CECL: The Importance of Data and Why

The Importance of Data and Why

As with every new standard or compliance standard there are questions about how to align with new requirements and practices. What are the best practices and how do organizations go about implementing them, how does an institution make sure they are not just checking the boxes? CECL is no different, from how institutions will work with their auditors, changes in analytics, board training, modeling support all need to have a renewed focus with such a disruptive accounting change. Because we still receive questions about data, lets revisit best practices.

At the June 17, 2019 FASB board meeting, the board voted to add an agenda item to extend the implementation date for all entities except for large public companies. Why did they do this? CECL is a difficult process change and early adopters are struggling with many aspects of the changes including data.

One reason we are addressing data is because we are still receiving questions about data and it is an important part of the CECL implementation and adoption process. There is still an enormous amount of confusion about how much data you need, what happens if you only have a few years of data, and what to do with the data you accumulate. Let’s get to the real answers!

First, every example FASB has issued including the WARM method and the first example in the standard utilizes data sets that match the contractual term of the pool the example utilizes. To be clear, 5 years of data in the WARM method example for a five-year contractual term pool and 10 years of data for a 10-year contractual term pool in the first example in the standard.

To be specific, you need data to cover the contractual term, not the average term.The root of the problem is starting with various CECL vendors that are selling the idea that only a few years of aggregated data are needed. What these vendors are not telling you is that they are going to have to fill in for what you don’t have! At ARCSys, we are about best practices and explaining how CECL works.

So here are the facts!

The standard states that your data is the best data for you to use, however, if you don’t have your data you can utilize third party data. Regardless of where your data comes from, there are three key steps in the CECL process.

Step One – Gather and understand your data, this is your unadjusted historical loss, unadjusted prepayment, unadjusted loss given default, and unadjusted probability of default cycles or historical data sets.

Understanding your data is an important part of the analysis. This understanding includes an analysis by pool of similar financial assets (class of loans or investments) comparing the historical pool to the current pool. This analysis would include comparing these minimum items as to composition of the portfolio:

  • Contractual Term
  • Risk Metrics such as FICO, LTV, Debt Service Coverage (DSC)
  • Prepayments
  • Underwriting
  • Charge-offs and Recoveries
  • Probability of default
  • Loss Given Default

The purpose of this analysis is to understand the differences between the historical data set selected for you calculation and your current data. This is important because this need to be done before you forecast and is part of the process of moving to step 2, adjusting your historical data.

Step Two – Analyze and adjust your historical data. Your data needs to be adjusted to ensure that it represents the correct adjusted data set for modeling. Therefore, you should consider adjusting the historical data for the following:

  • Contractual Term – If the contractual terms are different between the current and historical data sets, you may need more data to preform your calculation. Be careful that you include TDR’s in both data sets.
  • Risk Metrics – If your pools have not been broken out using credit quality indicators, then you have to access the credit quality of the historical data and compare to your current credit quality. Therefore, if the current population is more or less risky than the historical data then you will have to make adjustments to the historical data used in the calculation. Using static or migrated pools reduce the work effort here. If you don’t have any credit quality indicators in historical data, how do you do this assessment?
  • Prepayments – You will need to understand your prepayment speeds in your historical data and your current data set. If the historical data is different, you may need to adjust your historical data to as to the effects of prepayments on your loss data. This also means you need prepayment speeds historically.
  • Underwriting – You will need to do an analysis of changes in underwriting to make sure the loans are similar in status or credit quality. This can be similar to the risk metrics analysis.
  • Current and forecasted direction of the economic environment – You will need to compare the historical data to the current and forecasted direction of the economic cycle. For example, if you have data from 2007 to 2020, and your contractual term is 7 years, which seven-year period you select would be based on the current period you are in and the forecasted period you are forecasting into. Therefore, if you believe that your economic cycle is going to move towards an increasing unemployment environment, the 2007 to 2014 period may be the correct historical period to utilize. In addition, this data may require less adjusting for some of the factors above.
  • Unused Commitments – For those pools with lines of credit or open commitments, you will need to analyze the utilization and which period best fits the economic cycle you are forecasting into.

After your adjustments, this is your adjusted historical data that must be used in your calculation and forecasts

Issues with data will include but are not limited to the following:

  • Situations involving minimal loss history.
  • Losses that are sporadic with no predictive patterns. These are harder to model and forecast. Different models should be calculated to see which model provides the best predictive allowance.
  • Low numbers of loans in each pool. You may have to combine some pools you separate today or use third party data.
  • Data that is only available for a short historical period. You may have to supplement with third party data or use a similar pools data
  • A composition that varies significantly from historical pools of financial assets. You may have to make additional assumptions or use other data.
  • Changes due to in the economic environment.

Step Three – Forecast the data over the short term (reversion) or long-term (contractual term).

The standard allows for 2 ways to forecast. When defining this, FASB stated the following in the standard:

Contractual Term Forecasting

  • “Some entities may be able to develop reasonable and supportable forecasts over the contractual term of the financial asset or a group of financial assets.”

Reversion Forecasting (Short Term Forecasting: 1 to 2 years)

  • “However, an entity is not required to develop forecasts over the contractual term of the financial asset or group of financial assets. Rather, for periods beyond which the entity is able to make or obtain reasonable and supportable forecasts of expected credit losses, an entity shall revert to historical loss that is reflective of the contractual term of the financial asset or group of financial assets.”
  • “An entity shall not adjust historical loss information for existing economic conditions or expectations of future economic conditions for periods that are beyond the reasonable and supportable period.”
  • “An entity may revert to historical loss information at the input level or based on the entire estimate. An entity may revert to historical loss information immediately, on a straight-line basis, or using another rational and systematic basis.”

Data Options

You have 4 options for data. These are in order of best practices:

Option 1 – Get your detailed loan data through at least one economic cycle (2008) or through the longest contractual pool term including TDR’s

Option 2 – Use your summarized data such as call report data

Using summarized data poses additional work because the data sets are generally higher level and may not directly correspond to your class level poolings.

Option 3 – Use third party data

Using third party data poses additional work because the data set is not your data.

As discussed in the standard, historical credit loss experience of financial assets with similar risk characteristics generally provides a basis for an entity’s assessment of expected credit losses. Historical loss information can be internal or external historical loss information (or a combination of both). However, an entity shall consider adjustments to historical loss information for differences as discussed above in Step 2.
Third party data poses additional effort in this analysis because of the following:

  • Functionally analyzing the data may be difficult because of your access to the underlying data sets.
  • The economic environment the historical data went through may be different, such as data from different states or regions
  • The contractual term may be different
  • The underwriting may be different
  • The individual pool risk may be different such loan volume by LTV or FICO scores
  • The level or volume of TDR’s may be different
  • The prepayment history may be different
  • The charge-off and recovery levels may be different

Therefore, in order to use third party data, the user will have to continuously preform additional analysis and make adjustment to the data to adjust the data so it can be used in the users CECL calculation. This effort is significantly more work than adjusting a user’s own data.

Option 4 – Try and perform some kind of simulation (guess) to fill in for missing data. This is the most difficult and requires the most assumptions. We strongly recommend you don’t use this approach!

As organizations begin to develop their CECL programs and work through implementation, selecting a knowledgeable partner who has both experience with third party data and proprietary institution data will be critical. ARCSys has helped institutions of all sizes and complexity to implement their CECL programs confidently while deploying best practices.


About The Author