Performance Forecasting and Stress Testing: Selecting the Right Tool for the Job MICHAEL FADIL, CITIZENS BANK CRISTIAN DERITIS, MOODY S ANALYTICS

Transcription

1 Performance Forecasting and Stress Testing: Selecting the Right Tool for the Job MICHAEL FADIL, CITIZENS BANK CRISTIAN DERITIS, MOODY S ANALYTICS

2 The Great Duel: Banker vs. Economist 2

3 The Questions 1. Can one loss model serve two purposes? 2. Why should we care about segmentation? 3. Do we need to model both expected losses and net charge offs? 4. Net charge-off modeling: 1 or 2 steps? 5. What level of regional granularity is best? 6. Do loss and balance interactions matter? 7. What about idiosyncratic economic scenarios? 3

4 Can one loss model serve two purposes? Stress testing considers both severely adverse stress scenarios and baseline BAU scenarios (benign in this environment)» Can one model work for these two extremes?» Do we need separate models?» How about using the same model structure but with separate calibrations? 4

5 Example: Charge-off vs. Unemployment Rate Charge-off Rate, % (L) Unemployment Rate, % (L)

6 Overfitting Is A Bigger Risk than Underfitting Suppose there are 2 regimes. Models optimized on each will be more accurate than a single model. IF we know which regime we are in and IF the next stress period responds like the previous stress period REALITY ASSUME Best practice: BAU STRESS BAU OK UNDERSTATE LOSS STRESS OVERSTATE LOSS» Build a single model with stress factors and regime switches to better capture stress periods.» Consider multiple modeling approaches with sensitivity analysis OK 6

7 Why should we care about segmentation? What are the costs and benefits of segmentation?» Trade-offs on segmentation: pertinent segments vs data availability What are the segmentation requirements of top-down (vintage cohort) and bottoms-up (loan level) models? 7

8 Make Segmentation As Simple As Possible But No Simpler Costs/benefits exist on a continuum» No segmentation: Robust sample, smoother time series Ignores any heterogeneity Aggregation bias: throws out information» Account-level segmentation: Noisy data. Performance is discrete at account level (0/1) Maximize heterogeneity Attenuation bias: less sensitivity to macroeconomic factors» Best approach is usually somewhere in the middle. Depends on specific data set. 8

9 Do we need both expected loss and net charge off models? No, they are not the same thing Expected Loss is equivalent to MTM accounting» What is the discounted present value of future losses on a loan/cohort? NCO is equivalent to accrual accounting» What are the bank s losses over a specific time horizon?» Most banks use accrual accounting With EL modeling it is very difficult, if not impossible, to forecast NPLs. Does this mean we need to develop two different forecasting systems? 9

10 One system, two models Net Charge off and ELoss are different concepts» But they do share some similarities» Need to keep consistency between two Pricing and accounting assumptions need to align Insure consistency by:» Estimating on same historical data» Using similar inputs and drivers where possible e.g. previous delinquency status, economic variables, etc.» Estimate with similar model structure/methodology» Forecast in same platform» Use same economic scenarios and assumptions 10

11 Net charge-off modeling: 1 or 2 steps? If all we care about is net losses, why don t we model them directly?» Why bother with separate default and recovery equations? What s the benefit of modeling GCO and Recoveries separately?» Very important for wholesale, less important for retail» GCO$ f (AQ profile and lagged macro-economic stats)» Recovery$ f (primarily prior GCOs) 11

12 Occam s Razor: Are 5 blades better than 4? REGULATION REGULATOR 12

13 The Whole Isn t Always the Sum of the Parts In general, single model of target variable will have lower error than combining intermediate models» Avoid compound errors BUT this presumes stability in underlying components May not be true in consumer credit» Profile/mix of borrowers is shifting» PD is determined by the borrower» LGD is determined (in part) by the lender Matters more for some products than others (mortgage vs credit card) 13

14 What level of regional granularity is best? National? Regional? Metro? Zip code? 14

15 Macroeconomics Matter CCAR 2015 Scenarios: Unemployment rate % Analyzing Retail Credit Using Portfolio Analyzer & Moody s CreditCycle, October

16 But Local Economics Matter Too: Wide Range of UR Unemployment Rates by MSA, CCAR 2015 SA Scenario, % Analyzing Retail Credit Using Portfolio Analyzer & Moody s CreditCycle, October

17 Do loss and balance interactions matter? Future loss forecasts require estimates of new loan balances.» Need to go beyond run-off. What is best way to deal with interactive nature of modeling losses and modeling loan balances? How should we incorporate the price/quality elasticity of demand?» One really needs to forecast two of the three, and let the third fall out: Interest Rate Asset quality Loan volume 17

18 Taking the Broad View to Originations Econometric models of origination volumes insure consistency of price, quantity, quality Depends on:» Availability of data» Stability of historical relationships Two methods: 1. Internal: model historical relationships based on bank s own history 2. External: model historical relationships based on broader industry Value to both approaches 18

19 What about idiosyncratic economic scenarios? When are idiosyncratic economic scenarios used in the stress testing process? What s required? What are the challenges in constructing idiosyncratic scenarios? 19

20 Every Portfolio is Uniquely Above Average Managers and boards should identify unique risks to their institutions Large CCAR banks are required to run idiosyncratic scenarios through their stress testing models Smaller DFAST banks are encouraged to run idiosyncratic scenarios as a best practice Consider concentration of exposures: 1. Geography 2. Industry 3. Lending products Translating few specific shocks to broad, internally consistent economic scenario can be complicated 20

21 Moody s Analytics U.S. Structural Macro Model Banking sector Prices 10-yr yield Monetary policy rate Exchange rates Consumption Investment Government Exports Wages and salaries Labor force Employment Population Import prices Global prices Global GDP Imports GDP Unemployment rate Potential GDP Analyzing Retail Credit Using Portfolio Analyzer & Moody s CreditCycle, October

22 Moody s Analytics Forecast Models Macroeconomic Baseline & Alt Scenarios variables Results run through State models Results run through MSA-level models Results run through Housing models Results run through CreditForecast.com models Analyzing Retail Credit Using Portfolio Analyzer & Moody s CreditCycle, October

23 National Charge-off Rates Track Unemployment Rate and Unemployment Insurance Claims Charge-off Rate, % (L) Unemployment Rate, % (L) New UI Claims, mil (R ) Analyzing Retail Credit Using Portfolio Analyzer & Moody s CreditCycle, October

24 Key Takeaways Stress testing is an evolving process» Need flexible systems» Expect to test new products, risks and scenarios constantly Every modeling approach has strengths and weaknesses» Select the right tool for the problem» Think carefully before multi-purposing any given model Economic factors are central to forecasting and stress testing» Choose models that take the economy into account upfront» Select granularity and segmentation for today s -- and tomorrow s -- tests 24

25 United States 121 North Walnut Street Suite 500 West Chester PA Australia Level 10 1 O'Connell Street Sydney, NSW, 2000 Australia United Kingdom One Canada Square Canary Wharf London E14 5FA Prague Washingtonova Prague 1 Czech Republic

26 2013 Moody s Analytics, Inc. and/or its licensors and affiliates (collectively, MOODY S ). All rights reserved. ALL INFORMATION CONTAINED HEREIN IS PROTECTED BY COPYRIGHT LAW AND NONE OF SUCH INFORMATION MAY BE COPIED OR OTHERWISE REPRODUCED, REPACKAGED, FURTHER TRANSMITTED, TRANSFERRED, DISSEMINATED, REDISTRIBUTED OR RESOLD, OR STORED FOR SUBSEQUENT USE FOR ANY SUCH PURPOSE, IN WHOLE OR IN PART, IN ANY FORM OR MANNER OR BY ANY MEANS WHATSOEVER, BY ANY PERSON WITHOUT MOODY S PRIOR WRITTEN CONSENT. All information contained herein is obtained by MOODY S from sources believed by it to be accurate and reliable. Because of the possibility of human or mechanical error as well as other factors, however, all information contained herein is provided AS IS without warranty of any kind. Under no circumstances shall MOODY S have any liability to any person or entity for (a) any loss or damage in whole or in part caused by, resulting from, or relating to, any error (negligent or otherwise) or other circumstance or contingency within or outside the control of MOODY S or any of its directors, officers, employees or agents in connection with the procurement, collection, compilation, analysis, interpretation, communication, publication or delivery of any such information, or (b) any direct, indirect, special, consequential, compensatory or incidental damages whatsoever (including without limitation, lost profits), even if MOODY S is advised in advance of the possibility of such damages, resulting from the use of or inability to use, any such information. The ratings, financial reporting analysis, projections, and other observations, if any, constituting part of the information contained herein are, and must be construed solely as, statements of opinion and not statements of fact or recommendations to purchase, sell or hold any securities. NO WARRANTY, EXPRESS OR IMPLIED, AS TO THE ACCURACY, TIMELINESS, COMPLETENESS, MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE OF ANY SUCH RATING OR OTHER OPINION OR INFORMATION IS GIVEN OR MADE BY MOODY S IN ANY FORM OR MANNER WHATSOEVER. Each rating or other opinion must be weighed solely as one factor in any investment decision made by or on behalf of any user of the information contained herein, and each such user must accordingly make its own study and evaluation of each security and of each issuer and guarantor of, and each provider of credit support for, each security that it may consider purchasing, holding, or selling. 26