Expecting the Unexpected : Abnormal Loan Loss Provisions and Analyst Forecast Accuracy in the Transition to CECL

Sammanfattning

This thesis examines whether the adoption of the Current Expected Credit Losses standard (CECL) amplifies the negative association between abnormal loan loss provisions (ALLP) and analyst forecast accuracy for U.S. banks. While CECL was designed to improve the timeliness of credit loss recognition, its greater reliance on managerial judgment may make reported provisions more difficult to predict. Based on quarterly data for 137 publicly listed U.S. banks over 2015Q1–2024Q4, we examine analyst forecast errors in regressions with bank and year-quarter fixed effects. ALLP is estimated using the Kanagaretnam et al. (2010) model, following Hong et al. (2020). The results show that the association between abnormal provisioning and LLP forecast errors becomes significantly stronger following CECL adoption (β = 0.133, p < 0.01), while the stand-alone CECL coefficient is insignificant. This indicates that the effect varies with the magnitude of abnormal provisioning rather than reflecting a uniform shift in LLP forecast errors. For EPS forecast errors, the CECL × |ALLP| interaction is insignificant, while the stand-alone CECL coefficient is negative and significant at the 5% level. Taken together, our findings suggest that CECL is associated with lower LLP forecast accuracy for banks with larger abnormal provisions, while no similar amplification is found for EPS forecast accuracy.