Bank earnings smoothing, audit quality and procyclicality in Africa

Abstract

Purpose The purpose of this paper is to empirically examine whether the way African banks use loan loss provisions (LLP) to smooth earnings is influenced by capital market motivations and the type of auditor, after controlling for non-discretionary determinants of provisions and fluctuations in the business cycle. Design/methodology/approach To test the income smoothing hypothesis, the model was estimated using panel least square with White’s robust standard error correction, as well as, with and without period fixed effect. Findings The findings support the income smoothing hypothesis and indicate that African banks use LLP to smooth earnings; listed African banks use LLP to smooth earnings to a greater extent compared to non-listed African banks, possibly, for capital market reasons; income smoothing via LLP is not reduced among African banks with Big 4 auditors; and after controlling for macroeconomic fluctuation, there is evidence that bank provisioning is procyclical with fluctuations in the business cycle. Research limitations/implications The findings have three implications. One, listed African banks smooth income because they are more visible to investors; investors do not view stock price fluctuations as a good signal. Securities market regulators in African countries should enforce strict disclosure rules that reduce earnings smoothing practices to improve the transparency of bank earnings in the region. Two, the presence of a Big 4 auditor did not improve the informativeness of LLP estimates among African banks. Three, the evidence for procyclical provisioning suggest the need for dynamic LLP system in Africa. Originality/value This paper is the first cross-country African study to investigate whether provisions-based income smoothing decreases with the presence of a Big 4 auditor. The findings indicate that this is not the case among African banks.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 64 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.