CECL: Data, Scenarios and Cash Flow Thoughts

Transcription

1 CECL: Data, Scenarios and Cash Flow Thoughts H. Walter Young November 14, Risk Management Association Annual Risk Management Conference Dallas, Texas

2 Table of Contents I. Data: Not all data is created equal II. Scenarios... III. Cash Flow Generation IV. CECL Accounting Entries and Disclosures 2

3 Five Main Moving Parts/Impact of CECL Data Easily observable history, current conditions, and projections Scenarios Significant Fed data documented and published since 1976 for Stress Testing Cash Flow Generators Informed by ALM models for decades Link in new CCAR Loss Models CECL Accounting Magic Financial assets measured at amortized cost: Finance receivables, HTM and AFS securities, Accounts Receivable, Reinsurance Receivables, Repurchase Agreements, Leases, Off-Balance Sheet Loan Commitments, Standby LOCs Developing an Estimate of Credit Losses Impact on Capital (Possible Lower CET1; Yet Similar Capital + ALLL Reserves) 3

4 Exploring CCAR (and CeCL) Data Quality Dodd-Frank Act Stress Test 2016: Supervisory Stress Test Methodology and Results (June 2016) While BHC s are responsible for ensuring the completeness and accuracy of data reported on the FRY-14, the Federal Reserve made considerable efforts to validate BHC-reported data and requested resubmissions of data where errors were identified. If data quality remained deficient after resubmissions, conservative assumptions were applied to a particular portfolio or specific data, depending on the severity of deficiencies. If the quality of a BHC s submitted data were deemed too deficient to produce a supervisory model estimate for a particular portfolio, the Federal Reserve assigned a high loss rate (e.g.,90 th percentile) or a conservative PPNR rate (e.g., 10 th percentile) to the portfolio balances based on supervisory projections of portfolio losses or PPNR estimated for other BHCs. If data that are direct inputs to supervisory models were missing or reported erroneously but the problem was isolated in such as a way that the existing supervisory framework could still be used, a conservative value (e.g., 10 th or 90 th percentile) based on all available data reported by BHCs was assigned to the specific data. These assumptions are intended to reflect a conservative view of the risk characteristics of the portfolios given insufficient information to make more risk-sensitive projections. Capital Planning at Large Bank Holding Companies: Supervisory Expectations and Range of Current Practice (August 2013) BHC with weaker practices failed to compensate for data limitations or adequately demonstrate that external data reasonably reflect the BHC s actual exposures, often failing to capture geographic, industry, or lending-type concentrations. The level of segmentation used for modeling varied depending on the type and size of portfolio and estimation methods used. For example, BHCs often segmented the retail portfolio based on some combination of product; lien position; risk characteristics such as credit score, loan-tovalue ratio, and collateral; and underlying collateral information (e.g., single-family home versus condominium), though some models were estimated at the loan-level and others at the portfolio level. The Fed specifically called out certain data points in its August 2013 guidance; fields specifically mentioned by the Fed (and in the FRY-14M) may be critical to the Fed s own internal loss models, models used for CCAR and someday for CeCL 4

5 Data Quality Assessment by Criticality Possible Critical Fields 5 Original Credit Bureau Score Current Credit Bureau Score Debt to Income (DTI) Back-end at Origination Debt to Income (DTI) Front-end at Origination Original Combined LTV Refreshed Property Value Income Documentation Property Type

6 Sample CECL Loan Data (All Loan and Security Types) Macro-economic data (historical, current, future) Supportable allowance for loan and lease loss forecasts Actual full and partial losses (given default) by risk type Risk Drivers: Loan type, vintage, grades, days past due, current principal balance, collateral type, valuations, accounting classifications (troubled debt restructuring, impaired), modification date, modification amounts, amortization type, loan number history, fee history, location Accounting Information: Is a loan charged off? Is it partially charged off? What is loan life? Cash Flow Characteristics: Present value and future cash flows (to get fair value), lifetime expected cash flows and durations (by loan and security) at time of origination, sorted by similar risk types and performance characteristics (PD, LGD) Cash Flow Modeling/ALM Data: Origination date, maturity date, renewal date, repricing date, pricing spreads, fixed/variable, prepayment behavior 6

7 Sample Commercial Loan ALLL Inputs Compared to FR Y-14Qs 7

8 Select Mortgage ALLL Data Inputs Compared to FR Y-14Ms 8

9 Historical Data Paradigm to Measure CECL AFS/RAS Data = Bank + 'Market' PD, Delinquency & Other Loan Characteristic Data by Product by Region FR Y-14Q, M Data Internal ALLL Databases (Excel/Access) Official Fed CCAR Actual Macro-Economic Data Risk Driver, Cash Flow & PD/LGD Projections ALLL Loan Level for CCAR Banks Published Fed 'Baseline' Economic Forecasts 13 Quarters of CCAR Loan Level Cash Flow Forecasts 1976 Great Recession Recession Recession Recession Today < Future 13 Quarters > 9

10 In Goes: Loan Data SEC Banker Accountant 10

11 The RMA/AFS Weighted Average Expected Loss is Plotted (1-Year Lag) Showing Variance Between Expected and Actual Net Charge-Offs Industry Underestimating Losses Industry Overestimating Losses 11 Source: RMA/AFS Risk Analysis Service (RAS)

12 Examining C&I Loan Correlations, With and Without Owner-Occupied CRE Under-estimating Losses Over-estimating Losses Source: RMA/AFS Risk Analysis Service (RAS) 12

13 There is a High (70%) Correlation Between Nonaccruals in RAS and the FDIC Charge-Off Rate RAS Nonaccruals are a ~Proxy for Losses 13 Source: RMA/AFS Risk Analysis Service (RAS)

14 Losses by NAICS Industry, Size of Bubble ~ Sector s Contribution to the Total RAS Database (Exposure One of Many Views) 14 Source: RMA/AFS Risk Analysis Service (RAS)

15 Add In: Expected Scenario SEC Banker Accountant 15

16 Relevant Macro-Economic Information Relevant Information about past macro-economic events, evaluations of current conditions, and reasonable, documentable, and supportable forecasts are to be used in the estimation process. Qualitative and quantitative factors that relate to the current environment in which the company is currently operating that can be applied to the specific borrower are to be considered. This will have to be documented for history, current conditions and based on some reliable future forecast. It is suggested some measure of reasonable forecast should be used in the estimation process, to be able to estimate a measure of expected risk of credit losses even if that risk is remote: Why not use the CCAR Fed Base? Moody s? Blue Chip? 16

17 Sample Inputs to the Economic Reserve Analysis and Determination Quantitative Analysis (Starting Point) Sample economic indicators are identified and assessed quarterly (historical levels, recent trends). Macro: GDP Unemployment Asset Prices CRE Price Index HPI S&P 500 Business Centric 1 : ISM Manufacturing Factory Orders Durable Goods Etc. Consumer centric 1 : Housing Starts Retail Sales Consumer Spending Personal Income Etc. Qualitative Analysis Other considerations may include the following: Fiscal policy Interest rate levels and volatility Regulatory environment Inflation Monetary policy Global/international events SEC/FASB Administration in Power Etc. 1 Red items are CCAR forecasted for Fed Base Scenario, many Black colored macro variables are highly correlated to these Fed variables 17

18 M&T s Scenario Probability Tool (SPT) 1 SPT 1 evaluates macro economic trends and adds a likelihood to scenarios Knowing a scenario s likelihood helps understand if a scenario is reasonable Scenario likelihoods can be validated and documented The SPT tool was built for M&T s CCAR and Resolution Planning Scenario 18 SPT Tool Inputs All 16 Fed Macro Variables Probability Conditions will be Better Than SPT Tool Outputs Severity Index, i.e., Likelihood of the Scenario e.g., 50%, 99.95% Likely Sample SPT/BAU Use Fed Base (2016 CCAR) 54.3% Economic Capital (EC), Budgeting, Concentration Limits, CECL (ALLL) Blue Chip Forecast (Dec. 15) 51.8% CECL (ALLL), Budgeting Blue Chip Forecast (January 16) 55.0% CECL (ALLL), Budgeting Fed Severe Adverse ( 16 CCAR) 96.2% CCAR, EC, Concentration Limits Resolution Plans (2015) >99.99% Resolution Planning Analytics 1 M&T

19 Percentile of Severity Scenario Development Probability Tool 1 Severity Index* SPT is an Effective Challenge Benchmark for CCAR Scenarios, & a KRI or an Early Warning Indicator Expansion Moderate Recession Recession Stable 100% Severity Distribution 80% 60% 40% 20% 0% Severity Index Expansion 2003 ** *** Moody's S1 Expansion 1992 Stable 2012 Moody's S2 Stable 2005 Recession 2001 Recession 1990 Moody's S3 Recession 2008 Moody's S4 Resolution Plan Scenario Fed Base 2015CCAR 44.62% Blue Chip Dec % Fed Base 2016CCAR 54.28% Blue Chip Jan % Fed Adv 2015CCAR 90.68% Fed Adv 2016CCAR 91.01% Fed Sev Adv 2015CCAR 95.25% Fed Sev Adv 2016CCAR 96.22% Actual/historical scenario CCAR Scenario Blue Chip Scenario Moody s Scenario Resolution Planning Scenario *Severity Index = A score indicating the severity of a scenario **Underlined scenarios are actual/historical scenarios ***Moody s S1, S2, SD3, S4 scenarios come from Moody s Economy.com S1= Stronger Near-Term Rebound Scenario, S2=Slower Near-Term Recovery Scenario, S3=Moderate Recession Scenario, S4=Protracted Slump Scenario 1 M&T 19

20 Run Cash Flows SEC Banker Accountant 20

21 Reversion Periods (after the reasonable and supportable period) From: Current Expected Credit Loss (CECL) Model, at National AICPA Bank Conference. Slide by J.H. Caldwell (Deloitte) and Doug Smith (Wells Fargo), September Used with Permission 21

22 Some Granular ALM Models Can Do Lifetime Cash Flows by Loan ALM disclosures already in Annual Reports, some 10Qs Subject to Sox-like controls Covers full loan, securities; All on and off-balance sheet assets Should be loan/transaction level to link into CCAR models 22

23 CECL ALLL #s for CCAR + Disclosures for Financials SEC Banker Accountant 23

24 ALLL vs. Scorecard vs. CCAR vs. ALM CECL Evaluation Points 24

25 CECL Needs 1. Historical Expected Forecasts (Blue Chip, Fed?) 2. Historical Data, Current Data, Projected Data 3. Reliable Macro-Economic Forecasts 4. Granular Cash Flow Model 5. Linkage to CCAR Loss Models for Expected Scenario 6. Governance on Steroids Use CCARs ICAAP Process & SOX Guidance 7. Component Models will Need Validations Risk Management Department validations demand SR 11-7 reviews since results impact both CCAR and SEC Reporting 25

26 Thanks Q&A H. Walter Young Market Risk Officer, Regulatory Affairs and Capital Adequacy (RACA) M&T Bank (716)