Evaluating the credit risk of non-traditional Borrowers:

Non-traditional data can augment credit bureau-based scoring

techniques helping lenders discover profitable prospects and more
accurately evaluate the creditworthiness of applicants.

LexisNexis and the Knowledge Burst logo are trademarks of Reed Elsevier Properties Inc., used under license. ©2006 LexisNexis,
a division of Reed Elsevier Inc. All rights reserved.

September 5, 2006
Page 1 of 12
Introduction

Lenders today must evaluate the creditworthiness of individuals and businesses as

quickly and accurately as possible. Traditionally, they have relied on evaluations that
use a person’s credit history, as maintained by one of the major credit reporting
agencies to create a score that indicates how likely the person would be to default on
their loan. But the standard method sometimes falls short of doing the job because not
everyone has a robust history of using credit or alternatively may have a credit history
that is not reflective of their true creditworthiness. Consequently, more and more
lenders are turning to non-traditional sources, such as public records, to substitute for or
enhance traditional credit bureau histories – especially for applicants who have rarely
used credit in the past (in other words, those having “thin”1 or non-existent files). Non-
traditional credit scoring relies on the detailed analysis of positive and derogatory life
events, evidence of assets and address stability, presence of criminal convictions, liens,
judgments or bankruptcies, positive identify verification, and high-risk pattern
recognition.

While non-traditional credit scoring helps lenders weed out higher risk credit applicants
from pools of credit seekers, it also lets them find untapped pools of creditworthy
consumers who may be rejected by credit-bureau based scoring methods. For these
reasons, six of the top ten bankcard issuers, three of the top four wireless providers,
and many other lenders have added non-traditional credit scoring to their new account
evaluation process. They are finding that, much like the scoring tools derived from
traditional credit bureau data; it is an effective tool for predicting future consumer
behavior.

1
Most lenders consider a consumer with three or fewer credit tradelines as having a thin file.

September 5, 2006
Page 2 of 12
Market environment

Finding new creditworthy applicants is vital in order for a lender to achieve growth, but it
is increasingly difficult to do in today’s competitive lending environment. The average
credit user carries about nine cards in his or her wallet; nevertheless, each cardholder’s
household receives approximately 30 additional solicitations for credit cards annually.2
Students are becoming credit users at an early age: 83% of undergraduates have at
least one credit card, with 32% having four or more. By the time they graduate, they’ve
tripled the number of cards in their wallets and doubled their average credit card
balance.3

Of course, card balances represent only a fraction of total indebtedness: there is also
debt for cars, homes and education. Figures from 2004 reveal that nearly 75% of
Americans carry credit cards, with about half maintaining a balance on their cards. The
median number of bankcards per person is two, and the median credit limit is $13,500.

Total debt, including mortgage debt, reaches nearly 100% of post-tax annual
income.4

The pool of traditional creditworthy consumers is nearing a total saturation point, leaving
little to no room for borrowing. Therefore, credit issuers are looking for underserved or
untapped markets: people who traditionally have relied on cash. New immigrants, young
people new to credit and others with a cultural bias against credit offer the most
potential for lenders seeking growth. However, many of these applicants have “thin” or

2
“Digging Your Way Out of Debt,” Paul J. Lim and Matthew Benjamin, U.S. News and World Report,
3/19/2001 (from “Credit Card and Debt Statistics Database,” Scott Bilker, Debtsmart.com, April 2006).
3
“Undergraduate Students and Credit Cards: An Analysis of Usage Rates and Trends,” developed by
Nellie Mae, April 2002.
4
“Credit Card and Debt Statistics Database,” Scott Bilker, Debtsmart.com, April 2006.

September 5, 2006
Page 3 of 12
no files, making scoring methods based upon traditional credit bureau data inadequate
to determine their “true” creditworthiness.

Lenders hope these emerging populations, or even sub-prime segments with poor
traditional credit scores, will hold the key to profitable portfolio growth if properly
analyzed using non-traditional credit scoring. They want these scores to limit exposure
to risk by identifying profitable customers often missed by standard analysis that relies
on credit bureau history. The rewards for tapping into these markets, estimated to be
about 50 million creditworthy individuals,5 can be very high providing above market
returns.

Traditional Credit Scoring

Traditionally, managing credit risk has relied on the maxim that how a person has used
credit in the past is the best predictor of how they will use credit in the future. In other
words, if a person or business has paid its bills on time in the past, they are likely to pay
their bills on time in the future. The more data one has about a consumer’s credit history
and the longer their experience using credit, the more likely it is that a creditor can
predict future bill-paying behavior.

The scores that determine a consumer’s creditworthiness derive from the consideration
of many things: whether bills are paid on time, the amount of debt carried, the amount of
credit available, the frequency of seeking credit, debt-to-income ratios, and the
existence of any delinquent bills. What they can’t take into account in a timely fashion is
that circumstances change: a person can move or lose his or her job, a family can be
destabilized by illness or divorce, a wage earner can be imprisoned or die, and priorities
can change with new responsibilities (such as children or aging parents). While not
sensitive to all of these conditions, public records and other non-traditional data sources
can reveal aspects about a person or a business that can balance or enhance standard

5
“Lenders ‘Holy Grail’,” www.fairisaac.com, April 12, 2006

September 5, 2006
Page 4 of 12
credit histories, and can provide relevant information when credit bureaus know nothing
about a given consumer. For example, non-traditional credit data can tell a lender
about asset ownership, presence of derogatory filings, address stability and tenure,
utility listings, and other factors that differentiate the credit worthiness of otherwise
indistinguishable credit bureau no hits.

A score that relies on credit-bureau history is effective for most credit users, but it falls
short of clearly evaluating a few important groups: young consumers who have not yet
had access to credit, and consumers who do not have positive credit history. Although
these consumers may have sufficient income to be considered highly creditworthy, the
traditional method of predicting their likelihood to default may deny them access to
credit. The data demonstrating this potential credit worthiness reside in public records,
and non-traditional credit scoring makes this analysis possible.

But what is public record and non-traditional credit data, and what does it reveal that
credit bureau data might not? Public records constitute all the documentation that
reflects on an applicant’s life (or the life of a business) and is provided by a government
entity that is open to public scrutiny. They usually are recorded by a government agency
for purposes of identification, taxation and, when applicable, prosecution. They include
home addresses and phone numbers; birth certificates and death certificates; marriage
and divorce records; immigration or citizenship records; real property deeds, mortgages
and liens; personal property titles and tax records; military records, criminal records
and more. Privacy and safety concerns have recently reduced public access to vehicle
registration data, driver’s license data, and social security numbers, but they constitute
public data, as well. This data resides in the public domain and can round out
information about individuals and businesses missing from credit bureau data.

Public records and non-traditional data sources have always been available, but they
have not always been easily accessible. Reviewing them required travel and footwork,
manual research and a tremendous amount of patience. Although electronic

September 5, 2006
Page 5 of 12
documentation and the internet have made these data records more accessible, it is
LexisNexis that compiled, integrated, digitized and indexed them for precise,
comprehensive searching.

Enhancing Credit Scoring with LexisNexis Non-Traditional Credit Data

As a global leader in comprehensive and authoritative legal, news and business

information, LexisNexis is also the industry leader in compiling and searching public
record data. Using this data, LexisNexis developed scoring products to predict the
creditworthiness of individuals and businesses, especially those without strong credit
histories. The company’s databases contain more than 4.8 billion searchable
documents, with 7 million searchable documents added weekly. Its credit scoring
solutions are empirically derived from more than 35,000 nationwide data sources, which
document judicial, county courthouse, real and personal property ownership, and social
security issuance and deceased information.

In August 2004, LexisNexis gained additional data sources and an advanced

supercomputing platform when it acquired Seisint, Inc., of Boca Raton, Florida, along
with its preeminent database of public records, Accurint. Now, the LexisNexis arsenal
includes electronic directory assistance (updated daily); property, deed, mortgage, and
tax assessment data; criminal records, liens and judgments, professional licenses,
drivers license and voter registration (for certain permissible purposes); and business
data, such as Secretary of State, UCC Filings, and Yellow Page Listings.

Logic behind Non-Traditional Risk Assessment

The single most important factor in the success of any predictive score is the underlying
power of the data used to build it. LexisNexis has access to a wealth of dynamically
updated public record and non-traditional data that contribute to its predictive scores,

September 5, 2006
Page 6 of 12
making them exceptionally effective either as an alternative in the case of thin or no file
applicants or as an enhancement to traditional credit bureau-based credit scoring.

These non-credit data sources record historical information about an individual that is
analogous to traditional credit history. Just as credit reports record derogatory credit
events (delinquencies and charge-offs), public records chronicle certain life events that
often are correlated with poor credit worthiness. Liens (especially unreleased liens),
judgments, criminal convictions and bankruptcy filings are useful in differentiating higher
risk applicants. Similarly, evidence of property ownership, property value, and evidence
of vehicles and other licensed assets provide relevant information about the economic
lifestyle of an individual.

LexisNexis also documents the address history of an individual, as well as the phone
listings for those addresses. Having a history of stable addresses with phone listings is
evidence of responsible housing and utility payment – phone service will be
disconnected if unpaid, and unpaid housing results in eviction or foreclosure. Credit
bureaus do not typically record rent and utility payments, but non-traditional data
sources do record when phone service is disconnected. Address stability is also
evidence that a credit applicant pays their rent or mortgage responsibly.

Alternatively, residence in transient commercial addresses, such as hotels, motels, or

campgrounds, is often an indication of higher risk. Similarly, residence in an institution
or correctional facility also raises red flags for risk, as would an indication that a person
applying for credit is actually showing up as deceased in public records.

The LexisNexis scoring methodology resembles credit-scoring methods in that the

prospective borrower provides identifying information during the credit application
process. The creditor then crosschecks this data against known data about the
borrower. Instead of checking credit information, however, LexisNexis reviews public
record and other non-credit data, including driver’s license and motor vehicle data, and

September 5, 2006
Page 7 of 12
property information, looking for confirmation or inconsistencies in records that include
name, address, phone number, and social security numbers.

Other scrutinized data points include mortgage data, real property, tax assessments,
the purchase or sale of real property (including dollar amounts), and vehicle registration
(limited data). This scoring approach also utilizes evidence of positive and negative life
choices – such as frequent moves, disconnected utilities or a criminal record – that
might suggest a person’s level of stability and responsibility.

The LexisNexis’ methodology produces a three-digit score indicating the level of the
applicant’s credit risk. Scoring can be done machine-to-machine in real time, interactive
over a web-based application or in batch runs and can be completed in milliseconds.
The creditworthiness predicted by non-traditional data and analytics correlates well with
results returned by traditional credit scoring, so that a credit issuer can identify
creditworthiness with these scores about as confidently as with traditional credit bureau
scores.

Overview of LexisNexis Risk Scoring Products

LexisNexis’ solution for scoring non-traditional populations is LexisNexis RiskView,

which provides a three-digit credit risk score to predict an applicant’s likelihood of
default over the first two years of an account’s life cycle. RiskView can predict
creditworthiness for applicants that have little or no credit history, as well as those with
extensive credit history. It can serve both these groups of applicants because it does
not use traditional credit history data to assess risk. Instead, it relies on the detailed
analysis of positive and derogatory life events, evidence of assets and address stability,
presence of criminal convictions, liens, judgments or bankruptcies, positive identify
verification, and high-risk pattern recognition derived from non traditional data.

September 5, 2006
Page 8 of 12
Case Studies

LexisNexis works closely with many of the largest lenders and financial institutions to
implement non-traditional scoring solutions. LexisNexis has a suite of credit scoring
tools that effectively evaluates the creditworthiness of applicants by inputting only
standard identity elements (Name, Address, Phone, SSN & Date of Birth). These
scoring tools can be used in almost any environment such as retail card, credit card,
telecom and automotive finance, as the following case studies demonstrate.

The retail credit card market is traditionally known as high risk and high reward.
Creditors must weigh their losses against their opportunities for gain, factoring in the
profit they make on the goods they sell. This allows retail creditors to accept a much
higher level of risk than non-retail issuers of credit, such as most commercial bank
lending. But how can the retail creditor determine which group to continue offering
credit, and which group to deny credit? The LexisNexis’ scoring solutions provide clear-
cut insights into the decision (see Table 1).

Table 1: RETAIL CARD

RiskView Cumulative Bad
Score Accts % of Accts Rate K-S

< 653 9,331 9.3% 20.6% 16.2

653-659 8,866 18.2% 16.0% 25.9
660-665 10,244 28.4% 12.1% 31.8
666-670 10,265 38.7% 8.7% 32.9
671-674 8,509 47.2% 7.1% 31.8
675-680 11,480 58.7% 5.7% 28.3
681-686 9,762 68.5% 4.1% 23.2
687-695 10,597 79.1% 3.2% 16.3
696-708 10,119 89.2% 2.8% 9.2
709+ 10,827 100.0% 1.7% 0.0

Total 100,000 8.0% 32.9

Based on a population of 100,000 retail credit card users, 7,950 or 8% were considered
“bad” credit risks based on serious delinquency or write-off after eighteen months of
experience. The RiskView score indicates the predicted credit worthiness of an
account. High (good) scores are associated with lots of positive public records and

September 5, 2006
Page 9 of 12
predict good credit performance. Low scores are associated with derogatory public
records and predict poor credit performance. The table shows the very different actual
credit performance between those individuals with a good score and those with a poor
score. Good scoring accounts, above score 708, have a 1.7% credit default rate,
compared with the 20.6% default rate of poor scoring accounts below 653.

The premium bankcard market (see Table 2) is more risk averse, so it targets a
population having higher credit scores. Only 2.1% from an initial population of 100,000
consumers were considered bad risks following eighteen months of use. After reviewing
the data segmented by score, the credit card issuer can significantly reduce the risk of
loss to 1.6% just by raising the approval score for consumers to above 672.

Table 2: CREDIT CARD

RiskView Cumulative Bad
Score Accts % of Accts Rate K-S

< 672 9,937 9.9% 6.4% 21.0

672-679 9,959 19.9% 3.9% 29.6
680-685 8,368 28.3% 2.5% 31.4
686-690 10,209 38.5% 2.4% 32.7
691-695 11,192 49.7% 1.7% 30.6
696-699 9,571 59.2% 1.6% 28.1
700-704 8,811 68.0% 0.9% 22.8
705-710 11,076 79.1% 0.7% 15.3
711-717 10,406 89.5% 0.6% 7.7
718+ 10,471 100.0% 0.6% 0.0

Total 100,000 2.1% 32.7

The wireless phone industry experiences higher-than-usual delinquencies and risk

due to issues unique to its business, such as leftover contract terms recorded as
uncollected debt.

Telecom analyzes and segments the creditworthiness of its customer base according to
scores that reach 1000 instead of the more traditional 800 or 900 (see Table 3).

September 5, 2006
Page 10 of 12
 Table 3: TELECOM
RiskView Cumulative Bad
Score Accts % of Accts Rate K-S

<750 10,004 10.0% 38.9% 13.0

750-807 10,027 20.0% 31.7% 21.3
808-841 10,057 30.1% 25.0% 25.3
842-869 9,987 40.1% 24.6% 28.9
870-901 10,136 50.2% 20.2% 29.7
902-929 10,031 60.2% 14.9% 27.1
930-952 10,393 70.6% 13.5% 23.4
953-962 9,442 80.1% 9.7% 17.7
963-971 10,092 90.2% 5.6% 8.9
972+ 9,831 100.0% 5.0% 0.0

Total 100,000 19.0% 29.7

After following 100,000 new telecom customers for approximately eighteen months, a
total of 18,951 accounts, or 19%, were considered “bad” based on delinquency or write-
off. This loss rate could have been substantially reduced if the poor RiskView scoring
customers had been required to pay an upfront deposit (as is done for poor credit
scoring new customers). Those customers who scored below 808 had credit losses of
over 31%. Eliminating them from the portfolio would have dropped the overall loss rate
to less than 15%.

The K-S statistic shown on the tables is a measure of the statistical power of the scores
to predict credit loss rates. Traditional credit based risk scores typically have K-S
values that are comparable to these non-traditional credit scores.

Conclusion

Traditional methods of credit scoring rely on analyzing the way an individual or a

business has managed its credit usage in the past. This approach limits creditors to
approving only individuals and businesses that have a history of using credit. Therefore,
it limits a creditor’s ability to grow their market profitably by expanding offerings to thin

September 5, 2006
Page 11 of 12
file or no file consumers and/or businesses. It also may cause a creditor to reject a
currently creditworthy consumer or business based on historical credit problems.

The alternative approach uses non-traditional data for credit scoring instead of credit
history. This approach, fostered by LexisNexis, successfully enhances traditional credit
bureau-based solutions and performs as a viable alternative for scoring non-traditional
credit seekers. As with credit histories, however, predicting credit risk through public
records is only as effective as the information upon which the predictions are based.
Therefore, it is critical that the information be extensive, current, reliable and accurate, a
condition that LexisNexis, as the recognized leader in collecting, storing and searching
public record and non-credit data, can ensure.

For more information, visit http://www.lexisnexis.com/risk/.

September 5, 2006
Page 12 of 12