.png){kind=icon}
.png){kind=icon}
Return
Imperfect Expectations in Loan Loss Forecasts
.png){kind=icon}
PRE
AI
DOI：10.1016/j.jacceco.2026.101930.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Forecasts of future realized loan losses that deviate from rational expectations can impact banks’ loan loss provisions, lending procyclicality, and financial stability, especially under the forward-looking current expected credit losses (CECL) model. The behavioral finance literature suggests that bank forecasts are influenced by the representativeness heuristic, which posits that, in forming estimates of future events, forecasters overreact to current circumstances. Leveraging confidential supervisory data from FR Y-14A filings, we document the presence of such overreaction in banks’ forecasts of net loan charge-offs. We find that revisions of these forecasts (current forecast minus previous forecast) are negatively associated with the corresponding forecast errors (realized minus current forecast), and this pattern strengthens for longer-horizon forecasts. Successive forecasts of the same future net loan charge-offs exhibit negatively autocorrelated revisions and gradually converge to the realization, indicating the continued presence and gradual resolution of overreaction over time. The overreaction affects loan loss provision in the same direction as the forecast bias, especially for CECL adopters. This effect of overreaction under CECL is more pronounced for longer-horizon forecasts and banks with longer-maturity loans. Supporting the prevalence of this bias, we observe overreaction in banks’ forecasts across loan types and of regional macroeconomic conditions.
Journal
J
IF:
6.8
Papers:
19
Citations:
0
View more organizations