Nonlinear dynamics in discretionary accruals: An analysis of bank loan-loss provisions


Autoria(s): Balboa, Marina; López Espinosa, Germán; Rubia Serrano, Antonio
Contribuinte(s)

Universidad de Alicante. Departamento de Economía Financiera y Contabilidad

Finanzas de Mercado y Econometría Financiera

Data(s)

09/07/2014

09/07/2014

01/12/2013

Resumo

Several studies have analyzed discretionary accruals to address earnings-smoothing behaviors in the banking industry. We argue that the characteristic link between accruals and earnings may be nonlinear, since both the incentives to manipulate income and the practical way to do so depend partially on the relative size of earnings. Given a sample of 15,268 US banks over the period 1996–2011, the main results in this paper suggest that, depending on the size of earnings, bank managers tend to engage in earnings-decreasing strategies when earnings are negative (“big-bath”), use earnings-increasing strategies when earnings are positive, and use provisions as a smoothing device when earnings are positive and substantial (“cookie-jar” accounting). This evidence, which cannot be explained by the earnings-smoothing hypothesis, is consistent with the compensation theory. Neglecting nonlinear patterns in the econometric modeling of these accruals may lead to misleading conclusions regarding the characteristic strategies used in earnings management.

Spanish Department of Science and Innovation through the ECO2011-29751 and ECO2012-33619 projects.

Identificador

Journal of Banking & Finance. 2013, 37(12): 5186-5207. doi:10.1016/j.jbankfin.2013.05.020

0378-4266 (Print)

1872-6372 (Online)

http://hdl.handle.net/10045/38787

10.1016/j.jbankfin.2013.05.020

Idioma(s)

eng

Publicador

Elsevier

Relação

http://dx.doi.org/10.1016/j.jbankfin.2013.05.020

Direitos

info:eu-repo/semantics/restrictedAccess

Palavras-Chave #Earnings management #Income smoothing #Multi-way cluster #Panel data #Threshold regression #Economía Financiera y Contabilidad
Tipo

info:eu-repo/semantics/article