The future of credit card underwriting. Understanding the new normal



Similar documents
Understanding automotive loan charge-off patterns can help mitigate lender risk

SCORES OVERVIEW. TransUnion Scores

Bust-out fraud. Knowing what to look for can safeguard the bottom line. An Experian white paper

Battling for Balances: Targeting the Lucrative Card-to-Card Consumer Balance-Transfer Market. An Experian Perspective

Let there be light. The impact of positive energy-utility reporting on consumers. An Experian white paper

New Predictive Analytics for Measuring Consumer Capacity for Incremental Credit

Credit Score Basics, Part 1: What s Behind Credit Scores? October 2011

Existing Account Management: Building Effective Portfolio Management Tools May 2011

In Search of Credit Card Profitability: Finding and Retaining the Most Valuable Bank Card Customers

Risk-Based Pricing Back to the Basics

Unlocking the opportunity with Decision Analytics

Business Information Services. Product overview

Credit Profile Report for Tenant Screening

Tom Aliff Vice President, Modeling and Analytics Martin O Connor Senior Vice President, Global Analytics

Fraud detection in newly opened accounts. Connecting data helps predict identity theft

Utilizing Experian next generation decision management software to bring customer management to the next level of client experience and value creation

The small-business owner an in-depth B2B marketing study of owners of multiple small businesses. An Experian study

Summary. January 2013»» white paper

Alternative Data and Fair Lending

OCC and OTS Mortgage Metrics Report

UNDERSTANDING YOUR CREDIT REPORT (Part 1) By Bill Taylor

Improving decisions across the Customer Life Cycle

Application of SAS! Enterprise Miner in Credit Risk Analytics. Presented by Minakshi Srivastava, VP, Bank of America

Acquiring customers profitably. With Credit Bureau Scores

Small Business Credit Outlook

Understanding Your Credit Score

THE RISE OF THE COMPOSITE LENDING BUSINESS MODEL

How To Be Successful

STATISTICA. Financial Institutions. Case Study: Credit Scoring. and

Business Analytics and Credit Scoring

WHAT IS BUSINESS CREDIT?

Business Funding Evaluation YOUR BUSINESS NAME


The FPA Absolute Fixed Income Strategy (including FPA New Income, Inc.) Investment Policy Statement

A Guide to Mortgage Products. A Glossary of Lending Terms and. True. Know Before You Go...To Get A Mortgage. False. Federal Reserve Bank of Boston

Sankaty Advisors, LLC

Understanding Credit Scoring


CREDIT BASICS About your credit score

ConsumerViewSM. Insight on more than 235 million consumers and 113 million households to improve your marketing campaigns

Debt collection in lean times. An Experian briefing paper

Fraudulent accounts in collections: improve detection and reduce collector workload. An Experian briefing paper

DNBi Risk Management. Unparalleled Data Insight to Drive Profitable Growth. Insights from Data. Relationships from Insights

Expand & Grow: How to reach and educate more members using a new breed of credit scores

Implementing differentiated customer-centric strategies

Small/Mid-Cap Quality Strategy (including FPA Paramount Fund, Inc. and FPA Perennial Fund, Inc.)

Identity Theft Red Flags Rule

The Role of Credit Scores in Consumer Loan Securitization

6Billion. Over. VantageScore credit. and 7 of the 10 largest banks in scores used in

To Score or Not to Score?

The Road to Good. How to Start Building Your credit with In-House Auto Financing

High Yield Bonds A Primer

EFL Case Study: Using the EFL Score to Enhance Credit Bureau Data. Equifax Peru. Executive Summary:

CREDIT REPORTS WHAT EVERY CONSUMER SHOULD KNOW ABOUT MORTGAGE EQUITY P A R T N E R S

Office of Student Financial Management

Credit Suisse 2015 Financials Conference

Inquiry into Access of Small Business to Finance ANZ Submission to the Senate Economics References Committee

Student Loans and Credit Reports

New to Credit from Alternative Data

About your credit score. About FICO Score. Other names for FICO Score.

Risk, reward and the SME lending imperative. An Experian briefing paper

Credit Scores. Copyright 2009 How to Gain Wealth. All rights reserved.

Understanding Fixed Income

WHITEPAPER. Creating and Deploying Predictive Strategies that Drive Customer Value in Marketing, Sales and Risk

Understanding Your Credit Score

Appendix B D&B Rating & Score Explanations

Earning Extra Credit. Understanding what it takes to maintain and manage good credit now and for your future

The Business Case For SBA 7a Lending For Community Banks

Introducing Trended Credit Data Legislative & Regulatory Conference Sunday April 10, 2016

A.MORTGAGE LENDER B. CREDIT CARD ISSUER C.HOME INSURER E. ELECTRIC COMPANY F. LANDLORD G.ALL OF THE ABOVE D.CELL PHONE COMPANY

Transcription:

The future of credit card underwriting Understanding the new normal

The card lending community is facing a new normal a world of increasingly tighter regulation, restrictive lending criteria and continued economic challenges. This new normal will impact how the industry does business going forward. Deeper data assessment, fewer automatic approvals and a focus on pricing for the long term at the point of application will take center stage as lenders focus on how to grow within today s reality. Most card lenders spent much of 2009 focused on readying themselves for Credit Card Accountability Responsibility and Disclosure Act (Credit CARD Act) compliance and will continue these efforts throughout 2010. However, with the economy appearing to be recovering, they now are beginning to set their sights on growth putting increased emphasis on targeted marketing, primarily to the highest-credit-quality consumers. The key challenges the industry faces with regard to growth include: Developing a strategy that allows for profitable acquisitions to balance a portfolio Pulling forward into the marketing and application approval process policies that were formerly used in an account management process to ensure compliance Acquiring new data elements to comply with recent legislation and drive decisions that enable profitable portfolio growth Illustrating these difficulties, the chart below shows that new credit originations decreased 46 percent from December 2008 compared with December 2009. The previous year comparison resulted in a 16 percent decline. These significant dropoffs likely were due to factors such as rising on existing portfolios and 20,000 16% 15% 16% Thousands 15,000 10,000 42% 45% 46% 5,000 Q1 Q2 Q3 Q4 2007 2008 2009 Chart 1 new lending mandates in the Credit CARD Act. New bankcard origination volume trends Market Insight from Experian Page 3

Credit quality of new accounts is increasing In the current highly restrictive underwriting climate, Experian is seeing credit quality for newly opened bankcards increasing. In Chart 2, the risk score, VantageScore, * increased from an average of 793 to 810 for newly opened bankcards at the time of origination. This jump indicates that lenders are increasing the cutoff for new accounts to lessen risk exposure. As a result, consumers are finding it more difficult to obtain new bankcards unless they have a demonstrated history of excellent credit. Chart 3 reveals that consumers opening new bankcards have a credit history of nearly one year longer than what was seen the year before. Both trends clearly show that the criteria for new accounts have grown tighter and that lenders are seeking populations with a lower risk of default. New bankcard origination credit quality trends VantageScore Age of oldest trade Credit risk score 815 810 805 800 795 790 785 780 Credit risk score 815 810 805 800 795 790 785 780 Chart 2 Chart 3 Q4 2008 Q4 2009 Using predictive data to limit risk exposure To limit risk exposure while still keeping credit flowing with new accounts, lenders need more effective tools to evaluate consumers. Employing predictive data and scores to drive more granular segmentation can provide card lenders with the key metrics they require to make the most informed decisions. To demonstrate this point, Experian utilized two analytical approaches when examining consumers who received new bankcard accounts between January 2009 and March 2009, splitting the population by prime and subprime credit scores. In the first methodology, a score was built starting with VantageScore as the base and building a new score by adding new data components, including: Expanded credit attributes Short-term balance change attributes Estimated income and debt-to-income (DTI) ratio *VantageScore is owned by VantageScore Solutions, LLC. Page 4 The future of credit card underwriting

Macroeconomic data Personality clusters Score on propensity to open a new account As seen in charts 4 and 5, below, the results for the prime population reveal a significant lift in predictiveness compared with VantageScore alone. Experian s new expanded premier credit attributes, estimated income, DTI ratio and macroeconomic data each contributed measurable lift in predicting consumer behavior. The subprime population experienced an even larger lift, indicating that by using this additional data, lenders are significantly better able to identify consumers who will pay their bills. In addition to the data elements found to be predictive for the prime population, other behaviors such as short-term changes in balances, as well as data traditionally used for prospecting, like personality clusters and propensity-to-open models were surprisingly predictive of risk. Predictive power of new data Prime Subprime Stage in hierarchy KS Percentage KS lift over VantageScore Stage in hierarchy KS Percentage KS lift over VantageScore VantageScore 43.22 N/A VantageScore 25.66 N/A Credit attributes 46.86 8.42% Credit attributes 32.94 28.37% Estimated income and DTI 48.59 12.42% Estimated income and DTI 33.56 30.79% Macroeconomic 49.26 13.98% Shor-term balance changes 34.23 33.40% Macroeconomic 34.30 33.67% Personality clusters and propensity-to-open a new account KS refers to the Kolmogorov-Smirnov statistic Chart 4 Chart 5 35.40 37.96% Using a decision tree to analyze populations The second type of analysis employed relies on a Chi-squared Interaction Detector (CHAID) decision tree to examine the same prime and subprime populations. With this methodology, the lift is determined by defining smaller populations of consumers with similar characteristics; they are segregated into groups of those who are significantly more or less likely to pay their bills. As seen in the chart below, this portion of the analysis was broken into natural and forced parts. The natural decision tree was grown by inputting the data components and then letting Market Insight from Experian Page 5

the interaction of the variables determine the nodes. With the forced decision tree, VantageScore was used to drive the first level due to its high degree of predictiveness, but then the estimated income and DTI were forced as the second level. (Forcing the estimated income and DTI is a response to the recent mandate contained within the Credit CARD Act, which requires lenders to consider ability to pay when deciding to extend credit.) By doing a segmentation analysis using decision trees, lenders can reveal the populations where the largest difference exists from the entire business-as-usual (BAU) bad rate on both the high and low ends of the spectrum, thus allowing them to assign the correct rates and products commensurate with risk. Charts 6 and 7, below, show the BAU bad rate for the prime population was 2 percent, meaning 2 percent of those new accounts opened between January 2009 and March 2009 had at least two missed consecutive payments during the calendar year. An examination of the natural tree shows that the bad rate can be significantly spread across subsegments. A population with a 0.3 percent bad rate is 84 percent less risky compared with the BAU bad rate of 2 percent. Conversely, the riskiest population is identified with a 40 percent bad rate, which is more than 2000 percent worse than the BAU. The key drivers of this segmentation are VantageScore less than 689 and income less than or equal to $30,000, along with high previous on bills. Shifting attention to the prime forced tree, it shows that it spreads the segments even further so that the middle populations have a higher bad rate. By doing this, other useful data components are revealed. The Complacent Card User personality cluster helps identify the low risk, and a lower home price is indicative of higher risk. (The lower home price in this population is probably due to the relationship between lower income and lower home price and would be expected when income is forced into the segmentation.) Segmentation results by risk rates Prime natural tree Base 2% population Bad rate 0.3% 0.7% 15% 21% 40% Prime forced tree Base 2% population Bad rate 0.3% 0.6% 18% 23% 25% Lift: -84% >826 >$67K Low monthly bankcard payment Lift: -63% >826 $48K $66K Low total monthly payment Lift: 705% 689 710 High 60+ percentage Lift: 989% >$36K High external collections Lift: 2011% <=$30K High 60+ percentage Lift: -84% >826 >=$67K Complacent Card User Lift: -68% >826 >=$67K Credit-hungry Card Switchers, Cash-oriented Skeptics Lift: 832% <$27K Home price: >=$118K Lift: 1132% $27K $30K Propensity to open <677 Lift: 1205% <$27K Home price: <$118K Low risk High risk Low risk High risk Chart 6 Chart 7 Page 6 The future of credit card underwriting

Not surprisingly, when looking at subprime populations through the decision tree (see charts below), those with delinquencies in collections fall into considerably more risky ranges. Instead of a 2 percent bad rate for the entire population, as was seen for the prime population, the subprime population had a decidedly higher 28 percent bad rate. This means that of the new subprime accounts that opened between January 2009 and March 2009, 28 percent of them had at least two consecutive missed payments during the calendar year. Once again, the most predictive input is VantageScore. Medium to high external collections are also a significant indicator of risk in subprime, according to the natural tree analysis. (An external collection is when a lender sells the bad debt to an agency for recovery, usually when the account is 120 to 180 days past due.) Those below 531, considered deep subprime, have a 65 percent bad rate, or a more than 136 percent lift in predictiveness for this population. On the less-risky end of the spectrum, the segment with a VantageScore greater than 635, income of greater than $71,000 and a lower incidence of 90-plus days past due resulted in a 9 percent bad rate, which is 68 percent lower than the entire population. When the income estimate is forced in the subprime population, the new components that become predictive are changes in balances and utilization, as well as explicit high and low cutoffs for a propensity-to-open score. Through analysis, Experian carved out populations with a bad rate as low as 8 percent, which is a 73 percent decrease from the 28 percent BAU bad rate. On the upper end, the riskiest population had a VantageScore of less than 531, income of less than $36,000 and a small percent change in total balance. Much of this population has high utilization, which can result in fewer changes in balance. Segmentation results by risk rates Subprime natural tree Base 28% population Bad rate 9% 11% 55% 63% 65% Subprime forced tree Base 28% population Bad rate 8% 11% 52% 56% 61% Lift: -68% >=635 Low 90+ >=$71K Lift: -59% >=635 Medium 90+ $29K $70K Lift: 99% Medium 90+ Medium external collections Lift: 129% High 90+ Lift: 136% Medium 90+ High external collections Lift: -73% >=635 >=$61K Large percentage change in total balance Lift: -61% >=635 $29K $60K Propensity to open >=764 Lift: 90% Lift: 104% 531 575 >=$36K <$36K Low number of Propensity to trades with open <539 change in utilization Lift: 121% <$36K Small percentage change in total balance Low risk High risk Low risk High risk Chart 8 Chart 9 Market Insight from Experian Page 7

The take-away from these decision trees is that as scores go down and rates go up, the bad rate quickly exceeds the risk tolerance of most lenders. Lenders looking to expand their universe can confidently do so by incorporating new data elements to help stratify risk into their lending decision processes. Conclusion The study s results demonstrate that through the use of more insightful data, lenders can better separate and define populations of consumers who are more capable and willing to repay debt. This information will allow lenders to develop a strategy for profitable acquisitions to balance portfolios while simultaneously setting appropriate limits and interest rates commensurate with the consumer populations. Today s new normal is going to necessitate incorporating new data elements in marketing and application approval processes in order to comply with recent legislation and drive decisions that enable profitable account growth. Continuing with today s business-as-usual approach would leave a lender at a competitive disadvantage. For consumers, gaining a better understanding of how lenders evaluate credit usage and on-time payment data can only reinforce the importance of paying bills promptly and using credit wisely. In addition, this microsegmentation approach will provide consumers who have proven they use their credit responsibly with the best rates while opening the door for other populations to gain access to credit again. Methodology Experian used two distinct approaches to determining the risk or the contribution that new data could provide. One approach relied on VantageScore as a base and added the new data elements. Lift in the Kolmogorov-Smirnov (KS) statistic was used to measure value. The other approach relied on CHAID methodology, looking at both natural and forced income due to new Credit CARD Act requirements, where lift in the bad rate was used to illustrate the magnitude of value from the segmentation. Experian analyses and studies To review additional Experian information, including data analysis, research and related information, visit www.experian.com/insight. Page 8 The future of credit card underwriting

475 Anton Blvd. Costa Mesa, CA 92626 www.experian.com 2010 Experian Information Solutions, Inc. All rights reserved Experian and the marks used herein are service marks or registered trademarks of Experian Information Solutions, Inc. Other product and company names mentioned herein may be the trademarks of their respective owners. VantageScore is owned by VantageScore Solutions, LLC. 04/10 5527/5081 5424-CS