Is Institutional Ownership Related to Corporate Social Responsibility? The Nonlinear Relation and Its Implication for Stock Return Volatility

Available at http://​www.​unpri.​org/​signatories/​signatories/​.

Canadian Institute of Chartered Accountants, in their 2010 report of Environmental, Social, and Governance (ESG) issues in institutional investor decision making suggest, “… there has been a fundamental shift in consideration of ESG matters in investment decision making. In particular, in the past, trustees may have argued that it was beyond their fiduciary responsibilities to consider ESG matters in an investment decision. Today, it may be considered a breach of fiduciary duty not to consider such matters” (Canadian Institute of Chartered Accountants 2010).

Institutional ownership also influences accounting conservatism. Ramalingegowda and Yu (2012), for instance, find that higher ownership by institutions that are likely to monitor managers is associated with more conservative financial reporting, and ownership by monitoring institutions leads to more conservative reporting. Accounting conservatism, however, is not our main focus in this paper.

We examine the effect of one-year and two-year lagged CSR on the current year institutional ownership. In addition, to mitigate the problem of low statistical power arising from the stickiness of social scores in KLD data, we measure changes in CSR as changes from year t − 3 to year t − 1.

We conduct a robustness test based on CSR index incorporating governance net counts. Our untabulated results are largely consistent with those based on CSR index excluding governance scores.

We conduct a robustness test based on the CSR raw scores counts rather than CSR index and our untabulated results are largely consistent with those in Tables 4, 5, 6.

CSR may affect institutional stock return volatility directly and/or indirectly via its effect on institutional ownership. Path analysis allows us to examine direct and indirect effect separately. While Granger-causation test helps researchers identify a causal relationship, it does not allow researchers to distinguish direct and indirect effects. We provide Granger-causation analysis in Table 8.

Path coefficients are standardized regression coefficients. Because we need to compare effects across equations and between regressors, we need to standardize the effects. While path analysis is useful because it allows researchers to examine direct and indirect effects simultaneously with multiple independent and dependent variables, it cannot establish absolute direction of causality. Path analysis cannot distinguish which of two distinct path diagrams is more correct, nor can it distinguish whether the correlation between A and B represents a causal effect of A on B, a causal effect of B on A, mutual dependence on other variables C, D, and so forth (Lea 1997). Despite those limitations, the use of path analysis in social science research has allowed researchers to gain additional understanding and insights into various important issues, and therefore theoretical knowledge on the part of the researcher is critical to the successful application of path analysis (Stage et al. 2004).

We conduct a robustness test by excluding financial and utility firms and find qualitatively similar results.

The results without winsorization are qualitatively similar.

The level of institutional ownership may be correlated with the lagged level of institutional ownership. We include the lagged level of institutional ownership at the beginning of the year as an additional control variable in the regressions and rerun our analyses. Our untabulated results are largely consistent with those in Tables 4, 5, 6.

Statistical significance for the indirect effect in Panel B is determined based on statistical significance of the path coefficient on centered CSR index (CSRIDX_C(t − 1)) and centered CSR index squared (CSRIDX_C2(t − 1)) in the first regression and statistical significance of the path coefficient on institutional ownership in the second regression. We take the lesser of the two statistical significance levels.

We find that unlike in the level specification, the direct effect of changes in CSR on changes in stock return volatility is not statistically significant. It appears that the positive relation between the two-year lag of CSR level and the level of stock return volatility in Table 7 is only transitory (static) and therefore does not hold in the change specification.