Loan default analysis: A CECL case study and beyond

Loan default analysis: A CECL case study and beyond

By Guo Chen, PhD, Director, Quantitative Research, ZM Financial Systems

What challenges does CECL present to institutions?

The biggest challenge is the data. An institution’s data will tell a story, and the accuracy of the story will depend on the quality and quantity of data. By identifying and gathering all the data that could be used institutions will be able to segment to finer levels and determine where the losses really come from. The better data will also allow for the use of different methods.  Some methods may provide better accuracy with the loss rates and determination of the reserve amount.

Not having better data can make the cost or reserve amount to be higher than may be necessary. Also, without good data, it will be more difficult to defend assumptions and meet the reasonable and supportable component of the mandates.

How can institutions overcome challenges related to missing or errant data?

By using call reports and trial balances institutions can get the bare minimum to calculate loss rates.  That may be good to start.  However, each institution should identify data they will need for any of the calculation methods and for segmentation.  Then conduct a gap analysis to see what is missing.  Determine where and how this data can be gathered going forward and where it can be stored reliably and accessed.

What are some techniques to determine the probability of default using historical data?

Simple methods are a loan count method or balance method. There are increasing levels of complexity that can be used to increase the accuracy. Since CECL allows any reasonable and supportable method to acquire the life of loan loss rates institutions may use PDs but might also use simple loss rates. The method used may depend on the level of complexity of the institution and the portfolios as well as the data available. The real trick is to make sure you capture enough data that is representative of the meaningful or material portion of the historic loss curve.

How can a good ALM Model provide solutions to satisfy the challenges of CECL or IFSR-9?

A good ALM model that incorporates credit adjustments to cash flows when discounting will be able to adequately satisfy both mandates. Most ALM models already perform EV analysis and discount cash flows. Since the Discounted Cash Flow method (DCF) is a method available or mandated depending on the accounting body (FASB vs. IASB) ALM models can fit very nicely.  The question to ask is – does your ALM model credit adjust cash flows. Several do not.

What are the key components of building different statistical models that supports default forecasting?

An integral component of the mandates are the qualitative factors.  In other words, adjustments to the loss rates or PDs based on the forecasted economics.  By having adequate history and performing regression or correlation analysis between the losses and historic metrics institutions can get a better or more accurate assessment of the qualitative factors.  A key factor here comes back to the historical data in order to find out how loss rates behave in response to economic factors.

How can you use forecasted economics to project default-adjusted cash flows for CECL and IFSR-9?

By performing regression on the historical data, institutions can project adjustments to loss rates and subsequently adjusting cash flows based on the economic forecast with more confidence, making the assumptions more supportable and potentially reducing the impact to capital.

How can you use that forecast for DFAST stress testing?

Many DFAST filers are looking to leverage the DFAST process for CECL.  The forecast of economic variables for DFAST contains a most likely and stressed assumptions.  In conjunction with regression those forecasts can be used to fine tune adjustments and make more defendable assumptions.  Then layer on the adjusted loss rates to credit adjust cash flows for both CECL and DFAST.

How can you use that forecast to develop a better ALM model and better capital planning in general?

Ultimately this will be the big question.  Initially institutions will just be looking to meet the mandates. However, incorporating the results from CECL or IFSR-9 in the ALM process, then forecasting and stress testing them will help provide a better picture for capital planning.  None of the risk – IRR, liquidity, credit, capital, etc. – live in isolation of the others.  Incorporating the credit results will give a much more accurate picture and allow for stronger planning. As we saw in 2008, rates don’t have to move for a crisis to develop.  ALM and capital analysis should look beyond interest rate shocks and ramps.  Incorporate economic based stress tests which include a credit component.

As an overview, can you give a step-by-step, simplified roadmap to implement a CECL analysis?

Yes – as institutions move towards their respective filing dates they should already be taking action.  A common high level CECL roadmap or plan includes:
o  Establish a CECL committee.  This is important for a couple of reasons.  One is for transparency and visibility.  CECL may increase the reserve amount which will impact capital levels.  Executives should participate in the planning committee, so they have an idea of the potential impact and aren’t surprised on day 1.  Another reason is data.  Different members on the team will be able to provide insight as to data elements and where to find them.
o  Second is for data identification. The second step in planning and second point from above is about data.  Institutions need to assess the quality and quantity of their data.  What data do they think they need?  What are the gaps?  Where can they get data?  Where can they store and access it?  Members of the team may have insight into some of those fields that others do not.  Even if a team or committee isn’t or can’t be formed, data identification is an important part of the process.
o  Next evaluate methodologies.  What methods do you want to use, and which are supported by the data you have are important questions.  Better data will support different methods which will allow for better fine tuning of loss rates and adjustments.
o  Do you want to purchase an application?  Consider the data storage as well as the calculation and reporting applications.  Maintaining the amount of data for CECL as well as running calculations on that history can add a significant amount of overhead if done manually.  Purchasing a system which stores the date, calculates historic loss rates, and applies adjustments can free up time and resources for other strategic activities.
o  Run parallel with the current process so that you can work out the kinks in the new process and identify the magnitude of the potential change to your reserve.
o  Discuss the process being put in the place with your auditor and make sure nothing is being overlooked.

You may also be interested in…

Make your free account