Navigating the CECL Roadmap October 4, 2018

Transcription

1 Navigating the CECL Roadmap October 4, 2018 Presented By: Bill Astrab, CPA Senior Audit Manager

2 Agenda Part I: Understanding the CECL Model: What is CECL and Why? Difference Between Incurred Loss Model and CECL Model Financial Instruments Impacted Implementation Dates and Timing Models and Data

3 Agenda Part II: CECL - Estimating Loan Losses Preparing for Implementation Data Implementation Options Sample Calculations Environmental Factors

4 Understanding the CECL Model 4

5 What is CECL? CECL: Current Expected Credit Loss A credit loss measurement model for all financial instruments recorded at amortized cost Loans: v Current model past performance + current factors = estimated loss (impaired loans an event has occurred) v CECL past performance+ current factors + expected loss = estimated loss (all loans not only impaired loans) Current GAAP: It is inappropriate to consider possible or expected future trends that may lead to additional losses

6 Why the Change? Three main reasons for change to CECL Model: Alignment: A singular credit loss measurement model for all financial instruments Timing: Current model may delay loss recognition Presentation: Financial instruments booked at amortized cost will be presented at cash flows expected to be collected Amor%zed Cost Allowance for expected credit losses Cash flows expected to be collected

7 Timing - The Proof

8 Alignment 8

9 Alignment PCD Assets Debt Securi%es Troubled Debt Accretable Yield Other than Temporary Impairment Discounted Cash flow CECL CECL CECL

10 Purchase Credit Deteriorated Current Guidance CECL Guidance Credit Discount Accretable yield Accretable yield CECL Yield Discount Yield Discount

11 Debt Securities - AFS Interest Rate Changes Book Value Amor%zed Cost CECL AOCI Fair Value Credit losses will not impact the cost basis, such as with OTTI Allowance may be adjusted up and down based on estimates of credit losses

12 Troubled Debt Restructuring Current Guidance CECL CECL model is substantially similar to current guidance Economic concessions will be accounted for using discounted cash flow method through allowance

13 Impaired Loss Model Current Guidance: Allowance Loans Losses Only reserve for impaired loans Losses must be probable and estimable Loss emergence period 12 months (industry standard) Loan appears healthy, but is actually impaired Losses on Impaired Loans Allowance for loan losses 13

14 Impaired Loss Model vs. CECL Model Impaired Loss Model Incurred/probable recognition threshold Based on past events and current conditions Incurred/probable losses CECL Model No recognition threshold Based on past events, current conditions, and reasonable and supportable forecasts Expected lifetime credit losses Judgment is required 14

15 Collectively vs. Individually Collec%vely Individually Allowance for Loan Losses No Significant Impairment Noted Collec%vely Evaluated 12 Month Emergence Period New CECL Approach Impaired Loans Individually Evaluated Collateral Dependent Approach Collateral Dependent Approach 15

16 Impacts on Allowance Current Guidance: Allowance Loans Losses: Period Ending Auto Loans 12 Month Loss Rate CECL: Allowance Loans Losses: Impaired Loss Model YE 2020 $100,000, % $250,000 Period Ending Auto Loan Pool (Sta<c Pool) Period losses were incurred Total losses over the life of the pool Expected Credit Loss Model YE 2020 $100,000,000 $370,000 Future Period 1 YE 2021 $50,000,000 During 2021 $250,000 Future Period 2 YE 2022 $10,000,000 During 2022 $100,000 Future Period 3 YE 2023 $0 During 2023 $ 20,000 $370,000 16

17 Impacts on Allowance Loan Types Period Ending Balance Applicable Historical Loss Ratio Weighted- Average Life (Prepayment Adjusted) Lifetime Historical Loss Rates CECL Incurred Loss Model Unsecured $5,250, % % $85,833 $85,833 Share Secured $246, % % $0 $0 New Auto $15,200, % % $86,158 $49,233 Used Auto $29,090, % % $428,808 $214,404 Other secured $9,676, % % $59,968 $37,480 1st Mortgage Loans $10,000, % % $200,000 $28,571 2nd Mortgage Loans $4,250, % % $14,875 $4,250 Home Equity Lines of Credit $1,692, % % $19,737 $6,579 Total Loans to Members $75,405,000 $895,379 $426,350 17

18 Impacts on Financials Allowance $469, % Allowance increased due to change in accoun%ng Incurred Loss Model CECL Model Total Assets $100,000,000 $99,530,971 Total assets decreased by change in allowance Net Worth $10,000,000 $9,530,971 Net worth decreased by change in allowance Net Worth Ra%o 10.00% 9.58% Net worth ra%o decreased due to decrease in net worth FASB - ALL impact ~ 35% 18

19 CECL Implementation Dates Effective Dates Public Business Entities (SEC filers) v Fiscal years beginning after December 15, 2019, including interim periods PBE (non-filers) v Fiscal years beginning after December 15, 2020, including interim periods v First apply CECL on the March 31, 2021 Call Report (FASB excludes Credit Unions from the definition of a PBE)

20 CECL Implementation Dates Effective Dates Non PBE s (Credit Unions) v Fiscal years beginning after December 15, 2021 v Interim periods beginning after December 15, 2021 Calendar year Credit Unions v CECL reflected in December 31, 2022 Financials v CECL provisions are applied on January 1, 2022 v Balance sheet entry only

21 CECL Implementation Dates Effective Dates Fiscal year Credit Unions v Fiscal years beginning after December 15, 2021 For example a June 30 year-end, CECL would be first reflected on its June 30, 2023 financial statements. CECL would be applied on July 1, 2022

22 CECL Implementation Dates Effective Dates CALL Reporting (original guidance this will change) v December 31, 2021 Call Report - the amount of the CECL impact will be reflected for the whole year how? NCUA FAQ as stated currently: March September CALL reports, no change First report CECL impact on December 31, 2021 CALL report Apply to CECL as of January 1, 2021 Difference compared to December 31, 2020 is the cumulative adjustment. Cumulative adjustment is offset to retained earnings

23 Timeline - Credit Unions Early Planning Data: Quality Storage Access Loca%on Method: In-House Sodware Outsource Run Model(s) Parallel to Incurred Loss Model Go Live Date Fine Tune 23

24 Implementation Committee In order to assure a smooth transition, an Implementation Committee should be formed. Risk Credit Finance An effective CECL Model relies on crossdepartmental communication. Technology Accoun%ng 24

25 CECL Loss Models Historical loss rate approach Discounted cash flows Risk migration Static pool / vintage analysis Probability of Default (PD) & Loss Given Default (LGD) Regression analysis (correlates with economic data) 25

26 CECL Loss Models Which model? - depends on the complexity of your balance sheet PD/LGD provides an ALL model, plus a true indicator or risk analysis of the loan portfolio 26

27 Data Requirements Origina<on Data Current Data Historical Data Credit score Credit score Credit score migra%on Balance Balance Date granted Date granted Date of charge-off Maturity date Maturity date Date of close Interest rate Interest rate Recovery amount LTV LTV Charge-off amount DTI Loca%on DTI Loca%on Make/Model (e.g. Auto) Delinquency status # of %mes delinquent 27

28 Data Requirements Minimum of 5 years of historical data Data models and inputs Historical AIRES Data categorized and incorporated into the model Delinquency information 28

29 Data Requirements Standard AIRES loan file layout Critical fields Days delinquent Delinquency counter days Credit score Charge off amount Loan collateral code (Strip SSN numbers) 29

30 Data Requirements Standard AIRES loan file layout Additional data Delinquency status Pay history LTV ratio Insurance / guarantor Property type Loss migration status and dates Etc. 30

31 Third-Party Servicers Loans serviced by others Participations purchased (lead lenders) Credit cards Student loans Mortgage loans Participations sold (you are the servicer) 31

32 Implementation Options Implementation Options Available Option 1: In-House Build Option 2: Vendor Solution (hosted solution) Option 3: Third-Party Outsourcing 32

33 In-House Build Pros Possible long-term savings More customizable Results in good in-house understanding of CECL Favor simpler model Less flexibility (one model) Higher upfront costs and management costs In-house exper%se is requirement is high May bring auditor scru%ny Cons 33

34 Vendor Solution - Hosted Service Pros Prebuilt industry grade data model Extendibility to other solu%ons Up-to-date with regulatory changes Medium upfront and management costs Not all sodware op%ons are high quality Subscrip%on costs Employee exper%se requirement is medium Cons 34

35 Third-Party Outsourcing Pros Third-party provided calcula%on Least in-house exper%se needed Lowest solu%on set-up costs Lowest management aken%on costs May bring higher service costs than subscrip%ons or in-house set-ups (long term) Not all service op%ons are high quality Cons 35

36 Implementation Options Other Considerations: Most Credit Unions are outsourcing Simplified methods for smaller Credit Unions Vendors in the Market Model Validation Segregating the loan portfolio Auditing the model 36

37 Vendor Example Example vendor timeline and costs: 2 to 3 weeks to build model Ongoing turn-around time 48 hours Costs $10,000 initial set-up $1,000 monthly ($1billion Credit Union) 37

38 CECL Implementation Guidance and Resources FASB and NCUA Guidance is still emerging NCUA look for updated FAQ FASB Transition Resources Group (TRG) Other Webinars White papers 38

39 Thank You! Bill Astrab, CPA Senior Audit Manager

40 Preparing for Implementation 40

41 Timeline - Credit Unions Early Planning Data: Quality Storage Access Loca%on Method: In-House Sodware Outsource Run Model(s) Parallel to Incurred Loss Model Go Live Date Fine Tune 41

42 Implementation Committee In order to assure a smooth transition, an Implementation Committee should be formed. Risk Credit Finance An effective CECL Model relies on crossdepartmental communication. Technology Accoun%ng 42

43 Data Considerations Loca%on Quality Data Storage Access 43

44 Data Requirements Origina<on Data Current Data Historical Data Credit score Credit score Credit score migra%on Balance Balance Date granted Date granted Date of charge-off Maturity date Maturity date Date of close Interest rate Interest rate Recovery amount LTV LTV Charge-off amount DTI Loca%on DTI Loca%on Make/Model (e.g. Auto) Delinquency status # of %mes delinquent 44

45 Data Sophistication Data Sophistication Options Available Option 1: Loan type level Option 2: Loan or cohort level Cohort is a group of loans that share a defining characteristic, usually more granular than the loan type level. Note: The sophistication of loan level data can range significantly 45

46 Loan Class Level Pros Lower up front costs Similar to current ALL methodology May only use simple calcula%ons Outputs would be less sophis%cated May result in more subjec%ve methodologies Cons 46

47 Loan Class Results Loan Types Period Ending Balance Applicable Historical Loss Ratio Weighted- Average Life (Prepayment Adjusted) Lifetime Historical Loss Rates CECL Incurred Loss Model Unsecured $5,250, % % $85,833 $85,833 Share Secured $246, % % $0 $0 New Auto $15,200, % % $86,158 $49,233 Used Auto $29,090, % % $428,808 $214,404 Other secured $9,676, % % $59,968 $37,480 1st Mortgage Loans $10,000, % % $200,000 $28,571 2nd Mortgage Loans $4,250, % % $14,875 $4,250 Home Equity Lines of Credit $1,692, % % $19,737 $6,579 Total Loans to Members $75,405,000 $895,379 $426,350 47

48 Loan or Cohort Level Pros More sophis%cated methodologies More current and supportable to loan level Use beyond the allowance calcula%on Higher upfront costs Cons 48

49 Loan or Cohort Level Results Multiperiod Iteration Results Loan Type Credit Grade Delinquency Paid-Off Charged-Off (PD) Loss Factor (LGD) PD x LGD Current Balance CECL before Envir. Factors Migration Model(s) LGD Model(s) B x C ($ millions) D x E A B C D E F 3101 A % 1.5% 106% 1.6% $271.3 $ A % 5.5% 65% 3.6% $4.5 $ A % 0.0% 0% 0.0% $0.0 $ A % 0.0% 0% 0.0% $0.0 $ B % 4.5% 84% 3.8% $110.3 $ B % 6.8% 57% 3.9% $10.2 $ B % 53.4% 44% 23.5% $5.6 $ B % 0.0% 0% 0.0% $0.0 $ C % 0.1% 53% 0.0% $170.6 $ C % 0.1% 95% 0.1% $6.6 $ C % 0.0% 93% 0.0% $8.1 $ C % 0.0% 56% 0.0% $2.6 $ D % 0.5% 73% 0.4% $118.3 $ D % 0.5% 76% 0.4% $11.8 $ D % 0.5% 74% 0.4% $6.0 $ D % 0.5% 31% 0.2% $3.5 $ E % 5.4% 74% 4.0% $143.2 $ E % 5.9% 66% 3.9% $28.1 $ E % 4.8% 75% 3.6% $10.7 $ E % 52.7% 46% 24.2% $9.0 $ No Score % 0.1% 50% 0.1% $4.1 $ No Score % 0.1% 45% 0.0% $4.1 $ No Score % 0.0% 0% 0.0% $0.0 $ No Score % 0.0% 0% 0.0% $0.0 $0.0 49

50 Loan Level CECL Data Costs Estimated one-time costs range from 24K - 37K* Costs include: Data identification (available vs. missing) Cleansing internal and external data Working with system vendors Data to model streamline *Estimated costs were obtained from source Effective CECL Adoption Timelines Confirmed: Expected Cost of Implementation by Fintellix Solutions. This estimate is for preparing for a migration analysis model. 50

51 Data Flow Source Data Core System AIRES Consumer, Indirect, Mortgages, Business, Other 3 rd Party Database CUSOs Credit Cards, Mortgages, MBLs Participations Macroeconomic Data / Enviroment Legacy D.B. Mergers Core Conversions Mapping Data Loan Type Product Type Credit Scores Days Delinquent Charge-Offs Recoveries Roll-Up Level Loan Portfolio Mapped by Categories 51

52 Source Data - Matrix Core System AIRES Consumer, Indirect, Mortgages, Business, Other 3 rd Party CUSO s Credit Cards, Mortgages, MBL s Participations Macroeconomic Data Legacy D.B. Mergers Core Conversions Source / Format Timing Retention / Look back period Data Scrub Ancillary Needs Core system: e.g. AIRES file Excel, CSV, etc. Excel, CSV, etc. E.g. CPI, Unemployment, Home Price Index, Vacancy Rate Legacy systems (old AIRES files / other Annually, quarterly, monthly? Must ensure consistency (e.g. quarterly should on the same quarter every year: Mar and Mar. 2016) Updated credit scores / risk ratings, retention 3 years, 4 years, 5 years or equal to the average remaining life of the portfolio/segment? Consider statistically significant data Text format, number format, integers, loan type mapping, unique data fields, credit score versions, loan level detail Updated credit scores / risk ratings, retention Updated credit scores / risk ratings, retention Statistically significant correlation based don historical trends and future expectations Updated credit scores / risk ratings, retention 52

53 Zip Code Mapping ZipcodeTxt ZipcodeNumber City State County SMITHS CREEK MI SAINT CLAIR TIPTON MI LENAWEE WALDRON MI HILLSDALE STANWOOD MI MECOSTA SPARTA MI KENT CANNONSBURG MI KENT SPRINGPORT MI JACKSON HOWARD CITY MI MONTCALM WHITE PINE MI ONTONAGON Loan Type Mapping LoanTypeCode LoanTypeCode LoanTypeCode LoanTypeCode Loan Type Code Description- Description Description Description Description Loan Type Code DataSource Detail SummaryLevel1 SummaryLevel2 SummaryLevel3 SummaryLevel AIRES New Auto - Direct 36 Months New Auto - Direct New Auto Auto Loan Consumer Loan 2102 AIRES New Auto - Direct 48 Months New Auto - Direct New Auto Auto Loan Consumer Loan 2103 AIRES New Auto - Direct 60 Months New Auto - Direct New Auto Auto Loan Consumer Loan 2201 AIRES New Auto - Indirect 36 Months New Auto - Indirect New Auto Auto Loan Consumer Loan 2202 AIRES New Auto - Indirect 48 Months New Auto - Indirect New Auto Auto Loan Consumer Loan 2203 AIRES New Auto - Indirect 60 Months New Auto - Indirect New Auto Auto Loan Consumer Loan NA36D Legacy System New Auto - Direct 36 Months New Auto - Direct New Auto Auto Loan Consumer Loan NA48D Legacy System New Auto - Direct 48 Months New Auto - Direct New Auto Auto Loan Consumer Loan NA60D Legacy System New Auto - Direct 60 Months New Auto - Direct New Auto Auto Loan Consumer Loan NA1 Merged Entity System New Auto - Direct 36 Months New Auto - Direct New Auto Auto Loan Consumer Loan NA2 Merged Entity System New Auto - Direct 48 Months New Auto - Direct New Auto Auto Loan Consumer Loan NA3 Merged Entity System New Auto - Direct 60 Months New Auto - Direct New Auto Auto Loan Consumer Loan 3101 AIRES Used Auto - Direct 36 Months Used Auto - Direct Used Auto Auto Loan Consumer Loan 3102 AIRES Used Auto - Direct 48 Months Used Auto - Direct Used Auto Auto Loan Consumer Loan 3103 AIRES Used Auto - Direct 60 Months Used Auto - Direct Used Auto Auto Loan Consumer Loan 3201 AIRES Used Auto - Indirect 36 Months Used Auto - Indirect Used Auto Auto Loan Consumer Loan 3202 AIRES Used Auto - Indirect 48 Months Used Auto - Indirect Used Auto Auto Loan Consumer Loan 3203 AIRES Used Auto - Indirect 60 Months Used Auto - Indirect Used Auto Auto Loan Consumer Loan Credit Score Buckets Credit Score Credit Grade 800 A 799 A 798 A 689 B 688 B 687 B 659 C 658 C 657 C 599 E 598 E 597 E Delinquency Buckets Delinquency Days Delinquent Category 0 Current Days Days Days Days days days days Days Days Days Days Days High Level Indirect Auto Example: 3 year lookback / 12 Qtrs Core System (e.g. AIRES) C/O File Recovery File AsOfDate Loan# Loan Type Code Credit Score Orgination LTV at Origination Balance ZipCode Delinquency Days Credit Score Most Current LTV Most Current Dealer Code Other Data (See below) Charge off Recovery 3/31/ % $35, % ABC 6/30/ % $33, % ABC 9/30/ % $31, % ABC 12/31/ % $30, % ABC 3/31/ % $28, % ABC 6/30/ % $27, % ABC 9/30/ % $25, % ABC 12/31/ % $23, % ABC 3/31/ % $23, % ABC (23,726) 6/30/ % $ ABC 5,000 9/30/ % $ ABC 12/31/ % $ ABC AsOfDate Loan# Recoveries Loan Type Code Recovery Amount Notes 12/31/ ,500 12/31/ ,000 3/31/ ,000 3/31/ ,000 3/31/ ,500 Other Dealer Code Branch ID Loan Officer Etc. AsOfDate Loan# Charge-Offs Loan Type Code Charge-Off Amt Notes 12/31/ (2,500) 12/31/ (7,800) 3/31/ (27,000) 3/31/ (23,726) 3/31/ (12,500) 53

54 Points to Consider What rabbit hole do you want to go down? Stream lining the process Regulator expectations and auditability What happens as the economy changes? Other benefits of loan level data 54

55 Implementation Options Implementation Options Available Option 1: In-House Build Option 2: Vendor Solution (hosted solution) Option 3: Third-Party Outsourcing 55

56 In-House Build Pros Possible long-term savings More customizable Results in good in-house understanding of CECL Favor simpler model Less flexibility (one model) Higher upfront costs and management costs In-house exper%se is requirement is high May bring auditor scru%ny Cons 56

57 Vendor Solution - Hosted Service Pros Prebuilt industry grade data model Extendibility to other solu%ons Up-to-date with regulatory changes Medium upfront and management costs Not all sodware op%ons are high quality Subscrip%on costs Employee exper%se requirement is medium Cons 57

58 Third-Party Outsourcing Pros Third-party provided calcula%on Least in-house exper%se needed Lowest solu%on set-up costs Lowest management aken%on costs May bring higher service costs than subscrip%ons or in-house set-ups (long term) Not all service op%ons are high quality Cons 58

59 Model Examples 59

60 Loss Rate Approach Impaired Loss Model Loan Portfolio 12-month Year Balance Losses Loss Rate Year Static Pool Balances Static Pool Losses 2010 $ 223,500,000 $ 558, % 2010 $ 223,500,000 $ $ 239,890,000 $ 576, % 2011 $ 205,730,727 $ 530, $ 257,770,000 $ 616, % 2012 $ 187,072,991 $ 480, $ 275,650,000 $ 657, % 2013 $ 167,482,368 $ 450, $ 295,020,000 $ 762, % 2014 $ 146,912,214 $ 380, $ 315,880,000 $ 764, % 2015 $ 125,313,553 $ 280, $ 338,230,000 $ 818, % 2016 $ 102,634,958 $ 250, $ 362,070,000 $ 871, % 2017 $ 78,822,433 $ 180, $ 388,890,000 $ 975, % 2018 $ 53,819,282 $ 134, $ 417,200,000 $ 1,004, % 2019 $ 27,565,974 $ 35, $ 447,000,000 $ 1,117, % 2020 $ - $ - Impaired Loss Model Allowance $ 1,004,370 Life time losses 2010 static pool $ 2,719,100 $ 223,500,000 Historical lifetime loss rate 2010 static pool 1.22% Credit score changes 0.10% Environmental Loan-to-value changes 0.11% Factors Increase in unemployment 0.00% Adjusted historical lifetime loss rate 1.43% CECL CECL Allowance $ 6,376,900 60

61 Approaches Pros and Cons Loss Rate Approach Vintage Analysis PD and LGD Discounted Cash Flow May be calculated in Excel Not difficult to understand Improves credit risk management process Already in use for merger accoun<ng Similar to current ALLL approach for most ins<tu<ons More granular data lends beoer for forecas<ng Deeper insight into loss driving factors Only short term data is needed Difficult to forecast forward Long-term historical data required Significant amount of data required Simple versions may be calculated in Excel May be most subjec<ve Likely too complex for Excel Likely too complex for Excel Relies heavily on discoun<ng assump<ons 61

62 Vintage Year Approach Loss Experience in Years Following Origina<on Year of Origina<on Year 1 Year 2 Year 3 Year 4 Total Expected 20X1 $50 $120 $140 $30 $340-20X2 $40 $120 $140 $40 $340-20X3 $40 $110 $150 $30 $330-20X4 $60 $110 $150 $40 $360-20X5 $50 $130 $170 $50 $400-20X6 $70 $150 $180 $60 $460 $60 20X7 $80 $140 $190 $70 $480 $260 20X8 $70 $150 $200 $70 $500 $430 20X9 $70 $160 $200 $70 $510 $510 Allowance for credit losses $1,260 62

63 Approaches Pros and Cons Loss Rate Approach Vintage Analysis PD and LGD Discounted Cash Flow May be calculated in Excel Not difficult to understand Improves credit risk management process Already in use for merger accoun<ng Similar to current ALLL approach for most ins<tu<ons More granular data lends beoer for forecas<ng Deeper insight into loss driving factors Only short term data is needed Difficult to forecast forward Long-term historical data required Significant amount of data required Simple versions may be calculated in Excel May be most subjec<ve Likely too complex for Excel Likely too complex for Excel Relies heavily on discoun<ng assump<ons 63

64 Probability of Default (PD) Loss Given Default (LGD) Total Probability (count based): Loan Type 1 Loan Type 2 Loan Type 3 Loan Type 4 Loan Type 5 Loan Type 6 Loan Type 7 Loan Type 8 Risk Rating % 0.551% 0.007% 0.046% 0.000% 1.524% 3.794% 2.277% % 0.000% 0.027% 0.000% 0.000% 8.453% 1.459% % % 4.558% 4.410% 3.698% % 1.252% % % % % % % 8.280% 5.908% % % Probability of Default (count based): Loan Type 1 Loan Type 2 Loan Type 3 Loan Type 4 Loan Type 5 Loan Type 6 Loan Type 7 Loan Type 8 Risk Rating % 0.746% 0.217% 0.459% 0.000% 2.488% 4.412% 2.717% % 0.000% 3.419% 0.000% 0.000% 8.453% 1.459% % % % % % % % % % % % % % % % % % Loss Given Default (count based): Loan Type 1 Loan Type 2 Loan Type 3 Loan Type 4 Loan Type 5 Loan Type 6 Loan Type 7 Loan Type 8 Risk Rating % % 3.389% 9.944% 0.000% % % % % 0.000% 0.791% 0.000% 0.000% % % % % 8.740% % % % % % % % % % % % % % % 64

65 Approaches Pros and Cons Loss Rate Approach Vintage Analysis PD and LGD Discounted Cash Flow May be calculated in Excel Not difficult to understand Improves credit risk management process Already in use for merger accoun<ng Similar to current ALLL approach for most ins<tu<ons More granular data lends beoer for forecas<ng Deeper insight into loss driving factors Only short term data is needed Difficult to forecast forward Long-term historical data required Significant amount of data required Simple versions may be calculated in Excel May be most subjec<ve Likely too complex for Excel Likely too complex for Excel Relies heavily on discoun<ng assump<ons 65

66 Discounted Cash Flow Category (Credit Score Range) Current Balance Weighted Averages Contractual FICO Rate Remaining Score Prepay Remaining Term Market Rate Credit Risk Premium Credit Risk Adjusted Market Rate Market Adjustment Credit/ Collateral Adjustment Total Adjustment Fair Price Fair Value AIRES Loans <60 Days Delinquent 36 Mo. Ind. Auto 1,822, % % 105,749 (106,138) (389) 100.0% 1,822, , % % 0.10% 2.70% 5,215 (480) 4, % 551, , % % 1.41% 4.01% 4,279 (2,048) 2, % 157, , % % 3.53% 6.13% 7,975 (5,866) 2, % 183, , % % 7.37% 9.97% 19,312 (16,419) 2, % 240, , % % 12.17% 14.77% 15,621 (18,508) (2,887) 98.2% 158, , % % 13.61% 16.21% 38,603 (46,230) (7,626) 97.9% 355, , % % 15.00% 17.60% 5,690 (6,972) (1,282) 97.5% 50,788 No Score 124, % 28 N/A % 7.37% 9.97% 9,053 (9,616) (563) 99.5% 123,991 Updated credit scores as of valua%on date Result based on Condi%onal Prepayment Rate ( CPR ) Similar to Risk Based Pricing E.g. Life of loan ALLL adjustment (e.g. CECL) 66

67 Approaches Pros and Cons Loss Rate Approach Vintage Analysis PD and LGD Discounted Cash Flow May be calculated in Excel Not difficult to understand Improves credit risk management process Already in use for merger accoun<ng Similar to current ALLL approach for most ins<tu<ons More granular data lends beoer for forecas<ng Deeper insight into loss driving factors Only short term data is needed Difficult to forecast forward Long-term historical data required Significant amount of data required Simple versions may be calculated in Excel May be most subjec<ve Likely too complex for Excel Likely too complex for Excel Relies heavily on discoun<ng assump<ons 67

68 Elements of CECL Model Historical Events Current Conditions Reasonable and Supportable Forecasts 68

69 Environmental Factor Analysis Actual Loan Consumer Date Charge-Offs CPI Unemployment Sentiment Home Price Index Affordability Index Rate Rate 6/30/ % /31/ % /30/ % /31/ % /30/ % /31/ % /30/ % /31/ % /30/ % /31/ % Housing Home Ownership Homeowner Vacancy Correlation to Charge-Offs:

70 Environmental Factors Sources Factor CPI Unemployment Consumer Sentiment Home Price Index Housing Affordability Index Home Ownership Rate Homeowner Vacancy Rate Delinquency: Residential Real Estate Delinquency: All Consumer Delinquency: Commercial Real Estate C&I Charge-Off Rate Example Source: Consumer Price Index: Total All Items for the United States. Bureau of Labor Statistics (bls.gov) University of Michigan: Consumer Sentiment. Composit Home Price Index National Association of REALTORS Federalreserve.gov Data After 9/30/2012 is from

71 Thank You! Bill Astrab, CPA Senior Audit Manager Stephen LaBarbera, CPA Audit Manager