ALLL and the New Estimate of Loan Losses

Transcription

1 ALLL and the New Estimate of Loan Losses An update on the proposed impairment model and improving the measurement of credit losses MICH ARATEN, MANAGING DIRECTOR, CREDIT RISK CAPITAL ADVISORY CHRIS HENKEL, DIRECTOR, MOODY S ANALYTICS OCTOBER 2013

2 Provisioning for loan losses consumes a significant portion of the banking industry s net operating revenue Loan Loss Provision as % of Net Operating Revenue (all FDIC-Insured Institutions) 37.94% 40% 35% 30% 25% 20% 15% 5.18% 2Q % 10% 5% 0% Source: FDIC 2

3 Despite the rapid provisioning during the crisis, the ratio of reserves to noncurrent loans continued to fall 4.00% 200% 3.50% Reserves/TL Reserves/NCL 3.51% 180% 3.00% 160% 140% 2.50% 120% 2.00% 100% 1.50% 80% 1.00% 0.64% 60% 40% 0.50% 20% 0.00% 0% Source: FDIC 3

4 Agenda 1. Brief Review of Existing Guidance 2. Overview of FASB s Proposed Current Expected Credit Loss Model 3. Analytical Considerations and Loan Loss Reserves 4. Stress Testing and Reserves 4

5 1 Brief Review of Existing Guidance 5

6 An appropriate ALLL, in accordance with GAAP, should reflect an estimate of probable credit losses Estimated credit losses means an estimate of the current amount of loans that is probable the institution will be unable to collect given facts and circumstances as of the evaluation date. Thus, estimate credit losses represent charge-offs that are likely to be realized for a loan or group of loans. - Interagency guidance,

7 The principal sources of guidance on GAAP accounting for credit losses are FAS 5 and FAS 114 Measurement of Estimated Credit Losses Loan Portfolio Segmented Risk Pools Impaired? No FAS 5 Yes FAS 114 Unallocated Portion of the ALLL that is not attributed to specific segments of the loan portfolio PV of FCF Mkt. Price FV of Coll. 7

8 8 In 2009, FASB codified the accounting standards for recognition of credit losses ASC Loss Contingencies» Accrue an amount that appears to be a better estimate than others within a range of estimates» Accrual vs. Disclosure If it is probable that a loss will incur and the amount can be reasonably estimated, it should be accrued in the financial statements If it is reasonably possible that a loss will incur, it should be disclosed in the notes without recognition in the financial statements If the possibility of loss is remote, disclosure is not required ASC Receivables» A loan is impaired when it is probable that all amounts due from a loan are impaired» To determine whether a loan is impaired, the institution should apply its normal loan/credit review process» Impairment loss = Carrying amount of the loan, less: Fair value of the collateral (collateral dependent loans); or PV of expected future cash flows from a loan; or Observable market price of the loan 8

9 The incurred loss approach is believed to interfere with the timely recognition of credit losses Concerns Over the Current Incurred Loss Model» It prevents banks from provisioning for an impaired asset until a triggering event occurs» Banks must wait until the triggering event has already occurred before they recognize the loss» By waiting, the model precludes banks from provisioning for risks the bank can reasonably anticipate to occur» It leads to pro-cyclicality and delayed loss recognition» Changes in the probabilities of loss and of loss exposures should be reflected in the ALLL» The OCC supports FASB s proposed expected loss model over the current incurred loss impairment approach 9

10 2 Overview of FASB s proposed Current Expected Credit Loss Model (CECL) 10

11 Evolution of a new impairment model Over the last five years, the accounting community has worked to provide more actionable information about the expected credit losses on financial assets October 2008 Joint effort b/w FASB and IASB to address reporting issues arising from the global financial crisis Evolution of Subtopic , Financial Instruments Credit Losses (superseding ASC (SFAS 114) and (SFAS 5) - among others) November 2009 IASB published Exposure Draft, adding further support for a forward-looking measure of ECL January 2011 FASB and IASB published a supplementary document introducing Good Book and Bad Book distinction December 2012* FASB published the Exposure Draft Proposed Accounting Standards Update, Financial Instruments Credit Losses. Introduced the CECL. July 2009 Financial Crisis Advisory Group (FCAG) published report on delayed recognition of losses and complexity with different impairment approaches. Included forward-looking information. May 2010 FASB published a proposed ASU to ECL»Remaining life»cash flow based»economic conditions remain unchanged July 2012 FASB and IASB jointly released the threebucket impairment model whereby credit instruments would have had different measurement approaches and migration criteria across buckets *Current proposal; IASB had not concluded deliberation on credit losses at the time of release Source: FASB May 2013 Comment period ended 11

12 The proposed accounting standards update reflects several core objectives Objectives of the proposed update» More timely recognition of credit losses» Greater transparency regarding the expected credit losses» Improved understanding of the realizability of assets and the inherent credit risk in the portfolio» Improved understanding of credit risk changes that have taken place during the period» Improved understanding of purchased credit-impaired financial assets» Improved understanding and comparability of interest income» Enhanced consistency when credit impairment is measured at the individual asset level as compared with at the portfolio level Source: FASB 12

13 Working towards these standards will require a blend of judgment and empirical evidence» The allowance for credit losses (ACL) should be management s best estimate of the PV of all contractual cash flows that are not expected to be collected on an asset or group of like assets as of the financial statement date The timing and amount of the CFs is not required under the new proposal Management Judgment Empirical Evidence» The ECL should take into account: Historical loss experience (NCOs) with similar assets need to appropriately segment Current conditions prevailing credit cycle and business environment (including macroeconomic factors, collateral values, borrower behavior, underwriting standards, etc.) Reasonable and supportable forecasts (**New**) Time value of money, either explicitly or implicitly Source: FASB 13

14 The approach to estimating credit loss is not one-sizefits-all, but there are minimum requirements» Specific approaches are not mandated but should be consistent and appropriate for the portfolio it is applied to» Minimum requirements (for historical statistics): Consistent definition of default Definition of loss (i.e., amount charged off) Method for weighting historical experience (i.e., volumeweighted or equal-weighted) Method for adjusting loss statistics for recoveries How expected prepayments affect the allowance for ECL Incorporating the time value of money» Default probabilities and loss severities are not linear, therefore it is inappropriate to gross up a one-year measure over the remaining term Source: FASB 14

15 Example of non-linearity of default probabilities using cumulative measures Firm A Firm B A cumulative EDF credit measure gives the probability of default over that time period. For example, a five year cumulative EDF credit measure of 9.64% means that that company has a 9.64% chance of defaulting over that five year period (perhaps the remaining life of the loan). 15

16 A common measurement approach includes the use of PD, LGD, and EAD along with credit adjustments Performing Rated Loans ($) ( Pass ) PD LGD Credit Risk Adjustment RESERVES FOR PERFORMING LOANS Impaired Loans ($) (Pooled basis) EL Factor RESERVES FOR IMPAIRED LOANS Impaired Loans ($) (Individual) Uncollected Cash Flows RESERVES FOR IMPAIRED LOANS Special Reserves ($) (Mgmt Judgment) ADDITIONAL RESERVES TOTAL RESERVES 16

17 Institutions will need to estimate expected loss over the life of the loan, and also account for current conditions On average, the amount a lender could potentially lose depends on three things $45M how likely the borrower is to go into default the estimate of loss (1-recovery) should default occur = x x 3 % 30 likelihood on the dollar the exposure amount at the time of default $5MM of the $10MM originally lent Expected Loss Probability of Default Loss Given Default EL = PD x LGD x EAD* Exposure at Default» These estimates will need to be further adjusted for current economic conditions and the forecasted direction of the economy» In addition to time horizon, another dimension for consideration is the PD measurement (i.e., Point-in-Time (PIT) or Through-the-Cycle (TTC)) *For ALLL purposes, the EAD is typically the outstanding loan amount as of the financial reporting date. A different but related reserve is held for unfunded commitments 17

18 Moreover, risk measures can be expressed in terms of Point-in-Time or Through-the-Cycle 10.00% Feb. 09: 9.92% Median EDF for B rated companies 9.00% 8.00% 7.00% 6.00% 5.00% 4.00% 3.00% Median: 1.35% 2.00% 1.00% 0.00% Aug. 13: 0.31% Source: Moody s CreditEdge 18

19 Recent Developments Decisions reached to date during deliberations of CECL (through Sept. 27, 2013)» Revert to historical average loss experience for future periods beyond supportable forecasts» Consider prepayments but not extensions, renewals, and modifications (other than TDR)» Recognize risk of loss, even if remote, unless amount of loss would be zero» Can use loss-rate models, PD methods, or a provision matrix in addition to DCF models» Final guidance (TBA) will include guidance on reasonable and supportable forecasts Clarifications Regarding an Entity s Estimate of Expected Loss 19

20 While CECL brings noted improvements, FASB s new impairment model has been met with some dissenters Commonly Expressed Concerns» The operational impact could be significant» Stakeholders, such as regulators, accountants, investors, and the SEC, do not always share a common interest» Introduces a life-of-loan concept which is said to conflict with the conceptual framework» A forward-looking measure may be very difficult to support the estimates» The impact on current allowance levels An increase of 30% to 300% to the allowance, in addition to a potential one-time increase At a time when banks are adding capital in order to meet new regulatory requirements» Favor for an alternative, such as a Banking Impairment Model (BIM) 20

21 How does FASB s CECL align with the IASB s proposed Expected Credit Loss Model? Common Attributes» Removal of the incurred loss trigger for recognition» Lifetime ECL are the expected shortfalls in contractual cash flows» An estimate of ECL will reflect the probability that a credit loss might occur» The estimate will be based upon use of the same information» The amount of ECL should be the same for financial instruments that have deteriorated significantly in credit quality Divergent Attributes» The IASB s model includes three stages: 1. No significant deterioration (12 months ECL are recognized) 2. Significant deterioration (lifetime ECL are recognized) 3. Objective evidence of impairment (lifetime ECL are recognized)» The FASB CECL has no distinction for deterioration in credit quality; all measured at lifetime ECL» Timing difference in the recognition of ECL Note: The IASB issued an Exposure Draft, Financial Instruments: Expected Credit Losses, on March 7, The comment period ended on July 5, 2013 Remains a joint project between FASB and IASB, as they work together to deliberate on comment letters and potentially align on divergent views Source: IIFRS and IASB 21

22 3 Analytical Considerations for Loan Loss Reserves 22

23 As previously mentioned, the ALLL consists of three distinct components» Specific reserve for non performing loans» Expected losses for performing loans» Credit risk adjustment applied to expected losses 23

24 The impairment for nonperforming loans is usually on an asset-specific basis» Asset-Specific Reserve (ASC / FAS114)» Estimate periodic cash flows and discount at contract rate of interest. Large exposures: Estimate on a scenario basis Smaller exposures: Estimate conditional probability of remaining on nonaccrual, given amount of time already on non-accrual-from historical data 24

25 Considerations for estimating PD as it relates to the estimate of EL for allowance purposes» Expected Losses= PD x LGD x EAD over contractual term» Contractual term could be shortened based on expected prepayment ( expected life )» PD could be developed from historical data associated with current rating status incorporating transitions to other ratings» Evaluate Expected Default Frequency (EDFs) (PIT) vs. (TTC) PDs» Expressed as cumulative PD over expected life 25

26 Similarly, the estimate of LGD for loan loss reserving has unique attributes unto itself» LGDs for allowance are different from LGDs for regulatory capital» Not downturn LGDs» Exclude workout costs» Exclude AIR (accrued interest receivable)» Can use overall average discount rate, possibly based on contractual rate 26

27 There are also several treatment options for EAD» EAD for on-balance sheet exposures = outstanding balance» EAD for revolving credits based on unused portion» Add Loan Equivalent (LEQ) factor to EAD» Can either include this in the Allowance for Loan Loss Reserves or Allowance for Lending Related Commitments 27

28 Historical averages alone may not be sufficient, warranting a credit risk adjustment to the EL» Historical averages may not adequately consider the current point or forecasted direction of the economic cycle» Credit risk adjustment modifies the base EL to reflect reasonable and supportable forecasts about the collectability of future cash flows» Management evaluates the current point in the economic cycle, as well as other important current credit indicators such as borrower behavior and collateral values, how current underwriting standards compare with those in the base EL, and recent trends in economic conditions 28

29 Historical variability of PDs Historical Rating agency data Aaa Aa A Baa Ba B Caa Mean PD 0% 0.01% 0.03% 0.21% 1.12% 5.16% 22.56% One Standard Deviation 0% 0.08% 0.08% 0.31% 1.11% 3.47% 16.49% Coefficient of Variation CV= s/m) 0% 0% 200% 146% 99% 67% 73% LEQ Distribution 6% 4% 25% 30% 20% 10% 5% Weighted average upper bound (1s/m) 120% Weighted average lower bound (0.5s/m) 60% 29

30 Construct scorecard» Factors to be considered:» Portfolio regional and industry concentrations» Current point in economic cycle (PIT vs. TTC indicators---credit Edge and RiskCalc)» Forecast of macro factors» Underwriting characteristics of current portfolio 30

31 Apply scorecard result to range EL+120%EL a 1 F 1 a 2 F 2 a 2 F 3 a 4 F 4 a 5 F 5 a 6 F 6 EL-60%EL 31

32 1990Q2 1990Q4 1991Q2 1991Q4 1992Q2 1992Q4 1993Q2 1993Q4 1994Q2 1994Q4 1995Q2 1995Q4 1996Q2 1996Q4 1997Q2 1997Q4 1998Q2 1998Q4 1999Q2 1999Q4 2000Q2 2000Q4 2001Q2 2001Q4 2002Q2 2002Q4 2003Q2 2003Q4 2004Q2 2004Q4 2005Q2 2005Q4 2006Q2 2006Q4 2007Q2 2007Q4 2008Q2 2008Q4 2009Q2 2009Q4 2010Q2 2010Q4 2011Q2 2011Q4 2012Q2 Fed Reserve Survey Lending Standards Report covers the percentage of firms surveyed who state that their lending standards were tighter in the current quarter vs. prior quarter 100 Fed Survey of Net Tightening Standards Net(Tight-Loose) Stds Previous Qtr 32

33 1990Q2 1990Q4 1991Q2 1991Q4 1992Q2 1992Q4 1993Q2 1993Q4 1994Q2 1994Q4 1995Q2 1995Q4 1996Q2 1996Q4 1997Q2 1997Q4 1998Q2 1998Q4 1999Q2 1999Q4 2000Q2 2000Q4 2001Q2 2001Q4 2002Q2 2002Q4 2003Q2 2003Q4 2004Q2 2004Q4 2005Q2 2005Q4 2006Q2 2006Q4 2007Q2 2007Q4 2008Q2 2008Q4 2009Q2 2009Q4 2010Q2 2010Q4 2011Q2 2011Q4 2012Q2 Charge-offs follow tightening Tighter Lending Standards Lead C & I Chargeoffs by 1 year Adjusted R 2 = 82% Net(Tight-Loose) Stds Previous Qtr ChgOff Rates 1 Yr later 33

34 Covenant quality index» Historical information on covenant quality can also help determining underwriting standards embedded in current portfolio (e.g., change of control, structural subordination, cash leakage, leveraging)» Moody s covenant quality index: score summarizes protection to bond holders ranging - CQ1 (strong) to CQ5 (weak) 34

35 Jan-11 Feb-11 Mar-11 Apr-11 May-11 Jun-11 Jul-11 Aug-11 Sep-11 Oct-11 Nov-11 Dec-11 Jan-12 Feb-12 Mar-12 Apr-12 May-12 Jun-12 Jul-12 Aug-12 Sep-12 Oct-12 Nov-12 Dec-12 Jan-13 Feb-13 Mar-13 Apr-13 May-13 Jun-13 Jul-13 Aug-13 Moody s covenant quality 4.40 Covenant Quality Index

36 4 Stress Testing and Reserves 36

37 Stress testing and Reserves» Stress testing (CCAR) evaluates impact of macroeconomic factors on bank profitability and on regulatory capital ratios» Stress tests have an impact on increased PDs, accelerating downward rating transitions, higher loss severities, and increased likelihood of draw down on unused commitments» Maximum losses obtained from stress tests should be significantly above the maximum credit risk adjustment 37

38 Rating Distribution Credit portfolio migration under stress 0.25 Credit Portfolio Migration Under Stress Base Credit Expected Recession Risk Rating Loss 38

39 Reserves Reserve requirement under stress 2500 Reserve Requirement Under Stress Expected Recession Risk Rating

40 RAROC considerations» Economic capital and risk adjusted return on economic capital has been the guiding criteria since the mid-90s as portfolio measurement and management have advanced EDF measures (CreditEdge/Credit Monitor/RiskCalc) Portfolio models (Risk Frontier/Portfolio Manager)» Regulatory capital, stress tests, and liquidity measures now serve as constraints on return on economic capital objectives» As Loan Loss Reserve changes become implemented, care needs to be taken that these are not part of RAROC decisions 40

41 Questions? 41

42 moodys.com» Christian Henkel Director Moody s Analytics Enterprise Risk Solutions christian.henkel@moodys.com » Mich Araten Managing Director Credit Risk Capital Advisory araten@aol.com