bkd national financial services group€¦ · a+b+c 2.19% total cecl lifetime loss rate $ 1,650,000...
TRANSCRIPT
CECL Update: Where are We Now?
BUD HOLLENKAMP, DENVER, BKD, LLP
BKD NATIONAL FINANCIAL SERVICES GROUP
AGENDA
• We are NOT going to discuss how we got here!
• We will NOT spend much time on the standard
• We WILL talk about where institutions are today
• We WILL talk about where institutions should be and why
• We WILL walk through an example of analyzing the data needs, assumptions, pros and cons for a simple loss rate model
CURRENT EXPECTED CREDIT LOSS (CECL)
ASU 2016-13
New Impairment Guidance (CECL)
Incurred/Probable Expected/Lifetime
•Financing receivables
•Held to maturity debt (no more OTTI)
•Loan commitments, guarantees, standby L/C
•Lease receivables
•Reinsurance receivables
•Receivables on repurchase & securities lending agreements
Included
CECL Scope
• Financial assets at fair value
• Available for sale debt (updated model)
• Participant loans defined contribution benefit plans
• Insurance policy loans
• NFP pledges receivables
Excluded
CECL Calculation for Loan Pools
Historical lifetime
credit loss
Current conditions adjustment
Forecast adjustment
Current expected credit loss
By Loan Pool
Effective Date (for Entities with Calendar Year-Ends)
ASU 2016-13 Impairment (CECL)* Type 2020 2021 2022
PBE – SEC filers Interim & annual
PBE – small Interim & annual
All others Annual Interim
*Early adoption permitted for all entities for periods beginning after December 15, 2018
ASU 2016-13 Implementation Dates
CECL: WHERE ARE INSTITUTIONS NOW?
CECL: WHERE INSTITUTIONS SHOULD BE?
CECL Readiness & Implementation Process
Create committee & timeline
Education
Pool segmentation & credit risk
identification
Data inventory & gap analysis
Planning & Readiness
Model(s) selection &
development
Model(s) finalization & parallel run
Modify policies,
procedures, controls & disclosures
Implementation
Overall Implementation Considerations
Document conclusions & process followed in selecting models & complying with the standard
There will be setbacks & roadblocks – accept that fact & be adaptable
The overall implementation process will be scalable based on size & complexity of institutions
Make sure you have an executive sponsor & appropriate expertise involved
Pool Segmentation & Credit Risk Identification
One of the key principles of CECL is pooling based on similar risk characteristics
The standard provides example risk characteristics to consider (not all inclusive)
Institutions should re-evaluate current pooling for an expected lifetime loss approach
Institutions should take the time to document the following:
❖ Credit risk attributes deemed crucial in underwriting and loan policies
❖ How management monitors credit risk
❖ What macroeconomic conditions are considered drivers of credit risk for each pool
❖ Historic loss experience and if there is different loss experience in different risk characteristics
Pool Segmentation & Credit Risk Identification
This process should help with the following:
❖ Making sure pooling aligns with the institution’s credit risk
❖ Documentation of appropriateness of pooling based on the historic credit risk of the institution
❖ Create a list of models to consider based on credit risk driver in loan segments
❖ Understanding what models you may consider will narrow focus for the data inventory and gap analysis phase
Data Inventory
Create an inventory of where all historical loan information is maintained
❖ Look for detailed loan trial balances, loan level charge-off & recovery activity, loan origination activity✓ Need basic loan level data (loan number, origination balance, maturity, current
balance, interest rate, payment terms)✓ Need segment pooling characteristic, i.e., call code✓ Also risk indicators such as risk rating, LTV, FICO, DSCR, industry, property type
❖ Consider the following sources for historical data✓ Loan applications (core systems)✓ ALLL spreadsheets/software✓ ALM software✓ Network drives✓ Other
❖ Certain models will require more data✓ e.g., PD/LDG needs individual default events & exposure at default and DCF will
need prepayment rates
Gap Analysis
Based on models you are considering, what data gaps do you have?
❖ Data needs are highly dependent on model selection
❖ Chicken or the egg question
✓ Choose a method & then find the data you need or
✓ Find the data you have & then choose a method?
❖ Necessary data points & are they available for an adequate historical time frame to get meaningful results
❖ Sources for future economic forecast information
✓ Start with what economic data you look to currently
✓ National vs. regional
✓ Source of forecasts - Federal Reserve/OCC/Local Universities
❖ Don’t forget HTM securities, off-balance sheet commitments and disclosures requirements
Gap Analysis
Completeness & accuracy of the data you have
❖ Are data sets complete & can you prove it?❖ Is information from disparate systems in the same format?❖ Is information input timely & accurately (basic loan level info,
ratings, LTV, FICO, etc.)?
✓Accuracy of data input at origination & ongoing updates (updated risk ratings, renewed & modified loans, LTV, etc.)
✓Consider results of internal audits & loan reviews on data input accuracy at origination & on an ongoing basis
Common Data Gaps or Issues
Inability to access enough historical data
Inaccurate data fields
Data not captured consistently or updated after origination (i.e. FICO, LTV, DSCR)
Recording of loan renewals and extensions
Overwriting of data when using a report writer interface
❖ Example - only current loan grading may be able to be extracted
What to Do About Data Gaps?
Begin warehousing data on a monthly or quarterly basis in a usable format❖ Detailed loan trial balance
❖ Loan level charge-off & recovery detail
❖ Loan level origination activity
❖ Other information as necessary for models
Modify or increase the attributes input at the loan level
Modify internal controls over data input
Modify data retention policies
Consider the need for external data
EXAMPLE CECL MODEL ANALYSIS –OPEN POOL LOSS
Open Pool Loss Rate Models
Freezes all the loans in a segment pool at a particular point in time, then tracks the loss history on those loans over the remaining lives
Pros Cons• Less complex model
• Data collection less complex (no origination data)
• Q factor adjustment process will be similar to current practice
• Overall process is easier to understand
• Assumes historical pool has same credit risk & terms as current pool
• Maybe be reliant on older periods that are not relevant today
• Q factor & forecast adjustments are harder to support
Key Assumptions & Data Needs
Key Assumptions• Estimated CECL life of pools
Data Needs• Current & historical periodic (monthly, quarterly) loan level trial
balances with all necessary loan information, including current balance & risk segmentation characteristics, i.e., pooling category, risk grade, FICO, past due, etc.
• Historical loan level charge-offs & recovery data
• Understanding of how renewals, extensions & modifications are recorded in the loan systems
Open Pool Example
Year Amortized Cost
Losses on Loans as of
December 31, 2013 Comments
2013 1,010,000$ -$
2014 3,700
2015 7,600
2016 5,500
2017 1,650,000$ 3,900
20,700$ Cumulative lifetime losses on loans as of December 31, 2013
1,010,000$ Amortized cost balance as of December 31, 2013
a 2.05% Cumulative 4-year historical lifetime loss rate
Current Conditions Q-Factor Adjustments
0.05% - Real estate value decreased
0.03% - Unemployment rate increased
b 0.08% Total Current Q-Factor Rate Adjustment
Forecast Q-Factor Adjustments (over 2 years)
0.04% - Expect additional real estate value decreases
0.02% - Expect additional unemployment rate increases
c 0.06% Total Forecast Q-Factor Rate Adjustment
a+b+c 2.19% Total CECL Lifetime Loss Rate
1,650,000$ Amortized cost balance as of December 31, 2017
36,126.83$ Allowance for expected credit losses at December 31, 2017
Incorporating Risk Characteristics
Previous example does not consider changes in underlying risk in the portfolio
Next open pool example considers risk by sub-segmenting the pool by risk grade
To illustrate the vastly different result that could occur we kept all facts the same except we assumed the following:
❖ At December 31, 2013 special mention was 10% and substandard was 5% of the portfolio
❖ At December 31, 2017 special mention was 15% and substandard was 10% of the portfolio
❖ Subsequent losses on the December 31, 2013 balances, as graded on that date, were assumed to be 10% for pass, 30% for special mention and 60% for substandard
Open Pool Migration Excel Example
Rating 12/31/2013
CECL
Loss
Rate
CECL Allowance
12/31/17
Pass 858,500$ A 0.38% 4,716.34$
Special Mention 101,000 B 6.29% 15,564.07
Substandard 50,500 C 24.73% 40,811.20
1,010,000$ 61,091.61$
From previous example 36,126.83$
Additional reserve added by considering changes in concentrations in risk ratings 24,964.78$
Pass 12/31/2013 2014 2015 2016 2017 Totals
Net Charge Offs - 370 760 550 390 2,070$ D
Starting Loan Balance 858,500 A
Loss Percentage (D/A) 0.24%
Current & Forecat Q-Factor Adjustment 0.14% From previous example
Total Pass CECL Lifetime Loss Rate 0.38%
Special Mention 12/31/2013 2014 2015 2016 2017 Totals
Net Charge Offs 1,110 2,280 1,650 1,170 6,210$ E
Starting Loan Balance 101,000 B
Loss Percentage (E/B) 6.15%
Current & Forecat Q-Factor Adjustment 0.14% From previous example
Total SM CECL Lifetime Loss Rate 6.29%
Substandard 12/31/2013 2014 2015 2016 2017 Totals
Net Charge Offs 2,220 4,560 3,300 2,340 12,420$ F
Starting Loan Balance 50,500 C
Loss Percentage (F/C) 24.59%
Current & Forecat Q-Factor Adjustment 0.14% From previous example
Total Substandard CECL Lifetime Loss Rate 24.73%
12/31/2017
1,237,500$
247,500
165,000
1,650,000.00$
KEY TAKEAWAYS
• Now is the time to start! • You don’t have to pick a model today, but you do
need to understand your data gaps• As there is no required model to be used
institutions need solid documentation to support why they will choose their models
• Follow BKD’s thoughtware for additional articles on CECL
• Ask you BKD provider on how they can help you get on track with your CECL implementation