Exploring specialty finance data

Transcription

1 Exploring specialty finance data

2 Introducing: John Lodmell Advance America Paul DeSaulniers Experian Crissy Wallace Experian

3 Contents CFPB proposal Data analysis Business impact and the market ahead 3 Experian

4 Live poll 4 Experian

5 Consumer Finance Protection Bureau Proposed payday loan rules Paul DeSaulniers Experian 5 Experian

6 Summary of the 1,300+ pages Two categories of loans covered: Term of 45 days or less Term greater then 45 days that total cost of credit is greater than 36%, and: Repaid directly from borrowers bank account or secured by borrowers vehicle The proposed rule: Require lenders too make a determination that borrower has ability to repay loan Verify income Verify borrowers debt obligations using a credit report Verify borrowers housing cost Forecast reasonable living expenses Project borrowers net income, debt obligations and housing cost and ability to repay for the loan period Require lender to furnish information to registered information system (RIS) 6 Experian

7 Impact of rules, if implemented as written Significant reduction in pay day lending activity CFPB s own estimate would eliminate 84% of volume Online and FinTech lenders believe proposed rules will provide greater opportunity to fill void left by less sophisticated lenders (brick and mortar lenders Consumer groups have divergent views Represents an attempt to protect borrowers Does not go far enough Diminished liquidity options driving consumers to loan sharks, pawn shops or less desirable alternatives Over one million comments submitted to the CFPB 7 Experian

8 What we will explore today Does ability repay criteria as describe in rule predict consumer pay day loan performance How do traditional credit risk scores predict consumer performance in this space Profiles of consumers in the specialty finance consumers 8 Experian

9 Data analysis Crissy Wallace Experian 9 Experian

10 Additional data provides better visibility to consumers 10 Experian

11 Currently lenders decision heavily on generic score to define risk Low risk Super Prime Prime Near Prime Subprime High risk Deep Subprime Experian

12 The more we know about a consumer, the more accurately we can predict risk, using additional data Traditional data Specific models Trended data Thin file Aspirational data Specialty finance 12 Experian

13 Some data will increase predictability with positive data for subprime consumers Low risk High risk Positive examples: Rental, telecom, and utility 13 Experian

14 Some data will increase predictability with negative data for prime consumers Low risk Negative samples: Short-term loans, payday, rent-to-own High risk 14 Experian

15 This provides increase predictability, and this data is FCRA, disputable and CAN be used for decisioning Low risk Negative samples: Short-term loans, payday, rent-to-own Positive examples: Rental, telecom, and utility High risk 15 Experian

16 Analysis setup This analysis was designed to understand the new payday loan consumers which originated new loans in August 2015 Starting population: 38K Loans Open as of August 2015 (archive July 31, 2015) Performance August 2016 Performance flag bad = written off account 16 Experian

17 Data overview Waterfall counts Input records 38,490 Unable to identify (No Hit to bureau) 1.5% Address error <0.1% No Pin 0.56% Hit, but suppressed 0.91% Consumer identified (Hit to bureau) 98.5% Hit but no credit/scores (Collection, inquiry, public record or unscoreable) 11.5% Thin File (1-4 trades) 39% Thick File (5 or more trades) 48% 17 Experian

18 Cumulative % of bad accounts Model performance 100% KS Statistic 80% 60% Custom Model Score Telecom Risk Score TEC Connect VantageScore V3 ExtendedView Custom Model VantageScore V3 TEC Connect Telecom Risk EVS 40% % KS 0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95% 100% A statistical process called Kolmogorov-Smirnov (K-S) measures the separation of two populations, for example Goods versus Bads. The K-S statistic aims at measuring the maximum difference in two populations cumulative distributions. The K-S measure for the model shows the capability to distinguish Good Accounts from Bad Accounts. Cumulative % of accounts 18 Experian

19 Percent of bad accounts Model performance Worst Scoring 30% Custom Model Telecom Risk Score VantageScore V3 ExtendedView TEC Connect 29% 28% 26% 26% 26% 43% 41% 39% 38% 38% 15% 15% 13% 14% 14% 10% 20% 30% Lowest scoring ranges The ability of a model to push the largest percentage of bads to the lowest scoring ranges is a hallmark of scorecard performance. 19 Experian

20 Number of accounts Write-off rate Custom model 16,000 14, % 62.6% 65.6% 69.2% 67.0% 70% 60% 12,000 10, % 41.6% 50% 40% 8,000 6,000 4, % 30% 20% 2,000 10% Exclusion Score Loans Writeoff Writeoff Rate 0% 20 Experian

21 Profiling specialty finance consumers Total # auto loans = 1 trade; 448/mo Total # bankcard loans = 2 trades; $98/mo Total # personal loans = 1 trades; $21/mo What does the AVERAGE specialty finance (payday) consumer really look like? Total # traditional credit trades = 8 trades Total # mortgage loans = 0.27 trade; $961/mo Average Income Insight SM $39K Average Vantage Score V3 = 536 Average debt-toincome 16% 21 Experian

22 Number of accounts Write-off rate Income distributions Income Insight SM 18,000 70% 16,000 14,000 12, % 46.8% 60% 50% 10, % 40% 8,000 6,000 4,000 2, % 14.3% 30% 20% 10% - $1-35K $35-50K $50-100K $100K+ Exclusion Loans Write-off Write-off rate 0% 22 Experian

23 Generation segments and sample distribution Gen Z Gen Y Gen X Boomer Silent Age Younger: Age Age Age Age 70+ Older: Age Experian

24 Percent of accounts Loans by generation Percent of loans by generation Accounts by generation and VantageScore band Gen Z Gen YY Gen OY Gen X Boomer Silent Unknown Super Prime Prime Near Prime Subprime Deep Subprime Average VS % 19.6% 23.3% 29.0% Generation 22.1% 3.4% 0.0% Gen Z Gen YY Gen OY Gen X Boomer Silent Experian

25 File thickness overall and by generation No match / Not scoreable 1.5% Match but no trades 11.5% 7,000 6,000 No bureau match Match but no trade data 5,000 Thin 1-4 trades Thick >=5 trades 4,000 Thick >= 5 trades 48.0% Thin 1-4 trades 39.0% 3,000 2,000 1, Experian

26 Number of accounts Age of oldest trade 6,000 Age of oldest trade on bureau by generation 5,000 4,000 3,000 2,000 1,000 - Missing Gen Z Gen YY Gen OY Gen X Boomer Silent Missing < 2 yrs 2-5 yrs 5-10 yrs 10+ yrs 26 Experian

27 Number of payday loans by generation and score band 25,000 Number of payday loans by VantageScore band 20,000 15,000 10,000 5,000 - Super Prime Prime NearPrime Subprime Deep Subprime 1 Loan 2 Loans 3 Loans 4 Loans 5 Loans 6 to 10 Loans 11 to 15 Loans 16 to 20 Loans >20 Loans 27 Experian

28 More clear consumer view By adding additional data to the consumer profile the view of the consumer is now much more clear and the lenders can make more accurate credit decisions 28 Experian

29 Business impact and the market ahead John Lodmell Advance America 29 Experian

30 Thoughts on the market ahead 30 Experian

31 Questions and answers Experian contact: Paul DeSaulniers PaulDeSauliners Crissy 31 Experian

32 Share your thoughts about Vision 2017! Please take the time now to give us your feedback about this session. You can complete the survey at the kiosk outside. How would you rate both the Speaker and Content? 32 Experian

33