CECL's credit box curveball

With the Fed raising rates and long-term rates stabilizing, the yield curve is flattening, leaving financial institutions in a more vulnerable position. To offset the cost of funds and increase profit margins, many banks and credit unions are looking for ways to grab more market share, increase returns and appeal to a wider audience. Some lenders are taking on more risk by expanding credit boxes (appealing to underserved or complex borrowers) and loosening underwriting standards. With the ongoing rush to create and implement CECL models, how would expanding credit boxes impact your credit loss estimates?

Financial institutions that have recently expanded into new credit boxes need to watch out for the credit box curveball that may cause your CECL estimate to strike out. Since the recession of 2008, many financial institutions tightened their underwriting standards to minimize regulatory and financial risk. As a result, loan portfolios of the last 10 years may appear more homogenous and represent lower risk. In theory, lower risk, homogenous portfolios should make CECL modeling a bit easier for financial institutions, right? Not when the institution newly alters its underwriting strategy.

To comply with CECL, financial institutions have begun accumulating, organizing and warehousing years of loan data. While this significant amount of loan data will certainly improve a financial institution’s modeling success, it still may not be enough to generate an accurate estimate. Unless an institution has data going back to pre-recession days, the potential exists for bias in default and loss estimates. To address this bias, CECL credit modelers need to do at least one of three things when their institutions have expanded into new credit risk thresholds:

1. Incorporate their data back to pre-recession days. Relying only on data post-2008 crisis may not account for economic conditions that occurred during an economic downturn. This may understate their defaults, which may in turn cause institutions to underestimate their potential losses.

2. Supplement their data with peer data mimicking their institutions’ new portfolio credit composition. To do this, they need to know their institutions’ portfolio characteristics. If a peer institution portfolio is comparable, that information can be utilized in a model with appropriate sampling.

3. Institutions lacking data prior to 2008 should supplement additional data to leverage macroeconomic factors and assumptions to support loss forecasts by stressing credit metrics. By running multiple scenarios, financial institutions will get a more holistic view of potential risk.