A CECL Primer. About CECL

Size: px

Start display at page:

Download "A CECL Primer. About CECL"

Bernice Rice
5 years ago
Views:

1 A CECL Primer Introduction The purpose of this paper is to provide a brief overview of Visible Equity s solution to CECL (Current Expected Credit Loss). Many facets of our CECL solution, such as the methods used to collectively review loan portfolios, are the subjects of their own dedicated technical white papers. Therefore, the intention of this document is not to go into great detail on any one method or portion of CECL, but to offer a holistic view of the Visible Equity CECL solution. About CECL Why it exists In June of 2016, The Financial Accounting Standards Board (FASB) released Accounting Standards Update (ASU) , Measurement of Credit Losses on Financial Instrument (Topic 326), with the intent of improving current regulations on financial institutions treatment of credit losses. The current incurred loss requirements, by which credit losses may only be recognized if it is probable that losses have already been incurred, have proven insufficient. The global financial crisis was aggravated by the delayed recognition of credit losses, motivating the development of new regulation and a move to an expected loss requirement. CECL eliminates the incurred loss methodologies by requiring recognition of all expected credit losses, regardless of whether a probable threshold has been met. Thus, with CECL, Generally Accepted Accounting Principles (GAAP) requirements will align with financial statement users need for timely, forward-looking recording of credit losses. Key Principles While the ASU doesn t specify a required methodology for calculating the allowance, it does require financial institutions to use methodologies satisfying some core principles. The allowance for credit losses is a valuation account that is deducted from the amortized cost basis of the financial asset(s) to present the net amount expected to be collected on the financial asset. 1 So the allowance should reflect the portion of the financial asset(s) expected to not be collected due to credit loss. 1 FASB ASU

2 Expected losses should be measured on a pool basis when similar risk characteristics exist. Assets with unique risk characteristics should be considered on an individual basis. 2 The allowance should be estimated using relevant information about past events, current conditions, and reasonable and supportable forecasts. 3 It is left to the institution to determine which information is relevant, and how to incorporate that information in the loss estimate. Expected credit losses should be estimated over the contractual term of the financial asset, with prepayments considered either implicitly or explicitly in the estimate, and without regard to extensions. 4 Portfolio Segmentation Properly segmenting your portfolio is the first step in the CECL process. Referring to the collectively reviewed portion of CECL, the FASB says an entity should aggregate financial assets on the basis of similar risk characteristics. 5 Meaning, when you segment your portfolio, the spirit of the law requires that the segmentation result in pools of loans that are reasonably homogeneous in risk. This, of course, must be balanced with sample size considerations. Some loss estimation techniques have certain sample size requirements that might be violated if too much disaggregation is in place. Visible Equity recommends that each institution carefully consider how to create homogeneous pools without losing the benefits of healthy sample sizes. Generally, segmentation is composed of three levels: segment, class, and credit quality indicator (CQI). Segments are usually broad loan categories (e.g. residential real estate), classes are usually sub-categories (e.g. first mortgages), and CQIs are metrics of loan quality (e.g. credit score), although the ASU allows for other means of segmentation such as term, rate, etc. Methods for collectively reviewing loans are specified at the class level, and within each class, the institution may choose to have the methods carried out at a CQI level (i.e. the historical loans used in the method would be partitioned by CQI). The CQI-specific calculations are not necessary for the Discounted Cash Flow method (DCF-PD) because the probability of default, which is used in the discounted cash flow framework, accounts for CQI. Visible Equity recommends CQI-specific calculations for Vintage and Static Pool methods if the concentrations of credit quality within the class have materially changed over time and the pool formed is of sufficient size to accommodate the analysis. 2 FASB ASU FASB ASU FASB ASU FASB ASU

3 Individually Reviewed Loans The next step is to decide whether certain loans should be reviewed on an individual basis. Loans with unique risk characteristics that don t fit nicely into a class should be reviewed on an individual basis. This might include troubled debt restructurings (TDRs) or loans unique in type or circumstance. Visible Equity provides three methods for individually reviewing loans. The first method is what we call Related Allowance Provided. In this method, the user uploads the results of their off-system analysis to Visible Equity so that it can be included in the CECL reporting framework. The second method is Discounted Cash Flow. This method looks at future cash flows and discounts them to arrive at a present value. The allowance is the difference between present value of these future cash flow and the recorded investment (balance). Note that the Discounted Cash Flow for individually reviewed loans shouldn t be confused with the Discounted Cash Flow method (DCF-PD) for collectively reviewed loans described below. The final method is Fair Value Less Cost to Sell. This method is used for collateral dependent loans and takes the collateral value less costs to sell, as the name implies, and then subtracts these net proceeds from all outstanding balances, including senior liens. If the net proceeds are not sufficient to cover the outstanding balance an allowance is recorded. Collectively Reviewed Loans The final step is to collectively review loans. Remember that CECL requires an expected lifetime loss estimate, so a 1-yr charge-off ratio will not cut it anymore. However, Visible Equity feels strongly that there is not one single best loss rate method for CECL. Each institution s product mix, data availability, underwriting patterns, etc. all play a role in determining which method should be applied. For this reason, we offer multiple methods, each with a unique set of strengths. This section will outline each method, reserving specific details for individual white papers. Note that only a few methods are discussed below. Other methods (e.g. roll rate, state transition, etc.) are in the research phase and will likely be offered in the future. Static Pool Mathematically, this is the simplest method offered by Visible Equity, as it is meant to resemble a traditional lookback loss rate. Given a class of loans, we wish to obtain a rate which, when multiplied by the current balance of the class, results in the amount of the balance expected to not be collected due to credit loss. This rate is estimated by first determining the remaining contractual life of the class s active balance (say L years) and isolating a similar pool of loans that were active at some starting point at least L years ago from today. All losses from that pool of loans are added together and divided by the pool s balance at the starting point. The resulting rate is multiplied by the current balance to obtain a lifetime expected loss. The strength of this method is in its simplicity. Easily digestible and explainable to examiners and other stakeholders, the static pool method won t take much getting used to. On the other hand, it does require complete historical balance and charge off data for at least the term of the class to which you wish to apply it. Additionally, adjustments should be applied to the base 3

4 loss rate for current and forecasted economic conditions. The biggest drawback to this method is that it only uses a portion of your history. Following only a single pool of loans, it may ignore valuable information. Vintage This method is also motivated by a traditional loss rate, but it uses your entire available history, and a level of sophistication is added by consideration of loan age. Unlike the static pool method, the vintage method does not assume that the distribution of loan age is the same in your historic pool as it is in your current portfolio. Another difference is that the goal of the vintage method is to obtain, for a loan of a given age, loss rates for each year in the remaining life of the loan, which can be multiplied by the current balance to get yearly expected loss. Note that care must be taken in building these age-specific rates. For example, because balances of first-year loans are higher on average than balances of third-year loans, when estimating expected fourth-year losses, the rate applied to the third-year loan should be different than the rate applied to the first-year loan. To estimate the rates, the vintage method is like static pool in that it involves tracking historic groups of loans over time and considering losses from those groups. Rather than just a single group, however, the vintage method tracks multiple groups and takes averages. The groups are formed by vintage year (and optionally Credit Quality Indicator or CQI). For each vintage group, yearly loss amounts are tracked and divided by a balance to obtain yearly rates. Those yearly rates are then averaged across vintage groups (using a weighted average) to obtain average rates for each year after loan origination. The key is determining the appropriate balance for the denominator. Returning to our example from the previous paragraph, if we are building a rate that should represent the proportion of a third-year balance expected to be lost in the fourth year, then the formula for that rate should look like the following: fourth year losses third year balance Table 1 shows an example of rate estimation for loans in their first year. The denominator used to create the rates in the body of the table are the first-year balances of the corresponding vintage groups. To get rates for loans that are in their third year, we would have divided by the average third-year balances for each vintage. The vintage method is desirable if you prefer the concept of a loss rate but wish to make fewer assumptions than are present in the static pool method. It also does an excellent job of leveraging all historical data from your institution. With these benefits comes greater complexity, however, which may deter some from its use. Like static pool, the vintage loss rates should be adjusted for changes in current and forecasted environmental conditions. 4

Table 1: Vintage method rates for loans in their first year Discounted Cash Flows with Probability of Default (DCF-PD) The DCF-PD method takes a fundamentally different approach than do vintage and

Rather than producing a rate to multiply by current balance, DCF-PD leverages loan and economic factors to produce monthly projections of cash flows, which are in turn used to estimate monthly

5 Table 1: Vintage method rates for loans in their first year Discounted Cash Flows with Probability of Default (DCF-PD) The DCF-PD method takes a fundamentally different approach than do vintage and static pool. Rather than producing a rate to multiply by current balance, DCF-PD leverages loan and economic factors to produce monthly projections of cash flows, which are in turn used to estimate monthly projections of credit loss (or cashflow shortcomings). This method is truly a loan-level method, as it turns loan-level and economic characteristics into monthly probabilities of default. Visible Equity has fit discrete time survival models for residential real estate, auto, credit card, and student loans. Visible Equity s proprietary database of monthly, loan-level records was used to fit all models except for residential real estate, which uses a single-family, loan-level dataset from Freddie Mac. Given that each of these models has already been estimated, the only data required of you for this method is that of your current portfolio (i.e. no historical data are required). For a given loan, the DCF-PD method uses current loan-level and macroeconomic data, as well as reasonable and supportable forecasts, to produce monthly probabilities of default and prepay for each month in the remaining contractual life of the loan. Using projected loan-level balances in combination with Visible Equity s collateral value estimation techniques, we project the potential cash flows in each future month (from payments, prepayments, collateral recovery, etc.). Expected future cash flows are then estimated by weighting those potential 5

6 cash flows by their probabilities of occurring. We group all potential cash flows by the scenario under which they would occur (i.e. the default, prepay, or active scenarios), multiply each scenario s cash flows by the corresponding monthly probabilities, and add it all together to get monthly expected cash flows for each loan. The monthly cash flows are then discounted and subtracted from the current balance to get the allowance. The main appeal of this method is its ability to seamlessly incorporate loan quality and macroeconomic factors into your loss estimates. Additionally, because of Visible Equity s robust database, the discrete time survival models are fit with an accuracy that usually isn t possible with data from a single institution. Another benefit is the leniency in terms of data requirements. As previously mentioned, only current loan-level data is required, so those institutions with limited historical data for one or more class will find a friend in DCF-PD. Economic Forecasts Consideration of economic forecasts is one of the key changes that comes with CECL. Compliance with this aspect of CECL can be factored into two necessary steps: 1. Producing forecasts 2. Incorporating forecasts into expected loss Because both steps can be technical in nature, a separate white paper will be provided on this topic, and only high-level descriptions will be offered here. Producing Forecasts Visible Equity provides forecasts of unemployment and house prices at both state and Metropolitan Statistical Area (MSA) levels and uses either in-house statistical forecasts or modified industry-provided forecasts. The statistical forecasts utilize Gaussian Process Regression (GPR) to leverage the correlation between neighboring points in time. For example, because house prices between subsequent months are highly correlated, historical data can be used to quantify that correlation. The results are then used to project future house prices. When a user desires industry-produced forecasts, but the forecasts are only provided at a national level, Visible Equity will provide modified, geography-specific forecasts. Specifically, for each geographical area (state or MSA), historical data is used to find the correlation with the national metric. The correlation factor for each geographical area is then applied to national forecasts to achieve specific forecasts for each area. Incorporating Forecasts There are many ways forecasts can be factored into your loss allowance. The cleanest way to do so would be to use the DCF-PD method, which already accounts for economic and loan characteristics. The other methods will require some explicit adjustment for changing economic conditions. 6

7 Data Requirements As evidenced in prior sections, data requirements depend heavily on the chosen method. Table 2 communicates the required data fields for each method discussed in this paper. In each cell, a C indicates that the field is required (monthly) for all loans in your current portfolio, and H indicates that historical, monthly records for closed loans require that field. The length of history required depends on method. Static pool requires at least the term of the asset class, while vintage has no strict requirement, but will improve with longer history. It is important to note that Table 2 is meant to communicate the minimum necessary data for each method. If you plan to segment by credit quality, then, of course, whichever variables are relevant for that segmentation should also be included. Also, while charge-off data is not necessary for the DCF-PD method, we recommend it for reporting, back-testing, or the possibility you might change methodology in the future. Table 2: Data requirements by method Static Pool Vintage DCF-PD (Res. RE) DCF-PD (Auto) DCF-PD (Card) DCF-PD (Student) Loan ID C,H C,H C C C C Loan Type C,H C,H C C C C Orig. Date C,H C,H C C C C Maturity Date C,H C,H C C C C Term C,H C,H C C C C Rate C C C C Balance C,H C,H C C C C Orig. Balance C,H C C C Senior Balance C C Credit Limit C i C Orig. Cred. Score C C C C Credit Score C C C C DTI C ii Days Delq. C C C C Orig. Value C C Borr./Coll. Address C C C C Borr./Coll. City C C C C Borr./Coll. State C C C C Borr./Coll. Zip Code C C C C Auto Make C Auto Model C Auto Year C Direct/Indirect C Net Charge-Off H H Balance Charge-off Date H H Close Date H H 7

8 Reports and Disclosures CECL reports and disclosures should align with the following three purposes as outlined in the ASU: 1. The credit risk inherent in a portfolio and how management monitors the credit quality of the portfolio 2. Management s estimate of expected credit losses 3. Changes in the estimate of expected credit losses that have taken place during the period. 6 Specific disclosure sections include credit quality information, the allowance for credit loss itself, information on past due status and nonaccrual loans, as well as a handful of other disclosures that relate to less common, specific lending situations. Visible Equity will provide all required disclosures in addition to other useful reports. A set of standard reports will be automatically populated when the required steps are completed for your portfolio. In addition to the standard reports, users will have access to report builders for any desired custom reports. The standard reports will include the following: 1. Executive Summary 2. Individually Reviewed Summary 3. Collectively Reviewed Summary 4. Q&E Summary 5. Credit Quality 6. Delinquency Aging 7. Vintage Analysis 8. Charge-off Analysis 9. Non-accruals 10. TDR / Modifications 11. Stress Test Results 12. Data Quality Conclusion Visible Equity will provide a complete, turn-key, user-friendly CECL solution that allows a user to easily segment their portfolio, analyze individual loans for impairment and expected loss, collectively review loans using CECL-compliant loss rate methods, incorporate qualitative and 6 FASB ASU

9 environmental factors and reasonable and supportable forecasts, and access a comprehensive suite of standard and custom disclosure reports. i Only for HELOC ii Not necessary, but useful 9

CECL Overview and Analytics Catalyst Corporate Webinar

CECL Overview and Analytics Catalyst Corporate Webinar 8/31/2017 Sponsored by: 1 PRESENTERS Ian Dunn, EVP Products Rachel Messick, Data Scientist 1 Introduction Allowance for Expected Credit Loss The main