This is why CECL could be tough for credit card issuers

The Current Expected Credit Loss standard – or “CECL” as it is more commonly called – will fundamentally change how U.S. financial institutions (FIs) account for credit losses. In the pre-CECL world, smaller financial institutions, including credit unions, used relatively simple approaches to provision for losses. They used actual losses experienced during recent months (typically the last 12) as an estimate of future losses. When it takes effect for these institutions in 2020, CECL will require FIs to calculate expected loss over the lifetime of a loan, requiring substantially more sophisticated analyses.

Forecasting credit losses over the life of a loan requires more advance analytical approaches, powered by a much longer historical dataset to track loan performance, granular information about balances, payments, APR and origination vintages. Certain product lines, such as credit cards, will present greater challenges than other products when it comes to CECL calculation. That is because credit cards are fundamentally different from other loan products. They do not have a fixed term; they are revolving, balances fluctuate, and cardholders have options like making minimum, partial or full payments. Moreover, per CECL guidelines, in case of revolving lines, losses should only be considered from current drawn amount. This means that for any future default, FIs need to find a way to estimate the contribution of current snapshot balance in the final loss calculation. Due to the intricate nature of the credit card loans, issuers will need to apply extra care to CECL analysis.

Below are the four major challenges and solutions for credit card portfolios under CECL:

Life of loan estimation

  • Challenge: For typical installment loan products, the life of a loan and the life of an account are the same. Because credit cards are revolving by nature, it is difficult to determine the estimated life of a particular loan.
  • Suggested action: Retrieve historical data on balances and payments and analyze the balance pay down behavior for different kinds of accounts to determine the average life of credit card loans. In an ideal scenario, use granular information by balance segments (principal balance, interest & fees, promotions balance, etc.) to paint the true picture. Where data is incomplete or unavailable for portions of the portfolio, consider using third-party data for estimation.

Balance pay down curve creation

  • Challenge: To estimate the balance pay down curve for credit cards, FIs need to apply future payments on the current snapshot balance. For financial accounting of credit card payments and balances, FIs follow the CARD Act and allocate the received payments based on the APRs of the different balance components. For example, Purchase APR may be lower than Cash Advance APR; hence payments are first applied to the cash advance balance rather than the purchase balance. However, to simulate the true balance pay down curve at the time of the snapshot period, issuers will need granular and accurate forecasts for different balance segments, APRs and payments.
  • Suggested action: Different options are available for amortizing snapshot balances and estimating the pay down curve. These include “First In, First Out” (FIFO), minimum payment, CARD Act simulation or a hybrid of these approaches. Choose the approach based on the data requirement, complexity and effort needed for execution.

Portfolio segmentation

  • Challenge: To provide the best view of member behavior, the FASB has prescribed grouping the accounts into homogenous segments for analysis. Credit unions need to consider multiple aspects – riskiness, payment pattern, type of balance – when segmenting portfolios.
  • Suggested action: Segment the portfolio according to similar risk characteristics defined by risk score, delinquency buckets, transactor/revolver, etc. Monitor the performance (i.e. loss rate, average life of loan and shape of balance pay down curve) of the different segments over time. Each segment should have a statistically significant population for downstream analysis and reporting.

Event definition for model development

  • Challenge: Defining the dependent variable for model development purposes is particularly complex under CECL. Should the credit loss model treat an account that defaulted in the future, but fully paid off the current snapshot balance, as an event or a non-event?
  • Suggested action: When using current models built using the traditional “bad” definition (i.e. any kind of default), make adjustments for accounts that defaulted after the snapshot balance is fully paid off. Otherwise, define events based on accounts that have non-zero balance remaining from the snapshot balance at the time of default.

A successful CECL implementation for credit cards begins with the basics. The key is to meet the expectations of regulators and auditors in an efficient, sustainable and scalable manner. By using CECL as a catalyst to centralize and enhance processes, credit unions can improve long-term productivity, increase analytical capabilities and realize long-term value from their credit card portfolios.

Manish Jain

Manish Jain

Manish Jain is a Vice President for EXL. He has worked in the field of data analytics since 2007, primarily for banking and financial firms in the marketing and risk ... Web: https://www.exlservice.com Details