Bank Sentiment and Loan Loss Provisioning

45 Pages Posted: 19 Mar 2024

See all articles by Junsung Bae

Junsung Bae

College of Business, Korea Advanced Institute of Science and Technology (KAIST)

Allen N. Berger

University of South Carolina - Darla Moore School of Business

Hyun-Soo Choi

College of Business, Korea Advanced Institute of Science and Technology (KAIST)

Hugh Hoikwang Kim

University of South Carolina, Darla Moore School of Business

Date Written: March 2, 2024

Abstract

The recent regulatory change to forward-looking discretion-based loan loss provisioning relies on the assumption of objective optimizing behaviors of bank managers. In this paper, we challenge this assumption by testing the impact of bank managerial sentiment on loan loss provisioning. Utilizing various large-language models such as BERT and GPT, we extract banks' sentiment measures from their 10-K filings, which are distinct from fundamental-based beliefs and borrower-side sentiments. We find that banks with more negative sentiment increase their loan loss provisions above the level warranted by key economic fundamentals and their future loan charge-offs. And banks with more excess sentiment-driven loan loss provisions reduce their lending in the future. The impact of bank sentiment is more pronounced during the recessionary periods, suggesting that the sentiment can amplify the counter-cyclicality of loan loss provisions and the pro-cyclicality of bank lending. We address endogeneity concerns about bank sentiment by using exogenous weather conditions as instrumental variables. Overall, the findings suggest that the discretion-based loan loss provisioning driven by bank sentiment can exacerbate the pro-cyclicality of bank lending.

Keywords: bank sentiment, loan loss provision, lending, textual analysis, banking industry, financial disclosure

JEL Classification: G21, G40

Suggested Citation

Bae, Junsung and Berger, Allen N. and Choi, Hyun-Soo and Kim, Hugh Hoikwang, Bank Sentiment and Loan Loss Provisioning (March 2, 2024). Available at SSRN: https://ssrn.com/abstract=4745996 or http://dx.doi.org/10.2139/ssrn.4745996

Junsung Bae

College of Business, Korea Advanced Institute of Science and Technology (KAIST) ( email )

85 Hoegiro Dongdaemun-Gu
Seoul 02455
Korea, Republic of (South Korea)

Allen N. Berger

University of South Carolina - Darla Moore School of Business ( email )

1014 Greene St.
Columbia, SC 29208
United States
803-576-8440 (Phone)
803-777-6876 (Fax)

Hyun-Soo Choi

College of Business, Korea Advanced Institute of Science and Technology (KAIST) ( email )

85 Hoegiro Dongdaemun-Gu
Seoul 02455
Korea, Republic of (South Korea)

Hugh Hoikwang Kim (Contact Author)

University of South Carolina, Darla Moore School of Business ( email )

1014 Greene Street
Columbia, SC 29208
United States

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
123
Abstract Views
489
Rank
420,762
PlumX Metrics