ALLL Today: Challenges & Solutions. September 14 2016
ALLL Today: Challenges & Solutions. September 15, 2016 P R E S E N T E D B Y Tim McPeak Executive Risk Management Consultant Sageworks
Agenda. Current Market Overview Challenges & Solutions in Today s ALLL:» Data» Segmentation» Individual Impairments» Quantitative Issues» Qualitative Issues ALLL Looking Forward Key takeaways 3
Current Market: Reserve Levels. ALLL as % of total loans US Banks (to 6/30/2015) ALLL/Total Loans 1.9% 1.5% 2010 2011 2012 2013 2014 2015 1.2% 4
Current Market Over Reserved? Current reserve level high compared with annualized charge-offs We compared June 30, 2015 ALLL levels against the total of the next four quarters of charge offs for US banks More than 50% of banks had charge-offs totaling 10% or less of starting ALLL» More than 70% totaled 20% or less Average loan loss coverage ratio (ALLL/non-accruals) was 27.3 times at June 30, 2016 Strong case that US banks are over reserved today 5
Current Market Over Reserved? 6
Current Market Relying on the Subjective Low loss environment in recent years limits quantitative portion of calculation To maintain appropriate reserve levels, pressure falls on qualitative and environmental factors» Not forward looking (yet!)» How much embedded loss is really in the portfolio? Within Sageworks ALLL client base, Q factors make up 78% of total pooled reserves at June 30, 2016 7
Agenda. Current Market Overview Challenges & Solutions in Today s ALLL:» Data» Segmentation» Individual Impairments» Quantitative Issues» Qualitative Issues Looking forward Key takeaways 8
Challenges & Solutions 9
Challenges: Data. Not just a concern for CECL, issues are prevalent today:» Methodology options may be constrained by available data Common data issues include:» Disparate loan systems/loan servicers» Inconsistent loan coding through time pool jumpers» Handling of renewals and modified loans» Lack of accurate historical loan-level data 10
Challenges: Data. Issues in historical data often relate to changes in:» Core systems» Risk rating scales» Mergers/acquisitions» Changes in key personnel» General lack of attention/reconciliation 11
Solutions Data. Consider a data warehouse set-up to synthesize multiple loan systems and servicers:» Consistent source for internal/external reporting» Long term storage of loan-level history At a minimum, create and store loan-level extracts of month-end snapshots:» Usable formats (not PDF!)» Many core accounting systems store only 12 13 months Yes, this is a CECL issue as well:» Consider the different methodologies you may employ for CECL» Goal is to avoid lack of data as roadblock 12
Challenges: Pool Segmentation. Effective pooling of the portfolio is a balancing act Fewer, larger pools:» May lack needed granularity» Can dilute loss rates for loan types with unique characteristics Too many pools:» Can weaken statistical validity of pooled analysis» Individual charge-offs/recoveries can have a disproportionate effect on loss rate calculation Sub-segmentation needs to be considered 13
Solutions: Pool Segmentation. There is no one right way to segment portfolio Consider Federal Call Code as starting point:» Consistent, standardized format» Can break-out additional segments/sub-pools as needed Some additional segmentation to reflect risk is appropriate:» Risk rating/grade» Days past due buckets» FICO band 14
Solutions: Pool Segmentation. Loans with unique characteristics should be broken out:» Promotional lending programs» Shared national credits» Indirect and Asset Based Lending» Discontinued lines of business (run-offs) 15
Challenges: Individual Impairments. Levels of impaired loans have steadily declined since 2011 Despite lower levels, still a common area of scrutiny from auditors and regulators NPL/Total Loans 3.00% 2.50% 2.00% 1.50% 2010 2011 2012 2013 2014 2015 1.00% 16
Challenges: Impairments. For collateral dependent credits, scrutiny around:» Appraisal details and assumptions» Selling costs» Consistency in application of policies» Is it truly collateral dependent? For present value of cash flows based impairments:» Ongoing scrutiny of balloon payment assumptions» Justification of expected payments» Default assumptions? 17
Solutions: Impairments. Clear, consistently applied policies around key issues:» What constitutes collateral dependence» Appraisal management process» Selling/liquidation assumptions For cash flow based valuations:» Possible discounting of contractual payments (probability of default rates)» Lesser of calculated balloon payment or liquidation value of underlying collateral for terminal value» For larger credits, credit analysis supporting coverage of expected debt service 18
Challenges: Pooled Impaired? Often impractical to individually review all loans considered impaired by definition Common to see policy threshold on exposure amount» Materiality threshold for small balance loans» Loans above threshold individually analyzed for specific reserves Pooling small impaired may make practical sense, but:» Some confusion over current GAAP» Collect in one impaired pool, or pool of origin?» Loss rate method? 19
Solutions: Pooled Impaired. Applicable guidance on pooling impaired loans: ASC 310-10-35-21:» some impaired loans may have risk characteristics in common with other impaired loans. A creditor may aggregate those loans and may use historical statistics, such as average recovery period and average amount recovered, along with a composite effective interest rate as a means of measuring impairment of those loans. 20
Solutions: Pooled Impaired. OCC Bank Accounting Advisory Series, August 2016:» Question: Are large groups of smaller-balance homogeneous loans that are collectively evaluated for impairment within the scope of ASC 310-10-35?» Response: Generally, no ASC 310-10-35 would apply, however, if the terms of any of these loans were modified in a troubled debt restructuring, as defined by ASC 310-40-15. Otherwise, the relevant accounting guidance for these groups of smaller balance homogeneous loans is contained in ASC 450-20. 21
Solutions: Pooled Impaired. What does this guidance tell us?» ASC 310-10-35 does allow for pooling» Must have risk characteristics in common (beyond impairment status)» OCC indicates pooling is appropriate if the smaller loans were modified and are now considered TDRs» Otherwise, any pooled loans subject to ASC 450-20 What loss rate would apply?» Per ASC 310-10-35, average recovery rates and periods can be used as a basis for rates (composite interest rate as discount rate for cash flow valuations)» Otherwise, pool of origin historical loss rates may apply (under ASC 450-20) 22
Challenges: Quantitative Issues. Loss rate methodology matters, but mostly driven by charge-offs Low loss environment good for banking, not as good for your ALLL models Look-back periods can only be stretched so far» Effect of older losses is diluted by longer period averages Many institutions seeing negative loss rates» Recoveries continue to frequently outpace charge-offs 23
Solutions: Loss Rates. Possibly the biggest challenge in most institution s ALLL today Charge-off based models may understate required reserves during a turn in the business cycle» CECL anyone? Options to bolster loss rates include:» Extending look back periods» Period specific weightings» Loss Emergence Period» Loss rate floors? All of the above options have downsides as well 24
Challenges: Loss Emergence Period. Becoming more common in ALLL calculations» With help from some external audit firms Commonly applied as multiplier to average historical loss rates» Time based adjustment to annualized loss rates» Possible to gain additional quantitative reserves» Difficulty in calculating, documenting & defending 25
Challenges: Loss Emergence Period. Calculating LEP:» Must define emergence event (Downgrade, 90+ days, Non-accrual)» Pre-emergence period can be difficult to quantify How should it be applied to other methodologies?» Migration, PD/LGD? 26
Solutions: Loss emergence period. For calculation of LEP:» Downgrades to classified grade, non-accrual and 90+ days delinquent are common emergence events» Measuring length of time between emergence event and loss event (partial or full charge-off) Pre-emergence period:» Added to calculated LEP» Example: How long between job loss and delinquency?» Statistical analysis for Q factors can give insights here lag time for variables that correlate to charge-offs 27
Solutions: Loss emergence period. How does this apply to other loss rate methodologies?» Multiplying LEP by non-annual loss rates doesn t make sense» LEP should inform period length for migration, PD/LGD 28
Challenges: Qualitative Factors. Challenges here are not new, but reliance on Q-factors has increased scrutiny from auditors and regulators Make your qualitative factors more quantitative Requires identification of underlying drivers for risk factors» What data points can we measure that drive losses in the portfolio? Adjustment scales and ranges need to be justified & documented Current period only, not forward looking» What losses are embedded in the portfolio that are not yet recognized? 29
Solutions: Qualitative Factors. Make qualitative factors more quantitative» There is a spectrum here» More basic end is identifying big-picture drivers and ensuring directional consistency» More sophisticated end is statistical based analysis Multivariable regression against charge-offs/npls to identify drivers More quantitative models also risk losing flexibility» Setting appropriate ranges of adjustments is critical» In the end, this is still management s ability to adjust reserve levels 30
Agenda. Current Market Overview Challenges & Solutions in Today s ALLL:» Data» Segmentation» Individual Impairments» Quantitative Issues» Qualitative Issues ALLL Looking Forward Key takeaways 31
ALLL: Looking forward. 32
ALLL: Looking forward. Important to road-test different methodologies in lead up to CECL implementation» Many have applicability today with different inputs» Adopting new methodologies today may not make sense, but they may still provide insights into portfolio Reasonable & supportable forecasts under CECL» Driver based analysis seems somewhat inevitable» Current DFAST banks may have a head start here» Standards of practice will evolve in coming years 33
Key Takeaways. 34
Key Takeaways Focus on the transition to CECL is warranted, but challenges remain within today s incurred loss model Reserve levels are lower, but remain high relative to recent loss experience Led to heavy reliance on subjective aspects of reserve calculations» Increased scrutiny of Q-factors, particularly from auditors» More support and numerical basis for adjustments is becoming the norm» Balance to retain some flexibility for management Evaluating methodology changes for CECL can provide insights into your portfolio today 35
Contact Information & Questions. Tim McPeak Executive Risk Management Consultant Tim.Mcpeak@sageworks.com Websites sageworksanalyst.com ALLL.com 36