Cherry, Bekaert & Holl, L.L.P. The Allowance for Loan Losses Current Cid Hickman, Partner, Industry Leader Services Group chickman@cbh.com www.cbh.com 919.782.1040
Agenda Current Bank Performance Framework, Modeling Identification Questions & Comments 2
Bank Performance
Modeling Identification Earnings Improvement Trend Reaches Three-Year Mark The benefits of reduced provision for loan losses outweighed the drag from declining net interest margins, as insured institutions posted a 12 th consecutive year-over-year increase in quarterly net income. Banks earned $34.5 billion in Q2 2012, a $5.9 billion (20.7 percent) increase compared with Q2 2011. The average return on assets (ROA) rose to 0.99 percent from 0.85 percent a year earlier the third-highest quarterly ROA for the industry since second quarter 2007. Source: FDIC Quarterly Banking Profile, Second Quarter 2012 4
Modeling Identification Banks Reduce Loan Loss Provision to Five-Year Low Provisions for loan lease losses were $14.2 billion in the Q2 2012 representing a decline of $5 billion (26.2 percent) from Q2 2011 the smallest quarterly total in five years. Net charge-offs totaled $20.5 billion in the Q2 2012, an $8.4 billion (29.1 percent) reduction when compared to Q2 2011. This is the eighth consecutive quarter that charge-offs have declined from year-earlier levels represents the lowest quarterly charge-off total since first quarter 2008. Charge-offs were lower in all major loan categories with the largest yearover-year declines in credit cards. Noncurrent loan balances declined for the ninth consecutive quarter, falling by $12.9 billion (4.2 percent). Source: FDIC Quarterly Banking Profile, Second Quarter 2012 5
Quarterly Net Charge-Offs vs. Loan Loss Provisions All Institutions 2008 2012 Source: FDIC.gov 71 Loss Provisions ($ billions) 67 Net Charge offs ($ billions) 61 63 63 55 53 50 51 51 52 49 50 37 39 38 44 40 42 35 33 33 26 28 29 27 26 20 21 22 19 19 20 20 14 14 03/08 06/08 09/08 12/08 03/09 06/09 09/09 12/09 03/10 06/10 09/10 12/10 03/11 06/11 09/11 12/11 03/12 06/12 80 70 60 50 40 30 20 10 0 6
Condition Ratios By Asset Size to Loans Leases 2 nd Quarter 2012 vs. 2011 3.50% 3.00% 2.84% 2nd Quarter 2011 2nd Quarter 2012 3.10% 2.94% 3.16% 2.84% 2.50% 2.35% 2.13% 2.49% 2.58% 2.00% 1.50% 1.76% 1.80% 1.93% 1.84% 1.95% 1.00% 0.50% 0.00% All Insured Institutions Less than $100 Million $100 Million to $1 Billion $1 Billion to $10 Billion Greater than $10 Billion Southeast All Institutions Source: FDIC.gov North Carolina All Institutions 7
Condition Ratios By Asset Size Noncurrent Loans plus OREO to Assets 2 nd Quarter 2012 vs. 2011 4.50% 4.00% 2nd Quarter 2011 2nd Quarter 2012 3.79% 4.15% 4.29% 3.50% 3.34% 3.35% 3.62% 3.00% 2.50% 2.75% 2.40% 2.72% 2.86% 2.60% 2.00% 2.40% 2.21% 2.30% 1.50% 1.00% 0.50% 0.00% All Insured Institutions Less than $100 Million $100 Million to $1 Billion $1 Billion to $10 Billion Greater than $10 Billion Southeast All Institutions Source: FDIC.gov North Carolina All Institutions 8
Modeling Identification Relevant accounting guidance related to the 10
Modeling Identification Allowance for Loan Lease Losses An allowance methodology is a system that a bank designs implements to reasonably identify estimated credit losses as of the financial statement date. Traditional commercial banks incorporate segmentation to determine needed allowances related to classifications or internal loan grades. 11
Modeling Identification Allowance for Loan Lease Losses ONE OF THE MOST SIGNIFICANT ACCOUNTING ESTIMATES FOR FINANCIAL INSTITUTIONS An assessment of the appropriateness of allowances for loan losses is critical to the safety soundness of a financial institution. Allowance levels must be sufficient to absorb estimated credit losses within the portfolio. The term estimated credit losses means an estimate of the current amount of loans that it is probable the financial institution will be unable to collect; that is, net charge-offs that are likely to be realized for a loan or group of loans given facts circumstances as of the evaluation date. 12
Modeling Identification Allowance for Loan Lease Losses Methods to evaluate allowances vary are influenced by factors such as: The size of the financial institution Organizational structure Business environment strategies Management style Loan portfolio characteristics Administration procedures Management information systems Regulatory environment 13
Modeling Identification Allowance for Loan Lease Losses methodologies should be accurate, credible, adaptable, executable, supportable, should generally include: Detailed portfolio analyses performed on a regular basis Consideration of all loans, whether current or delinquent For loans not reviewed on an individual basis, segmentation of the portfolio into groups of loans with similar risk characteristics Consideration of all known relevant factors affecting collectability Consistently applied but appropriately modified for new collectability factors Consideration of the particular risks inherent in different kinds of products Consideration of collateral values (less costs to sell), where applicable A requirement that analyses, estimates, reviews other methodology functions are to be performed by competent well-trained personnel The use of current reliable data Validation on a regular basis 14
Modeling
Modeling Identification Use of Models Creates New Risk Development Risks Dependence on estimation data Underlying theoretical assumptions may not always hold Outlier years may get discarded based on management judgment Implementation Risks Increased documentation Quality control over model input Model validation back testing Model Usage Risks Over reliance on model Inadequate model maintenance to adjust for changes Model predicts results different than actual Documentation of qualitative subjective inputs 16
Modeling Identification Modeling Issues Challenges Policies Procedures Lack of communication between business units Lack of written polices processes Lack of monitoring reporting Inadequate risk rating model coverage maintenance Ineffective loss mitigation collection strategies Accuracy of collateral valuations 17
Modeling Identification Modeling Issues Challenges Data Data that cannot be linked or aggregated Lack of internal loss history to support model Assumed relationships to external data Lack of granular data Stale credit ratings borrower financial information Historical data may not be representative of the future Lack of review monitoring by individuals that own the process 18
Modeling Identification Modeling Issues Challenges Methodology Inconsistent definitions of risk factors model components Unidentified untested assumptions (assumed relationships to past experience) Inadequate segmentation of loan portfolio Inclusion of inappropriate balances (unfunded commitments) Lack of written polices processes Lack of model validation, back-testing stress testing 19
Modeling Identification Modeling Issues Challenges Risk Rating Systems Inaccurate untimely assignment of risk ratings Insufficient granularity of risk rating categories Lack of independence in the risk rating process Lack of clear rating definitions criteria Inconsistency of ratings over time Assumed mapping with external ratings Insufficient calibration validation 20
Modeling Identification Modeling Issues Challenges Documentation Insufficient or ambiguous policies procedures Lack of performance measurement Unclear underlying or implied assumptions in the application of the methodology Management adjustments including qualitative factors not supported 21
Modeling Identification Modeling Issues Challenges Adhering to GAAP while ensuring that reserves are sufficient to cover inherent losses. Both GAAP the regulators allow management to use discretion in estimating credit losses. Management is allowed to include all known qualitative factors impacting credit risk as of the balance sheet date. It is appropriate for management to identify include all the known qualitative factors impacting credit risk test adequacy of model at least quarterly. Dramatic deviations between expected actual portfolio performance may justify a more aggressive application of subjective inputs into the provision process. Source: Enterprise Risk Management, November 2008, The RMA Journal. 22
Modeling Identification Modeling Issues Challenges Determine an appropriate balance between objective subjective inputs in accordance with GAAP Dialogue between finance, accounting, credit risk management functions within the bank. Adjustments to historical loss rates as a result of subjective inputs needs to be supported documented. Documentation should include a quantitative qualitative analysis with underlying supportive information. Based on relevant facts circumstances as of the reporting date. Supporting evidence could include relevant articles from newspapers other publications that describe economic events affecting a particular geographic area, economic reports data, notes from discussions with borrowers. Source: 2006 Interagency Policy Statement on Allowance for Loan Lease Losses 23
Qualitative Factors Illustrative Metrics Supporting Qualitative Factors Economic Trends National regional GDP National regional unemployment rates Housing starts permits Durable goods sales Manufacturing index Portfolio Concentration Risk Portfolio Quality Characteristics Industry segments Commercial real estate Residential real estate Acquisition development Overall delinquency rates trends Benchmarking vs. peers Comparison to historical performance in similar credit cycles Regional foreclosure REO trends Quality growth of new originations Model Risk Strength of governance framework Consistency of output Level of segregation Sophistication of modeling techniques Data quality Source: Enterprise Risk Management, November 2008, The RMA Journal 24
Modeling Identification Modeling Issues Challenges Common reasons for poor model performance are: Infrequent stress testing Model errors not caught due to lack of model validation review by independent reviewer Failure to improve or recalibrate models for known explanatory factors To assess whether the model is appropriate for the current environment, management should consider stress testing validating current models. Source: Enterprise Risk Management, November 2008, The RMA Journal. 25
Modeling Identification Modeling Issues Challenges Stress Testing Industry best practice Underst extreme case scenarios Estimates range of possible losses for risk factors impacting the portfolio (Interest rate changes, borrower credit ratings, collateral values) Sensitivity of loss estimates to various model assumptions provides comfort about the reasonableness of the estimates Adjustments to model assumptions should be directionally consistent with underlying portfolio characteristics external factors effecting loss estimates Source: Enterprise Risk Management, November 2008, The RMA Journal. 26
Modeling Identification Modeling Issues Challenges Model Validation -OCC Bulletin 2000-16, the best defense against model risk is the implementation of a sound model validation framework that includes a robust validation policy appropriate independent review More than just testing the mathematics of the model Appropriateness of the model methodologies Accuracy of data inputs assumptions Controls around data model processing reporting Validation of data model security as well as model version control. Independently tested when there is deterioration in credit quality or significant changes in quantitative qualitative factors Source: Enterprise Risk Management, November 2008, The RMA Journal. 27
Modeling Identification Methodologies Roll-Rate Models (Migration or Flow models) Predicts losses based on delinquency Most roll-rate models assume delinquency is the only loss event Less effective at providing an estimation of loss exposure in the loans that are not delinquent Typically used for consumer loans, credit cards, residential mortgage 28
Modeling Identification Methodologies Average Charge-Off Method Provides an estimate of annual charge-offs based on past performance Easy to apply when portfolio is appropriately segregated (risk rating, industry, vintage, etc.) Assumes future loss rates will be similar to historical experience Potential for bias if management relies on longer-run averages when conditions are deteriorating on short-run trends at the earliest signs of recovery Charge-Off rates are periodically adjusted to take into consideration current portfolio characteristics other factors effecting future loss rates 29
Modeling Identification Methodologies Vintage Analysis Projected losses are determined based on the age of the loan Common approach for credit cards residential mortgages Patterns or curves are generally predictive for future vintages provided adjustments are made for changes in underwriting criteria economic conditions Reserve is adjusted as the loan ages through the loss curve 30
Modeling Identification Methodologies Regression Analysis Uses economic data internal data to estimate losses Trained statisticians needed to synthesis data forecast loss rates Model validation is critical for complex models due to transparency issues Not widely utilized Ratio Analysis Not a sufficient basis for determining an adequate allowance level Encouraged as a supplemental check for evaluating the overall reasonableness of the allowance 31
Identification
Modeling Identification Identification Impaired loans ASC 310 (formerly FAS 114) Definition of Impaired: A loan is impaired if it is probable that a creditor will not be able to collect all amounts when due according to the contractual terms of the loan agreement All amounts due means both contractual interest payments contractual principal payments will be collected as scheduled in the loan agreement Identification measuring impaired loans is a two step process 33
Modeling Identification Identification Impaired loans ASC 310 (formerly FAS 114) How should an institution identify loans that should be individually evaluated for impairment under ACS 310: Set an objective consistent criteria to identify potentially impaired loans Apply normal review procedures when identifying which loans should be individually evaluated Sources of information to identify loans for evaluation include a specific materiality criterion internally generated listings such as: watch lists, past due reports, overdraft listings, compliance exception reports; loan files lacking current financial data related to borrowers guarantors; loans secured by collateral that is not readily marketable or that is susceptible to deterioration in realizable value; loans to borrowers in industries experiencing economic instability. Source: 2006 Interagency Policy Statement on Allowance for Loan Lease Losses (Question Answers) 34
Modeling Identification Identification Impaired loans ASC 310 (formerly FAS 114) Substard classification does not automatically meet the definition of impaired in accordance with ASC 310. Loans classified as substard are characterized by the distinct possibility that the bank will sustain a loss if the weakness deficiencies are not corrected Loss potential, while existing in the majority of substard credits, does not have to exist in an individual credit. Generally, however, a bank would typically be unable to collect all amounts due in accordance with the contractual terms for loans significantly past due, on nonaccrual status, classified doubtful. Therefore, these loans should most certainly be evaluated under ASC 310 Source: SRC Insights: First Quarter 2011 35
Modeling Identification Identification Impaired loans ASC 310 (formerly FAS 114) If a bank concludes that an individual loan specifically identified for evaluation is not impaired under ASC 310 that loan should be included in the assessment of the under ASC 450 (formerly FAS 5) Bank should estimate credit losses under ASC 310 only for loans individually evaluated determined to be impaired. If a bank assess an individual loan under ASC 310 determines that it is impaired, but it measures the amount of impairment as zero, NO additional loss recognition is appropriate under ASC 450 Source: 2006 Interagency Policy Statement on Allowance for Loan Lease Losses (Question Answers) 36
Modeling Identification Acceptable Methods for Impairment ASC 310 (formerly FAS 114) When measuring impairment of a loan in accordance with ASC 310 when is it appropriate to utilize the Present value of future cash flows, the Fair market value of collateral or the Observable market price Present value of future cash flows is the preferred methodology, except as a practical expedient, the bank can use an observable market price for the loan If the loan is collateral dependent (repayment is expected to be provided solely by the underlying collateral not from the borrower) the fair market value of the collateral is acceptable When the bank determines that foreclosure is probable, a re-measure of the loan at the fair value of the collateral is required so that loss recognition is not delayed until actual foreclosure 37
Modeling Identification Present Value of Expected Future Cash Flows ASC 310 (formerly FAS 114) Estimate the expected future cash flows of the impaired loan Can estimate a range of either the amount timing of possible cash flows utilize the likelihood of the possible outcomes in determining the best estimate of expected future cash flows Discount the expected future cash flows at the loan s effective interest rate If the loan's contractual interest is variable, the loan's effective interest rate may be calculated based on the factor as it changes over the life of the loan or may be fixed at the rate in effect at the date the loan is impaired. Projections of changes in the factor shall not be made for purposes of determining the effective interest rate or estimating expected future cash flows. 38
Modeling Identification Fair Value of Collateral Method ASC 310 (formerly FAS 114) Consider the appraised value of the collateral as the starting point. The appraised value is only one data point, albeit an important one Consider other factors affecting the current fair value of the collateral since the appraisal was performed Consider whether appraised values on impaired loans are actually being realized when property disposition or loan sales occur Track actual results based upon property type, geographical locations, etc. to develop an appropriate valuation adjustment Avoid broadly applying valuation adjustments in a uniform manner Adjustments should be well documented supported through comparative data information Source: SRC Insights: First Quarter 2011 39
Modeling Identification Documenting Impairment Maintain documents that outline the decision process, valuation techniques utilized, other considerations used to support the level of impairment Include the following items in the impairment analysis: o o o o o An explanation of why the loan was selected for individual review The impairment measurement technique used The measurement calculation A comparison to the current recorded investment The amount of reserve to be included in the Source: SRC Insights: First Quarter 2011 40
Modeling Identification Documenting Impairment Method Present Value of Expected Future Cash Flows Supporting Documentation The amount timing of cash flows The effective rate of interest used to discount the cash flows The basis for determining cash flows, including consideration of current environmental other information reflecting past events current conditions Fair Value of Collateral Method How the fair value was determined, including the use of appraisals, valuation assumptions calculations The supporting rationale for adjustments to appraised values The determination of cost to sell, if applicable The appraisal quality, expertise, independence of the appraiser Observable Market Price The amount, source, date of the observable market price Source: SRC Insights: First Quarter 2011 41
Modeling Identification Charge-off Impairment or Add to Reserve When an incurred loss in confirmed, charge-off is appropriate If it is determined that liquidation of the underlying collateral is the only source of repayment, the loss has most likely been confirmed, the impairment should be charged off If there is a possibility that impairment could be collected through improvement of existing cash flows establishment of a specific reserve would be appropriate 42
Modeling Identification Credit Losses Related to Off-Balance Sheet Credit Risk ASC 825 Instruments An accrual for credit loss on a financial instruments with off-balance-sheet risk shall be recorded separate from a valuation account related to a recognized financial instrument (not included in the but a separate liability account) Credit losses for off-balance-sheet financial instruments shall be deducted from the liability for credit losses in the period in which the liability is settled Off-balance-sheet financial instruments refers to off-balance-sheet loan commitments, stby letters of credit, financial guarantees, other similar instruments with off-balance-sheet credit 43
Modeling Identification Impact of TDRs for Impairment on the Loans whose terms have been modified in TDR transactions should be evaluated for impairment in accordance with ASC 310 (Formerly FAS 114) Includes loans that were originally not subject to that stard prior to the restructuring, such as individual loans that were evaluated in a group of smaller-balance, homogeneous loans collectively evaluated for impairment in accordance with ASC 450 (Formerly FAS 5) If impairment is measured using an estimate of the expected future cash flows, the interest rate used to calculate the present value of those cash flows is based on the original effective interest rate on the loan, not the rate specified in the restructuring agreement 45
Modeling Identification Impact of TDRs for Impairment on the Example 1: Borrower A cannot service his $100,000 loan from the bank. The bears interest at 10 percent, which is also the current market rate. On June 1, the loan is restructured, with interest-only payments of 5 percent required for two years a final payment of $105,000 required at the end of the third year. The present value of the expected payments under the restructured terms, discounted at 10 percent (the original loan interest rate), is $87,500. The loan is neither collateral dependent nor readily marketable. In this example, the difference between the present value of the payments ($87,500) of the restructured loan, discounted at the loan s original rate of interest, the recorded value ($100,000) is the measurement of impairment is recognized through a valuation allowance ($12,500) The loan would subsequently be measured for impairment on a periodic basis the valuation allowance adjusted 46
Modeling Identification Impact of TDRs for Impairment on the Example 2: Consider the same facts as Example 1, except that Borrower A transfers the collateral to a new borrower (Borrower B) not related to Borrower A. The bank accepts Borrower B as the new debtor. The loan with Borrower B provides for interest-only payments of 5 percent for two years a final payment of $105,000 (principal plus interest at 5 percent) at the end of the third year. The fair value of the loan, discounted at a current market rate of interest, is $87,500. ASC 310 requires that the receipt of a loan from a new borrower be accounted for as an exchange of assets. Accordingly, the asset received (new loan) is recorded at its fair value of $87,500. In this example a loss is recognized in the amount of $12,500 the new loan recorded at fair market value The loan would subsequently be evaluated through the bank s normal analysis process 47
Modeling Identification Impact of TDRs for Impairment on the Can a TDR to be collateral dependent immediately following the loan modification (on day 1)? A TDR can be collateral dependent at the time of or immediately after the loan modification A modified loan requiring only a nominal monthly payment from the borrower with no support that the borrower can repay the recorded loan balance may result in a loan that ultimately is repaid only through the liquidation of the underlying collateral If the facts circumstances indicate that the borrower does not have the ability to repay the modified loan or if the terms of the loan are based on future, uncertain events, the loan may be deemed collateral dependent at the time of modification 48
Modeling Identification Impact of TDRs for Impairment on the When would a charge-off be required at the time of restructuring? If a portion of the TDR loan is uncollectible (including forgiveness of principal), the uncollectible amount should be charged off Although a restructuring is designed to improve the collectability of a loan, the bank is still responsible for promptly identifying charging-off identified losses If the difference between the present value of the payments of the restructured loan, discounted at the loan s original rate of interest, the recorded value of the loan ($12,500 in Example 1) was insufficient based on the bank s normal analysis process additional provisions would need to be recognized 49
Modeling Identification Insignificant Delay with Concession Example 3: Bank A makes a acquisition development loan to a real estate developer. The developer is experiencing financial difficulty will default on the loan if the project cannot be refinance or sold. Since the value of the project is less than the loan balance, the loan cannot be refinanced. To facilitate a sale the developer requests a three month extension which was granted by the Bank. In this example, the delay in timing of the payment may be considered insignificant, but the amount of restructured payment subject to delay is significant Creditor expects a significant shortfall in cash flows relative to the contractual amount due when the property is sold because the property is the sole source of repayment there is shortfall 50
Modeling Identification Accounting For Instruments Credit Impairment Joint Project of the FASB IASB Credit Impairment Impairment for financial assets that are debt instruments would follow a three-bucket approach based on deterioration in credit quality The Bucket 1 measurement approach expected losses for those financial assets on which a loss event is expected in the next 12 months. The Bucket 2 3 measurement approach financial assets that have meet the threshold for recognition of lifetime expected credit losses there has been a more than insignificant deterioration in credit quality it is at least reasonably possible that some or all the contractual cash flows may not be collected 52
Modeling Identification Accounting For Instruments Credit Impairment Joint Project of the FASB IASB At the August 1, 2012 FASB Board Meeting the Board: Asked the staff to address stakeholders significant concerns about the understability, operability, auditability of the three-bucket credit impairment model under development Expressed concern that the dual-measurement approach the threebucket model is difficult for users to underst Directed the staff to explore an alternative expected loss model that (a) does not utilize a dual-measurement approach (b) reflects all credit risk in the portfolio Such an approach would create an impairment model that is more understable, operable, auditable 53
Modeling Identification The bank has adequate data capture reporting systems to supply the information necessary to support document its estimate of an appropriate Evaluate any loss estimation models before they are employed modifies the models assumptions, as needed, to ensure that the resulting loss estimates are consistent with GAAP Document support any adjustments made to the models or to the output of the models in determining the estimated credit losses Promptly charge off loans, or portions of loans, that available information confirms to be uncollectible Periodic validation of the methodology should be performed by a party who is independent of the process in order to confirm effectiveness 55
Modeling Identification (Continued) An unallocated loan loss allowance is appropriate when it reflects an estimate of probable losses, determined in accordance with GAAP, is properly supported Changes in the level of the should be directionally consistent with changes in internal external factors, taken as a whole Documentation should include a quantitative qualitative analysis based on facts circumstances as of the reporting date include relevant supporting evidence For impaired loans that are collateral dependent thoroughly document valuation for fair value of collateral 56
Modeling Identification Questions & Comments Contact. Cid Hickman Partner, Industry Leader Services Group chickman@cbh.com 919.825.4202 Todd Batchelor Partner, Services Group tbatchelor@cbh.com 919.825.4264 Robert Hawley Senior Manager, Services Group rhawley@cbh.com 919.825.4240 Cherry, Bekaert & Holl, L.L.P. www.cbh.com 57