Loan loss provision models in Brazilian banks

Abstract

Purpose This paper investigates which loan loss provision (LLP) model [International Accounting Standards39 (IAS39) based on incurred losses and Brazilian Central Bank Generally Accepted Accounting Principles (GAAP) based on a mixed model] presents higher quality in terms of predictability, and which model is less susceptible to earnings management practices using LLP. Design/methodology/approach To test the difference between the explanatory power of the mixed model and incurred loss model in explaining the LLP, this paper runs a two-stage fixed-effect panel regression model to evaluate the association between LLP of each model and variables representatives of non-discretionary aspects related to the quality of the loan portfolio, business cycles and qualitative evidence indicated in each GAAP. Then, this paper tests the relationship between the errors generated in each regression and the discretion of bank managers and banks’ characteristics. Findings This paper finds that the mixed model results in higher R2 demonstrating that the number produced under this regime is more related to observable variables than the number produced under the incurred losses model. Further, this paper finds no evidence that there is a difference in earnings management between the two standards and this paper does not find that banks manage earnings through regulatory capital. Nevertheless, this paper finds that earnings management is higher in private than in listed banks. Originality/value This paper takes advantage of the unique feature of the Brazilian Central Bank regulation to investigate the impact of two different accounting standards on LLP in a perfect setting.

Managerial discretions and loan loss provisions in Nigerian banks

“…The same practice was also observable in the provisioning practices of troubled Nigerian DMBs during IFRS given similar behaviour of IFRS*SVR*EBTL. This evidence of no improvement in the use of LLPs to smooth/manage earnings during IFRS found in this study is consistent with the findings of Eneje et al (2016), Atoyebi and Simon (2018), Ozili and Outa (2018) and Galdi et al (2021). However, this study's findings invalidate the conclusion of Ozili and Outa (2019) on reduction in the use of LLPs to smooth earnings subsequent to the adoption of IFRSs in Nigeria.…”

“…Based on the derivation of MDI, findings obtained from regression models presented in Table 8 are exclusive to this study and represent a contribution to LLPs literature. However, an increase in the discretionary use of LLPs for managerial decisions as identifiable with coefficient of IFRS*MDI is comparable to the findings of Attia et al (2013), Ashraf et al (2015), Ozili and Outa (2018), Ashraf et al (2019) and Galdi et al (2021). In contrast, a reduction in the managerial discretionary use of LLPs by Nigerian DMBs in solvency crisis during IFRS found in this study given the significantly negative coefficient of IFRS*SVR*MDI in relevant models can be compared to the findings of Leventis et al (2011Leventis et al ( , 2012.…”

“…A similar improvement in the discretionary use of LLPs is also noticeable with insignificantly positive coefficient of SIGN when interacted with principles-based accounting standards indicating imminent use of LLPs to signal. Although higher coefficient of determination is found for models with Brazilian Central Bank's accounting rules, Galdi et al (2021) found no difference in the use of LLPs for earnings smoothing by Brazilian banks when distinction is made between reporting in IAS 39 and Brazilain Central Bank's accounting principles which are based on mixed loan loss model.…”

“…Although the influence of index of managerial discretionary decisions (MDI) on LLPs is the focus of this study, the continued test of use of LLPs for each of the components of MDI in the literature (Muriu, Josea 2020;Galdi et al 2021;Le et al 2021) necessitated separate test of the use of LLPs for earnings smoothing, capital management and earnings signalling in addition to MDI. The results of regression analyses are presented in Table 9 for earnings smoothing, Tables 10 and 11 for capital management and Table 12 for earnings signalling.…”

Avoiding financial instability: essays on bank performance and riskiness

“…Dantas, de Medeiros, & Lustosa (2013), comparing nine models of LLP determinants, find that macroeconomic variables and characteristics of the quality of the loan portfolio improved the measure of LLP measures in Brazil. In a recent study, Galdi, De Moura, & França (2021) conclude that the Brazilian GAAP for financial institutions, a mixed model, presents higher quality in terms of predictability than the IAS 39. However, they find no evidence of earning management in these two systems.…”