of the RFI, our comments are organized as follows:

• Observations. From review of the CFPB’s RFI and based on prior concerns the Institute has raised, we have included a few observations covering a consumer’s ability to correct mistakes on their credit report, regulatory access to a platform’s proprietary information, and subjecting small businesses borrowers to the same level of protections consumer borrowers currently have.

• Risks due to alternative data and models. The lack of understanding as to how alternative data and models will react in a downward credit cycle and whether there is enough transparency in the lending marketplace are concerns.

• Regardless, there is a need for alternative data and modeling techniques in the credit process. It’s not a question of whether we need alternative data, but how the data can be utilized to drive credit to the unbanked and underbanked portion of the U.S. population in a responsible manner.

• The opportunities driven by the FinTech community to address current pain points in the credit process. We provide a look into recent research and on-the-ground evidence that suggest alternative data can and is playing a role in enabling platforms to meet consumer and small business credit needs.

Observations
The Institute has brought forward prior comments and provided additional insight on certain issues pertinent to innovative platforms operating in the lending space. They are:

• First, as the Bureau recognized in its release and others have noted, there is a persistent problem in the U.S. in the ability of a consumer to correct a mistake on a credit report that could have significant ramifications in the ability of said consumer to access credit. You do not need John Oliver6 to tell you how painstaking the process is to rectify bad data and/or mistakes that can affect the credit score of an individual. Add that to the fact that most Americans lack basic knowledge as to what makes up a credit score,7 and you have a potential recipe for disaster when it comes to accessing credit. This is not just an issue for only creditors engaged in lending to consumers and small businesses using alternative data, but this represents a significant policy gap covering a wide variety of industries that certainly needs to be addressed to ensure borrowers are able to correct their reports in a streamlined and more efficient manner, rather than waiting months, if not years, for the mistake to be corrected.

• Second, we are very concerned about any regulatory effort to seek unfettered access to a firm’s proprietary information related to the underlying algorithms used in a platform’s underwriting processes.8 We would note the amount of concern the CFTC’s recent actions regarding the source code behind algorithmic trading raised not only among industry stakeholders, but with regulators, policymakers, and other industries, as well. Showing a loan book to regulators is one thing, but giving up a firm’s underlying code is completely different and potentially reckless.

• Third, we would caution against any efforts to subject small business lenders offering $100,000 or less in financing to the same thresholds that apply to the consumer lending space. Although additional transparency is certainly needed in the small business lending marketplace, which we discuss further below, we believe the consumer and small business lending markets are separate and distinct from one another and efforts to apply consumer protections to small business lending risks restricting capital to small businesses.

Risks due to Alternative Data and Models

• There are risks in using alternative data and modeling techniques. First, the lack of understanding as to how the various innovative credit platforms will perform in a credit downturn is of concern, especially when most alternative credit models have yet to go through even one credit cycle. Market gyrations in early 2016 exposed certain platforms’ vulnerabilities to capital flight, in particular.

•  A second risk is transparency. Are platforms both in the consumer and small business lending space adhering to current state and federal regulation? Based on prior comment letters from stakeholders, there seems to be consensus that there are sufficient protections in the consumer lending marketplace, but several commenters have pointed out the lack of transparency in the small business lending marketplace.
  
  We would note, however, the industry efforts being made9 to provide for more clarity related to the costs of various financial products provided by certain creditors, and would encourage the CFPB to engage with stakeholders from the Innovative Lending Platform Association and the Responsible Business Lending Coalition on these initiatives.

Regardless, there is a need for alternative data and modeling techniques in the credit process
There would not be a need for this discussion if the vast majority of U.S. consumers and small businesses were being served by the traditional financial services ecosystem in a verifiably fair manner according to the Equal Credit Opportunity Act (ECOA) and Regulation B, the Fair Credit Reporting Act (FCRA) and Regulation V, and the prohibitions on UDAAP. In reality, that is not the case and has been so for some time.

One could even argue that we are well past the debate about whether there is a need for alternative data and modeling techniques given the following:

• The number of Americans classified as “unbanked” or “underbanked” is still far too high
  In a 2015 report10, which used 2010 census data, the CFPB found that 19.4 million Americans (or 8.3 percent of the population) have credit records that cannot be scored, while a further 26 million Americans (or 11 percent of the population) are credit invisible.
  
  The Federal Deposit Insurance Corporation’s biannual study of underbanked and unbanked households found roughly 9 million households were unbanked and 20 million underbanked. If we breakout the households into individuals, as the FDIC did in its 2015 report,11 15.6 million adults and 7.6 million children were unbanked, while more than 51 million adults and 16 million children were underbanked in 2015.

• The ability to bank and access credit continues to be a struggle for both consumers and small businesse
  
  More than 4,800 community branches between 2009 and 2014 were shut, equating to about 5 percent of all branches in the U.S. Mergers and acquisitions particularly among small, community banks continues and, last year, reached a seven-year high, compounded by a low interest rate environment, regulations following the most recent financial crisis, among other reasons.

Similarly, the reduced numbers of small banks, in particular, across the United States, not only threatens physical access to the formal financial system, but has replaced the traditional, relationship-driven underwriting models with more automated methods of assessing credit worthiness, particularly among larger financial institutions. These models often fail to take into account the on-the-ground realities of local economies and dispense with the traditional relationship-based models of assessment where local banks and institutions were able to address the credit needs of their communities that larger institutions and more automated models could or would not take into account.

Beyond credit assessment, there are also issues concerning the supply of credit. For instance, it is well known that many banks have retreated from financing small businesses given how uneconomical it is to lend at such levels, despite the demand.14 For instance, while total C&I lending has recovered since the financial crisis, C&I lending under $1 million is down roughly 14 percent from its peak back in 2008.

In 2013, then Secretary of the Treasury Jack Lew gave prepared remarks where he stated that there are roughly 8,000 small business declines every day—more than 2 million small business declines a year.16 Yet, as the Association for Enterprise Opportunity points out, roughly one- third of those businesses are credit-worthy using current underwriting methodologies available in the market.

As a result, individuals, small businesses, and whole communities can be locked out from traditional financial services and products, and, in some cases, the industry in its entirety.18 Consumers and small businesses are left with reduced choice, more expensive options for credit, and few options to build and rebuild their credit profiles.

• Minorities face a widening credit gap in the wake of the financial crisis
  
  While minorities have faced persistent access to credit issues in the past, including existing racial biases, minority communities were hit the hardest during the housing market collapse. For African-American and Hispanic communities in particular, more than half (53 percent and 66 percent, respectively) of their net worth was wiped out. For most small businesses, household equity is a key source of collateral for business financing. With collateral wiped out and household wealth in decline, minority-owned small businesses that are viable borrowers are often unable to meet the credit standards employed by traditional banks.

• Generational shift in preferences and capacities
  
  There are between 70-80 million Millennials in the United States. Advancements in technology have profoundly influenced how this generation, against all others, transacts in and interacts with financial services and products. This is a generation that continues to stray away from previously held financial and societal norms, and, most importantly, previously held views on how to build a strong credit profile.
  
  For instance, only one-third of adults between the ages of 18 to 29 have credit cards. This generation should more aptly be called the renter’s generation with homeownership levels for Americans aged 18 to 34 around 35 percent. Importantly, less than 60 percent of Millennials view maintaining a good credit score as important.22 Put all this together, among other aversions to the traditional financial system, and you have a generation that continues to be difficult to score utilizing traditional methods, thereby making it more difficult and/or costly to access credit.

The FinTech community offers opportunities to address current pain points in the credit process
Clearly, we have come to a point where the traditional ways of assessing credit and the current financial ecosystem itself are unable to effectively serve significant portions of the U.S. population. Through the use of alternative data and modeling techniques, FinTech offers the opportunity to break down the silos and barriers currently inhibiting consumers and small businesses from obtaining capital, and the right kind of capital, in a responsible manner.

And we’re seeing promising reports as well as early signs pointing to some of the advantages in platforms leveraging alternative data to produce results. For instance:

• FinTech can act as an alternative to short-term, small-dollar credit. A recent study found that certain FinTech companies are currently providing products and services that can act as a superior substitute for current short-term, small-dollar credit, thereby allowing low-income families to see meaningful improvements to their financial wellbeing. FinTech products can help address income volatility—a persistent problem among low-income working families—and, if made widely available, could address the utility needs of roughly 16 million full-time workers in low-income working families.

• FinTech can provide access to capital to SMEs in areas abandoned by traditional finance, with particular focus on minority- and women-owned small businesses. In a study conducted by PayPal and Kiva covering PayPal’s Working Capital (PPWC) and Kiva Zip portfolios, roughly one- quarter of PPWC loans were disbursed in the three percent of counties that experienced the loss of ten or more banks since the financial crisis. In addition, more than one-third of PPWC loans went to low- and moderate-income businesses, while Kiva Zip saw more than half of its lending portfolio go towards women- and minority-owned small businesses.

• FinTech can provide meaningful growth opportunities for small businesses and for economies overall. In a study published by OnDeck Capital, the first $3 billion in financing provided to small businesses resulted in more than $11 billion in additional economic output and the creation of 74,000 jobs. Alternatively, $1 in lending from OnDeck generates $3.62 in economic output. A similar study by Funding Circle UK, which has operations in the U.S., found that since 2010, financing small businesses has contributed $3.6 billion to economic growth and supported the creation of 40,000 jobs.

• FinTech can reach the unbankable. FinTech lending in the consumer segment is particularly focused on borrowers with high FICO scores, or the crème of the crop, so to speak. That said, there are a number of platforms in the U.S. that incorporate alternative data to reach individuals and households who are locked out or face difficulty accessing the current financial services system. For instance, consumer lending platform LendUp has detailed their efforts on building the credit profiles of individuals locked in the subprime, or deep subprime credit marketplace. According to a recent LendUp study, borrowers on the platform with credit scores below 500 have a 62 percent likelihood of a 50 point VantageScore increase after two years with LendUp.

The Milken Institute would again like to thank the Consumer Financial Protection Bureau for providing the opportunity to comment. Alternative data and modeling techniques certainly have a role to play in an increasingly digital economy, and we look forward to hearing more from the CFPB on this topic.

Please let us know if we can provide any additional information, and we would be honored to have the opportunity to continue this discussion in person.