Credit Modeling, CECL, Concentration, and Capital Stress Testing

Transcription

1 Credit Modeling, CECL, Concentration, and Capital Stress Testing Presented by Wilary Winn Douglas Winn, President Brenda Lidke, Director Frank Wilary, Principal Matt Erickson, Director September 26,

2 Douglas Winn President Today s Presenters Mr. Winn co-founded Wilary Winn in the summer of 2003 and his primary responsibility is to set the firm's strategic direction. Mr. Winn is a nationally recognized expert in financial institution accounting and regulatory reporting and has led seminars on the subject for many of the country's largest public accounting firms, the AICPA, the FDIC, and the NCUA. Mr. Winn began his career as a practicing CPA for Arthur Young & Company - now Ernst & Young. 2

3 Brenda Lidke - Director Today s Presenters Ms. Lidke has over 15 years of experience in the financial services industry and has been with the firm since Her areas of expertise include modeling of complex cash flows and financial analysis. Brenda manages and leads Wilary Winn s fair value business line and is one of the country's foremost experts regarding purchase accounting. 3

4 Today s Presenters Frank Wilary Principal Mr. Wilary co-founded Wilary Winn in 2003 and has over twenty years of diversified experience in the financial services industry. His areas of expertise include asset-liability management, capital markets, derivatives, information systems and valuation of illiquid financial instruments. Frank s primary responsibility is to lead the research, development and implementation of Wilary Winn's new business lines. He works to ensure that new products and services meet our firms' high standards and makes the critical determination of whether to buy or build valuation software and how to best utilize the system selected. 4

5 Matt Erickson Director Today s Presenters Mr. Erickson leads Wilary Winn s asset liability management (ALM) business line. He consults with ALM clients on interest rate, credit, concentration and liquidity risks as well as capital stress testing, risk-based pricing and real return optimization. Matt uses his knowledge of credit risk analytics and quantitative analysis skills to strengthen the firm s proprietary valuation models, develop assumption input databases, and track industry-wide performance trends on loans and deposits. 5

6 Topics Covered Today Balance sheet risk Credit loss models focus on discounted cash flow Loan stratification Predictive credit indicators Capital stress testing Concentration limits and sub-limits Risk-based pricing and real return analyses CECL implementation 6

7 Balance Sheet Risk Primary risks are credit, interest rate and liquidity which are ideally measured on an integrated basis Credit is the most critical because losses incurred on loans and investments have been key factors in banking crises and failures Credit risk is best measured from the bottom up 7

8 Estimating Credit Losses Modeling should be based on type of loan (CECL allows various methods) For example, MBLs that are CRE re-underwrite lower rated loans and use migration analysis for higher rated loans Residential real estate and consumer loans are best modeled statistically: Roll rate analyses Probability of default analyses Discounted cash flow analyses 8

9 Static Pools Annual Loss 15 yr mortgage, FICO, 80% LTV Actual Annual Loss by Year Since Origination Losses Losses Origination Original to Date to Date Year Balance $ % ,000,000 10, % 0.0% 0.0% 0.0% 0.0% 0.1% 0.1% 0.1% 0.1% 0.1% 0.1% 0.1% 0.1% 0.1% 0.1% 0.0% ,000,000 13, % 0.0% 0.0% 0.0% 0.0% 0.1% 0.1% 0.1% 0.1% 0.2% 0.3% 0.2% 0.1% 0.1% 0.1% 0.0% ,000,000 21, % 0.0% 0.0% 0.0% 0.0% 0.1% 0.1% 0.1% 0.4% 0.5% 0.5% 0.1% 0.1% 0.1% 0.1% ,000,000 30, % 0.0% 0.0% 0.0% 0.1% 0.1% 0.4% 0.5% 0.6% 0.5% 0.4% 0.2% 0.2% 0.2% ,000,000 40, % 0.0% 0.0% 0.0% 0.1% 0.4% 1.0% 1.1% 0.8% 0.3% 0.2% 0.2% 0.1% ,000,000 86, % 0.0% 0.0% 0.1% 0.8% 1.8% 2.3% 1.6% 0.9% 0.6% 0.4% 0.2% ,000,000 84, % 0.0% 0.0% 0.8% 1.6% 2.0% 1.8% 0.9% 0.8% 0.3% 0.3% ,000,000 74, % 0.0% 1.0% 1.3% 1.8% 1.8% 0.6% 0.5% 0.3% 0.2% ,000,000 66, % 0.4% 1.0% 1.3% 1.5% 0.9% 1.2% 0.3% 0.2% ,000,000 31, % 0.5% 0.9% 0.9% 0.3% 0.3% 0.3% 0.1% ,000,000 3, % 0.0% 0.0% 0.0% 0.1% 0.1% 0.1% ,000,000 1, % 0.0% 0.0% 0.0% 0.0% 0.1% ,000, % 0.0% 0.0% 0.0% 0.0% ,000, % 0.0% 0.0% 0.0% ,000, % 0.0% 0.0% ,000, % 0.0% Average 0.0% 0.1% 0.2% 0.3% 0.4% 0.4% 0.3% 0.3% 0.2% 0.2% 0.1% 0.1% 0.1% 0.1% 0.0% 9

10 Static Pools Annual Loss 2.50% Annual Losses by Vintage Pool by Age % % Loss % 1.00% % % Age of Loan (in years)

11 Discounted Cash Flow Analysis Examples of DCF analysis: OTTI non-agency MBS Fair value determinations in mergers ASC TDRs MSRs Economic Value of Equity in ALM 11

12 Discounted Cash Flow Analysis Key Valuation Inputs: Conditional Repayment Rate (CRR) Conditional Default Rate (CDR) Conditional Prepayment Rate (CPR = CRR + CDR) Loss Severity Discount Rate depends on accounting context. For CECL it is original yield 12

13 Loan Example FICO Loan Example FICO group Discounted Annual Annual Annual Sched. P&I payment $ 37, Losses CRR% CDR% Severity% $ 75, % 1.0% 20% Loan Remaining Repo Total Total Repo Valuation Payment Loan Actual Voluntary Prin Prin P&I DQ Repo Prin Monthly Monthly Monthly Month Month Balance Amort Prepays Recoveries Collected Interest Collected Balance Balance Liquidations Losses CRR% CDR% Severity% ,525, ,478,309 13,201 33,490-46,691 24,448 71,139 5, % 0.08% 20% ,431,916 13,171 33,222-46,393 24,253 70,646 10, % 0.08% 20% ,385,819 13,141 32,956-46,097 24,059 70,156 16, % 0.08% 20% ,340,016 13,112 32,691-45,803 23,866 69,669 21, % 0.08% 20% ,294,505 13,082 32,428-45,511 23,674 69,185 26, % 0.08% 20% ,249,285 13,053 32,167-45,220 23,484 68,704 32, % 0.08% 20% ,204,354 13,023 31,908-44,931 23,295 68,226 31,869 5, % 0.08% 20% ,159,710 12,994 31,650-44,644 23,107 67,751 31,613 10, % 0.08% 20% ,115,351 12,965 31,394-44,359 22,920 67,279 31,358 16, % 0.08% 20% ,065,813 12,935 31,140 4,370 48,445 22,735 71,180 31,105 16,127 5,463 1, % 0.08% 20% ,016,601 12,906 30,887 4,335 48,129 22,551 70,679 30,854 15,998 5,419 1, % 0.08% 20% ,967,712 12,877 30,636 4,300 47,814 22,368 70,182 30,605 15,869 5,376 1, % 0.08% 20% ,919,144 12,848 30,387 4,266 47,501 22,186 69,687 30,357 15,742 5,332 1, % 0.08% 20% ,870,896 12,819 30,139 4,232 47,190 22,005 69,196 30,111 15,615 5,290 1, % 0.08% 20% ,822,966 12,790 29,893 4,198 46,881 21,826 68,707 29,867 15,489 5,247 1, % 0.08% 20% ,775,351 12,762 29,649 4,164 46,574 21,648 68,222 29,624 15,364 5,205 1, % 0.08% 20% ,728,049 12,733 29,406 4,130 46,269 21,471 67,740 29,383 15,240 5,163 1, % 0.08% 20% ,681,059 12,704 29,165 4,097 45,966 21,295 67,261 29,143 15,116 5,121 1, % 0.08% 20% ,507,970 2,676, ,993 5,581,811 1,966,882 7,548, ,241 99, % 0.08% 20% Total 2,741,087 3,240, ,085 6,416,229 2,378,073 8,794, , , % 0.08% 20% 13

14 Discounted Cash Flow Analysis Discounted Discounted Annual Annual Gross Discount Discounted Lifetime Annual Loan Payment Credit LTV LTV Ending Prepay % Default % Loss Avg Future Rate Future Future Future Type Status Score Status % Balance (CRR) (CDR) Severity % Life Losses (WAC) Losses Losses % Losses % Fixed Current 720+ Under 50% 45% 13,500, % 0.0% 0% % - 0.0% 0.0% Current % - 75% 65% 9,450, % 0.1% 0% % - 0.0% 0.0% Current % - 100% 85% 5,400, % 0.1% 6% 7.6 2, % 1, % 0.0% Current % - 120% 115% 3,150, % 0.4% 30% , % 22, % 0.1% Current % - 150% 140% 1,350, % 1.3% 43% , % 49, % 0.4% Current 720+ Over 150% 175% 450, % 1.8% 54% , % 27, % 0.7% Repeat for FICO Buckets Current by LTV bucket 101% 6,525, % 1.0% 20% , % 75, % 0.1% Current by LTV bucket 70% 2,115, % 3.5% 0% % - 0.0% 0.0% Current by LTV bucket 88% 1,350, % 13.0% 9% , % 65, % 0.8% Current Under 500 by LTV bucket 85% 1,462, % 20.0% 6% , % 66, % 0.9% Delinquent days 70% 45, % 30.0% 0% % - 0.0% 0.0% Delinquent days 88% 135, % 50.0% 9% , % 16, % 3.8% Delinquent 90+ days 85% 67, % 75.0% 6% 2.7 7, % 7, % 4.0% ARM repeat all FICO & LTV buckets above 125% 30,000, % 2.5% 36% 6.0 1,620, % 1,269, % 0.7% Total Mortgages 95% 75,000, % 2.1% 19% 6.8 2,077, % 1,602, % 0.3% 14

15 Assumptions by LTV and FICO CRR% and CDR% by LTV and FICO CRR 12.0% 3.0% CRR 10.0% 2.5% CDR 8.0% 2.0% CDR CRR % 6.0% 1.5% CDR% 4.0% 1.0% 2.0% 0.5% 0.0% 0.0% LTV Category 15

16 Discounted Cash Flow Analysis It is very important to note that while we are applying our statistical inputs at the loan level in order to achieve a more accurate result for the aggregated cash flows, we do not for a moment believe our results are accurate for any given loan. In fact, we show a small percentage of each loan prepaying and defaulting each year the latter, of course, being impossible. We are not re-underwriting individual loans, we are applying inputs prepayment rates, default rates and loss given defaults, which we have derived from our statistical analysis to a pool of loans. Our results are intended to be accurate and to be used only in the aggregate. 16

17 CECL Basics Why GAAP did not properly reflect risk before the financial crisis because of the delayed recognition of credit losses What ASU Measurement of Credit Losses on Financial Instruments When Issued June 16 th, effective 2021 for credit unions, early adoption permitted beginning 2019 How Adopt a cumulative effect adjustment to retained earnings Special rules for OTTI, PCD assets and beneficial interests 17

18 CECL Applicability CECL applies to: Loans HTM securities OTTI no longer exists Net investments in leases Off balance sheet credit exposures: Loan commitments Standby letters of credit Financial guarantees/similar instruments 18

19 CECL Applicability CECL does not apply to: Assets measured at fair value, including AFS securities ASC concerning AFS debt securities issued in conjunction with CECL Loans held for sale will be continued to measured at the lower of amortized cost or fair value 19

20 AFS Securities Concept of Other than Temporary Impairment ( OTTI ) goes away Assess impairment at the individual security level Recognize impairment related to credit losses through an allowance For new securities the debit is to expense For PCD securities the debit is to purchase price Credit loss is amortized costs minus the present value of expected cash flows discounted at the effective interest rate Changes in the allowance positive and negative flow through credit loss expense Allowance is limited to excess of amortized cost to fair value cannot have a negative allowance 20

21 AFS Securities If the intent or the more likely than not requirement to sell exists write security down to fair value through expense Expected cash flows in excess of this new amortized cost accreted into interest income Cannot use length of time provision to avoid loss recognition No longer required to consider volatility and postbalance sheet date changes in fair value 21

22 Existing OTTI Securities Prospective treatment for OTTI No change to existing adjusted cost basis for both AFS and HTM Amounts previously recognized in OCI related to previous improvements in cash flow will continue to be accreted into interest income on a level yield Recoveries of amounts previously written off relating to improvements in cash flow after adoption are recorded to income in the period received 22

23 CECL Major Provisions Departs from the incurred loss model the probable threshold was removed and CECL results in day one life of asset loss recognition Loss is recognized through an allowance for financial assets, including HTM debt securities, and through a liability for off balance sheet exposures Changes in the allowance positive and negative are recorded immediately through credit loss expense 23

24 Measuring Credit Loss Net carrying amount should be based on the cash flows an entity expects to collect Contractual cash flows are adjusted for expected prepayments and defaults Cash flows should not be adjusted for extensions, renewals, or modifications unless a TDR is reasonably expected Cash flows expected to be collected are discounted at the effective interest rate when using a discounted cash flow method Credit loss is carrying amount less present value of expected cash flows Measure expected losses on a pool basis whenever similar risk characteristics exist 24

25 Modeling Techniques Permits allowance calculation to be based on methods which implicitly include the time value of money DCF explicitly considers time value of money Loss-rate, roll-rates, probability of default methods, and provision matrices implicitly consider time value Contemplates use of mean and not mode if using statistical modeling 25

26 Miscellaneous CECL Items Guidance states that on collateral-dependent assets, the reserve is measured as the difference between the collateral s fair value (less selling costs) and the amortized cost basis of the asset Guidance states that an entity is not required to recognize a loss on a financial asset for which the risk of nonpayment is greater than zero, yet the amount of the loss would be zero - Example have a CDR, but have a zero loss severity 26

27 Regulatory Perspective Standard does not specify a single method for measuring expected credit losses Smaller and less complex institutions do not have to use costly and complex models Institutions may apply different modeling methods to different groups of financial assets 27

28 Overall Perspective Big opportunity software companies are creating a big hullabaloo as a result Credit unions can apply methodologies used by their lenders and best practices from other industries External auditors are considering the auditing implications given various calculation methodologies being discussed 28

29 Estimating Expected Credit Losses Consider relevant information internal and external Do not rely solely on past events adjust historical loss information for: Current asset specific risk characteristics Current conditions Reasonable and supportable forecasts Life of loan estimate to estimate losses after reasonable forecast time period, revert to historical loss rates 29

30 Predictive Credit Indicators Research conducted by others Ratings agencies approach by type of securitization Credit reporting bureaus Appropriate level to model loan or cohort Need to have cohorts that perform similarly FICO distribution Correlation is not causation 30

31 Correlation is not Causation Source: Spurious Media LLC 31

32 FICO Score Distribution 32

33 Loan Stratification Cohort Collateral Type FICO Cohort CRR% CDR% Severity % Future Loss % New Vehicle - Direct % 0.07% 33.56% 0.04% Used Vehicle - Direct % 0.16% 34.78% 0.08% New Vehicle - Indirect % 0.39% 36.95% 0.22% Used Vehicle - Indirect % 0.48% 35.48% 0.26% 33

34 Loan Stratification Cohort Collateral Type CRR% CDR% Severity % Used Vehicle - Direct Current % 0.02% 35.48% Used Vehicle - Direct Current % 0.05% 36.23% Used Vehicle - Direct Current % 0.16% 34.78% Used Vehicle - Direct Current % 0.43% 39.95% Used Vehicle - Direct Current % 4.63% 37.72% Used Vehicle - Direct Current under % 20.09% 39.42% Used Vehicle - Direct Delinquent % 45.00% 35.26% 34

35 Predictive Inputs Examples: Performance of auto loans is highly correlated to type of loan, FICO score and unemployment rate Performance of residential real estate loans is highly correlated to FICO and CLTV 35

36 Predictive Inputs 36

37 Expected vs. Actual Annual Losses 1st Mortgages 1.50% 1.25% 1.00% 0.75% 0.50% 0.25% Expected Annual Loss % Actual Annual Loss % 0.00% Year 37

38 Predictive Inputs Quantifying the relationship between unemployment and defaults: Perform regression analysis to determine best fit trend line including beta and R-squared Perform roll rate analysis to determine estimated default rates for any given unemployment rate Utilize changes between scenarios to determine default factors 38

39 Predictive Inputs Unemployment 14.00% 12.00% 10.00% 8.00% National - Unemployment CA - Unemployment Los Angeles - Unemployment 6.00% 4.00% 2.00% 0.00% 39

40 Predictive Inputs DQ Status vs Unemployment 14.00% National - Unemployment Correlation Coefficients National: 93.4% CA: 97.0% Los Angeles: 94.5% 12.00% 10.00% 8.00% CA - Unemployment Los Angeles - Unemployment 6.00% 4.00% 2.00% 0.00% National - Total 1st Mortgages DQ 30+ CA - Total 1st Mortgages DQ 30+ Los Angeles County - Total 1st Mortgages DQ

41 Predictive Inputs Unemployment and Default Factors by Year - Cyclical Assumptions Yr 1 Yr 2 Yr 3 Yr 4 Yr 5 Yr 6 Yr 7 Yr 8 Yr 9 Yr 10 Market Loan Category Unemployment Los Angeles County - Cyclical 5.80% 5.86% 9.70% 11.59% 11.75% 11.06% 9.70% 8.27% 7.06% 5.89% Estimated Default Factors Los Angeles County - Cyclical 1st Mortgage - Fixed 109% 114% 451% 617% 631% 571% 451% 326% 219% 117% Los Angeles County - Cyclical 1st Mortgage - Adjust 116% 127% 760% 1070% 1098% 984% 760% 524% 324% 131% Los Angeles County - Cyclical Other RE - Fixed 107% 111% 373% 502% 513% 466% 373% 275% 193% 113% Los Angeles County - Cyclical Other RE - Adjust 103% 104% 203% 251% 255% 237% 203% 166% 135% 105% Los Angeles County - Cyclical Credit Card 101% 102% 149% 173% 175% 166% 149% 132% 117% 102% Los Angeles County - Cyclical Other Consumer 101% 102% 139% 158% 160% 153% 139% 125% 113% 102% Estimated Default Rates Los Angeles County - Cyclical 1st Mortgage - Fixed 0.22% 0.23% 0.92% 1.25% 1.28% 1.16% 0.92% 0.66% 0.44% 0.24% Los Angeles County - Cyclical 1st Mortgage - Adjust 0.11% 0.12% 0.74% 1.04% 1.07% 0.96% 0.74% 0.51% 0.32% 0.13% Los Angeles County - Cyclical Other RE - Fixed 0.13% 0.14% 0.46% 0.62% 0.63% 0.57% 0.46% 0.34% 0.24% 0.14% Los Angeles County - Cyclical Other RE - Adjust 0.32% 0.33% 0.64% 0.79% 0.81% 0.75% 0.64% 0.52% 0.43% 0.33% Los Angeles County - Cyclical Credit Card 1.10% 1.11% 1.62% 1.87% 1.90% 1.80% 1.62% 1.43% 1.27% 1.11% Los Angeles County - Cyclical Other Consumer 1.04% 1.05% 1.44% 1.63% 1.65% 1.58% 1.44% 1.29% 1.17% 1.05% 41

42 Predictive Inputs Home Price Index National Actual California Actual Los Angeles, CA

43 Predictive Inputs Los Angeles County - HPI Impact by Year - Cyclical Assumptions Yr 0 Yr 1 Yr 2 Yr 3 Yr 4 Yr 5 Yr 6 Yr 7 Yr 8 Yr 9 Yr Appreciation/(Depreciation) % 0.9% -9.2% -21.3% -1.4% -2.5% -4.4% 10.1% 19.4% 7.7% 6.2% LTV% 90% 89% 91% 106% 115% 110% 109% 101% 82% 71% 64% Severity % 14% 16% 29% 35% 32% 31% 26% 8% 0% 0% CDR% 0.5% 0.8% 2.0% 2.5% 2.4% 2.4% 2.2% 1.0% 0.5% 0.4% Losses % 0.0% 0.1% 0.4% 0.8% 0.7% 0.6% 0.5% 0.1% 0.0% 0.0% 43

44 Statistical Significance and Creditability Margin for Error (M) 3% Confidence Level (1-α) 95% Required Proportion Sample Size , , , , , ,200 1, Required Sample Size Proportion Source: Edward (Jed) Frees, Professor Risk and Insurance, Hickman-Larson Chair of Actuarial Science, University of Wisconsin Madison

45 Statistical Significance and Creditability Materiality Example 500,000,000 Asset Size 200,000,000 Fixed Rate Mortgages 250,000 Average Loan Size 800 Number of Loans in Portfolio 140,000 Materiality Threshold FICO Balance Balance % Number of Loans Proportion / CDR% Severity Estimated Loss Amount Materiality Threshold Confidence Level (1-α) Margin for Error as a Proportion (M/π ) Margin for Error (M) Required Sample Size Estimated # of defaulted loans ,397, % % 23% 6,858 13, % 3, ,208, % % 23% 14,685 22, % 1, ,670, % % 23% 36,587 18, % 2, ,852, % % 23% 60,687 27, % ,541, % % 23% 206,648 51, % under , % % 23% 17,498 4, % ,000, % % 23% 342, , % 8, ,000 Estimated Average Balance 140,000 Materiality Threshold 800 Estimated Count of Loans Pass 8,249 Required Sample Size Fail 45

46 Creditability Estimator New Estimator = Z Company Estimator + (1 Z) Prior (Industry) Estimator Z = n/(n+k) where for some quantity k and company sample size n k = 4/(L^2 * Prior Estimator) Here, "L" is the proportion desired (margin for error as a proportion or M/π per the previous slide). 46

47 Creditability Estimator Results 620 to 659 FICO Band Industry CDR is 4.51% and Credit Union s CDR is 3.20% We want to be 95% confident our sample is within 45% of true default probability k= 4/(L^2 * Prior Estimator) = 4/(0.45^2 * ) = Thus the credibility factor Z = 23/( ) = Our final CDR estimate for the 620 to 659 FICO band is equal to our company input (4.99% * 3.20%) + (1 4.99%) * 4.51% or 4.44%. 47

48 Incorporating Your Credit Union s Loss Experience Data to collect: Unemployment rate Change in housing prices Change in used auto prices Delinquency rates by loan cohort by quarter Balance of defaulted loan and date of default Proceeds from liquidation of defaulted loan FICO and combined LTV at time of default Balance of prepaid loan and date of prepayment 48

49 DCF Modeling Advantages Uses key credit indicators that credit union uses to make loans Current economic conditions relatively easily implemented o Base case run with current FICOs and updated CLTVs Change in economic conditions relatively easily implemented o Near-term forecasts for unemployment and change in housing prices Reversion to historic trends can be implemented by reverting to national forecasts Technique explicitly discounts modified contractual cash flows 49

50 Other Uses of DCF Modeling Technique can be used for capital stress testing Results of capital stress testing can be used to set concentration limits Technique relatively easily adapted to risk-based pricing and real return analyses Inputs can be integrated into ongoing ALM modeling to determine interrelated risks 50