Econometric framework
To test our hypotheses, we estimate a treatment effect model with the cost of equity capital as the outcome, and whether a firm is included in any SRI fund’s portfolio as the treatment. Let
\(Y_{ijt}^0\) denote the potential cost of equity capital for firm
i in industry
j when the firm is not invested by any SRI funds at time
t, and let
\(Y_{ijt}^1\) denote the potential cost of equity capital when the firm is invested by at least one SRI fund. Then,
\(Y_{ijt}^1 - Y_{ijt}^0\) is the treatment effect of SRI investment. Let
\(X_{ijt}\) be a set of observable firm characteristics that influence a firm’s cost of equity capital, then
\(Y_{ijt}^0\) and
\(Y_{ijt}^1\) can be decomposed into the mean given firm characteristics,
\(\mu _0(X_{ijt})\) and
\(\mu _1(X_{ijt})\), and the deviation from the mean,
\(U_{ijt}^0\) and
\(U_{ijt}^1\):
$$\begin{aligned} Y_{ijt}^0= & {} \mu _0(X_{ijt})+U_{ijt}^0,\nonumber \\ Y_{ijt}^1= & {} \mu _1(X_{ijt})+U_{ijt}^1. \end{aligned}$$
(1)
Our goal is to learn about the effect of the treatment,
D, on the cost of equity capital. We define the treated group as the firm-year observations that receive SRI investment, and the control group as those that do not. Define
\(D_{ijt-1} = 1\) if a firm is invested by SRI funds (treated) at time
\(t-1\), and
\(D_{ijt-1} = 0\) otherwise. Since each firm is observed only in one state, either invested or not invested by SRI funds, the observed outcome,
\(Y_{ijt}\), is
\(Y_{ijt} = D_{ijt-1}Y_{ijt}^1 + (1-D_{ijt-1})Y_{ijt}^0\). Substituting Eq. (
1) into this expression, we get
$$\begin{aligned} Y_{ijt}= & {} Y_{ijt}^0+(Y_{ijt}^1-Y_{ijt}^0)D_{ijt-1} \nonumber \\= & {} \mu _0+(\mu _1-\mu _0+U_{ijt}^1-U_{ijt}^0) D_{ijt-1}+U_{ijt}^0. \end{aligned}$$
(2)
Assuming a linear in parameters function for
\(\mu _l, l=\left\{ 0,1\right\}\) and adding fixed effects, Eq. (
2) implies the regression:
$$\begin{aligned} Y_{ijt} =X_{ijt} \beta ^0 +[X_{ijt} (\beta ^1-\beta ^0) + (U_{ijt}^1-U_{ijt}^0)]D_{ijt-1} + \tau _t + \phi _j + U_{ijt}^0. \end{aligned}$$
(3)
That is, we define the cost of equity capital as a function of time-varying firm characteristics
\(X_{ijt}\), and state-specific coefficients
\(\beta ^l\) for
\(l=\left\{ 0,1\right\}\). In addition, we account for an unobservable year-specific effect
\(\tau _t\), an unobservable time-invariant effect at the industry level
\(\phi _j\), and state-specific time-varying unobservables
\(U_{ijt}^l\)8.
In the base model we assume a common unobservable effect and that \(D_{ijt-1}\) is conditionally exogenous, i.e. \(U_{ijt}^1=U_{ijt}^0 = U_{ijt}\), and \(D_{ijt-1} \bot U_{ijt} | X_{ijt}, \tau _t, \phi _j\). Our parameter of interest \(\gamma \equiv E[Y_{ijt}^1-Y_{ijt}^0] = E[X_{ijt} (\beta ^1-\beta ^0)]\), is the average treatment effect of SRI investment on firm cost of equity capital. A negative estimate of \(\gamma\) indicates that on average, SRI investment decreases firm cost of equity capital, which is in line with our main hypothesis. Identification of \(\gamma\) relies on adequately controlling for the factors that lead to variation in the cost of equity capital, so that conditional on the control variables, firm cost of equity capital only varies through the receipt of SRI investment. We include industry fixed effects to control for the variation in cost of equity capital due to industry-specific and time-invariant characteristics, and year fixed effects to control for changes over time, such as economic shocks or the evolution of investor (equilibrium) preferences, that affect firm cost of equity capital.
As a robustness check of the results from our baseline regressions, we estimate the effect of SRI investment on cost of equity capital in a difference-in-differences model. This model allows us to compare the average change in cost of equity capital after a firm receives SRI investment to the change for a firm that never receives SRI investment. For this analysis, we use a subset of firms that are either never included in the portfolios of SRI mutual funds in our sample, or are included in the portfolio after the start of our study period.
Selection issues
While each SRI fund determines in which firms to invest, the set of SRI eligible firms is not likely to be random. In other words, a firm may self-select into being eligible for SRI based on an array of intrinsic and extrinsic factors. Some of these factors may be common within industries or geographical regions. Industry characteristics may drive a firm’s environmental behavior, since undertaking environmentally responsible activities may be more beneficial or less costly for a firm in certain industries than others. Bagnoli & Watts (
2003) show in a theoretical framework that in competition for socially responsible consumers, the level of environmental and social responsibility provided by a firm depends on the market competitiveness of an industry. Empirical evidence also shows that for a firm that sells final goods to consumers, environmental responsibility may be a product differentiation strategy to attract customers that care about the environment (Henriques & Sadorsky,
1996; Anton et al.,
2004; Eccles et al.,
2014). The industry fixed effects in Eq. (
3) capture many types of selection issues that we believe exist at the industry level. Firm environmental behavior also depends on the environmental preferences of the community members and regulators in the geographic region in which a firm is located. We use a state fixed effect to account for the spatial variation in the regulatory pressure on a firm to behave environmentally responsibly. Moreover, firm financial characteristics, which determine the affordability of environmental activities to a firm, influences the firm’s decision to undertake environmentally responsible activities. If by conditioning on firm characteristics and the fixed effects,
\(U_{ijt}^1=U_{ijt}^0 = U_{ijt}\) and
\(D_{ijt-1} \bot U_{ijt}\), then we can identify
\(\gamma\).
To the extent that there still exist idiosyncratic and time-varying unobservables that are not captured by the fixed effects, for example, if the unobservables lead to distinct cost of equity capital between SRI eligible and SRI ineligible firms without SRI investment, i.e.
\(Cov(D_{ijt-1}, U_{ijt}) \ne 0\) after controlling for
\(X_{ijt}\) and the fixed effects, the estimate of
\(\gamma\) may still be biased. The concept of stakeholder influence capacity (Barnett,
2007) and its role in determining the financial outcomes of CSR may be a potential source of time-varying unobservables at the firm level. According to Barnett (
2007), a firm achieves the financial outcomes of CSR activities through its ability to use CSR to improve stakeholder relationships. These relationships could be dynamic in light of, for example, the intensity of media scrutiny of the firm, which changes over time. In the case that the unobserved stakeholder influence capacity is correlated with cost of equity capital through channels other than SRI investment, while also correlated with firm eligibility for SRI investment, then we cannot use the fixed effects to capture this (unobservable) variability in cost of equity capital. For example, by studying events of conflicts and cooperation with stakeholders of public firms in gold mining, Henisz et al. (
2014) find that greater stakeholder support leads to higher firm valuation. If some of the stakeholders events influence firm eligibility of SRI investment, for example, events concerning environmental compliance, then we are faced with selection bias in the sense that firms that are eligible for SRI investments are those faced with less stringent environmental regulations. Although this relationship may be more relevant to industries that transform natural resources to shareholder wealth (Henisz et al.,
2014), since the level of shareholder cooperation is dynamic, an industry fixed effect may be insufficient to identify the effect of SRI investment.
To address the issue of a non-random sample due to selection bias, we use a set of instrumental variables to predict the propensity that a firm behaves environmentally responsibly, which qualifies the firm for SRI investment, before estimating the impact of SRI investment on cost of equity capital. Formally, define
\(D^*_{ijt}\) as a latent variable that generates
\(D_{ijt}\) according to a threshold crossing rule:
$$\begin{aligned} D_{ijt} = \varvec{1} [D^{*}_{ijt}>0], \end{aligned}$$
(4)
where
\(\varvec{1}[A]\) is an indicator function (
\(\varvec{1}[A] = 1\) if
A is true; 0 otherwise). Specifically, define
$$\begin{aligned} D^*_{ijt}=\mu _{D_{ijt}}(Z_{ijt}) - V_{ijt}, \end{aligned}$$
(5)
where
\(Z_{ijt}\) is a vector of firm and industry specific, time-varying characteristics that influence the firm’s decision to become eligible for SRI, and
\(\mu _{D_{ijt}}(Z_{ijt}) - V_{ijt}\) can be interpreted as the net benefit for a firm with characteristics
\((Z_{ijt}, V_{ijt})\). Identification requires that some element
\(Z_{ijt}^k\) in
\(Z_{ijt}\) is excluded from
\(X_{ijt}\), so that by varying
\(Z_{ijt}^k\), we can recover the probability that a firm is eligible to receive treatment without affecting the outcome.
Our first instrumental variable is the ratio of independent directors over the total number of directors. It is excluded from
\(X_{ijt}\) as a firm’s board composition is not likely to directly influence firm financial performance. Although some may argue that shareholders may be willing to accept lower returns from firms with better corporate governance, of which board independence is an aspect, several studies that examine the relation between firm corporate governance and financial performance do not find evidence of a significant correlation between board independence and firm performance which include the cost of equity capital (Ashbaugh et al.,
2004; Pham et al.,
2011; Lima & Sanvicente,
2013). In particular, Ashbaugh et al. (
2004) find that the majority of the governance attributes considered in their study, including board independence, are significantly associated with market risk (market beta), and no significant relation exists between governance attributes and financial performance when beta is controlled for. A conclusion follows that board independence affects cost of equity capital only through the effect on market risk. Since we control for beta in our analysis, the potential correlation between board independence and cost of equity capital would be subsumed in beta.
The motivation underlying this instrument is twofold. First, while the external factors may influence firm behavior with respect to the environment, they are unlikely to have a homogeneous effect because these factors may take effect through the board of directors, which vary in philosophies and styles, and therefore attitude towards stakeholder interests
9. To what extent a firm takes account of the interests of its stakeholders and fulfills their demands depends on firm engagement with the stakeholders. As is argued by Kassinis & Vafeas (
2002) and Kock et al. (
2012), stakeholders have greater influence over the board when the board is composed of fewer insiders (employees of and individuals affiliated with the firm), because non-affiliated directors are more likely to address stakeholder interests. We therefore use the fraction of independent directors – non-affiliated directors who are independent of management and tend to be friendly to stakeholders – over the total number of directors to measure firm responsiveness to pressure from external stakeholders.
Second, as is argued in several studies (e.g., Barnea and Rubin,
2010; Barnett,
2007), certain socially responsible activities, such as those that are purely altruistic or out of the manager’s personal benefits, do not substantially contribute to improving stakeholder relationships, and therefore do not improve firm financial performance. These activities are especially likely to occur when the manager makes environmental decisions without going through the board. Therefore, these activities create agency problems since they benefit the manager or society, but not shareholders (Jensen & Meckling,
1976)
10. Barnett (
2007) argues that when agency problems confound with CSR, it may create a downward bias in the estimate of the financial outcome of CSR (i.e., an upward bias in the effect on cost of equity capital). (Krüger
2015) provide further empirical evidence that CSR that reflect agency problems harm shareholder wealth. Given the evidence from Byrd & Hickman (
1992) that independent directors can monitor management decisions on the behalf of shareholders, and therefore mitigate the agency problem, we use the fraction of independent directors as an instrument to account for the potential downward bias in the estimate of the financial benefit of SRI investment due to agency costs associated with firm environmental behaviors. Data for this instrumental variable are from the Institutional Shareholder Services (ISS) database.
Our second instrumental variable addresses the case that social and environmental preference may translate into actions that influence corporate strategy regarding environmental performance (Kitzmueller & Shimshack,
2012). The variable is constructed as the number of Sierra Club members in the state in which a firm is headquartered per 1,000 residents of the state. Sierra Club is a nation-wide environmental organization in the United States which promotes green policies in areas such as green energy and climate change by lobbying politicians. Membership of the Sierra Club is argued to represent the environmental preferences of the population of a state, or the marginal value the state residents place on environmental quality, which influences the pressure of behaving environmentally responsibly received by a firm in that state (Maxwell et al.,
2000; Delmas & Montes-Sancho,
2010)
11. This instrument is correlated with firm cost of equity capital only through SRI investment, which satisfies the exclusion restriction for instrumental variables
12.
Industry heterogeneity
In addition to controlling for industry-level variation in the cost of equity capital, we also expect that the effect of SRI investment may vary across industries. Several studies indicate that the motivation for undertaking environmentally responsible activities and the costs and benefits vary across industries, which may in turn lead to a heterogeneous financial impact (e.g., King
2001; Eccles et al.,
2014; Lyon and Maxwell,
2011). In particular, industry environmental performance and closeness to final consumers are two important factors relevant to the financial outcomes of firm environmental responsibility. Since polluting industries may have stronger incentives pursue environmental responsibility, it is important to correct for industry type and industry-level environmental performance (Lyon & Maxwell,
2011; Bénabou & Tirole,
2010; Dam
2015). The effect of SRI investment on the cost of equity capital may therefore depend on the industry’s environmental performance.
In terms of closeness to final consumers, Anton et al. (
2004) find that for industries that deliver final goods or services to consumers and individual customers, environmental responsibility is positively related to firm financial performance. Dimson et al. (
2015) find that SRI fund managers’ engagement with firm environmental and governance issues are especially concentrated in certain industries, such as manufacturing and advertising-intensive industries, and the financial impact of such engagement also varies across industries. Since an increasing number of consumers are willing to pay higher prices for goods and services with environmental features (Kitzmueller & Shimshack,
2012), environmental activities in these industries are likely considered as value-relevant by both SRI and neutral investors. When neutral investors do not require a risk premium for trading firms in these industries, the firms benefit from a reduction in the cost of equity capital through SRI investment (Vanwalleghem,
2013). On the other hand, the clients of the intermediary industries (i.e., the downstream industries) are likely not willing to pay for the price premium associated with environmental responsibility if the premium cannot be passed along to final consumers; in this case, environmental activities are seen as an unnecessary cost.
Since there is variation in the scope of industries within the industry groups, the effect of SRI investment may also vary in industry diversity. For less diversified industry groups such as Chemicals, Construction, and Automobiles and Trucks, there more likely exists a homogeneous effect of SRI investment
13. However, for other industry groups such as Retail, Wholesale, and Business Services, the sub-industry groups are highly diversified, it is possible that SRI investment has a positive effect on the cost of equity capital of certain sub-industries, and a negative effect on that of other sub-industries
14.
To explore the potentially heterogeneous effect of SRI investment on the cost of equity capital of firms across different industry groups, we estimate a model with interactions of SRI investment and the three industry group characteristics discussed above – industry group diversity, environmental performance, and closeness to final consumer. We account for industry group diversity with the number of sub-industry groups in each industry group. For industry environmental characteristics, we construct an index of environmental concerns relative to environmental strength for each industry group. Data for this index are from the MSCI ESG STATS dataset (previously KLD STATS). We group the firms by the Fama-French 48 industries, and construct the index as the ratio of number of environmental concerns in four categories – Regulatory Compliance, Toxic Spills & Releases, Climate Change, and Other Concerns – to environmental strengths in four respective categories - Environmental Opportunities, Waste Management, Climate Change, and Other Strength – among firms in each industry group over the available time period
15. We calculate the fraction of intermediate industries in each industry group based on four-digit SIC codes to measure an industry group’s closeness to final consumers.
To address the endogeneity of SRI investment and its interactions with the industry characteristics, we use a control function approach by first regressing the SRI investment indicator on firm characteristics and the two instrumental variables, then including the error term from the first stage in the estimation of the effect of SRI investment on firm cost of equity capital.