Score migration strategies for turbulent times

Transcription

1 Score migration strategies for turbulent times Chuck Robida, Experian 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. No part of this copyrighted work may be reproduced, modified, or distributed in any form or manner without the prior written permission of Experian Information Solutions, Inc.

2 Score migration What is score migration? Shift in score over time for aggregate credit bureau population Drift in the interpretation of the score Decrease in performance between refreshes Change in scores of individual consumers What should you be concerned about? 2

3 Average VantageScore over time

4 Average VantageScore over time By product type 4

5 Bad rate (90+ DPD) of new accounts 5

6 Bad rate (90+ DPD) of existing accounts 6

7 Interpretation of score There is not a fixed bad rate at a given score Score interpretation varies Product type Lender to lender for the same product Environmental factors Unemployment Asset values 7

8 New real estate good bad odds 8

9 Existing real estate good bad odds 9

10 New bankcard good bad odds 10

11 Existing bankcard good bad odds 11

12 Average score issue 12

13 Score drift Interpretation of scores drifts over time including custom models Portfolio risk may be increasing without any change in score distribution A consistent strategy may not yield consist risk Average score is not a robust measurement of portfolio risk Actions Monitor, monitor, monitor to understand, adjust and execute Loss forecasting using macro economic variables Advanced tools for provisioning in a new age Tuesday 9:45 Use interval bad rate for averaging to account for exponential relationship of bad rate with score 13

14 Analyze refresh intervals Organizations may increase time between score refreshes to cut costs Assess impact on score performance VantageScore 90+ DPD t -6 t -3 t -2 t -1 t 0 t

15 Degradation of KS over time 15

16 Drop in bad capture rate in lowest scoring 10% 16

17 Refresh intervals Performance of score begins to deteriorate immediately and linearly Internal score will accurately assess risk of the some of the accounts Greatest exposure is on accounts that appear to be low risk on internal scores, but whose is increasing on other accounts Actions Evaluate cost / benefit analysis of update cycles Evaluate cost / benefit analysis of risk triggers 17

18 Does knowledge of score trend increase predictive power? Regress VantageScore against good bad flag with 1, 2, 3, 6 and 12 month change in score Least squares trending of 1, 2, 3, 6 and 12 scores, projecting score 1-12 months after outcome at consumer level Use exponential smoothing of score and repeat least squares trending analysis Smoothed_Score t0 = Score t0 Smoothed_Score t-1 = 0.9*Smoothed_Score t *Score t-1 Smoothed_Score t-2 = 0.9*Smoothed_Score t *Score t-2 Smoothed_Score t-1 = 0.9*Smoothed_Score t *Score t-3 18

19 Score trending 19

20 Existing bankcard accounts 20

21 Existing bankcard accounts 21

22 Existing bankcard accounts Good bad odds 22

23 Value of score trending None of the trending analysis methods resulted in any improvement in predictive power Positive changes not indicative of lower risk Current score is the best predictor Scores already incorporate time dependent variables such as delinquency and inquiries Score is a composite of many factors and is too high a level to accurately discriminate between trends and noise 23

24 How much do consumer scores change over time Look at progression of VantageScores after: One year (June 2004 June 2005) Three years (June 2004 June 2007) Five years (June 2004 June 2009) 24

25 Consumer scores over time One year 65.0%-70.0% 60.0%-65.0% 55.0%-60.0% 50.0%-55.0% 45.0%-50.0% / %-45.0% 35.0%-40.0% 30.0%-35.0% 25.0%-30.0% 20.0%-25.0% 15.0%- 20.0% 10.0%- 15.0% 5.0%-10.0% 0.0%-5.0% 6/

26 Consumer scores over time Three years 60.0%-65.0% 55.0%-60.0% 50.0%-55.0% 45.0%-50.0% 40.0%-45.0% / %-40.0% 30.0%-35.0% 25.0%-30.0% 20.0%-25.0% 15.0%-20.0% 10.0%-15.0% 5.0%-10.0% 0.0%-5.0% 6/

27 Consumer scores over time Five years 50.0%-55.0% 45.0%-50.0% 40.0%-45.0% 35.0%-40.0% / / %-35.0% 25.0%-30.0% 20.0%-25.0% 15.0%-20.0% 10.0%-15.0% 5.0%-10.0% 0.0%-5.0% 27

28 Consumer scores over time Consumer scores are relatively stable over time Consumer behavior not dynamic although unfortunate events can happen to any consumer Unlikely to rehabilitate low scoring consumers, especially in short term (one to two years) 28

29 Conclusions A score is not an absolute predictor of bad rate Portfolio risk may increase with out a change in score distribution Average scores are not a good indicator of portfolio risk Score interpretation drifts over time Portfolio risk can rise with no change in score distribution or strategy Monitoring is essential Predictive power of scores begins to deteriorate immediately and linearly Need to assess cost benefit of refresh cycles Look to real time updates Trends in scores do not add lift in predictive power Consumers behavior is relatively consistent over time 29

30 Learn more in the Vision Expert Annex and online Meet Network Research Learn Vision Expert Annex Open every morning and afternoon as well as during session breaks Interact with Vision speakers and seek-out information on products and services Schedule one-on-one meetings with Experian experts Network with peers Review and request Experian thought leadership materials and research Visit the Internet café or relax in the Expert Annex lounge Vision Community Site Join the conversation at 30

31 For additional information, please contact: