Heterogeneity in CSR activities: is CSR investment monotonically associated with earnings quality?

Abstract

Extant studies have investigated the relation between corporate social responsibility (CSR) endeavours and earnings quality based on monotonic models, showing mixed and inconclusive empirical evidence. By identifying heterogeneity in CSR investments, we extend prior literature to explore the potentially nonmonotonic nature of this relation. Specifically, we classify firms into two sub-groups, entities underinvesting and overinvesting in CSR activities, in which the levels of CSR investments are lower and higher than the theoretically optimal point respectively. Our empirical results show that the level of CSR underinvestment is positively associated with the magnitude of both accrual-based earnings management (AEM) and real earnings management (REM) and, hence, negatively related to earnings quality. For firms overinvesting in CSR activities, we do not find a significant relation between CSR overinvestment and AEM. The empirical analyses for real activities manipulation exhibit inconsistent results throughout our four REM proxies. However, the mixed evidence for firms with CSR overinvestment cannot fully exclude the possibility that overinvesting in CSR activities has a significant impact on future financial reporting quality. Varying incentives for CSR overinvestment in different firms could drive the inconsistent results. The positive effect of CSR overinvestment by some firms may offset the negative effect brought about by other entities, making the overall effect minor and unnoticeable. Our empirical results, together with some other CSR-related research, emphasise the need for more transparent reporting regarding the detailed nature, aim, and strategy of relevant CSR investments to help investing communities and other constituents better understand the incentives behind CSR activities.

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 — 24 citations exist upstream.

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