Corporate Social Responsibility, Investor Sentiment, and Stock Returns

Abstract

In this study, we show that corporate social responsibility (CSR) increases volatility, since it creates noise in financial markets. Some outstanding studies related to the impact of investor sentiment state that companies with higher volatility exercise lower returns following high sentiment periods. Using environmental, social, and governance (ESG) research data, we sort a large number of US firms into high, medium, and low groups along with their social and environmental scores. We then predict the return of the high average minus the low average with investor sentiment , which has the tendency to act based on cognitive biases rather than the information at hand. Investor sentiment is proxied by direct sentiment surveys and the Baker and Wurgler (2006) composite sentiment index. We find that companies with higher environment-focused CSR activities have lower subsequent returns following high sentiment periods. We capture this evidence also for social performance with the composite sentiment index. The study introduces new insight to corporate social responsibility literature and extends return predictability literature. It also contributes a behavioral view to CSR–company performance relations.

Appendix 1

1.1 1.1 ASSET4 Database

The ASSET4 database is shown in Table 18.3.

Table 18.3 ASSET4 pillars and categories

1.2 1.2 Sentiment Surveys

Consumer confidence or sentiment surveys are often used as sentiment proxies. Lemmon and Ni (2008) consider the consumer confidence index as retailer investor sentiment since it bases the perception about households’ current and expected financial situation on a survey. US-based studies cite The University of Michigan Consumer Sentiment Index and The Conference Board Consumer Confidence Index because they are important measures indicating the strength of the US economy via consumers. These indexes are constructed by surveys comprised of the answers of many households regarding their financial situations, expectations for the US economy, and propensity of basic goods’ consumption (Lemmon and Portniaguina 2006, p. 1500). The surveys are shown in Table 18.4.

Table 18.4 Direct sentiment surveys

1.3 1.3 Baker and Wurgler (2006) Composite Sentiment Index

Baker and Wurgler (2006) estimate the first principle component of six measure described in the table below. They construct an index to proxy sentiment involving level of closed-end fund discounts, number of initial public offerings (IPOs), and equity share in new issues while considering lagged values of NYSE turnover, dividend premium and first-day returns on IPOs. Then, the authors regress raw values of six proxies on macroeconomic variables and use residuals to obtain the pure sentiment component. However, for a comparison with direct sentiment measures, we first use index as the sentiment proxy, which explains sentiment effect as independent from common business cycle component. The components are shown in Table 18.5.

Table 18.5 Components of the Baker and Wurgler (2006) composite sentiment index

1.4 1.4 Panel Regression Tests

We perform some tests for model choice, serial correlation, cross-sectional dependence, and heteroscedasticity. Then, we use Driscoll and Kraay’s (1998) standard errors, which are robust to heteroscedasticity, autocorrelation, cross-sectional dependence, and cluster (Table 18.6).

Table 18.6 Panel regression tests