“Alternative Data,” and “Big Data,” have become
buzzwords throughout financial services over the past few years. While
there are varying positions one can take on the value of incorporating
Alternative and/or Big Data into credit reporting and credit underwriting,
it is clear that there are now new data sources that did not exist (or
were less accessible) in the past.
What is Alternative Data?
The term “Alternative Data” refers to data that is structured
and resides in databases outside of the Nationwide Consumer Reporting
Agencies (NCRAs). This data is generally not visible to most financial
lenders when they are evaluating the consumer’s credit worthiness.
The data is generally accessed by association members such as telecommunications
companies, utilities, etc.
What is Big Data?
Big data refers to high volume and high velocity data which tends to be
more unstructured and can be collected through streaming, or recording
information about a consumer invisibly. Often this information includes
web browsing, GPS locations, internet preferences, television preferences
through companies like Netflix, and Amazon, channel surfing collected
by cable providers, computer cookies, etc. There are also efforts to mine
information in Facebook, Twitter, Instagram and other applications.
The mining and use of this data holds some attraction for lenders as it
can help to tailor products to specific customer needs and preferences,
or can potentially assist in making credit decisions that better assess
risk for populations that may have historically had thin or no credit
files at all (known as the credit invisibles).
However, using these data sources pose challenges to financial institutions
as the existing regulatory structure does not yet explicitly consider
use of these different data sources. There are also challenges with using
these sources in meeting current regulatory compliance requirements. This
has created uncertainty. Some larger established lenders have chosen to
“stay away” for the time being. The logic is ‘the pioneers
are the ones with the targets on their backs’. Other established
lenders are experimenting but not implementing models using these new
data sources.
This blog post will explore some of the specific regulatory considerations
that are associated with the utilization of Alternative and Big Data –
specifically, touching on requirements related to the following items:
- Transparency
- Accuracy of Consumer Information and Right to Due Process (FCRA)
- ECOA Non-Discrimination
- Privacy
Transparency & Accuracy of Consumer Information and Right to Due Process (FCRA)
The Fair Credit Reporting Act (Regulation B) requires adverse action notices
to be sent if adverse action (application decline or increased rate/ terms)
was taken based in whole or in part on information in a consumer report.
Included in the adverse action notice are reasons for which the adverse
action was taken (e.g. “proportion of loan balances to loan amounts
is too high”), to provide the consumer with a specific reason as
to why they were denied, or, offered less favorable terms, on the credit
product they applied for. The reason codes provide transparency to the
consumer. They also help the consumer to recognize and correct errors
in their report that may have led to the adverse action. When firms use
unusual types of data for underwriting credit, it raises compliance questions
in credit decisioning when there is an adverse action. It may also lead
to situations where the consumer is unable to dispute the information
or have it corrected.
An example of challenging adverse action reasons can be found with the
use of mobile phone data to make credit decisions. In Sub-Saharan Africa,
a lender has determined that the count of cell phone towers that an applicant’s
phone hits in a given month can be a predictive variable for loan repayment.
The number of phone towers in that model serves as proxy or correlate
for mobility of the applicant, which is interpreted as predictive of credit
worthiness. If a lender were to try and use this model in the current
US regulatory regime, would a decline reason code of “insufficient
cell phone tower pings” be acceptable? How would a U.S. consumer
respond to such a reason code?
This example highlights one of the inherent challenges that come with using
Alternative or Big Data for credit decisioning with regards to transparency
principles that are contained in US consumer protection laws. The same
example described above also raises questions with regards to accuracy
of consumer information and the right to due process principles, which
are also included in the FCRA.
What happens in the Sub-Saharan Africa example above when the consumer
disputes the information the credit evaluation was based on? In this scenario,
where a consumer was denied access to credit based on the “insufficient
cell phone tower pings,” what could the consumer do if, for example,
they have two mobile phones or if they traveled to areas that did not
have any towers at all? How would they be able to dispute this information
and how would it get corrected? Who would own the correction and how would
the relationship between the telecom company and the lender work? Presumably,
does providing the data to the lender make the telecom company a Consumer
Reporting Agency subject to all associated legal requirements?
Non-Discrimination
Perhaps the most straightforward of the requirements to understand is the
applicability of non-discrimination principles – or fair lending
requirements – with Alternative or Big Data usage. The Equal Credit
Opportunity Act prohibits discrimination in credit transactions based
upon prohibited factors (e.g. race, national origin, sex, etc.). Discrimination
does not necessarily need to be overt, as disparate treatment (basing
a lending decision on a prohibited factor) and disparate impact (a lending
policy having a disproportionate impact on a specific population) also
create fair lending risks.
When developing models which use Alternative or Big Data (or any data for
that matter), lenders should validate that their models are not creating
disparate impact. This could be overlooked by smaller, leaner organizations,
who may be more likely to use Alternative or Big Data sources in their
desire to gain a competitive edge in the market.
Privacy
Lastly, privacy concerns come into play with the utilization of alternative
models in credit decisioning. Some marketplace lenders are experimenting
with social media as a data source for credit models. Here the question
is raised: Do consumers have an opportunity to define what they consider
being personal, private, and out of reach data for these models? While
certain data can be found outside of the CRAs is clearly public record
(e.g. Court Cases, Criminal Records), the usage of Alternative or Big
Data begins to blur the lines as to what is public and what is private.
Social and Mobile data come to mind as good examples. While disclosures
in many cases clearly define how data is used, as data capture and data
sharing increases in the future, privacy matters are likely to become
increasingly scrutinized.
There is also a broader concept regarding privacy. Is privacy an area that
can be regulated, or is privacy a social construct that changes by generation,
location and culture?
Moving Forward
Does this mean that use of Alternative or Big Data is not acceptable or
valuable in credit decisioning? Consider emerging Millennials who pay
their cellular, utility, cable and rent regularly and on time. These individuals
often tend to have no traditional credit record at the NCRAs to demonstrate
their reliable payment behavior. They are part of a group referred to
as the “Credit Invisibles”. In addition, these industries
collect payment data for self-use and only use the NCRAs through third
party collection agencies to report late bills in collections. Leveraging
this alternative positive payment data in the traditional NCRAs could
effectively open this market to potentially 26 million consumers.
In the U.S. there is interest by regulators and government agencies to
find safe and sound ways for the Credit Invisibles to enter the credit
market deservingly, supported by demonstrated good payment behavior.
The focus on use of these types of data will likely rise from our standpoint,
with the ultimate goal of ensuring Alternative and Big Data is used in
a way that complies with the current laws. There is opportunity for Lenders
looking to use Alternative and/or Big Data to expand into new credit segments
such as the Credit Invisibles. We recommend that they should do so carefully,
with thoughtful testing plans, compliant disclosures, well-documented
processes, and support of their mission and goals from their specific
regulators.