Skip to main content

Looking for Valuant? You are in the right place!

Valuant is now Abrigo, giving you a single source to Manage Risk and Drive Growth

Make yourself at home – we hope you enjoy your new web experience.

Looking for DiCOM? You are in the right place!

DiCOM Software is now part of Abrigo, giving you a single source to Manage Risk and Drive Growth. Make yourself at home – we hope you enjoy your new web experience.

CECL: Practical Modeling Examples

February 24, 2018
Read Time: 0 min

Recap of the webinar by Regan Camp, Managing Director, MST Advisory Services

FASB’s CECL Guidance tells us that “the Board did not prescribe one type of methodology for measuring expected credit losses.” The update is conceptual with few specific rules, which presents both opportunity and challenges to deciding on a model. There is typically not only one “right” method for a particular institution, often any of several will work. We advise our clients to hone in on those that are the most feasible, most suitable to their portfolios, to their loan practices, to their accessible loan data. You merely need to determine which is best and be able to support that decision. 

Let’s examine a few of the models being most widely considered by institutions for CECL adoption: Cohort, Vintage, PD/LGD (transition matrix), and Discounted Cash Flow. 


The model captures all the loans that qualify for a particular pool segment as of a particular point in time to form a cohort, then tracks the cohort over the remaining lives of the loans to determine their behavior. Loans are pooled by similar characteristics. 

The Cohort method is simple and straightforward, but still sophisticated enough that it is being adopted by many large institutions. In essence, you’re taking a snapshot at a particular point in time and rolling it forward over a period when enough of your portfolio has had enough experience to determine a loss rate. You determine where to set that bar – for example, when 90 percent of loans have been exhausted – and compile those years’ loss rates for a historical average. Then you layer over with qualitative factors for your future expected losses. Prepays are implicitly considered. 

Advantages of Cohort Methodology

  • Requires less data to calculate than some other methods
  • Are less prone to spikes in losses in a particular year
  • Can be run on pools with broader risk characteristics than some other methods

Disadvantages of Cohort Methodology

  • Difficult to isolate losses expected in early or older years of a loan
  • More difficult than some other methods to revert to mean for a non-forecastable period
  • Could require more qualitative assumptions than others


The method stratifies loans within a pool segment by years of origination, then calculates expected loss for future periods based on historical experience with loans of a similar vintage of life cycle. It uses homogeneous loans and predicts future performance based on historic performance. Vintage modeling is generally limited to a small subset of a portfolio, such as auto loans or other consumer loans. 

To estimate a coming year with a Vintage method, begin with your historical averages, examine any trends in recent vintage loss rates, and factor in changes to current conditions and any reasonable and supportable forecasts. When you have determined that expected loss rate, you apply it to the current balance of each vintage cohort as of the reporting data to determine your expected loss allowance.

Advantages of Vintage

  • Easy to differentiate between reserve needs on loans of different vintages
  • Easy to see the shift in reserve needs for loans at different points in their lives
  • Easy to adjust for changing forecasts in different future years (conceptually).
  • Easy to revert to the historical mean for years beyond the forecastable period.

Disadvantages of Vintage

  • Requires a large amount of data
  • Year-by-year loss rates can be thrown off by outlier events
  • Necessitates lots of individual adjustments for future forecasts
  • Fairly complicated to set up and run on quarterly basis
  • Difficult to implement in a pool with multiple terms

PD/LGD (transition matrix model)

Probability of default (PD) tracks rates of movement between loan states, using Markov chains to determine long-term default rates. The steps to determining a transition matrix or PD are:

  • Determine your allowance segments by common characteristics
  • Determine your loan states – active or terminal (defaulted, charge-off or paid off) by rating
  • Establish movement rates from active to terminal from beginning of a period to the end of a period – optimally multiple rating periods are used for best-case averages
  • Apply to the current period, until the pool is exhausted – that is, until the vast majority of loans have reach a terminal state – and determine what percent at that point have defaulted, which is the PD for that pool of loans.

The loss given default (LGD) is the magnitude of the default, which is calculated by capturing the loans which have defaulted and using their balance prior to default as the denominator and their charge-off amounts as the numerator.

Advantages of PD/LGD

  • Don’t need as much data as some other methods
  • Allows for easier adjusting for forecasting economic conditions in future years based on the year-by-year default projections
  • Easier to revert to the mean for non-forecastable periods than some other methods

Disadvantages of PD/LGD

  • Requires consistent historical application and tracking of risk ratings or other loan state criteria
  • Needs considerable volume of loan state movements historically to establish meaningful movement rates
  • Only produces half the expected loss picture; requires a separate LGD calculation
  • Is labor intensive and cumbersome to run manually

Discounted Cash Flow

Under CECL the expectation is to take contractual cash flows and incorporate assumptions to determine expected cash flows. Calculations would include the processes required for PD/LGD, plus curtailment rates, prepayment rates, recovery delays and funding rates. Essentially the method involves taking the contracted cash flows, applying the model assumptions to arrive at your expected cash flows, which are then discounted at the effective interest rate to arrive at discounted cash flows.

Advantages of Discounted Cash Flow

  • Explicit considerations for instrument contractual terms and prepayments
  • Might reduce volatility in the allowance
  • Easier to monitor changes, since loss projection is periodic and not one life-of-loan number
  • Allows for back-testing assumptions and appropriate adjustments

Disadvantages of Discounted Cash Flow

  • More complex than other methods
  • Requires a second method to generate loss rate of PD/LGD assumptions
  • Extensive data requirements
  • Could require significant investment in systems
  • Requires additional layers of modeling and controls

To listen to the entire one-hour webinar, click here. 

About Abrigo

Abrigo enables U.S. financial institutions to support their communities through technology that fights financial crime, grows loans and deposits, and optimizes risk. Abrigo's platform centralizes the institution's data, creates a digital user experience, ensures compliance, and delivers efficiency for scale and profitable growth.

Make Big Things Happen.