The Financial Accounting Standards Board’s new current expected credit loss impairment standards require timely, forward-looking measurement of lifetime risk using credible models. We answer the leading questions related to data challenges.
What if we don't have data from the 2008 downturn?
Having data going back to the downturn is preferable for creating "reasonable and supportable" forecasts. Clients’ historical data can be augmented with industry performance data then calibrated to resolve gaps and limitations. Although this is not a perfect method, considering the short history of data out there for this sector, this would be the best approach until more data accumulate.
Moody’s Analytics provides a wide array of industry performance data, covering multiple asset classes. Our solutions can fill gaps and alleviate the need to aggregate and clean records from various data sources. We gather historical data from numerous sources and add value by including series and estimates that address limitations in the as-reported data such as short history, low frequency, long lag, limited granularity, and changes in definitions or classifications. We supplement these historical data with forecasts generated by our experts.
For lifetime credit loss estimate, if historical data cover the past 10 years, can a 10-year loss forecast be used as a proxy for lifetime?
Even with 10 years of history, forecasts can extend out for more than 10 years using input variables that have longer forecasts. Depending on the line of business in question and effective lifetime of the loans in a portfolio and the mean reversion concept in CECL, longer periods might not be needed. This requires analysis of the portfolio.
For financial technology firms that are fairly new to the market and do not have historical data, how does CECL apply?
The best proxy for fintech companies is to leverage industry forecasts for personal loans that can be calibrated with the limited data from the institution. Although not a perfect method, considering the short history of data out there, it would be the best approach until more data accumulate.
What is a good external data source for estimating effective life of loans for mortgages?
Effective life of loans for mortgages would be much smaller than the contractual maturity in this low-rate plus high house-price indexes environment. CreditForecast.com is a good source for estimating effective life of loans for all consumer credit products, including mortgages.
Would securitized data sources be appropriate for small community banks?
Yes. The securitized data can either be used as a proxy for internal data where wholesale data are not available or, if the bank has securitized portfolios, they could also be used as the main data source. They could be used as a proxy and calibrated to the bank's own portfolio or, if the bank doesn't have any data, they can be the main data source.
Are there publicly available sources for asset-back securities data?
The monthly reporting on publicly registered ABS deals is generally available; however, the depth of data and format of the data varies from issuer to issuer. Moody’s Analytics aggregates and standardizes these data for customer use. Per Regulation AB II, publicly registered auto loan and lease ABS transactions provide monthly loan-level data in a standardized format from origination of the security. This data universe has been available via the Securities and Exchange Commission’s Electronic Data Gathering, Analysis and Retrieval system since January 2017. Each transaction provides data on approximately 60,000 loans or leases and follows them on a monthly basis until the earlier of ABS maturity, full repayment or default/recovery. The database currently covers about 50 deals, and Moody’s Analytics maintains the full database, which is available for customized selection per loan or lease selection criteria. For privately issued deals (Rule 144a), the data are available only to investors or platforms that collect the data for qualified institutional buyers (and generally is only available based upon pool performance).
How do you link macro variables and loss rates, assuming a lag/lead in macro indicators?
Macro variables can have a direct impact on behavior of institutions’ portfolios, and the impact might come with a lag, since it might take time for credit behavior to change. For instance consumers don't become delinquent right away when they lose a job; they might have savings or unemployment insurance to still pay debt for some more months.
How do you factor macroeconomics variables into loss rates?
Usually there are intuitive relationships between macro variables and loss rates. For instance, as the unemployment rate increases probability of default rates increase for consumer credit portfolios. To quantify these relationships, a regression analysis is usually done and econometric methods are utilized. Moody’s Analytics economic data enable you to identify historical and current relationships between macro variables and credit, improving forward-looking estimates.
How should the sale of repossessed equipment be reported for recoveries, as sale of equipment or as a reduction to write-off?
Generally, this is part of recoveries and reported separately from gross charge-offs. But for CECL purposes, data can also be saved as net write-offs if separating modeling would be complex for the institution.