Stress Testing Webinar Series:Macroeconomic Conditional Loss Forecasting
Originally presented and recorded on October 29, 2013
Presented by: Moody’s Analytics
2
Agenda
1. Introductions
2. Overview
3. Consumer Loss Modeling
4. Structured Product Loss Modeling
5. Commercial/Wholesale Loss Modeling (non-public)
6. Commercial/Wholesale Loss Modeling (public)
7. Next Webinar: PPNR Models
8. Questions
3
Introductions1
4
Presenters for Today’s WebinarThomas Day, Senior Director, Regulatory and Risk SolutionsThomas works to solve difficult stress testing, capital planning, and risk management problems across complex portfolios and product sets for financial services institutions worldwide. His areas of focus include CCAR/DFA stress testing, pre-provision net revenue (PPNR) calculations, systems and methodologies, advanced liquidity risk quantification and reporting, capital planning, performance and balance sheet management.
Cristian de Ritis, PhD, Senior Director, Consumer Credit AnalyticsCristian leads a team of economists focused on consumer credit modeling and analysis for banks and other financial institutions. He provides regular commentary to clients and the media on the state of consumer credit markets and small business.
Luis Amador, Senior Director, Valuations & ConsultingLuis leads the Structured Finance Valuations and Advisory Team at Moody’s Analytics. His team is responsible for analyzing secured products globally and develops risk and regulatory software solutions. His team’s clients include banks, asset managers and insurance companies seeking credit analysis and market valuations for structured portfolios, including ABS, CLOs, TruPs CDOs, RMBS and CMBS securities and private deals.
5
Presenters for Today’s Webinar
Chris Henkel, Director, Enterprise Risk SolutionsChris leads risk management engagements throughout North America. He has extensive experience in commercial credit and financial analysis, portfolio management, asset quality, loan loss reserve methodologies, credit administration, process redesign, and credit risk modeling. He has served as a credit risk instructor and is a frequent lecturer at industry conferences.
Danielle H. Ferry, PhD, Associate Director, Capital Markets Research GroupDanielle leads the development of Stressed EDF measures, a corporate credit risk metric providing probability of default forecasts conditioned on varying macroeconomic scenarios. Her experience in helping financial institutions manage risk stemming from macroeconomic factors includes the full-cycle development of numerous proprietary quantitative models.
6
Overview2
7
Stress-Testing Complexities» The stress-testing exercise is one of the biggest challenges undertaken
by the industry and regulatory community– Impacts numerous business processes and functional areas
– Achieving “best-practice” remains a work in progress
» The national policy objective – increase the loss absorbing capacity of banks for:– Losses under severe stress Higher capital, and higher quality capital
– Ensure a resilient pool of unencumbered liquidity to reduce over-reliance on the “lender of last resort”
» The objective of the banks: Satisfy the regulators, but also ensure that any firm infrastructure and/or reporting investments improve business processes and create firm-value
8
Stress-Testing Complexities Create Many Questions
» Modeling losses under given economic scenarios is one of the most difficult aspects of the stress-testing exercise– Do you have enough reference default data by the right dimension?
» Industry
» Geography
» Product type
– Are you conditioning the models on the most important economic drivers of risk?
– Are the economic variables selected at the right level of granularity?» National v. Local markets?
– Are the models validated? Do you have required resources?
– Do you have challenger approaches that help you “triangulate” your loss estimates?
» How do you translate loss estimates to charge-offs and the ALLL?
» How do you reduce net interest income for growth in non-performing loans?
9
Our Objectives Today?
» Given the criticality of loss estimation, and the need for different models by asset class, we will cover loss estimation for:
– Retail Exposures (non-mortgage)
– Structured Portfolios
– Wholesale C&I (non-public)
– Wholesale (public)
» During future webinars, we may revisit other asset classes:
– Mortgage
– CRE
– Municipal
10
Macroeconomic Conditional Loss Forecasting of Retail Exposures3
11
» Retail credit risk models for regulatory stress testing need to be sound, transparent and well understood by banks
» Estimate losses as a function of the probability of default (PD), loss given default (LGD) and exposure at default (EAD)EL = EAD * PD * LGD
» Variety of models are available – Panel, competing risk, transition matrices
– Loan or segment level models
– Choice depends on data availability and objectives
» Documentation is critical– Validation groups focus on each model's conceptual underpinnings.
– Validators review modeling decisions and assumptions in addition to forecast results
Best Practices in Consumer Credit Risk Modeling
12
Macroeconomics Matter… Models Should Consider Broader Trends, Feedback Loops, etc.
13
But Local Economics Can Matter Even More…Idiosyncratic Scenarios Stress Geographic or Industry Concentrations
2014Q1 Severely Adverse
14
Economic Sensitivity Impacts Choice of Modeling Framework… Consider Economic Dynamics and Impact of Future Originations
15
The Fed Provides 14 CCAR Variables at the US Level –but Other Variables for Credit Models May Be Necessary» Employment
» Unemployment Insurance claims
» Bankruptcy filings by chapter
» Consumer credit debt outstanding (revolving and non-revolving)
» Used car prices
» Sales volumes (car, truck, housing, retail)
» Oil prices
» Prime rate, LIBOR, other rate indices
» ABA/MBA delinquency rates
» Personal savings rate
» Debt service and financial obligations ratios
» Credit Forecast delinquency rates by product, vintage, geo, score
» Regionality of house prices, employment, etc. is critical
16
Support Forecasts with External Data
» Champion/challenger models give a broader view– Over-reliance on single model technologies during last recession– Leverage strengths of multiple approaches– Fully transparent, back-tested and documented econometric loss forecasting models customized to
specific portfolios
» Benchmarking– Several sources of industry data exist across individual consumer credit products
» Credit bureaus, consortiums, ABS/MBS securities data
» Credit variables including volume, delinquencies, default, prepayment, etc.
– Industry data can fill in portfolio data deficiencies for modeling
» Size the approach to meet the needs of the institution– Banks of all sizes will need some sort of stress testing– Need to balance model complexity with institutional plans
17
Macroeconomic Conditional Loss Forecasting of Structured Portfolios4
18
Securitization: A Microcosm (and, sometimes, Macrocosm) of Institutional Balance Sheets
ResidentialMortgages
Corporates
CRE
Retail Auto
19
Challenges in Stress Testing Structured Instruments
» Integration– Full Integration of all the data and models needed for Structured Finance is the biggest challenge– Requires multiple models and dedicated development resources
» Coverage– Credit and Cashflow models for certain asset classes may not exist
» Data– Requires addition of supplementary data, cleansing and validation– Loan Level not available for all asset classes
» Consistency – Models developed by different companies, with varying academic approaches, are inherently
different
» Transparency– Higher level of documentation requirements for all parts of the process - (bank has to “own” the
models)– Explaining strength of macro variables, non intuitive results, all analytic assumptions – Ongoing model validation to ensure model is applicable in new macro/credit environment
20
Scenario Outcomes often seem Counterintuitive
» Static approaches can expose “thin tranche” and “cliff” effects– Seemingly innocuous changes in assumptions (e.g. default and prepayment vectors) can drive
significant movements in cash flows– Price can jump higher in more Adverse scenarios vs. Baseline– Main cause: overcollateralization-based triggers cause cash lock-up– Magnitude of the effects varies depending on position in the capital structure
0%
20%
40%
60%
80%
100%
1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
Collateral Loss
Tran
che
Pric
e Collateral Loss ∆1%
Tranche loss ∆100%
21
Our Approach
» Macro Economic Scenarios– Fed Scenarios, Moody’s Analytics Scenarios, Custom Scenarios
» Credit models– Retail, Consumer, Residential Mortgage
– Commercial, Corporate» Cashflow Libraries
– Full Global Waterfall Coverage
– Loan and Pool Level Data» Portfolio Risk Management Software
– WSADesktop» Regulatory and Risk Management Metrics
– Estimated PD’s, SSFA, OTTI, Interest Rate Sensitivities Cashflows» API’s
– Integrate all content programmatically» Advisory & Valuations
– “Moody’s Analytics Opinion”
– Help prepare financial entity for regulatory review
Fully Integrated Solution for Stress Testing & Risk Management
22
Stress Testing Structured Finance Requires Many Steps
Step 1 – Collateral Validation Stage
Assign PD to Guarantors
Non Cash Reserve Funds
Liquidity Facilities
Hedges-F/X, interest rates, basis swaps
Servicers
Custodian Banks
Step 3 – Credit Models Stage
Step 4 -Counterparty Risk Stage
Step 5 –Optionality Stage
Step 6 – Loss & Income Cashflows
Step 7 – Estimated OTTI and SSFA
Incorporate Economic Forecasting into PD framework
Custom Scenarios
Fed Scenarios
Resi
Commercial
Corporate
Consumer
Esoteric
Static / Stochastic capabilities
Collateral PD’s
Validate and Clean Trustee Loan Level & Performance Data
Append Data points required by Credit Models
Map Identifiers for Credit Models
Attachment, Detachment, Delinquency Data, for SSFA
Issuer Call
Investor Call
Income and Losses
Collateral & Tranche Losses
For All Fed & Custom Scenarios
Market Color Integration
Discount Rates
All scenarios
FR Y-14Q
FR Y -14M
Step 8 – Regulatory Reporting Stage
Step 2 – Economic Forecasting Stage
23
Analytical tool for stress testing structured portfolios
24
Macroeconomic Conditional Loss Forecasting of Wholesale C&I (Non-Public) Portfolios5
25
Our Stress Testing Framework Links Macroeconomic Factors to Credit Risk Measures – and Charge Offs
Scenario
Δ in Expected
Loss Δin
10-
yr
Trea
sury
Yie
ld
Δin
1-y
earF
ed
Fund
s Ta
rget
Δin
Cor
e G
oods
C
PI
Δin
Con
sum
er
Con
fiden
ce
Δin
Spe
c G
rade
Sp
read
s
Δin
Non
-Far
mBi
z Pr
oduc
tivity
Oth
ers
…
S1 ?
S2 ?
S3 ?
S4 ?
S5 ?
ScenarioConditions
External Impacts
Internal Impacts
Financial Impacts
CapitalImpacts
Δ inProbability of Default Δ
in 1
0-yr
Tr
easu
ry Y
ield
Δin
Cor
pora
te
Tax
Rat
e
Δin
1-y
earF
ed
Fund
s Ta
rget
Δin
Cor
e G
oods
C
PI
Δin
Wag
es a
nd
Sala
ries
Δin
Con
sum
er
Con
fiden
ce
Δin
Spe
c G
rade
Sp
read
s
Δin
Non
-Far
mBi
z Pr
oduc
tivity
Oth
ers
…
--- … … … … … … … … …
• The macroeconomic variables are drawn from those specified by the Federal Reserve in CCAR process. Moody’s and the client will jointly determine the macro-variables to be considered
• In advance of modeling, segmentation is performed for appropriate granularity (e.g., geography, industry, etc.)
• The PD, LGD, and EAD models are used to calculate the EL –and translate those to charge-offs at the segment level
• The output will also be used to calculate rating transitions and future portfolio balances
HISTORICAL DATA PREDICTIONS (Via regression model)
Independent “explanatory” variables(macroeconomic factors) Regression modeled
Predictions Values of macro factors from
forecast scenarios
iii
i XFactor εβα +∆×+=∆ ∑ ]%[%
Our goal is to translate the relationship between scenario conditions and their impact
Dependent variables(credit risk measures,
such as PD)
FOR ILLUSTRATION
26
A typical stress testing engagement tends to follow a five-phase process
Phase 1:Inputs (PD)
» Segment the portfolio
» Start with initial PDs
» Cycle-adjusted PD (e.g., RiskCalc CCA)
» Rating to PD Mapping
» Develop initial PD models
Phase 4: Back-Testing
» Test for fit, level, stability, and intuition
» Compare with historical experience
» Document methodology, assumptions, and findings
Phase 2:EL Components
» Construct facility-level LGD and EAD models
» Estimate changes in PD, LGD, and EAD based on macro variable forecasts
» Decide from candidate models
» Expected loss = PD*LGD*EAD
» Distribute EL across a future time horizon (aka “loss emergence”) to capture NCOs
» Forecast balance projections
Phase 5:Support
» Support during model validation
» Often an iterative process
» Generally extends through submission or “sign-off”
» May include implementation support
Phase 3:Loss Forecast
Validation and SupportEstimation of EL and NCOInitial PDs
Phases of a Typical Engagement
27
0
200
400
600
800
1,000
1,200
1,400
1,600
1,800
1Q90
4Q90
3Q91
2Q92
1Q93
4Q93
3Q94
2Q95
1Q96
4Q96
3Q97
2Q98
1Q99
4Q99
3Q00
2Q01
1Q02
4Q02
3Q03
2Q04
1Q05
4Q05
3Q06
2Q07
1Q08
4Q08
3Q09
2Q10
1Q11
4Q11
3Q12
2Q13
1Q14
4Q14
3Q15
Adverse Severely Adverse Base
Time Series for the S&P 500 Index
Projected period: 2Q13 to 4Q15
28
0
10
20
30
40
50
60
70
80
Rec
over
y %
def_price_adj
predicted
Comparing a Time Series of Actual to Predicted Performance is a Critical Validation Exercise
Recovery % = 1 – LGD %; e.g., 70% = 1 – 30%
Actual to Predicted Recovery Rates (1-LGD)
29
Combined Framework: Time Series Expected Loss Projection (dollar value)
$-
$1,000,000
$2,000,000
$3,000,000
$4,000,000
$5,000,000
$6,000,000
$7,000,000
$8,000,000
$9,000,000
$10,000,000
Dec
-05
Apr-
06Au
g-06
Dec
-06
Apr-
07Au
g-07
Dec
-07
Apr-
08Au
g-08
Dec
-08
Apr-
09Au
g-09
Dec
-09
Apr-
10Au
g-10
Dec
-10
Apr-
11Au
g-11
Dec
-11
Apr-
12Au
g-12
Dec
-12
Apr-
13Au
g-13
Dec
-13
Apr-
14Au
g-14
Dec
-14
Apr-
15Au
g-15
Dec
-15
Severely Adverse
Adverse
Base
30
Macroeconomic Conditional Loss Forecasting of Wholesale C&I (Public) Portfolios6
31
A Structural Model Relating PDs to Macroeconomic Scenarios Has the Advantage of Highly Intuitive Results
BL S1 S2 S3 S4
Low GDP Growth High PD
High GDP Growth Low PD
32
Loan Level Model or Pool Level Model? How About Both?
Firm-level PDs conditioned on macroeconomic variables
Macroeconomic factors affect the economy-wide distribution of default risk:
We model the discrete-ized distribution of economy-wide PDs
Macroeconomic factors affect sectors and individual firms differently:
We model sector- and firm-level default risk on idiosyncratic and macroeconomic factors
Moody’s Analytics’ Stressed EDF model consists of two sub-models: one to capture the economy-wide effects and another to capture sector- and firm-specific effects
33
Choice of Macroeconomic Variables Should Be Based On Theoretical Groundings Specific to the Asset Class
MACROECONOMIC
Driver North America
Western Europe
Real GDP growth X X
Real consumption growth X
Real investment growth X
Real export growth X X
Unemployment rate X X
CPI inflation rate X X
PPI inflation rate X X
Corporate profit growth X
FINANCIAL
Driver North America
Western Europe
Stock index growth X X
Yield curve X X
Short-term interest rate X
Baa spread X X
Ted spread X X
S&P 500 volatility X
* Wherever possible, firms are matched up with macroeconomic or financial data specific to their country of incorporation. The W. Europe models also include US real GDP growth, to proxy for global growth. In the aggregate-level models, we use weighted averages of the constituent countries’ macro drivers, where the weights are based on each country’s representation in the Public Firm EDF universe. The yield curve is defined as the long-term less the short-term government bond rate. The Baa spread is defined as the Moody’s Baa yield less the 10-year Treasury yield. The Ted spread is defined as 3-month LIBOR less the 3-month T-bill yield. The 30-day moving average of the standard deviation of the percent change in the S&P 500 is used to measure volatility.
34
In a CCAR-Style Stress Test, PDs Are Often the Most Difficult Piece of the Puzzle
[ ] EADLGDPDLossE ××=
40% / 45% / 50%
−⋅⋅⋅
−
−−= ++
121
2412
1
112
1
1001
1001
10011 ttt
tSEDFSEDFSEDFCEDF
$1
35
Hypothetical Stress Test of Financial C&I Exposures
Baseline Scenario “Second Recession” Scenario “Protracted Slump” Scenario
Source: Moody’s Analytics
05
10
15
20
25
2013m
1
2014m
1
2015m
1
2016m
1
2017m
1
2018m
1
U S
05
10
15
20
25
2013m
1
2014m
1
2015m
1
2016m
1
2017m
1
2018m
1
C o r e E u r o p e
05
10
15
20
25
2013m
1
2014m
1
2015m
1
2016m
1
2017m
1
2018m
1
P e r i p h e r a l E u r o p e
A s s u m e s 4 5 % L G D
A v g C u m u l a t i v e E x p e c t e d L o s s - F i n a n c i a l s
36
Simulated Stress Test for C&I Loans at DFAST/CCAR Banks
BHC Default Rate
Historical Default Rates vs. EDFs
Target BHC EDF
Simulated BHC Portfolio
PD = Stressed EDF
LGD = 45%
EAD = 1
BHC EL
Aggregate: 7.0% (MA) vs. 6.8% (Fed)
0
2
4
6
8
10
Ally Financial Bank of America
BNY Mellon Citigroup Goldman Keycorp PNC Financial State Street U.S. Bancorp
Fed MA Fed Aggregate MA Aggregate
Source: Board of Governors of the Federal Reserve; Moody’s AnalyticsNotes: Not displayed is the Fed’s estimate of 49.8% for The Goldman Sachs Group.
37
Summary
» There are many difficult choices and practical limitations when building a conditional expected loss model
» When modeling post-stress PDs for public C&I exposures, we take a unique and highly intuitive approach that explicitly captures both economy-wide, sector level, and firm level effects of changes in macroeconomic variables
» The model is flexible enough to accommodate any macroeconomic scenario, including the Fed’s supervisory scenarios and user-defined scenarios
» As the results of our 2012 & 2013 CCAR simulations for C&I loans show, this approach produces stressed PDs that are well-suited for CCAR-style stress tests, whether used as a primary or challenger/benchmark model
38
Next Webinar7
39
Moody’s Analytics Stress Testing Webinar SeriesMacroeconomic Conditional PPNR Forecasting
January 28, 2014 at 12:00pm EST
Join Thomas Day and other Moody’s Analytics experts for a webinar covering:
» The primary challenges confronting banks when forecasting macroeconomic conditional pre-provision net revenue (PPNR).
» Best practices for forecasting macroeconomic conditional PPNR.
» Tools and techniques used by Moody’s Analytics to address the challenges and/or close any gaps between best practices and current challenges.
Register at: http://www.cvent.com/d/64q8kn/4W
40
Questions?8
41moodysanalytics.com
Thomas DaySenior Director
Direct: [email protected]
7 World Trade Center at250 Greenwich StreetNew York, NY 10007www.moodys.com
Cris de Ritis, PhDSenior Director
Direct: [email protected]
7 World Trade Center at250 Greenwich StreetNew York, NY 10007www.moodys.com
Chris HenkelDirector
Direct: [email protected]
7 World Trade Center at250 Greenwich StreetNew York, NY 10007www.moodys.com
Luis AmadorSenior Director
Direct: [email protected]
7 World Trade Center at250 Greenwich StreetNew York, NY 10007www.moodys.com
Danielle Ferry, PhDAssociate Director
Direct: [email protected]
7 World Trade Center at250 Greenwich StreetNew York, NY 10007www.moodys.com
42
Find out more about our award-winning solutions
www.moodysanalytics.com
43
@MoodysAnalyticsStay current with the latest risk management and assessment news, insights, events, and more.
@dismalscientistView global economic data, analysis and commentary by Mark Zandi and the Moody's Analytics’ economics team.
@CSIGlobalEdRead the latest financial services education information
@MA_CapitalMktsKeep up to date on credit and equity market signals reflecting investment risk and opportunities for issuers and sectors.
7 World Trade Center250 Greenwich StreetNew York, NY 10007(212) 553-1653
121 North Walnut StreetSuite 500West Chester PA 19380(610) 235-5299
405 Howard StreetSuite 300San Francisco, CA 94105(415) 874-6000
www.moodysanalytics.com
Moody's AnalyticsFollow our company page to view risk management content, such as white papers, articles, webinars, and other insightful content and news.
The Economic OutlookThis group features insightful discussions and knowledge sharing among business, economics, and policy professionals regarding the economic outlook.
Risk Practitioner CommunityThis group brings together risk management practitioners from around the world to discuss best practices, share ideas and insights, and gain networking opportunities.
44
© 2013 Moody’s Analytics, Inc. and/or its licensors and affiliates (collectively, “MOODY’S”). All rights reserved.
ALL INFORMATION CONTAINED HEREIN IS PROTECTED BY LAW, INCLUDING BUT NOT LIMITED TO, 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.
Any publication into Australia of this document is pursuant to the Australian Financial Services License of Moody’s Analytics Australia Pty Ltd ABN 94 105 136 972 AFSL 383569. This document is intended to be provided only to “wholesale clients” within the meaning of section 761G of the Corporations Act 2001. By continuing to access this document from within Australia, you represent to MOODY’S that you are, or are accessing the document as a representative of, a “wholesale client” and that neither you nor the entity you represent will directly or indirectly disseminate this document or its contents to “retail clients” within the meaning of section 761G of the Corporations Act 2001.