Charleston, WV: Field Hearing on Alternative Data 2/16/2017


Welcome. Welcome to the Consumer Financial Protection
Bureau’s field hearing in Charleston, West Virginia, at the University of Charleston. At today’s field hearing, you will hear from
Director Richard Cordray and a panel of distinguished experts, who will discuss issues related to
developments, risks, and benefits of using alternative sources of financial information
developed through new technologies to weigh a consumer’s creditworthiness. Today, the Bureau issued a Request for Information
about this issue. With this RFI, the Bureau is seeking comments
from the public on whether unconventional sources of information, new ways to analyze
the data, and the use of new technologies could open up more access to credit for many
Americans who are currently outside the mainstream credit system. The Consumer Financial Protection Bureau,
or the CFPB, is an independent Federal agency whose mission is to help consumer finance
markets work by making rules more effective, by consistently and fairly enforcing those
rules, and by empowering consumers to take more control over their economic lives. As part of the Bureau’s mission to protect
consumers, to date, we have handled over 1 million complaints and actions resulting in
nearly $12 billion in relief to over 27 million consumers. My name is Zixta Martinez. I am the Associate Director for the External
Affairs Division at the CFPB. Our audience today includes consumer advocates,
industry representatives, state and local officials, and, of course, consumers. We’re delighted that you’re here. We’re also grateful that the Honorable Patrick
Morrisey, Attorney General for West Virginia, is here with us today and will provide remarks. We’re also grateful that several members of
the West Virginia State Legislature have joined us at today’s field hearing. Let me spend just a few minutes telling you
about what you can expect. First, you will hear from the Attorney General,
Patrick Morrisey; then from CFPB’s Director, Richard Cordray, who will provide remarks
about alternative data and the Bureau’s RFI. Following the Director’s remarks, David Silberman,
Acting Deputy Director and the Associate Director for the Bureau’s Research, Markets, and Regulations
Division, will frame a discussion with a panel of experts. After the discussion, there will be an opportunity
to hear from members of the public. Today’s field hearing is being livestreamed
at consumerfinance.gov, and you can follow CFPB on Facebook and Twitter. So let’s get started. Patrick Morrisey was elected as the Attorney
General for the State of West Virginia on November 6, 2012, and reelected to a second
term November 8, 2016. Among the Attorney General’s many impressive
accomplishments over two terms, he secured a $160 million Internet settlement in December
of 2015, which marked the largest independently negotiated consumer protection settlement
in West Virginia’s history. He also strengthened the Office’s Consumer
Protection Division, enabling it to vigorously enforce the state’s laws and proactively educate
citizens about scams and ways to protect their identities. Attorney General Morrisey, you have the floor. Well, thank you very much. And I’m grateful that everyone has come here
today. And I would like to do a special shout-out
to Director Cordray for coming in from Washington, D.C. I think the topic that you’re going to hear
about today is incredibly important because all of us in the State of West Virginia care
very passionately about enhancing consumers’ credit opportunities, and so I look forward
to learning a little bit more about some of the alternative data and the innovative new
ways that we can expand opportunities for West Virginians. So a hearty welcome to you, Director, for
coming. We appreciate it. Now, as previously mentioned, my name is Patrick
Morrisey. I’m the Attorney General of the great State
of West Virginia. And it’s pleasure to welcome a number of you
to our great state. I trust Director Cordray because he’s a neighbor
from Ohio, is familiar with our state’s natural beauty. We have an incredible place, one of the most
beautiful states in the Nation. I ask everyone who is coming in to come back,
ski our mountainsides, come to the summer, go to the New River Gorge, go down to Greenbrier
County. You’re not going to find a better place to
come back and visit. I’m also grateful for Director Cordray’s commitment
to public service, and a lot of the work he did in Ohio, in particular, some of the $2
billion he secured for Ohio retirees. So I know he’s been very aggressive protecting
consumers, and I have a lot of respect for that. I also know that there are times that we might
have different views on the roles of government and the role of the CFPB. It’s probably no secret that West Virginia
was part of a collection of states that filed suit over the CFPB because we had a number
of significant legal questions, questions pertaining to how the agency was funded and
the constitutionality of it, but despite those differences, I can say that we share a common
interest in protecting consumers, and that has to be all of our priority. I can say that our offices have worked together
on a number of fronts. Together we have monitored the conduct of
banks in the wake of the national mortgage settlement from 2012. That settlement sought to remedy misconduct,
predatory lending, servicing issues, and compliance failures, and we believe that that partnership
has resulted in West Virginia consumers receiving at least $3 million in payment and mortgage
modifications since 2014. Beyond that, our offices have actually worked
together to successfully sue Sun Trust Bank, HSBC, and Morgan Drexen, a company that provided
outsourced administrative support services to attorneys and debt settlement practices. The Sun Trust settlement, well, that provided
more than $260,000 to West Virginia borrowers whose loans were serviced by the Bank. People lost their homes to the terrible foreclosures
that occurred in 2008 through 2012. The HSBC settlement allowed more than 2,000
West Virginia consumers to receive payments or rate reductions, modifications, and some
decrease on their loan. That was a good result for the consumers of
our state. And like the national mortgage settlement,
the Sun Trust and HSBC lawsuits really sought to remedy that misconduct, that predatory
loan behavior, and the servicing and compliance failures that we saw in place for a long period
of time. So the object isn’t just to secure money for
the consumers, the object is to fix the problems so they don’t occur again in the future. We look at the Morgan Drexen case. That focused on misrepresentations as to the
provisions of legal services and debt settlement. I can tell you, one of the issues we spend
the most amount of time on from a consumer protection perspective is ensuring that our
debt collection laws are upheld vigorously, whether we’re dealing with the issues pertaining
to how the debt is collected so that there are reasonable tactics for collection, or
ensuring that the entities are licensed in the State of West Virginia before they collect
the resources. We take our responsibility over that very,
very seriously. We’ve also been pleased to work with CFPB
on other issues of importance for the state, including the closure of ITT Technical Institute,
which had provided courses to hundreds of West Virginians. And I think there are many other examples
of how both agencies have worked to try to enforce the laws that we’re responsible for. In my office, we try to attack every violation
of the law aggressively. We don’t care about political affiliation
or economic status as we’re enforcing the law. Our job is to make sure that consumers are
protected. Since 2013, my Office’s Consumer Protection
Division has brought in over $84 million through lawsuits, assurances of discontinuances, and
it’s also secured almost $30 million in debt cancellation for consumers. It was also mentioned earlier we were able
to secure a separate $160 million settlement from Frontier Communications, over $10 million
of which went to help consumers to address the issues that came from promises that weren’t
met. $150 million went back to invest to ensure
that the Internet speeds were going to be increased because, once again, it’s not enough
to just collect a settlement, we actually have to change behavior and make sure that
the underlying problems that gave rise to the consumer protection violations in the
first place are addressed. That’s part of our underlying philosophy within
the West Virginia Attorney General’s Office. Beyond the actual settlements, the lawsuits,
and the investigations, we also spend a lot of time educating West Virginians on scams. Every day someone is being ripped off across
the country, and within the State of West Virginia. So we have consumer advocates that are spread
out throughout the state meeting with people, collecting consumer complaints, and, most
importantly, educating people about what’s going on. I can tell you that consumer scams are a real
problem because fraud is a very real problem. And I’ll give you a personal experience that
occurred just the other day. I was down in Greenbrier County, and I was
talking to an 86-year-old elderly woman. She was recently ripped off by the grandparent
scam, and she talked through how the scammers called and they were able to reach out to
her, and how they set up this manipulative system in order to rip her off. Scams affect the elderly, but they also affect
people regardless of demographics, regardless of age. All of us have a responsibility to protect
the most vulnerable and to ensure that people know more about these types of scams so they
don’t fall prey to it. It’s easier than people think to fall victim
to a scam, and I can personally attest to that. I can tell you about many people who have
dealt with technology scams, especially in a state like West Virginia, when you’re dealing
with slower Internet, when people say, “Does your computer have a virus? Does it have a problem? Would you mind if we access that? Because we can fix your problem.” Those are issues that come up all the time,
but then you’re surrendering your precious personal identification away. We have to educate people, and that has to
be done every single day by everyone here in this room. This is critical. So we’re busy working on a lot of consumer
protection issues, but I also want to mention another matter that I think is critical as
consumer protection. While we have a great topic today in terms
of alternative data, I want to emphasize that consumer protection matters also include fights
on substance abuse, and we’ve worked very hard to educate people in West Virginia about
the nature of this devastating problem in the state. West Virginia has the highest drug overdose
death rate in the Nation, at 41-1/2 people per 100,000. No office has done more to be more aggressive. We’ve been very holistic, going after it from
a supply, a demand, and an educational perspective. And just recently my office announced the
largest pharmaceutical settlement ever in the history of the state, $36 million. It’s my hope that a lot of that money can
go to treatment and to address this terrible problem. We have to hold everyone accountable within
the pharmaceutical supply chain and change that problem. It’s a real public health crisis, and I’m
committed to continue working on that. In closing, every part of consumer protection
demands our daily attention. And I look forward to working with all of
the relevant agencies, people within the state, and in Washington to make sure that we achieve
our common goals of protecting consumers, going after consumer fraud, and respecting
our boundaries, making sure that Federal and state agencies comply with the Constitution
and our laws. But, Director Cordray, I’m particularly grateful
that you’re here today to grace your presence in our incredible state. Thank you and to your team for coming here
today. And we’ll look forward to hearing about this
exciting new topic. Thank you. Thank you, Attorney General Morrisey, for
the remarks and for the warm and gracious welcome. It is reciprocated, and we look forward to
working with you in the future. I am now pleased to introduce Richard Cordray. Prior to his current role as the CFPB’s first
Director, he led the CFPB’s Enforcement Office. Before that, he served on the front lines
of consumer protection as Ohio’s Attorney General. In this role, he recovered more than $2 billion
from Ohio’s retirees—for Ohio’s retirees— —for Ohio’s retirees, investors, and business
owners, and took major steps to help protect its consumers from fraudulent foreclosures
and financial predators. Before serving as Attorney General, he also
served as an Ohio State Representative, Ohio Treasurer, and Franklin County Treasurer. Director Cordray? Actually, Zixta, your unexpected and unusual
stumble there reminded me of when my children came home from school, they were in elementary
school at the time, and they had learned that day that commas save lives, we could say prepositions
save lives. They were talking about the difference between
a t-shirt that said, “Let’s eat, Grandma,” as opposed to “Let’s eat Grandma.” I also want to take a moment to really thank
General Morrisey. He and I had a chance to talk a little bit
ahead of time, and we share a common view that consumer protection is a form of law
and order, and people who break promises to the citizens of this state or those of the
United States or know that there are laws in place and violate them in order to get
an advantage or get money in their own pocket, those are people who need to be enforced against
vigorously. We try to do that at the Consumer Bureau,
and the West Virginia Attorney General’s Office definitely tries to do that. And, General Morrisey, I also appreciated
your comments about opioids, although there are many issues that go well beyond what we
do at the Consumer Financial Protection Bureau, it’s a reminder that the state Attorneys General
deal with the entire menu of public policy problems in the states across the country. And we’re happy to have the chance to work
should-to-shoulder on them in our particular area of consumer finance. So thank you all for joining us. And we’re glad to be in Charleston as we explore
some new frontiers for consumer access to credit. As many of you know, the Consumer Financial
Protection Bureau is the single Federal agency with the sole mission of protecting consumers
in the financial marketplace. We’re working to ensure that consumers can
gain access to financial products and services that are fair, transparent, and competitive. In this spirit, we continue to encourage consumer-friendly
innovation, such as through our Project Catalyst. So today we’re announcing a Request for Information
about unconventional sources of information, new ways to analyze this data, and how new
technologies can help in assessing people’s creditworthiness. We want to learn more about whether this kind
of alternative data could open up greater access to credit for many Americans who are
currently stranded outside the mainstream credit system. We also want to understand how market participants
are or could be mitigating certain risks to consumers that may arise from these innovations. Let us begin by reviewing how our mainstream
credit system generally works. Until the rise of the modern credit reporting
industry, many loans were made based on personal relationships of longstanding that developed
between creditors and their customers. Someone who knows all about your personal
financial story, including your way of making a living, your accumulated wealth, your spending
habits, and your family background has an excellent vantage point for deciding whether
it’s a good risk to extend credit to you. Based on everything they know about you, they
can size up your creditworthiness, including any collateral you may be able to post to
security. Thus, they can make a pretty careful determination
as to whether you’re likely to recover what they decide to lend to you. Although this framework still describes some
fairly vigorous modes of local lending in this country, particularly at community banks
or credit unions, where it’s quite successful, we’ve also developed another credit framework
in our society. It uses automated underwriting systems, and
it’s built on extensive data about people’s credit histories and algorithms that are used
to analyze that data. This newer approach reflects changes in our
society, such as increased mobility and the growth of national banks and online financial
firms. These companies are not in the same position
to know all the detailed history of local communities and individual customers at a
personal level. This approach also reflects new technological
capabilities that can mine huge mountains of data and determine mathematically which
elements are most closely correlated with future performance. To get a loan under this more automated framework,
a consumer typically needs to have a credit score. An individual credit score is fashioned from
the information contained in individual files that are managed by nationwide credit reporting
companies. They typically have files on 200 million Americans
or more. This is the product of the modern era, now
greatly bolstered by computerized databases. Each file, known as a credit report, tells
the story of a consumer’s credit history and current credit usage, at least what can be
known from the information that’s actually in the file. It records the size and type of loans made
to the consumer, what is owed, how much credit is available, and whether prior debts were
paid on time. It may list personal loans and car loans,
credit card balances, student loans, and mortgages. It may also note unpaid bills and debt collection
and list court judgments, liens, or bankruptcies. This credit history is then used to determine
how likely consumers are to repay existing debts and to gauge the prospects for repayment
of any new debts they may take on. Some of the limitations of this system derive
from historical and contingent circumstances. For example, consumers often try just as hard
to make their monthly rent payments—I know I did when I was a renter—as they do their
monthly mortgage payments, but rent is often omitted from credit files, unlike mortgage
payments. This may be because rent is not typically
viewed as credit or it may be because mortgage loans are made by banks and financial companies
that have mechanisms for keeping careful records of them, which result in more regular categories
of reportable data. By contrast, rents are collected by millions
of individual landlords scattered all over the country, and data on those payments is
not collected in any systematic way. To take another example, debt collectors often
report data on the debts they are collecting, including debts arising from unpaid medical
bills, for example, but the billers themselves, such as medical providers, do not report such
information. Credit files thus may include information
about bills you failed to pay but not necessarily about all the bills you did pay. In automated underwriting systems, and even
in many manual underwriting systems, decisions to grant credit and set interest rates on
loans are based on credit scores to a large degree. These familiar three-digit scores are drawn
from the information contained in individual credit files. As such, credit scores play a central role
in the financial lives of American consumers. They can determine whether people will be
granted credit at all, the terms or conditions for doing so, and including the interest rate. The availability of credit scores and the
accuracy and completeness of the underlying data have thus become increasingly important
to almost all Americans. Unfortunately, one of the reasons we’re here
today, for many consumers with a limited or nonexistent credit history, a credit score
is out of reach. The Consumer Bureau has run the numbers and
estimates that 26 million Americans—26 million—are credit invisible, meaning they have no credit
history at all in these files. Under the most widely used scoring models,
another 19 million people have credit histories that are too limited or have been inactive
for too long to generate any credit score. Here in West Virginia, nearly 180,000 residents
are credit invisible, and nearly 130,000 more residents have too little credit history or
histories that are too inactive to have a credit score. Added up, about one in five adults here in
the Mountain State are hampered in their financial lives by a lack of a credit score. The same story could be told virtually anywhere
in the country, since 45 million adults, Americans, fall into these categories nationwide. People with little or no credit history or
who lack a credit score have fewer opportunities to borrow money in order to build a future,
and any credit that is available usually costs more. That only deepens their economic vulnerability. Among them are those living in lower income
neighborhoods, young people just starting out in life, and many who are recently widowed
or divorced and may not have yet built sufficient credit history on their own. Many people without credit records or credit
scores work hard and strive to pay their bills on time. They may live paycheck to paycheck straining
to make ends meet. They often are caught in a catch-22, unable
to get credit because they have not had credit before. They cannot seize meaningful opportunities,
such as borrowing to start a business or buy a house. For these consumers, the use of unconventional
sources of information, known as alternative data, may allow them to build a credit history
and gain access to credit. Alternative data may draw from sources such
as rent or utility payments. These obligations may not qualify under more
traditional definitions of credit, and, as a result, would not be factored into the credit
decisioning process. Alternative data may also draw from electronic
transactions, such as deposits, withdrawals, or transfers from a checking account, and
it can encompass the kinds of information that relationship lenders typically know as
a matter of course, such as the consumer’s occupation, educational attainment, and various
other personal accomplishments. New forms of alternative data may come from
sources that never existed before, such as the way we use our mobile phones or the Internet. By filling in more details of a consumer’s
financial life, this information may paint a broader and more accurate picture of their
creditworthiness. Adding this kind of alternative data into
the mix thus holds out the promise of opening up credit for millions of additional consumers. Alternative data holds out further promise
as well. Credit scores, by their very nature, are backward-looking
indicators. Consumers who experience a financial hardship,
such as the loss of a job or a large medical expense—and many people experience such
hardships—may fall behind in making credit payments. This may tag them with a low credit score
long after their financial situation has turned around. Alternative data may help lenders identify
more precisely from those who currently carry so-called subprime credit scores a substantial
subset of consumers who are, in fact, good credit risks. These people should not be held back simply
by their retrospective credit score. The Request for Information we’re issuing
today looks into the pros and cons of these uses of unconventional sources of information. We’re examining what data are already available
for use today and we’re looking into what the future may hold as technologies continue
to evolve. We’re seeking to study how these data are
being gathered and analyzed in underwriting models now used by banks and other financial
companies, including the so-called fintech companies, and we’re seeking to better understand
how these models and modeling techniques are evolving. This Request for Information focuses on four
main issues. First, it looks at the potential risks and
benefits for consumers of using this additional information to better assess their likelihood
of repaying a loan. Second, it looks at how introducing new alternative
data sources into the credit decisioning process might add to its complexity. Among other things, we want to find out if
this will make credit decisions more difficult for people to understand, and thus make it
harder for them to control their financial lives. Third, the Request for Information looks at
how the use and interpretation of these data may affect privacy and transparency. And, finally, it looks at whether reliance
on some types of alternative data could result in discrimination, whether inadvertent or
otherwise, against certain consumers. Let me start with the first point, access
to credit. As I mentioned, a key question for the Consumer
Bureau is how people without a credit score could begin building a credit history. We want to learn more about how we could promote
the responsible use of alternative data even as we continue to protect consumers’ interests. For instance, someone with no credit history
might nonetheless be quite reliable in paying their cell phone bill or their rent on time,
or they may have a history of checking account deposits and have made good use of a debit
card. This might make them a very viable credit
risk. We know that some lenders will not loan money
to consumers with a credit score that is less than, say, 620, according to traditional measures,
but they might do so if alternative data suggests that a particular consumer with such a score
would be less likely to default on the loan as based on this other type of information. This leads us to the second issue. Even as alternative data may shed more light
on a consumer’s creditworthiness, the sheer volume of new data that may be streaming into
the system could have other effects. On the one hand, new analytical methods based
on unconventional information could produce a faster, less complicated application process
with lower operating costs for lenders, and, thus, lower loan costs for borrowers. On the other hand, the accumulation of more
and more alternative data could create a tangle of information that is harder for people to
understand and unravel. The credit process can already be somewhat
murky, so we want to learn whether folding in alternative data could complicate the decisions
facing consumers. The harder is it for consumers to understand
their credit record or whether they’re likely to qualify for certain loans, the harder it
will be for them to master their finances. The same complexity could also burden lenders,
who must explain adverse credit decisions to consumers. And it may bog down financial educators and
counselors who are trying to help people understand their credit standing and take more control
of their financial lives. The third issue we’re raising today concerns
how alternative data is shared, by, and to whom, and whether these interactions are safe
and secure. We want to know whether this information is
reliable and whether its use is transparent to consumers. Some consumers may not even know that the
information was collected and shared, let alone how it may be used in the credit process. We’re also exploring whether some information
is more prone to errors because it was collected under weaker standards in place at the time. Another question is whether consumers can
correct any mistakes that turn up. As part of our inquiry, we’re looking into
how the credit reporting laws may apply to these and other issues. And, finally, we’re looking into how this
information, even if entirely accurate, may be applied or interpreted. If the use and analysis of alternative data
leads to certain consumers being needlessly penalized, we want to know that. For example, some newer underwriting algorithms
use measures of residential stability. These measures may help predict creditworthiness,
they may well do so, and may identify consumers who make their rent payments on time. Yet members of the military, to take one example,
are required to move frequently as their duty stations change. As a result, this particular measure could
hinder access to credit for service members, even if they are in fact a good credit risk. Other data may be strongly correlated with
characteristics such as race or gender, which could enable lenders to do indirectly what
they’re forbidden from doing directly, drawing conclusions about whether to make a loan based
on a person’s race, gender, or other prohibited categories. Similarly, data tied to a consumer’s place
in the economic ladder may hinder those trying to climb it. This may be especially true for those who
are already struggling financially and facing a system that’s full of obstacles. So we’re looking into how fair lending laws
might apply to these and other issues. As we consider how the risks of alternative
data may give rise to the potential for discrimination, I want to pause for a moment and make clear
our intentions with the Request for Information. The fair lending laws are designed to promote
equal access to credit for all Americans without regard to race, sex, ethnic background, or
a variety of other personal characteristics. The reason for these laws is to eliminate
such credit discrimination in the financial marketplace, but if fair lending concerns
cast a large enough shadow, they may prevent people from considering and using alternative
data that might open up more credit for minority and underserved consumers. This could interfere with progress for the
very people these laws are intended to protect. Equal access to credit means even more if
overall access to credit is expanded and not constrained by lingering uncertainty about
how regulators intend to apply fair lending laws. So we’ve crafted this Request for Information
to help us better understand whether and how such uncertainty may be hindering credit access
for disadvantaged populations. We also want to learn more about how the Consumer
Bureau might reduce that uncertainty while holding fast to the anti-discrimination principles
that are the cornerstones of Federal law. That would help market participants go about
their business with more confidence that they can better assess the creditworthiness of
particular consumers without running afoul of legal requirements. In short, we see alternative data as holding
out the promise to the very populations that may be most disadvantaged by excessive reliance
on traditional credit reports and credit scores. And we’re committed to having a full and frank
discussion about how we can minimize the risks and maximize the potential benefits. With the Request for Information that we’re
issuing today, the Consumer Bureau invites all who are interested in these developments
to share their views on this rapidly evolving aspect of financial services. We strongly encourage affordable, responsible
lending to more people who may be already deserving of the opportunities that credit
can bring to their lives. At the same time, we want to make sure that
all lenders are playing by the same rules. This evenhanded oversight both protects consumers
and ensures a level playing field for the financial industry, and it applies to both
big banks and small startups. We want to learn more about how the use of
this data affects consumers and how it’s being analyzed and interpreted, and we want to know
whether it can help more of our neighbors gain control of their financial decisions,
enjoy more options, and achieve their own vision of the American dream. Thank you. Thank you, Director Cordray. At this time, I would like to invite the panelists
to take the stage. While they are doing so, I will briefly introduce
CFPB and guest panelists. David Silberman serves as the Bureau’s Acting
Deputy Director and Associate Director for the Bureau’s Research, Markets, and Regulations
Division. Gail Hillebrand serves as the Associate Director
for the Bureau’s Consumer Education and Engagement Division. Keo Chea serves as the Assistant Director
for the Bureau’s Office of Community Affairs. Our guest panelists are Chi Chi Wu, Attorney
with the National Consumer Law Center; Aaron Rieke, Principal of Upturn; Amanda Jackson,
Organizing and Outreach Manager, Americans for Financial Reform; Michael Gardner, Senior
Vice President of Specialized Services and Initiatives, Equifax; Nipun Goel, Senior Market
and Portfolio Strategist with Kabbage; Francis Creighton, Executive Vice President of Government
Affairs, Financial Services Roundtable. David, you have the floor. Thank you, Zixta. Good morning, everyone. As Zixta said, I am David Silberman. I am the Acting Deputy Director of the CFPB
and the Associate Director of the Bureau’s Division of Research, Markets, and Regulations. It’s my pleasure to be with you to moderate
this panel discussion portion of our hearing about the use of alternative data and modeling
techniques in the credit process. As Zixta indicated, we’re going to hear today
from a number of respected panelists, including consumer advocates and industry participants. Each panel member will give us some background
and provide their perspective. We’ll then pose questions to our panelists
and engage in a discussion. The panel discussion will then be followed
by the public comments component of the hearing, where we will hear from members of the public
who have signed up to share their observations. As Director Cordray noted in his remarks,
the Bureau estimates that approximately 26 million Americans have no traditional credit
history, and are considered credit invisible, and another 19 million Americans do not have
sufficient or recent credit history to generate a credit score under commonly used scoring
models. For these 45 million Americans, obtaining
credit can be difficult, if not impossible. As Director Cordray has explained, the use
of alternative data could expand access to credit for these consumers, but certain types
of alternative data presents significant risks for consumers. So as you’ve heard, today the Bureau has published
a Request for Information, or RFI, to seek information about the use of alternative data
and modeling techniques in the credit process. The purpose of this RFI is to assist the Bureau
and market participants in better understanding what’s happening in the market and what issues
are raised. We want to be able to assess whether there
are steps the Bureau can and should take to facilitate practices that enable responsible
innovations, allowing consumers to realize the benefits of alternative data while providing
necessary consumer protections and safeguards to mitigate any consumer risks. In the past 10 months, Bureau staff have been
meeting with market participants, including incumbent financial institutions and fintech
firms, with consumer advocates, with academics, and with our sister agencies to understand
the benefits and risks of various types of alternative data and modeling techniques. While we’ve learned much, there is still a
great deal to be learned. Today’s RFI and this field hearing are the
next steps in the process as we move forward to seek to ensure that consumer benefits from
the use of alternative data are realized and that risks are addressed. So with that in mind as a framing, I would
like to invite our panelists to present their opening remarks. Each panelist will have 3 minutes to make
a brief statement. Following this, Gail, Keo, and I will moderate
a discussion. So we’ll start, we’ll go down the line, and
we’ll start with Chi Chi Wu, from the National Consumer Law Center. Thank you, David, for both the introduction
and those remarks. So alternative data, there is a lot of buzz
about alternative data. That’s why we’re here. Right? It’s seen by many as a potential solution
for this issue of credit invisibility. And for some, it’s seen as a panacea. For us, though, on the consumer advocacy side,
it’s definitely not a panacea. It could be a solution, but more likely it’s
a tool, and like any tool, it could be good or it could be bad. It depends, from our perspective, on what
kind of data is being used and how that data is being used. For us, the devil is always in the details. And so I’m going to talk a little bit about
the kinds of alternative data that have been bandied about, and more right now the more
conventional alternative data. So, for example, Director Cordray mentioned
rental data. It seems from some of the pilot studies, rental
data could be very promising, especially given that the rental data that’s being added is
predominantly positive data. The negative data is already in there in terms
of there being collection items. And new information, negative information,
in terms of late payments isn’t being added. So that’s one that seems to have promise. On the other hand, a type of data that we’ve
been strongly opposed to using, because we think it’s potentially harmful, is gas and
electric utility data. This is an industry that is very heavily regulated
because it’s a natural monopoly. There are a lot of strong consumer protections,
and we think that monthly reporting of gas and electric utility data will undermine those
protections. Plus, it’s a payment obligation that really
fluctuates greatly. In the North, during the winter, you get high
bills; in the South, high bills during the summer; but then as the seasons roll along,
the obligations drop. And so people struggle to pay their bills
during high usage areas, and it depends on the weather, which is one thing that consumers
definitely can’t control. So that’s one we have concerns about. Telecom, cable, and cell phone data, on the
other hand, doesn’t raise those same concerns, because it’s not as heavily regulated. The issue with that is we want to make sure
that when such data is used, consumers’ rights to dispute errors or problems is properly
protected, because, as folks know, people do have problems with their cell phone and
cable companies every once in a while. But there is potential there. And then how it’s used is also important. We are more concerned about when a bunch of
new data is dumped into the files of the big three credit reporting agencies because while
there are maybe 50 million people who are credit visible, there are 200 million who
have existing files. And will that data help or hurt? For the credit invisibles, will it create
a bad score? Credit invisibility may be bad for credit
purposes, but those big three files are often used for other purposes, such as employment
and insurance, where invisibility isn’t such a bad thing, where a no hit is better than
a bad score. So the way the data are used is very important. And then just a couple seconds on less conventional
types of alternative data: social media, Internet searches, things like that. Those, they have potential, but they also
have risk; most importantly, accuracy and predictiveness. How do dispute whether your Internet search
result accuracy is being properly recorded? Whether it’s my searches on my laptop or my
14-year-old son is engaged in some interesting searches. So we’ll leave it at that. Thank you. Aaron Rieke, from Upturn. Hi, there. First of all, thank you very much to the Bureau
for holding this hearing and for looking into this issue and its past research on this issue. And I want to start, I’m I think here partially
to be the public interest computer nerd on the panel, but I want to start with what makes
this interesting beyond the technology and beyond the data, and that’s what Director
Cordray mentioned in his remarks, that there are about 45 million people in this country
that are currently underserved by this traditional system of credit reporting, and many of those
people have various other vulnerabilities that contribute to their absence in the system. And so I just want to keep that as the starting
touchstone and the touchstone we return to as we continue this discussion. Without access to credit, what’s there is
pretty ugly. I did a report last fall where we looked at
the sorts of search advertisements that came up when you searched for “I need money fast,”
and it’s really not pretty. So I think part of this is actually defending
consumers against the really atrocious products that await them when they don’t have other
options. I agree with everything that Chi Chi just
said in terms of data potentially being a tool that could be part of the answer in improving
the situation. I want to start with a broad theme, and I’ll
go into more details as the questions emerge throughout the hearing, which is that the
goal here is not just to improve lenders’ and banks’ ability to predict creditworthiness. That’s certainly an important piece of the
solution, and having some sort of good prediction is oftentimes better than having no prediction,
but I think it’s really important that we not lose sight of the fact that prediction
is half of what we’re looking for here, and the other half of what we’re looking for here
is fairness. We’re not just looking at new big datasets
to see what kind of combinations of things correlate with likelihood to repay versus
likelihood not to, that’s a starting place, but that we need to think carefully about
whether or not the predictiveness and the correlations that we’re seeing in the data
are actually serving the people, the 45 million people, that we’re setting out to help here. We shouldn’t allow protected class to sneak
in somehow and be something that’s driving predictions. Where you live shouldn’t sneak in somehow
and be driving predictions. Who your friends are shouldn’t sneak in somehow
and be driving predictions. So we need to be really, really careful, especially
as we grow the datasets, especially if we ever get to the point where we’re thinking
about something like social media data. We always need to ask, Why is this driving
predictions? And is this the kind of score and the kind
of data we want driving these decisions in our society? We don’t want to get past the point where
these decisions aren’t explainable anymore. I think “murky” is a great word for our current
credit system. It’s going to get harder the more data we
add. And so I think we need to keep explainability
both for consumers and for regulators in the puzzle here. And, finally, I think we need to look at the
bigger questions of, Are we driving virtuous cycles rather than spirals down? Are the sorts of data we’re adding in likely
to elevate people, or are they going to suffer from the same issues of punishing people that
fall on hard times? So I think that, again, the theme I just want
to start with here, prediction is half of the equation, but we can’t lose sight of asking,
Why is it predictive? Is it predictive in a way that we want our
credit scores to be predictive? And is this actually going to lift up the
people that we care about, that kind of core measure of success? Thank you. Amanda Jackson, from Americans for Financial
Reform. Thank you. And hi, everyone. Thank you for the opportunity for Americans
for Financial Reform to join this important panel on the use of alternative data to help
make credit scoring decisions. As you may know, Americans for Financial Reform
is a project of the Leadership Conference on Civil and Human Rights, and continues to
support and seek insight into data conversations. Alternative data is just part of the larger
universe of big data. Of course, the credit bureaus which collect,
aggregate, and then sell credit reports based on billions of bits of information concerning
whether and how millions of consumers pay their bills and whether they pay them on time
have been using big data for years. Credit scores, whether sold by the bureaus
or others, are based on credit reports. But with the growing power of network computers
and the ubiquitous data collection system that has grown up in the digital economy,
more and more data are being collected, and more and more algorithms are being proposed
to aid in corporate decision-making. Recognizing the importance of looking at big
data through a civil rights and consumer protection lens, several years ago, the Leadership Conference
on Civil and Human Rights, leading members of Americans for Financial Reform, aided by
the technologies here at team Upturn, worked with a broad group of organizations to form
civil rights principles in the area that is big data. We aim to stop high-tech profiling, ensure
fairness in automated decisions, preserve constitutionality, enhance individual control
of personal information, and protect people from inaccurate data. These principles represent the first time
the national civil and human rights organizations have spoken publicly about the importance
of privacy and big data for communities of color, women, and other historically disadvantaged
groups. Through these principles, we and the other
signatory organizations highlight the growing need to protect and strengthen key civil rights
protections in the face of technological change. Today, discrimination is not just a product
of biased human decision-making; rather, as the Obama White House noted at the conclusion
of its review on big data, discrimination can result from the way big data technologies
are structured and used. The data we have reflects our history, which
is in part a history of systemic unfairness towards some consumers in the consumer credit
marketplace and systemic economic exclusion in the broader economy overall. Whether big or small, more data and more kinds
of data will play an even bigger role in the future of lending. Some of this data may help more Americans
join the financial mainstream, helping to identify where individuals in protected status
groups aren’t enjoying the same access to credit as similarly qualified non-minority
bars. So while I’m hopeful that more uses of big
data will unlock new benefits, I am also concerned about its risk. The same is true with smaller alternative
datasets. The benefits of adding rental and utility
data to credit reports on credit scores for some consumers must be weighed against its
risk, as my colleague at the National Consumer Law Center explained today. But one thing is very clear: as we move forward
to understand the implications of automated decision-making on financial opportunity,
we are grateful to CFPB, an effective independent agency that has had the best interests of
consumers at its core mission, is on the job. The area of big data is one in which we particularly
need rigorous oversight and standard setting in the public interest. Otherwise, the corporate users of data will
have all the information, and members of the public will have no way to see the big picture. This is just one more reason we think it’s
important that the CFPB keep on the path of its independence and rigor in tact. Thank you. Thank you, Amanda. So now moving to my left, Michael Gardner,
from Equifax. David, thank you, and thank you for having
us as a part of this panel. Equifax is obviously inextricably linked to
the credit report, as we have traditionally defined it, and we take that responsibility
very seriously. But we have also been in the game of what
we call additional data. Here we call it alternative data—right?—which
is, what additional data can be found in the marketplace that can be used to predict a
consumer’s creditworthiness? And that interest in additional data or alternative
data goes back really almost 15 years for Equifax. And we believe that that alternative data
benefits both consumers and the financial system, and we approach how we use that data
looking through both of those lenses. One of the other things that is perhaps a
bit beyond the scope of this particular hearing, but I would like to get it out there, is we
also look at how alternative or additional data can be used in use cases prior to getting
to the financial risk decision that a lender might make, and facilitating those decisions
that get to that credit risk decision are equally as important to consumers as the final
credit risk decision. If I can’t be identified appropriately, et
cetera, then I’m not even going to get to the point of having a risk decision made. We at Equifax are also known for housing the
National Consumer Telephone and Utility Exchange database. That is a voluntary database that includes
utility, telecommunications, pay TV, and additional data for consumers. That database now approaches 215 million consumers. At any given time, that database has somewhere
between 25 and 30 million consumers in it that we do not find present on our credit
files, so correlating very closely, Director Cordray, to the numbers you mentioned in your
comments. So we leverage this data and other additional
data along with the traditional credit data in models that we use, or that we sell, obviously,
and that our customers, financial institutions, and others use in the marketplace to assess
creditworthiness. Those models and new modeling techniques are
a big part of how you leverage this data. Other panelists have mentioned that some alternative
data does not have the same coverage, and, therefore, you have to identify modeling techniques
that overcome that coverage issue. The most notable one of those solutions that
is in the market today is our FICO XD solution where we partnered with FICO as the modeling
arm. We leveraged the NCTUE data and traditional
credit data, and we also leveraged public record data sourced from LexisNexis Risk Solutions. So a lot of alternative data going into a
single credit score, with the purpose of that score being specifically targeted to allow
non-scorable consumers on a traditional credit risk score to be scored by the credit card
industry. And we do all of this with a very, very strong
focus on maintaining the same types of consumer protections for any data that is used in a
risk decision that we apply to the core credit data. Thank you. Thank you. Nipun Goel? Hey, good morning, everyone, and thank you
so much for the opportunity to talk about a subject that is truly near and dear to our
hearts over at Kabbage. Just to give you a quick overview of what
we do, we are a tech and data platform, or, if you will, a fintech company, that’s focused
on delivering credit to small businesses all over the U.S. And so while we’re not squarely in the consumer
space, I think where we are squarely focused is leveraging alternative data to broaden
access and break down barriers to traditionally underserved markets. Right before we got on the stage, I took a
quick look, and in the State of West Virginia, to date, we’ve delivered $7 million in capital
to over 200 small businesses. And so our company was founded sort of on
this question and notion of, How do we use e‑commerce seller data to provide loans
to businesses selling on e-commerce platforms? And that sort of notion has been what’s truly
been the foundation of our growth and trajectory over the last few years. And so for the purpose of this discussion,
I would like to offer some of the guideposts that we’ve used as we’ve sort of looked for
alternative data sources and have served both us as well as our customers pretty well. The first being we look to leverage data sources
that our small business—again, caveating that we operate in the small business space—we
try to look for data sources that our customers are already familiar with. Some examples here include connect your QuickBooks
account if that’s what you use to manage your accounting. If you use Stripe to process payments, we
look to use that sort of data. And so what that does for us is it gives us
a very in-depth and real-time understanding of the financial and operating health of a
company. What it does for the consumer themselves,
it actually engages them in a more in-depth and sort of engaged experience with the lending
process and really gives them a stronger voice in the lending process itself. The second guidepost that I would sort of
throw out there is one around transparency and reinforcing the use of how we’re using
that data. And so that’s not to the level of sort of
breaking down the machine learning algorithms and narrowing out every single feature, but
really trying to understand for our customers, for example, using financial accounts to understand
cash flows, or if they are a seller on an e-commerce platform, using that to understand
sales activity and getting, again, a more in-depth picture of what we can use to deem
creditworthiness. So if borrowers the alternative data source
and how we’re using it, and we, as financial service providers, are both transparent in
our use and extremely rigorous in our diligence, as I think a few folks here have mentioned,
what we enable is just a more informed, straightforward lending process. In summary, whether we’re using alternative
data as a single model or embedding it on top of existing credit features, what I want
to highlight is not striving just for transparency, but effective engagement of the customer in
the process as well. Thank you. Thank you. Finally, Francis Creighton, from the Financial
Services Roundtable. Thank you very much for having me here this
morning, all of you from the CFPB, especially Director Cordray. I work at the Financial Services Roundtable,
a trade association of about 100 of the 150 largest financial services companies in the
country, including banks, insurance companies, payments companies, investment firms, and
others. This is a really exciting issue for us, and
it’s exciting because there is so much innovation and good thinking going on here. And that innovation should help more people
get access to the kinds of financial services products that they need from the regulated
lending community. First let me note that financial institutions
are both furnishers of information and consumers of that information, and I’m here today to
talk mainly about how financial institutions use the information that they obtain from
others to make lending and other financial decisions. Now, financial institutions use data and information
to assess how people have handled their bills in the past because that can be predictive
of how they might handle their bills in the future. We use this information to judge whether a
potential customer will be able to meet their future obligations. So from that perspective, the more information,
the better. Having more information gives us a better
ability to determine whether we should make a loan or not. We welcome more data and are working with
our partners in the consumer data industry to get more because it helps us better serve
our customers, both current and prospective. However, we need to make sure that the data
we’re using for lending decisions are fair, accurate, and verifiable, and that consumers
have the ability to dispute information that they believe is inaccurate. The Fair Credit Reporting Act is a strong
legal protector of fairness, security, accuracy, and data integrity. But given that more data is more helpful,
we should also remember that we get data through a system where furnishers voluntarily provide
the information and consumers of information voluntarily use it. Individual credit bureaus work hard to improve
their data offerings, and that innovation and competition among them makes the process
better. Therefore, we would be very concerned about
any efforts to mandate what data we could and could not use. Further, what makes our system strong is that
when we use data, we have access to both positive information and negative information about
a consumer’s payment history. Both types of data play a role in assessing
a consumer’s situation, and limiting the type of data we can use we think would be counterproductive. We need both sides of the coin. We also want to note that as new data become
available, we have to example the implications of that data, not only on its predictiveness,
but also for implications on fair lending and other regulatory grounds. For example, if some new data were more available
for urban rather than rural families, that could result in thinner files for rural dwellers
with the attendant impacts that could have on underwriting decisions impacting them. As the Bureau examines the implications of
using new types of data, we just ask that these potential unintended consequences be
kept in mind. Finally, I would like to just note that data
are used for reasons beyond whether we should grant credit. For example, our members may use data not
covered by FCRA, the Fair Credit Reporting Act, including some forms of alternative data,
these new kinds of data that some of you have referenced, for purposes like fraud protection,
identity resolution and verification. Using data for lending and other FCRA-covered
purposes should not be conflated with data that may be used for these non-FCRA purposes. Again, thank you all very much for having
me at this hearing. I look forward to working with the CFPB, our
colleagues in industry and in the advocacy community, to get this right, because if we
do, we’ll be able to help more of those people that the Director referenced in his earlier
statement. Thank you. Thank you. I want to thank all the panelists for your
thoughtful remarks. We now have Gail, Keo, and I now have an opportunity
to engage in some Q&A and try to engender some interesting dialogue. I’ll ask the first question, and I’ll ask
it of you, Francis, if I may. So what effect do you think alternative data
will have on the 26 million credit-invisible Americans and the 19 million who don’t have
enough data in their credit files to have a credit score under the commonly used scoring
models? What effect will alternative data have on
those? And also what effect will it have on those
with existing credit profiles? So I know we’re far from Washington, but I’ll
give you a typically Washington phrase first: it depends. And that’s a really important point because
the strength of our system is that we look at people’s individual situations. So those 45 million people that you’re referring
to, the 300,000 people here, what does their individual circumstance show? We would probably be able to serve more people
as a result of that, but what we don’t know until we really do some research and have
some historical data and everything else, is what are the implications of who we serve,
and, therefore, who we don’t serve, on fair lending and other questions? And my members would be very unlikely to use
information until they really can understand that because while they think that they’re
doing everything correctly and according to the law when they make the decision, we know
from past experience with the Bureau and other agencies that making the decision right on
an individual basis in the past won’t necessarily be viewed when you look at the book of business
in aggregate at some point in the future. So we want to be very careful of that, and
we really owe that to our fair lending and other regulatory obligations that we do so. Keo? [Speaking off mic] So I think that’s an enormously important
question. By no fault of the Bureau, alternative data
is an enormously broad term, and I kind of want to make a binary distinction about what
we’re talking about. On one hand, we’ve heard things like utility,
rental, telecom bill, payment-related behavior. We heard Kabbage talk about payment-related
behavior for businesses on e-commerce platforms. That’s something that I’ve just started calling
mainstream alternative data. And I say mainstream because this is the kinds
of data that the system and regulators are used to seeing. How have you paid your bills? What does your cash flow look like? And I want to hold that in one hand as something
that I have some cautious optimism about with the giant asterisk of all the things that
Chi Chi said in her opening statement about making sure that consumer protection laws
related to utility companies are respected at the state level, making sure that people
aren’t otherwise coerced with even these basic data sources. But I just to want to like call that mainstream
alternative data, how you pay your bills, and I think we have a better sense of how
to deal with that, some of the big credit scoring models are already ready to consider
that. We know how that works. On the other end of the spectrum is probably
the stuff you read about in the media a lot more, which is the social media data, the
web data, the quote/unquote big data, and that’s an entirely different ballgame, an
entirely different ballgame. No one knows—I don’t think I can find a
data scientist in the world that is ready to come and sit here and say if you throw
Facebook data into a consumer credit profile and make a credit score off of that, that
we have any idea how to make sure that that’s in any way, shape, or form fair, that it isn’t
proxying for race, that there is actually some explainable relationship between who
your friend network is and how creditworthy someone thinks you are. We’re not even close to being ready in my
opinion to throw social network data into the mix. And I don’t think that anyone sitting up here
would disagree with that. But I just want to point that out because
sometimes the specter of social media data, of all the big marketing data on the Internet,
is kind of the tail that wags the dog of this discussion in a way that I think distracts
us from the potentially more kind of silver and concrete near-term steps. And I just want to point out that I think
it was early last year, Facebook—many of you are probably on Facebook—put a policy
into place that prohibits any outside party using Facebook data to make eligibility decisions,
and that includes credit and all the other FCRA-covered purposes. So you have, at least in Facebook’s judgment,
for whatever purpose, for whatever their internal reasons are, which I don’t know, but no one
should be using Facebook data to make these decisions. And so if you see startup companies claiming
to do so, think twice before investing. And the very last point I just want to make
here is that one concern that many of us have on the advocacy side of the table is even
when you’re in that comparatively safer territory of bill repayment behavior that we understand,
I still pause because even when that goes into the FCRA-regulated framework, that still
means that whether I pay my cell phone bill or my utility bill is something that could
be visible and used by my employer, and that’s not the topic of today’s conversation, but
there is this kind of shadow of non-credit uses that gets cast across all of this that
deserves some thought. And, David, I believe my microphone wasn’t
on when I asked the question, so I’m just going to restate it for the record. The question was, How do you define alternative
data as it is used in the credit process? And what is the impact that alternative data
has had and could have on consumers? Thanks. You’re not going to restate your answer, right? I hope not. Gail? Thank you. Michael Gardner, my question is for you. What data quality standards for issues such
as accuracy, review, correction, should companies supplying alternative data and companies using
alternative data be utilizing? Great. Thank you for that question, Gail. At Equifax, we take the approach that any
data that will be used in any FCRA-regulated decision should meet very, very similar data
standards. I intentionally did not say “exact” because
the source of the data and the data elements in there are not going to be consistent. Right? So that starts with working with our furnisher
community, those that voluntarily provide the data, making sure they understand the
data that we would want them to be providing on a consistent basis, and giving them the
feedback when that data is not provided in an appropriate format. Then we look internally at Equifax itself,
and then, how do we manage our data? Right? How do we make sure that the data is appropriately
refreshed? How do we make sure the data that is stale
is removed? How do we make sure all of the legal requirements
for FCRA are met on each of our FCRA-regulated databases? Then internally, you also have to look at
use case assessment for how that data will be used. Right? That’s both in building our own products,
and I think Aaron just kind of alluded to potentially how that data might be used as
raw data out in the marketplace. And we manage that in multiple ways. Most notably, we do not commingle alternative
data—so the NCTUE database, as an example—with the core credit file. So those are distinct databases, and anytime
that data will be commingled in a solution, that goes through extensive internal review. And then, lastly, we make sure that any data
that is used in an FCRA use case can be properly disclosed to the consumer and can be appropriately
disputed by the consumer. And so we make sure that those CRAs are managed
under the FCRA. So thank you. Thank you. Amanda, what consumer protections should be
considered when alternative data is incorporated into predictive credit scoring modeling techniques? Thank you for that question. So the CFPB already has two tools in its authority,
the ECOA or—more “D.C.-speak” here, acronyms—but the Equal Credit Opportunity Act and the FCRA,
the Fair Credit Reporting Act. And adding a third, is authority to oversee
and examine banks, and I think that that’s critical for ensuring that the Bureau leverage
that authority to ensure that banks aren’t—or see what the secret of sauce is, to see what
people are developing with the algorithms to see how various consumers perhaps are targeted
or factored into those algorithms. Lastly, I think the CFPB should ensure that
all companies are marketing their products as they should be marketed and aren’t using
them to promote financial opportunity through misleading norms. I think the Bureau should make sure that all
aspects of marketing comply with the FCRA via ECOA and also make sure that various data
aggregators aren’t tapping into or trying to avoid the FCRA by lumping in data based
on myself with a neighbor or others in my community, just making sure that there is
accuracy and privacy with respect to the FCRA. Thank you. Keo. Thanks, David. This question is for Nipun Goel. What role do you think alternative data has
in responsible consumer financial innovation? What are the key factors to ensure that use
is responsible? Thanks for that question. I think the role of alternative data in responsible
consumer financial innovation is finding more ways to responsibly say yes. Right? And whether it’s to open access to folks who
have credit-thin files or credit-invisible files, or to just make a more informed decision
on a standard credit file. What I think alternative data does—and I
think, Francis, you alluded to this as well—is to understand the nuances that exist in a
financial situation that inherently exists for all consumers, and as we see, for small
businesses. And that’s what alternative data and some
of the other sort of buzzwords you hear around these types of things, like machine learning
and Random Forest algorithms, et cetera, I think what it really does, and sort of one
of the sort of comments we always make is it gives you sort of information about not
just the data points themselves, but what’s going on between those data points and understanding
the nuances of those financial situations. As we all know, credit decisions aren’t black
and white, and I think what the alternative data opportunity lies is around sort of helping
reveal some of that gray and being able to penetrate some of those underserved and underrepresented
areas. In terms of from key factors to ensure responsible
use—so I’ll sort of start top down—I think, Michael, you mentioned a lot of the stuff
around accuracy and veracity of the data furnishers, I think we and sort of, Aaron, as you mentioned,
one of the pieces that we’re responsible for, as a financial institution or financial service
provider, is really being rigorous in our diligence before even using a single data
point in a credit decision. It’s very easy for us to onboard a data source
and start pulling that information, like you mentioned. You’ll see marketing campaigns all over the
place saying, “We’re doing this, we’re doing that.” That part is the easy part. The developing and researching takes millions
upon millions of data points and thousands of different model builds before you see an
even single feature used in that. So it’s really around just driving more diligence
and rigorous analysis in that what is still a murky process. But I think doing those things together just
really—and the last piece being just engaging the customer in the process itself. One of the big opportunities we see is that
this does give more control back to the consumer or the end customer as it’s involved in that
lending process. So if we can get all of those things together,
I think it drives again a more informed and a better lending experience. Thank you. Thanks. Gail. Chi Chi Wu, my question is for you. How should companies using alternative data
take into account fair lending considerations? Thank you, Gail, for that question, and it’s
a very important one obviously. If alternative data is being used to underwrite
credit decisions, it is subject to the ECOA, and as you know, the ECOA doesn’t just prohibit
intentional discrimination, it prohibits disparate impact, and it doesn’t just apply to banks
or financial institutions, it applies to all lenders and even some forms of small business
credit. And the disparate impact is whether a policy
has or creates disparities for a protected class—race, gender, national origin, et
cetera. One of the important things to note is that
credit scoring itself actually has a disparate impact. There are many, many, many studies that show
that as a group, certain minority groups have lower credit scores, and the reason shouldn’t
be surprising, it’s what Amanda talked about. Credit scores are a reflection and a measurement,
and one of the things they measure is the disparities in economic health of minority
communities, which has been impacted by 400 years of slavery and Jim Crow and redlining
and legalized economic discrimination, and that has an impact on these communities. That’s why you have the racial wealth gap
where African Americans have 7 cents on the dollar in assets to white families, and the
Latinos have 8 cents on the dollar. And that makes a big difference in paying
your bills because if you have a financial crisis, someone with $100,000 in financial
assets, which is your average white family, will be able to overcome it and pay their
bills a lot better than someone with $7,000 in assets. So credit scores—and so the thing is credit
scores aren’t the only thing that is going to reflect those historical inequities, a
lot of data sources will reflect that. If you have trouble paying your credit card,
you’re going to have trouble paying your cell phone bill. And there’s this concept Amanda alluded to,
structural racism, where the very institutions and systems in our society carry forward that
racism, not because of any animus, although unfortunately I’ve seen a lot of that recently,
not because of any animus, but because systems replicate themselves, and so it will get reflected
in the data. So is this hopeless? No. I mean, credit scoring is legally used for
credit, and that’s because the ECOA test is a three-part test. And the second part is whether there is a
legitimate business justification for a policy that creates a disparate impact. And for credit scoring, the justification
is it’s proven to be predictive, it’s empirically sound, statistically derived. There’s actually a test in Regulation B. And
so any alternative data source is going to have to pass that same sort of rigorous level
of being proven to be predictive and empirically derived and statistically sound. And by the way, there is a third part. Is there a less discriminatory alternative?
which we don’t really focus on as much, and there should be more focus. Is there a way to get rid of this disparate
impact? And really that’s the brass ring here for
alternative data and any data. Can you find the data source that doesn’t
have this structural racism baked in? That’s I think what we should be looking for. I think, Director Cordray, you had a great
point talking about looking at data that’s forward looking and not so much a measurement
of the past. I think that, because the past is infused
with this historical discrimination. Thank you. So let me ask one final question for all the
panelists. This is really a summary kind of wrap-up kind
of question, and ask you for any additional thoughts you want to add as to some of the
benefits and the risks that using alternative data has in the credit decision process. And we’ll just go in reverse order from when
we started, so we’ll start with you, Francis, and go down the line this way. Great. Thank you. Just to sum up, I think what I would say is
that the possibility of alternative data should be looked into two different areas, the kind
of traditional pieces, the way Aaron described it, and the more innovative pieces. On the more traditional pieces, we have to
have a better understanding of what the long-term impacts are going to be in fair lending and
other areas before we’ll use it. We have obligations under the law, and we
want to follow those obligations. That competes with this idea that we want
to serve more people. We want to get the unbanked into the system,
and those are at tension, and we have resolve that tension together if we’re going to make
this worthwhile. 45 million people is both a problem for our
society, that we’re not serving them, and it’s also an opportunity that we can go out
and help those people get the financial services products that we need. We have to overcome that struggle, though. Yeah, I would sort of echo a lot of the same
points. And I’ll go back to I still think at its simplest,
it’s being able to provide consumers and other end customers with a voice that’s in the lending
process that traditionally has not existed there before. From a risk side, again, as I mentioned, just
ensuring the rigorous analysis and diligence that’s needed when onboarding a new data source
or continuing to assess and regulate existing data sources that are in the market now. Thank you. Michael. Thank you. Benefits I think have been well stated by
my colleagues here. I won’t reiterate those. I think one of the risks that we need to be
vigilant about, beyond the ones that have been brought forward already, is this idea
of really understanding the long-term impacts of leveraging new alternative data in any
decision. I mentioned earlier in my comments that we
have the privilege of access to the NCTUE database at Equifax. We have had that data and have been modeling
that data for over 10 years, and really only in the last couple of years moved that data
into solutions that are in the marketplace, in part, because we want to understand those
longer term impacts. We want to understand how a consumer who is
present in the NCTUE data file becomes present in the credit file, and understanding that
consumer journey and being able to facilitate that. So that’s one risk I think we need to be vigilant
about, is understanding longer term impacts because, as Nipun said, getting the data and
throwing it into an algorithm is easy, we all have great IT departments that can crunch
the numbers for us—right?—but it’s really understanding the impacts there. And we, as an NCRA, have sort of a dual obligation
because we need to be able to educate our end customers, financial institutions, and
others on appropriate ways to use the data, and, of course, we have our own FCRA requirements. So thank you. Amanda. So the CFPB should be commended for its two
reports in the credit marketplace, and one, which has been referenced here today via the
Director’s remarks and by some of us on the panel, on credit invisibles. And if I could, I’ll just read a line from
that report, which also the Director cited as well in his opening remarks. “In 2015, we published a report finding that
26 million Americans are credit invisible. This figure indicates that 1 in every 10 adults
does not have any credit history with one of the three nationwide credit reporting companies. The report also found that black and Hispanic
consumers and consumers in low-income neighborhoods are more likely to have no credit history
or not enough current credit history to produce a credit score.” Clearly, the primary benefit of alternative
data, if works as promised, would be to raise credit scores of credit invisibles and open
up a financial opportunity, but if alternative data is implemented unfairly or inappropriately,
it could harm the consumer. Thank you. Thank you. Aaron? So I just wanted to pick up and highlight
a few threads of conversation in closing. The first is what Chi Chi just said about
even today’s credit scoring system kind of baking in uncomfortable gaps in wealth and
privilege in our society. I think that’s a really important place to
start that’s tolerable as a matter of policy for a lot of the reasons you mentioned, but
I think it’s just important to keep in mind that that’s the foundation we’re building
from. And even if you just take the easier datasets
like bill repayment behavior that we understand and throw it into the mix, that isn’t necessarily
immediately going to help the 45 million people that don’t have good visibility in the system. And so I think we need to be really thoughtful
about how we do that. Again, the touchstone is, What helps the people
we’re trying to help? And this is a situation in which I think some
careful additions of new data could be really useful, but it’s certainly not the case that
more is better. More data doesn’t equal more innovation in
a helpful way. Like I said, there’s a whole spectrum of stuff
that people are talking about that’s just all smoke and no fire. Anytime someone talks about social media data
or web browsing behavior or your online shopping behavior going in the credit scores, please
hold that separately, that’s a whole separate world. I think it’s an interesting question of, if
we bring new data into the formal credit reporting infrastructure, whether or not it’s possible
for there to be an industry standard of holding that data separate for credit decisions. I think that there’s been a lot of talk about
how all of this data is studied really, really carefully before it’s applied to these decisions. I think that if we’re going to bring new data
into the conversation, it makes sense to say let’s start by using this for credit decisions,
and maybe we’re not ready for employment, rental applications, and insurance yet, and
I think that would provide some comfort to at least me personally. And the final point I want to make is there
has been discussion about making algorithms transparent and being able to audit algorithms. The easiest way to get your arms around what
an automated decision-making system is doing is understanding the data that’s going into
it and what you’re trying to measure. And there does become a point where the complexity
gets so high or the data becomes so tangential that it becomes really, really hard to ask
rational questions of that system, which is I think another reason to just take this slow
one step at a time. Thank you. Okay. Chi Chi, you get the last word. Oh, cool. Thank you, David. So I want to actually take a step back from
the question. You know, the question is risks and benefits
to credit decisions, and I want to talk about the credit itself because I think it’s really
important. We have been talking about expanding access
to credit without qualifying what kind of credit. And it’s I think really important to remember
we want the credit to be affordable. Some of the fintech lenders, not anybody at
this table, but some of the ones we’ve seen have triple digit APRs, and really I’m not
sure that’s going to benefit anyone. Right? And also even affordable credit, you want
to make sure the consumer can afford that credit. Let’s go back in the “way back” machine, 10,
15 years ago, or early 2000s, mid-2000s, there was lots of access to credit. People could get credit without stating their
income or their assets. And when we, consumer advocates, said, “Hey,
this is not a good thing,” we were told we were against the democratization of credit,
and how dare we be against democracy itself? Well, we know how that story ended, right? So the touchstone for credit should always
be ability to repay and creating alternative data to create a credit score, whether it’s
a traditional alternative credit score, that’s only one part of the equation. The other is the consumer has to be able to
afford that credit. And one of the interesting things about some
of the data sources we were talking about is it might be able to incorporate that. Director Cordray, you mentioned bank account
information, and that incorporates both payment data, but also income data, so there might
be some—and Nipun Goel mentioned the cash flow. Cash flow obviously is important to repay
credit obligations. Now, that’s why it’s great that the Bureau
has taken this interest in letting consumers use third-party aggregators, because I’m not
sure it’s a great idea always to have lenders looking directly at the transaction level
data because there could be some privacy concerns over there. So having some sort of intermediary to protect
the privacy of consumers, also to give them control so they can opt in and decide to do
this or not to do this is important. And, finally, I think if we’re going to use
things like bank account data, we really do have to deal with the overdraft issue. But I think that there is a potential there,
and certainly I think having 6 months or a year of bank account data is probably more
useful information than a 3-year-old collection account for a medical debt where you disputed
with the provider over the copay. Thank you. So this concludes the panel portion of our
program. And I’m going to ask you to join me in thanking
all of our panelists for a thoughtful discussion. And let me now invite the panelists to retake
their seats and turn the program back over to Zixta Martinez, our Associate Director
for External Affairs, who will moderate the next portion of the field hearing. Thank you, David. And again thank you to the panel of experts. An important part of how the Bureau helps
consumer finance markets work is to hear directly from consumers, from industry, from state
and local partners, and, of course, from community advocates across the U.S. One of the ways that we gather public feedback
is through events like these. We’ve held field hearings, town halls, and
other public events across the U.S. from Miami, Florida, to Itta Bena, Mississippi, to Seattle,
Washington, and at these events, we not only hear from experts in the field, we also invite
the public to participate. But before I open the floor up for public
comments, I want to remind folks that there are several other ways to communicate your
observations, your concerns, or complaints to the CFPB. You can submit a consumer complaint with the
CFPB through our website at consumerfinance.gov. Our website will walk you through that process. Or you can call 1‑855‑411-2372. We take complaints about mortgages, car loans
or leases, payday loans, student loans, or other consumer loans. We also take complaints about credit cards,
prepaid cards, credit reporting, debt collections, money transfers, bank accounts and services,
and other financial services. If you don’t have a specific complaint but
would like to share your story with us, we have a feature on our website called Tell
Your Story, where you can tell us your story, good or bad, about your experience with consumer
financial products or services. Your story will help inform the work that
we do to protect consumers and create a fair marketplace. We have another feature called Ask CFPB, where
you can find answers to over 1,000 frequently asked questions about consumer financial issues
as well as additional resources. We also have a Spanish language website called
CFPB en Español, where you can find answers to consumers’ frequently asked questions and
additional consumer resources. So I encourage you to visit consumerfinance.gov
to learn more about the resources and tools the Bureau has developed to help consumers
make the best decisions for themselves and for their families. So now it’s time to hear from members of the
public that are here today. A number of you signed up to share comments
and observations about today’s discussion. The public comment portion of the field hearing
is also an important opportunity for the Bureau to hear about what’s happening in consumer
finance markets in your community. What we hear from you is invaluable. We would like everyone who signed up to be
able to speak, so we encourage you to take about 2 minutes to share your thoughts and
observations with us. And I will call our first public commenter,
and that is Jonathan Marshall. Peter will bring you
a microphone. So my name is Jonathan Marshall. I’m a consumer rights lawyer here in West
Virginia, in Charleston. And I found this discussion pretty interesting. Kind of my role over the last couple years,
at least here, has been down at the legislature. We’ve faced some pretty unprecedented attacks
on existing debt collection regulations that have been in place for many, many, many years
here. And over the last couple years, we’ve been
able to work with industry, work with banks, to come to reasonable compromises. But unfortunately again this year folks are
back again and asking for more. As I see what the CFPB is trying to do with
respect to alternative credit, I understand the need for the availability of credit, and
that’s important, but I think that that needs to be balanced with existing both Federal
and state debt collection protections. I think the last panelist or the last commenter,
commentator, here talked about it’s not just about alternative data, and you have to look
at those three C’s, right? It’s capacity, too, right? It’s character and it’s collateral. And I would hope that the CFPB, as it moves
forward in this area, does keep in mind those important consumer protections that exist,
both at a Federal level and at a state level. Thank you, Mr. Marshall. Sam Vallandingham? All right. Patrick Walker? Thank you. Just a few comments. I would like to underscore the need for alternative
data and new solutions. It was alluded to a little bit, but I think
that the numbers are stark. From the CFPB’s own data point on credit invisibility,
while the overall rate of unscorability was 19 percent nationwide, in the lowest income
census tracts, it’s 45 percent. So it’s not kind of a small or minor problem
in some parts of the country, it’s actually a very large issue. And so I would just like to point that out
to note that the status quo has some very important gaps that we definitely need new
solutions. Secondly, PERC has done some of our own research
where we looked at the lowest income census tracts, and what we find is that when you
do add in the alternative data, that the 30, 35, 40 percent rates of unscorability falls
greatly. Depending on what score you use and what solutions,
it can fall as low as a few percentage points. And those individuals are not just coming
into the system with subprime scores, they’re coming in, in many cases, with prime scores,
620s or above. I would also like to comment on one of the
notes that we need to look at the longer term impacts of this data, and, of course, it is
completely the case that we need to be prudent with new data and new solutions. I would like to point out, though, that there
have been a number of utilities that have been reporting to the CRAs for decades now,
so the data is out there. We don’t need to have an experiment that will
go into the future for many, many years and then look at the outcomes. We actually have individuals that were new
to the credit system with alternative data back in the late ’90s, in the early 2000s. So you could actually look at those individuals. PERC has done a little bit of that work and
we didn’t find any decrease in credit scores after individuals access credit from alternative
data; in fact, we found score increases over time similar to a control group. So if that is an area of concern, that need
not hold up the transition to new solutions. We can look at data that’s already been reported
for years. Thank you. Thank you, Mr. Walker. Chris Arthur? First of all, I want to thank everyone for
taking the time to come to the great State of West Virginia. My name is Chris Arthur. I’m general counsel for the West Virginia
Division of Financial Institutions. I see benefits and I also see concerns regarding
the use of alternative data. In particular, I’m worried about discriminatory
practices and how massive this information may be. I even think something like utility—not
utility bills, but cable bills or telephone bills is something very concerning to me because
a lot of people look at that as it’s not a must, it’s a luxury, and I know people that
live close to me that they may not pay their phone bill because they would rather use that
money to do something for their children, but the rest of their credit, they pay their
rent, they pay their other bills, but that’s a decision. They may do the exact same thing with a cable
bill, they may have something that they find is more important, so they’ll miss their cable. And I could see that being more of a detriment
and a negative to people who are really trying to build a good credit. The other thing, let’s face it, technology
is really moving fast, and the amount of data that you can get, especially like Facebook
or social media, I can see a lot of discriminatory practices. So I think we have to weigh this and limit
its use and make sure that we make good decisions when alternative data is useful to build credit
versus alternative data that may be used in a very negative way, including a discriminatory
practice. The State of West Virginia is very poor, and
there are a lot of poor people, and a lot of poor people who have worked very hard,
but, like I mentioned, they make decisions based on what’s in the best interest of their
children or their family, and some of the things that were discussed today may be a
detriment to them building credit. But, again, thank you for your time. Thank you, Mr. Arthur. Bren Pomponio? Thank you. I wanted to first express our appreciation
for the work that the Bureau has done in protecting consumers. I work at a nonprofit legal service organization,
and many of our clients have seen the benefits of the Bureau’s actions, whether it be consent
decrees that give them new rights and enforcement actions that show tangible benefits to them
in terms of returning money. And so we’re thankful for the work that the
Bureau does. My comments on the alternative data issue
come from representing low-income West Virginia consumers for more than 15 years in a variety
of contexts. And West Virginia has the highest incidence
of homeownership in the country, but it’s also one of the lowest states in terms of
economic and poverty. And so at this intersection of high homeownership
incidence and low socioeconomic incidence is a fertile ground for predatory lenders
in the past. And I would like to hope that when considering
the alternative data, that we make sure that the safeguards are in place that doesn’t allow
in alternative facts. I’m thinking about going back to before 2008,
the no doc loans where a lot of West Virginians’ income was falsely inflated because there
wasn’t sufficient safeguards in the underwriting process to ensure that the information that
was being used to make the credit decision was accurate. And this hurt people because it put them into
loans that were secured by their home, converted unsecured credit to secured credit, in which
they couldn’t pay because their income source was not sufficient to cover the monthly payments. So I would just ask that the accuracy of that
data be considered when the underwriting decisions are made. Thank you, Mr. Pomponio. Margot Saunders? Hi. As you know, I’m Margot Saunders, with the
National Consumer Law Center. I have seen a number of instances where the
promise of the use of alternative data, particularly rent payments, through one of these new rent
reporting agencies for rent has been used as the premise for pulling people into very
dangerous predatory credit transactions. And I think we’ve seen it also with some payday
loans. So I would just urge you to, while you investigate
the viability of using alternative data to boost credit scores, that you ensure that
you don’t allow this vehicle of using alternative data to improve credit scores to be the means
by which people are pulled into dangerous credit. Thank you, Ms. Saunders. Linda Frame? Good afternoon. My name is Linda Frame, and I work for the
West Virginia Center on Budget and Policy. We research economic policies to determine
what can be done at both the state and Federal level to give regular people in our communities
a shot at a decent job that earns a decent wage and allows for a decent quality of life. I would like to thank the CFPB for visiting
Charleston today. And we are here to express our appreciation
for its hard work and also the hard work of the Attorney General’s Office here to protect
our families, in particular, from payday lenders. Today’s field hearing is a great explanation
and exploration of how to expand credit opportunities for people. Protecting those who are credit invisible,
however, from payday lending is also a very important component of the CFPB’s work. My organization is working with states across
the Nation where payday lending is illegal, as it is here in West Virginia. Last year, Director Cordray may recall, we
all gathered in D.C. to present him with a whole pile of cards, postcards, signatures,
petitions, from all of our states where we do not have payday lending. All total, there are 90 million people in
the United States who live in states where we do not have payday lending, and we hope
that this can grow. Experience from our states—well, the states
that do have payday lending—has shown that allowing payday lenders to do whatever they
want does not benefit people. In fact, according to the Center for Responsible
Lending, keeping payday lenders out of West Virginia saves our residents $48 million every
year in payday lending fees. This has helped our families from falling
into the debt trap caused by payday loans. So thank you for allowing me to stray a little
bit off topic to thank you all for coming to Charleston, and we hope you will continue
your good work and help us preserve West Virginia’s strong tradition of consumer protection and
banning payday lending here. Thank you. Thank you, Ms. Frame. And thank you to all that provided thoughtful
public testimony today. Thank you to the audience, to the panelists,
and to all those watching via livestream at consumerfinance.gov. This concludes the CFPB’s field hearing in
Charleston, West Virginia. Have a great afternoon.

Leave a Reply

Your email address will not be published. Required fields are marked *