Introduction

Recent years have seen growing attention on corporate social responsibility (CSR) in the business world and in the academic literature, with a particular emphasis on building a business case for corporate engagement in CSR (e.g., Margolis and Walsh 2003; Orlitzky et al. 2003). Researchers have begun to explore institutional determinants of CSR activities (e.g., Campbell 2007; Marquis et al. 2007; Matten and Moon 2008). Initial empirical evidence has identified influences of nation-level institutions and business systems on cross-country variations in corporate engagement in CSR (e.g., Jackson and Apostolakou 2010; Ioannou and Serafeim 2012). Since corporations are embedded within social structures in local, small-scaled, geographically bounded communities, local contexts should also drive corporate practices (e.g., Marquis and Battilana 2009).

A better understanding of the social antecedents in local communities that either enable or constrain CSR activities is important for two reasons. First, it promises a larger tool kit from which policy makers, non-governmental organizations, and public policy think tanks can rely upon when seeking to influence the respective corporate activity so as to achieve desirable societal objectives. Second, it provides another frame of reference to build the business case for CSR. Yet, despite the early efforts of Navarro (1988) and Galaskiewicz (1997) on corporate charitable giving activities, we still lack rigorous, large-scale evidence that enhances our understanding of whether, and in what ways, social environments in local communities affect the adoption and implementation of CSR activities (e.g., Aguilera et al. 2007). Indeed, Brammer et al. (2012, p. 4) argued that although “C‘S’R includes the aspect of ‘society’ already in its very label …. [i]t is fair to say that the literature on CSR, most of it published in management or business or business studies journals, has neglected the ‘societal’ aspects of CSR by and large. Most of the literature has treated the ‘social’ element as a black box.” This study fills the gap by exploring the distinct effects of community-level social environments, namely community social capital, on CSR.

Firms can engage in CSR by either increasing positive CSR activities or by decreasing negative CSR activities. These two forms of CSR activities are “empirically and conceptually distinct constructs” (Mattingly and Berman 2006), and they embody the kinds of “demonstrable corporate behaviors” that Aguilera et al. (2007) suggested researchers should focus on when examining CSR. Positive CSR activities are corporate social actions that “extend beyond immediate profit maximization goals and are intended to increase benefits or mitigate social problems for constituencies external to the firm” (Marquis et al. 2007, p. 926). Negative CSR activities are “irresponsible and minimally responsible corporate behaviors” that do harm to one or more non-shareholder stakeholders, including employees, suppliers, customers, communities, and government (Campbell 2007). We measure these two forms of CSR activities separately and treat the corresponding measures as distinct constructs.

Specifically, we explore whether social capital at the county level in the US is systematically and respectively related to adoption and implementation of positive CSR activities and negative CSR activities of resident corporations with organizational headquarters located in the county.Footnote 1 We focus on the US setting because CSR practices are particularly prevalent among US corporations (Matten and Moon 2008). Additionally, prior studies in finance and accounting find that religious norms and networks in US counties where corporate headquarters are located affect resident firms’ investment and tax planning practices (e.g., Hilary and Hui 2009; Boone et al. 2013). Accordingly, it is logical to explore the effects of social capital in US counties, a construct that captures non-religious social environments in US counties, on CSR practices.

Social capital has been a subject of extensive investigation across various fields in the social sciences. A central theme in this diverse literature is that the level of social capital in a geographically bounded community, as captured by social norms and social networks in the community, facilitates norm-consistent and constrains norm-deviant behaviors of individuals and organizations embedded in that community. Following this tradition, we use density of social networks and strength of civic norms in US counties to build the community social capital construct. Civic norms are non-religious social norms that constrain narrow self-interest (Knack and Keefer 1997), prescribe behaviors that are cooperative in nature (Fukuyama 1995), and place emphasis on the interests of the collectivity rather than one’s self-interest (Coleman 1988). Social networks capture horizontal social relations that exist in associations and organizations in the community which provide closures in social relations (Coleman 1988) and instill “habits of cooperation, solidarity, and public-spiritedness” (Putnam 1993).Footnote 2

Based on the social capital research, we hypothesized that a community’s social capital, as captured by the strength of civic norms and the density of social networks, facilitates positive CSR activities but constrains negative CSR activities, leading to an overall positive relation (overall negative relation) between community social capital and positive CSR activities (negative CSR activities). Using longitudinal data from 3688 unique US public corporations in 518 unique US counties during 1997–2009, we found strong evidence to support our hypothesis. Firms headquartered in US counties with higher levels of community social capital have significantly more positive CSR activities and significantly fewer negative CSR activities. These findings were based on panel regressions that control for firm and industry characteristics, county-level demographical factors, political influences, and year and industry fixed effects; they were confirmed in a range of robustness checks and sensitivity analyses.

Our conceptual framework highlights the extent to which a specific external institutional logic (i.e., community social capital) infiltrates local corporations to enable positive CSR activities and constrain negative CSR activities. Corporations are embedded within diverse communities where multiple institutional logics coexist (Kraatz and Block 2008; Glynn and Raffaelli 2013). Community isomorphism, a process where firms mimic other local firms’ CSR activities, could also be an influential community-level institutional logic (e.g., Campbell 2007; Marquis et al. 2007). Accordingly, we also explore the effects of institutional logics arising from community isomorphism on positive and negative CSR activities, respectively.

This article is organized as follows. In the next section, we review the related CSR literature that examines institutional determinants of CSR activities. Next, we build on the literatures to develop our hypotheses. We then test our predictions using publicly listed corporations with headquarters located in US counties between 1997 and 2012 for which the requisite data on CSR and social capital are available. Finally, we discuss the results, limitations, and implications of the study.

Institutional Determinants of CSR Activities

This study explores whether, and in what way, community-level social institutions affect the adoption and implementation of “demonstrable corporate behaviors” of resident firms that are closely related to CSR (Aguilera et al. 2007), namely positive CSR activities and negative CSR activities. Given the focus, we found a limited number of prior studies that have examined institutional antecedents of CSR activities.

Empirical evidence suggests that local institutions, as captured by norms and networks among local corporate executives, have significant influences on corporate charitable giving. In particular, local executive social networks in US cities, such as “tithing clubs” (Navarro 1988) and social ties to local philanthropic corporate leaders (Galaskiewicz 1997), promote corporate charitable giving practices of resident corporations.

As well, evidence suggests that variations in nation-level institutions drive cross-country variations in CSR activities as captured by the firm’s overall social performance in all types of CSR activities, including both positive CSR activities and negative CSR activities (i.e., CSP). Jackson and Apostolakou (2010) found that European firms operating in more liberal market economies have higher levels of CSP when compared to comparable firms operating in more coordinated market economy of Continental Europe. Using a sample over 12,000 firm-year observations from 42 countries spanning seven years, Ioannou and Serafeim (2012) provided large-scale evidence to show that nation-level institutions, including political, labor, education, and cultural systems, have significant impact on the levels of CSP.

A shortcoming of these prior empirical investigations is that they provide little insight on social influences of local communities on CSR activities (Aguilera et al. 2007; Brammer et al. 2012). In particular, they neglected the effects of community social capital on either positive CSR activities or negative CSR activities. Yet, recent theoretical advances in the CSR literature pointed to community-level social environments as important institutional antecedents of CSR activities. Focusing on corporate undertaking of positive CSR activities, Marquis et al. (2007, p. 937) predicted that “community-level social and normative institutional forces will affect the level of corporate social action.” Focusing on corporate undertaking of irresponsible and minimally responsible corporate activities (i.e., negative CSR activities), Campbell (2007, p. 959) predicted that “corporations will be more likely to act in socially responsible ways if they operate in an environment where normative calls for such behavior are institutionalized.”

Community Social Capital as an Institutional Antecedent of CSR Activities

Motivated by the aforementioned gap in the CSR literature, in this section, we build on the social capital research to develop a hypothesis that expounds the respective effects of community social capital on positive CSR activities and negative CSR activities. We define community social capital as the manifestation of the effects of civic norms and social networks arising from the local, small-scaled, geographically bounded community surrounding a firm’s headquarter. Coleman (1988, p. 104) argues that “an especially important form of social capital is the norm that one should forgo self-interest and act in the interests of the collectivity.” Civic norms promote trust and provide commonly shared frameworks or mental models that community members use when they communicate and judge observable behaviors of members in the community. However, dense social network ties and interactions among persons are required to facilitate effective communications and enforcement of civic norms (e.g., Coleman 1988; Putnam 1993; Spagnolo 1999). Consequently, social capital encapsulates “features of social life—networks, norms, and trust—that enable participants to act together more effectively to pursue shared interests (Putnam 1995, pp. 664–665).” Following these pioneers in social capital research, we take density of a community’s social networks and strength of a community’s civic norms as the two pillars that form a community’s social capital.

Researches that focused on community social capital have emphasized its function as a public good (e.g., Coleman 1988; Putnam et al. 2000; Guiso et al. 2004). These researchers argued that community social capital produces diffused influences “not only to those who possess social capital but also to people living in regions with a high level of social capital.” and it promotes “community cohesion and information flow that accrue to community members who do not have high levels of personal social capital themselves” (Kwon et al. 2013, p. 981). Prior empirical findings showed that community social capital has salient effects on individuals and organizations in the community (e.g., Knack and Keefer 1997; Rupasingha et al. 2000; Guiso et al. 2010). In particular, Rupasingha and Goetz (2007) and Buonanno et al. (2009) measured community social capital as the joint effect of civic norms and horizontal social networks across counties in the US and across provinces in Italy, respectively. They found that community social capital lowers poverty rates and property crime rates in the community, respectively.

A central theme of these prior studies is that community social capital facilitates civic-minded, socially cooperative actions and constrains behaviors that are inconsistent with the prescribed values associated with civic norms. In this context, positive CSR activities and negative CSR activities are two distinct forms of corporate actions that fit the analytical framework particularly well. On one end, positive CSR activities include corporate philanthropy, clean energy, profit sharing programs, corporate volunteering programs, diversity-enhancing work rules, and business practices that promote human rights. On the other end, negative CSR activities include environmental pollution, discriminatory human resource practices, failure to recall defective and dangerous products, child labor violations, investments that disregard negative impact on communities, and corporate tax avoidance. People in communities with high community social capital should view positive CSR activities as conforming to the prescribed values associated with civic norms and perceive negative CSR activities as norm-deviant. Consequently, one would expect that community social capital facilitates positive CSR activities and constrains negative CSR activities.

CSR practices are strategic corporate decisions (e.g., Porter and Kramer 2006) that are likely made in corporate headquarters rather than branches. Moreover, local norms and networks surrounding communities where corporate headquarters are located have significant effects on resident firms’ investment, tax planning practices, and debt contracting (Hilary and Hui 2009; Boone et al. 2013; Hasan et al. 2016). Therefore, we propose the following refutable hypothesis.

Hypothesis 1

Firms headquartered in communities with higher levels of social capital undertake more positive CSR activities but fewer negative CSR activities.

Community Isomorphism and CSR Activities

Our conceptual framework highlights how a specific external institutional logic (at a more macro level), as captured by community social capital, infiltrates corporations (at a more micro level) to enable positive CSR activities and constrain negative CSR activities. However, corporations and their stakeholders are embedded within diverse communities characterized by increasing institutional pluralism where multiple institutional logics coexist (Kraatz and Block 2008; Glynn and Raffaelli 2013), and these other institutional logics could also affect CSR activities.

Among these other institutional logics, the idea that community isomorphism facilitates corporate philanthropic practices is well documented. Navarro (1988), Galaskiewicz (1997), Russo and Fouts (1997), and Wang and Qian (2011) found that firms mimic corporate charitable giving practices of other local corporations to maintain their legitimacy and gain positive stakeholder responses. Extending the mimicking effect to all positive CSR activities, Marquis et al. (2007) theorized that “community isomorphism, that is, the resemblance of a corporation’s social practices to those of other corporations within its community” facilitates positive CSR practices within communities. Building these arguments, we propose the following refutable hypothesis.

Hypothesis 2

The level of positive CSR activities a firm undertakes is positively related to the level of positive CSR activities among other local corporations headquartered in the same community.

In theory, firms can improve legitimacy and gain favorable stakeholder responses by increasing positive CSR activities, by decreasing negative CSR activities, or by doing both. However, in practical terms, firms have a greater degree of freedom in increasing positive CSR activities than in decreasing negative CSR activities. Accordingly, we do not expect similar systematic relation between negative CSR activities a firm undertakes and the level of engagement in negative CSR activities among other local corporations headquartered in the same community. Consistent with this expectation, Zyglidopoulos et al. (2012) found that firms respond to increased media attention by increasing positive CSR activities, but corporate undertaking of negative CSR activities is not systematically related to the level of media attention.

As do Zyglidopoulos et al. (2012), we conjecture that a firm is less apt to mimic by decreasing its negative CSR activities just because negative CSR activities are less prevalent among other local corporations. This is so because there are significant costs associated with decreasing negative CSR activities, and these costs could vary widely from firm to firm depending on “structural” elements such as the firm’s lines of business and its business model or mode of operations. For example, the costs a manufacturing firm faces in decreasing child labor violations depend on whether the firm has domestic or international operations, and if it is the latter, they depend on the reach of the firm’s supply chain and the countries in which the firm’s supply chain is established. Varying idiosyncratic costs associated with decreasing negative CSR activities will mitigate the incentives to mimic other local corporations’ efforts in reducing their negative CSR activities.

Additionally, we conjecture that a firm is also less apt to mimic by increasing its negative CSR activities just because negative CSR activities are more prevalent among other local corporations. This is so because many negative CSR activities are outcomes of traditional or conventional corporate practices that lag behind the increasing demand for higher standards of appropriateness concerning business behaviors and activities (e.g., concern in union relations, concern in defined-benefit pension plan, substantial emission, antitrust.). Firms operating in different industries, and even firms operating within the same industry, could face very different opportunities when it comes to undertaking negative CSR activities. While automobile manufacturers could undercut unionized labor to increase profit, this opportunity is not available to firms in a non-unionized environment. While traditional automobile manufacturers could undermine their vehicles’ exhaust emission and fuel efficiency to enhance profit, such opportunity is unavailable to non-traditional players such as Tesla as the company produces only electric cars. If firms face varying opportunity sets in terms of the negative CSR activities they can undertake, then their ability to mimic the negative CSR activities undertaken by other local firms in the community could be limited. Collectively, the foregoing arguments imply that there is a negligible mimicking effect of community isomorphism in terms of negative CSR activities.

Financial Performance and Social Performance: The Role of Community Social Capital

Improvements in corporate social performance could enhance a firm’s financial performance and benefit shareholders through legitimization, branding, employee loyalty, reputation building, and access to finance (e.g., Hillman and Keim 2001; Wang and Qian 2011; Cheng et al. 2014). Accordingly, the relationship between corporate financial performance and corporate social performance should be positive, that is, there is a positive CFP–CSP relationship.

However, empirical investigations of the CFP–CSP relationship have produced inconclusive results (e.g., Margolis and Walsh 2003; Orlitzky et al. 2003). Garcia-Castro et al. (2010) contended that this empirical inconclusiveness is due to contingent circumstances that moderates the CFP–CSP relationship. Indeed, recent studies found a range of moderating factors such as customer satisfaction (Luo and Bhattacharya 2006), customer perception (Lev et al. 2010), and political access (Wang and Qian 2011) that affect the CFP–CSP relationship.

Juxtaposing insights from these CSR studies with those in social capital research, we conjecture that community social capital could moderate the CFP–CSP relationship. On one hand, community social capital could amplify the positive CFP–CSP relationship because non-shareholder stakeholders in local areas with high levels of community social capital are naturally more attentive, responsive, and receptive to CSR activities that improve a firm’s social performance. In this case, proactive responses to the demand of non-shareholder stakeholders in the community could position the firm as a local leader in CSR, potentially resulting in strong, beneficial effects on its status as a preferred employer and improves its relations with local governments, which, in turn, allows the firm to obtain additional resources, in the form of subsidies and other supports, from the governments.

On the other hand, based on the agency paradigm (Friedman 1962; Haley 1991), it is also plausible that community social capital could attenuate the positive CFP–CSP relationship. In particular, based on the agency paradigm, Benabou and Tirole (2010) contended that non-shareholder stakeholder influences do not always benefit the firm or its shareholders as their influences could lead to overinvestment in CSR because stakeholders could seek “direct values” through corporate engagement in CSR, but they do not bear the full cost of such engagement. Di Giuli and Kostovetsky (2014) found evidence consistent with this “direct values” hypothesis,Footnote 3 implying that “direct values” seeking stakeholders are indeed instrumental in inducing overinvestment in CSR when the external political environment exerts influences conducive to their causes. Indeed, if “direct values” seeking stakeholders are more instrumental in inducing overinvestment in CSR among firms whose organizational headquarters are located in counties with high levels of community social capital, social performance should have a less pronounced effect on firm's financial performance among corporations with headquarters located in local areas with high levels of community social capital.

Firms can enhance their social performance by either increasing positive CSR activities or by decreasing negative CSR activities. Consequently, based on foregoing arguments, one could expect community social capital to either amplify or attenuate the effects of these distinct forms of CSR activities on corporate financial performance. In particular, based on the “direct values” theory of the agency paradigm, one would expect that a high level of community social capital induces local firms to adopt and implement too many positive CSR activities or avoid too many negative CSR activities. In both cases, the result is a less pronounced relationship between the respective CSR activities and firm's financial performance among firms headquartered in counties with high community social capital. Based on the “direct values” theory, we propose the following refutable hypothesis.

Hypothesis 3

Positive and negative CSR activities have a less pronounced effect on corporate financial performance among firms headquartered in local areas with high community social capital.

Empirical Design: Sample, Data, Variables, and Method

To examine the hypotheses, we conducted a multi-level analysis that relates firm-level CSR outcomes to county-level social capital measures while controlling for firm-level, industry-level, and county-level factors that affect the implementation and adoption of CSR activities.

Sample and Data

The starting point of our sample selection was all firms in the KLD Research & Analytics, Inc., hereafter KLD, between 1997 and 2009.Footnote 4 We gathered KLD social ratings data across the six categories of corporate activities including employee relations, environment, community, diversity, human rights, and product quality and safety. We excluded corporate activities belonging to the corporate governance category because activities directed toward shareholders do not strictly belong to and can conflict with a firm’s “social” responsibility (Friedman 1962). Our results are unchanged, if we include the corporate governance category in all the ensuing analyses.

We consulted Standard and Poor’s Compustat database to obtain the corresponding financial and accounting information for the firms. We obtained institutional ownership data from the Thomson Reuters Ownership Database and financial analyst coverage data from the Institutional Brokersʼ Estimate System. Additionally, we complemented the resulting firm-level data with county-level data obtained from various sources. We obtained county-level social capital data from the Northeast Regional Center for Rural Development at the Pennsylvania State University, hereafter NRCRD. We obtained county-level demographic information from the Bureau of Economic Analysis. The final sample contains 19,389 firm-year observations for 3688 unique firms with headquarters located in 518 unique US counties between 1997 and 2009 for which complete data are available from all the aforementioned sources.

Dependent Variables: Firm-Level CSR Activities

We used two firm-level CSR measures because our hypotheses were predicated on positive CSR activities and negative CSR activities, respectively.

Positive CSR

We used KLD’s positive social ratings to measure positive CSR, which are corporate social actions intended to increase social welfare beyond immediate profit maximization considerations (McWilliams and Siegel 2001). For our analysis, KLD positive social ratings include 35 binary scores across the six categories of corporate activities that have a positive impact on non-shareholder stakeholders such as employees, suppliers, customers, communities, government, and society-at-large. In each case, the social rating takes on the value of one, if KLD, based on its proprietary information and assessment, recognizes the firm as maintaining strength in that specific activity; it equals zero otherwise. We excluded two positive social ratings that are related to minority representation on the board of directors and the CEO office, ,,because we judged that the underlying activities are directed toward shareholders. Our results are unchanged, if we include these two positive social ratings. In the end, Positive CSR is the sum of the remaining 33 binary positive social rating scores for a firm in a given year during the sample period between 1997 and 2009. Positive CSR measures the total number of corporate actions that positively affect one or more of the firm’s non-shareholder stakeholders. Appendix A provides definitions of all variables used in the study.

Negative CSR

We used KLD’s negative social ratings to measure negative CSR activities, which include corporate practices that impose externalized costs on one or more of the firm’s non-shareholder stakeholders (Hoi et al. 2013). In each case, the negative social rating takes on the value of one, if KLD discovers a weakness in a specific activity; it equals zero otherwise. In this case, our procedure produced a Negative CSR variable for each firm-year that spans the same six categories we used to calculate Positive CSR. In all, Negative CSR considers up to 27 different corporate actions that have a negative impact on the firm’s non-shareholder stakeholders. Negative CSR is the sum of these 27 binary negative social rating scores for a firm in a given year during the sample period between 1997 and 2009.

Independent Variables: Community Social Capital and Community Isomorphism

Community Social Capital

NRCRD provides two datasets reporting the social capital index and its constituent components, namely the corresponding data on civic norms and social networks that make up the index, for all the US counties. The old dataset reports data for 1990, 1997, and 2005. The new dataset, NEW_NRCRD, reports data for 1997, 2005, and 2009. The major difference between these datasets is that NEW_NRCRD uses a different methodology to account for non-profit organizations in the county, resulting in different estimates for social capital. We used the more recent data in the NEW_NRCRD to construct the county-level social capital measure.

The NEW_NRCRD provides three variables, sk97, sk05, and sk09, which are the estimates of the levels of social capital across all the US counties in three different years (i.e., 1997, 2005, and 2009, respectively). As described in Rupasingha et al. (2006), the NRCRD’s social capital index in each year is the first principal component from a principal component analysis based on four factors in that year: percentage of voters who voted in presidential elections (Pvote), response rate to the Census Bureau’s decennial census (Respn), number of non-profit organizations in the county (Nccs), and number of social organizations and associations in the county (Assn). The first two factors are proxies that reflect the extent to which civic norms manifest themselves in individuals through their actions (Guiso et al. 2004). The latter two factors are proxies for density of horizontal social networks through people’s participation in associational, voluntary social groups and organizations (Coleman 1988; Putnam 1993), including non-profit organizations, social organizations such as sports clubs, public golf courses, bowling and fitness centers, and associations with a professional, business, political, religious, or other orientation.

Our test variable, Social capital, was directly based on the NEW_NRCRD’s social capital index sk09, sk05, and sk97. To create a complete longitudinal sequence of social capital between 1997 and 2009, we filled in the data for the missing years using the social capital index in the preceding year for which data are available. For instance, we filled in the missing data for 1998 to 2004 using sk97, the social capital index in 1997. This procedure is reasonable because social capital is relatively persistent overtime. Nevertheless, we performed several sensitivity analyses to ease the concern that this back-filling procedure drives our findings. In particular, we ran the regressions using a reduced sample limited to the three years in which social capital data are in fact available, namely 1997, 2005, and 2009. We used a linear interpolation method to generate the social capital index in the missing years and re-estimate our models using linearly interpolated social capital data. We found the same results in these alternative analyses. These results are not tabulated.

Community Isomorphism

Since we treat positive and negative CSR activities as distinct corporate activities, we used two separate measures to capture the respective effects of community isomorphism on positive CSR activities and negative CSR activities. Positive isomorphism for a firm in a given year, say firm X, is the mean number of positive CSR activities for all other firms headquartered in the same county as firm X in that year, excluding the positive CSR activities of firm X. Negative isomorphism for a firm in a given year, say firm Y, is the mean number of negative CSR activities for all other firms headquartered in the same county as firm Y in that year, excluding the negative CSR activities of firm Y.

Control Variables

Following Ioannou and Serafeim (2012), we included a range of firm-level control variables to isolate the effects of firm size (Size), growth potential (M/B), firm performance (ROA), leverage (Leverage), research intensity (R&D intensity), institutional ownership (IO), diversification (Segments), firm risk (Earnings volatility), and firm visibility (Analyst coverage). In particular, the two variables M/B and ROA are included to capture the effects of market-based and accounting-based corporate financial performance on CSR. As did Ioannou and Serafeim (2012), we also included industry-level factors to control for market competition (Herfindahl index) and industry fixed effects based on two-digit SIC classification to account for systematic differences in regulations and self-regulations across industries (Campbell 2007). Additionally, we included a control variable to capture the effect of political influences on CSR (Political strength) and a range of county-level control variables to isolate the effects of income inequality, rural/metro classification, fraction of minorities, and fraction of female labor force. Di Giuli and Kostovetsky (2014) found that political tilt toward Democratic values drive corporate social performance. Whereas county demographical factors could affect social capital (Alesina and La Ferrara 2002; Putnam 1995; Rupasingha et al. 2006), there is little a priori reason to expect that they are systematically associated with positive CSR activities or negative CSR activities. The idea here is to show that the effects of social capital are not just picking up the effects of these county-level demographical factors or political tilt in the social environment. Lastly, we included year fixed effects to capture potential time trends in the data. All these variables are defined in Table 5 in Appendix.

Empirical Model

We used the following baseline regression model to test our hypotheses:

$$\begin{aligned} {\text{CSR}}_{\text{it}} = \, \beta_{0} + \, \beta_{ 1} \,{\text{Social capital}}_{\text{it}} + \, \beta_{ 2} \,{\text{Size}}_{\text{it}} + \, \beta_{ 3} \,M/B_{\text{it}} + \, \beta_{ 4} \,ROA_{\text{it}} + \, \beta_{ 5} \,{\text{Leverage}}_{\text{it}} \hfill \\ \quad \quad \quad + \, \beta_{ 6} \,{\text{R}}\& {\text{D intensity}}_{\text{it}} + \, \beta_{ 7} \,{\text{IO}}_{\text{it}} + \, \beta_{ 8} \,{\text{Herfindahl index}}_{\text{it}} + \, \beta_{ 9} \,{\text{Segments}}_{\text{it}} \hfill \\ \quad \quad \quad + \, \beta_{ 10} \,{\text{Earnings volatility}}_{\text{it}} + \, \beta_{ 1 1} \,{\text{Analyst coverage}}_{\text{it}} + \, \beta_{ 1 2} \,{\text{Political strength}}_{\text{it}} \hfill \\ \quad \quad \quad + \, \beta_{ 1 3} \,{\text{Income inequality}}_{\text{it}} + \, \beta_{ 1 4} \,{\text{Dummy}}\left( {\text{Urban}} \right)_{\text{it}} + \, \beta_{ 1 5} \,{\text{Minority}}_{\text{it}} \hfill \\ \quad \quad \quad + \, \beta_{ 1 7} \,{\text{Female labor}}_{\text{it}} + {\text{ Industry dummies }} + {\text{ Year dummies }} + \, \varepsilon_{\text{it}} \hfill \\ \end{aligned},$$
(1)

where CSR could be either Positive CSR or Negative CSR, and the subscript it stands for a given firm or county in a given year during the sample period between 1997 and 2009. For ease of exposition, we drop the it subscript from this point forward.

Results

Table 1 reports means, the first and the third quartile values, standard deviations, and correlations of the variables. Social capital and Positive CSR are positively and significantly correlated (0.06, p < 0.01). In contrast, Negative CSR and Social capital are significantly and negatively correlated (−0.03, p < 0.01). These correlations provide tentative support for our hypotheses. Additionally, we found that Positive CSR and Negative CSR are positively and significantly correlated (0.33, p < 0.01), suggesting that some companies in our sample used positive CSR activities to deflate the detrimental consequences of negative CSR on the firm’s overall CSR reputation (Hoi et al. 2013).

Table 1 Descriptive statistics and correlations
Full size table

Effects of community social capital

Table 2 presents the results from the baseline regression model of Eq. (1). Our sample involves panel data in which firm-level observations could be serially dependent across time. Accordingly, we estimated the baseline model using standard errors that were clustered at the firm level. Petersen (2009) showed that the firm-level clustering method produces standard errors that are robust to correlation within a firm across time. We obtained similar results when we clustered standard errors at the county level; these alternate results are not tabulated. In Table 2, model 1, we used Positive CSR as the dependent variable and ran Poisson regressions because the dependent variable is count, rather than continuous, data. In model 2, we ran Poisson regression with Negative CSR as the dependent variable.

Table 2 Effects of community social capital and community isomorphism on positive/negative CSR activities
Full size table

Hypothesis 1 predicted a positive relation between community social capital and positive CSR activities a firm undertakes and a negative relation between community social capital and negative CSR activities. The results in Table 2, models 1 and 2, support these predictions. In model 1, where Positive CSR is the dependent variable, the estimate on Social capital is positive and significant (0.09, p < 0.01). In model 2, where Negative CSR is the dependent variable, the estimate on Social capital is negative and significant (−0.05, p < 0.01).

Effects of Community Isomorphism

Table 2, models 3 and 4, present the results of two additional regressions based on the baseline model, except that in each of these models we included an additional independent variable to estimate the effect of community isomorphism on positive CSR activities and negative CSR activities, respectively. In model 3, where Positive CSR is the dependent variable, we added the Positive isomorphism variable to examine the effect of community isomorphism on positive CSR activities. In model 4, where Negative CSR is the dependent variable, we added the Negative isomorphism variable.

Hypothesis 2 predicted that the number of positive CSR activities a firm undertakes is directly related to the level of engagement in positive CSR activities among other local corporations headquartered in the same community. The evidence from Table 2, model 3, support this prediction. In model 3, the estimate on Positive isomorphism is positive and significant (0.08, p < 0.01). On the other hand, we do not expect similar systematic relation between negative CSR activities a firm undertakes and the level of engagement in negative CSR activities among other local corporations headquartered in the same community. In model 4, the estimate on Negative isomorphism is insignificant at conventional levels (p = 0.18).

More importantly, across both models, the estimates on Social capital remained significant and they retained the same sign and same magnitude as before: in model 3, the estimate on Social capital is positive (0.09, p < 0.01), and in model 4, the corresponding estimate is negative (−0.05, p < 0.01). These empirical regularities suggest that the effects of community social capital on CSR activities remain unchanged after controlling for the effects of community isomorphism.

The Moderating Effects of Community Social Capital on the CFP–CSP Relation

Based on the agency paradigm, the “direct values” theory of Benabou and Tirole (2010) and its corollaries predicted that community social capital will attenuate the positive CFP–CSP relationship. Alternatively, one would expect community social capital to amplify the financial benefit that a firm obtains from CSR activities, if non-shareholder stakeholders in communities with higher levels of social capital are naturally more attentive, receptive, and responsive to such activities. This section explores the validity of these competing arguments.

Following Di Giuli and Kostovetsky (2014), we used return on assets (ROA) to measure corporate financial performance (CFP). ROA is income before extraordinary items in year t divided by lagged total assets in year t − 1. It is an accounting-based financial performance measure that has been widely used to explore the CFP–CSP relation (e.g., Waddock and Graves 1997; Orlitzky et al. 2003). Because there is a lag between social performance and its impacts on accounting-based performance, the effect of social performance was evaluated in terms of subsequent ROA. We used an ROA measure with a 1-year lead, ROA +1 year, to measure subsequent accounting-based performance in the year immediately following the year in which corporate engagement in social responsibility was measured.

To create a global measure that captures a firm’s overall corporate social performance (CSP), one must first decide the proper ranking of priorities (i.e., weights) to be assigned to each category of corporate activities that either positively or negatively affect the firm’s stakeholders. Our theoretical argument implies that activities directed toward shareholders, namely activities belonging to the category of corporate governance, should be assigned no weight as they are not strictly related to corporate “social” responsibility. However, we do not have additional a priori expectations to further guide the construction of the CSP measure. Thus, we followed the convention established by Sharfman (1996) and Waddock and Graves (1997), followed by Hillman and Keim (2001) and Ioannou and Serafeim (2012) among others and assigned equal weight to each of the six categories of corporate activities we included in constructing positive CSR and negative CSR. This aggregation method treats each social rating in each category, positive or negative, as equally weighted. It is easy to operationalize: CSP equals to Positive CSR minus Negative CSR for a firm in a given year during the sample period between 1997 and 2009.

We performed OLS regressions with ROA +1 year as the dependent variable. As it is common in the CFP–CSP literature (e.g., Orlitzky et al. 2003), we included lagged independent variables to isolate the effects of firm size, firm risk, leverage, industry, and year. Following McWilliams and Siegel (2001) and Surroca et al. (2010), we included additional controls to mitigate model misspecification problems arising from omitted variables. Specifically, we isolated the effects of managerial ability (Demerjian et al. 2012), intangibility as measured by research intensity (e.g., McWilliams and Siegel 2001), and external monitoring of institutional owners (Chaganti and Damanpour 1991). Hypothesis 3 and its corollaries predicted that the effects of corporate engagement in social responsibility will be attenuated among firms headquartered in counties with high community social capital. Accordingly, we included a dummy variable, High social capital, to capture the influence of high community social capital on subsequent firm's financial performance. High social capital equals one, if social capital of the county in which a firm is headquartered belongs to the top-quartile of the social capital distribution in that given year; it equals zero otherwise. Lastly, we included county-level demographical factors in the empirical model.

Table 3 presents the findings from three pairs of OLS regressions. The three pairs of regressions differ in terms of the specific measure used to gauge corporate engagement in social responsibility. Models 1 and 2 used CSP. Models 3 and 4 used Positive CSR. Models 5 and 6 used Negative CSR. In the second model in each pair of regressions, namely models 2, 4, and 6, we included an interaction term between the specific measure of social responsibility (i.e., CSP, Positive CSR or Negative CSR) and High social capital. We are particularly interested in the estimates of these interaction terms because they provide direct evidence on whether and how a social environment with high community social capital influences the effect of corporate engagement in social responsibility on a firm’s future financial performance.

Table 3 Effects of social capital on the CFP–CSP and CFP–CSR relations
Full size table

The estimates on CSP in model 1 and Positive CSR in model 3 were both positive and significant. These empirical regularities are consistent with those in Wang and Qian (2011) and Erhemjamts et al. (2013). More importantly, the estimates on the interaction terms in models 2 and 4, namely CSP × High social capital and Positive CSR × High social capital, were both positive and significant (0.73, p < 0.01 in model 2 and 0.44, p < 0.10 in model 4). These results show that both positive CSR activities and overall corporate social performance have a more prominent, positive effect on firm's future financial performance among firms headquartered in counties with high community social capital.

Next, we examined whether and how a social environment with high community social capital moderates the effect of corporate engagement in negative CSR activities on a firm’s financial performance. The findings in models 5 and 6 contain these results. The estimates on Negative CSR in both models 5 and 6 were insignificant at conventional levels; these results are consistent with Erhemjamts et al. (2013) which reported that there is no systematic relation between negative CSR activities and firm performance. More importantly, the estimate on the interaction term in model 6, namely Negative CSR × High social capital, was negative and significant (−0.61, p < 0.05), indicating that negative CSR activities have a more pronounced, negative effect on future firm performance among firms headquartered in counties with high community social capital.

Taken together, our findings provided little evidence to support the “direct values” theory of Benabou and Tirole (2010) for which Hypothesis 3 and its corollaries were predicated on. In other words, there is no evidence to support the agency paradigm argument that stakeholders seeking “direct values” are instrumental in affecting overinvestment in the adoption and implementation of either positive CSR activities or negative CSR activities. In contrast, our findings lend credence to the notion that because non-shareholder stakeholders in local areas with high levels of community social capital are naturally more attentive, responsive, and receptive to corporate engagement that enhances its social performance, firms can reap greater financial benefits from their engagement in such CSR activities.

Robustness Checks

We performed a range of sensitivity analyses to show that our main findings concerning the effects of community social capital and community isomorphism on positive CSR activities and negative CSR activities were robust. This section discusses two such analyses that are particularly germane to our analysis.

Instrumental-Variable Two-Stage Regressions

To buttress the causal interpretation of our results, we followed Cheng et al. (2014) and used an instrumental-variable two-stage regression method to estimate the data. Putnam (2001, p. 48) argued that closeness to the Canadian border is “the best single predictor of the level of social capital in American states.” Accordingly, we used the natural logarithm of the closest distance between a county and the Canadian border as an instrument to predict social capital. Also, based on the evidence in social capital research (e.g., Alesina and La Ferrara 2000; Rupasingha et al. 2006), we used the percentage of families with children in the county as another instrument .

In the first-stage regression, Social capital was the dependent variable, and the independent variables include the two instrumental variables and all control variables as specified in the baseline model of Eq. (1). As expected, the coefficients on the instrumental variables are statistically significant with p < 0.01, suggesting these two variables are not weak instruments.

The second-stage regressions were modified versions of regression models reported in Table 2. Specifically, we used the predicted social capital from the first-stage regression, Fitted social capital, to replace the original Social capital variable in the regressions. Table 3, Panel A, reports these results. The estimates on Fitted social capital remained positive and significant (0.22, p < 0.01 and 0.19, p < 0.01) when we used Positive CSR as the dependent variable in models 1 and 3, respectively. When we used Negative CSR as the dependent variable in models 2 and 4, the analogous estimates were negative and significant (−0.10, p < 0.05 and −0.09, p < 0.05). These results provide additional evidence to support the inference that community social capital in US counties facilitates positive CSR activities and constrains negative CSR practices of firms headquartered in the counties.

Reverse Causality

Corporate engagement in CSR activities could enhance the vibrancy of local civic communities. For instance, this reverse causality effect could be at play in our analysis in that a higher level of positive CSR activities leads to an ensuing increase in social capital which in turn increases future levels of corporate engagement in CSR activities in the community. Adopting the method used by Hillman and Keim (2001), we evaluated the effect of this reverse causality by estimating regressions that predict subsequent change in social capital or community engagement in CSR activities using variables that capture CSR practices while holding other factors constant. Specifically, we performed regressions with the change in Social capital (or change in Positive isomorphism or change in Negative isomorphism) between 2005 and 2009 as the dependent variable and Positive CSR or Negative CSR from 2005 as the independent variable while including all control variables as specified in the baseline regression model.

Table 4, Panel B, reports the results from this analysis. Across the models, the coefficients on the CSR variables were insignificant, with p values ranging from 0.13 to 0.75 across the models. Nevertheless, the overall models were significant because control variables, particularly county-level demographical factors, have explanatory power toward subsequent change in social capital. We repeated the same analysis with change in Social capital (or change in Positive isomorphism or change in Negative isomorphism) between 1997 and 2005 and CSR variables from 1997, and we found similar results. These results indicate that the reverse causality is not supported, suggesting that the relations between social capital and CSR activities, regardless of whether they are positive or negative CSR activities, are unlikely to be driven by reverse causality.

Table 4 Robustness and sensitivity analyses
Full size table

Discussion

The study’s primary contribution is to put the aspect of “social” influences back into the C“S”R literature. Despite years of research, the extant CSR literature has yet to provide large-scale empirical evidence documenting the influences of social environments within geographically bounded local communities on either positive CSR activities or negative CSR activities. Galaskiewicz (1997) showed that local social networks among executives in Minneapolis-St. Paul create an isomorphism process that institutionalizes corporate charitable giving practices among local corporations residing in the metropolitan area. Navarro (1988) and Guthrie et al. (2008) provided additional evidence to support this effect. We go beyond these researchers to examine the effects arising from social environments in geographical communities in which the firm’s headquarters are located. In particular, we provide evidence to pin down the extent to which community social capital—a local, social institution—facilitates positive CSR activities and constrains negative CSR activities of resident corporations embedded in the community.

As well, we provide fresh evidence to confirm and extend the theory of Marquis et al. (2007) that community isomorphism facilitates corporate social actions (i.e., positive CSR). We find that, holding other constant, both community social capital and community isomorphism have a positive effect on positive CSR activities. Specifically, holding the level of community social capital constant, we find that firms undertake more positive CSR activities when the level of engagement in positive CSR practices among other local corporations is higher. These results echo and extend the findings of Navarro (1988), Galaskiewicz (1997), and Guthrie et al. (2008) on corporate philanthropy.

Additionally, we conjectured and found novel evidence that there is negligible mimicking effect arising from community isomorphism with respect to negative CSR activities; specifically, we found that negative CSR activities undertaken by a corporation is not systematically related to the level of negative CSR activities among other local corporations headquartered in the county. More importantly, we found that the documented effects of community social capital on positive CSR activities and negative CSR activities remain unchanged after controlling for the mimicking effect arising from community isomorphism.

Community social capital could increase the benefit that a firm obtains from CSR activities because non-shareholder stakeholders in communities with higher levels of social capital are naturally more attentive, receptive, and responsive to such activities. However, since stakeholders do not bear the full cost of CSR activities, they could induce overinvestment in CSR (Benabou and Tirole 2010). If stakeholders are more instrumental in effecting CSR activities in communities with high levels of social capital, by way of overinvestment in CSR, then their influences could potentially reduce the benefit a firm obtains from its CSR engagement. These arguments imply that community social capital is an external institutional logic that moderates, either positively or negatively, the relationship between corporate financial performance and corporate social performance. Our findings suggest that both positive CSR activities and a firm’s overall engagement in social responsibility (i.e., CSP) enhance a firm’s future financial performance, and these positive effects are more prominent among firms headquartered in counties with high community social capital. These results provide little evidence to support the “direct values” theory of Benabou and Tirole (2010). In contrast, they lend credence to the notion that firms with headquarters located in local areas with higher levels of community social capital reap greater financial benefits from their CSR activities. Additionally, we found that negative CSR activities dampen a firm’s future financial performance when the firm is headquartered in a county with high community social capital; but they do not affect firm performance among firms headquartered in counties with lower levels of community social capital. These results provide some support for the social capital theory.

Our research promises additional contributions to two other disparate literatures. First, the extant social capital research in management has focused on private effects of own-social-capital that arises from an actor’s network ties, where the network ties could be dyadic or not and the actor could be a firm, a subordinate unit within a firm, or an executive/director serving the firm (e.g., Nahapiet and Ghoshal 1998; Tsai and Ghoshal 1998; Lester et al. 2008). While informative, these researchers have ignored social capital in geographical communities. Our study complements these researches by showing that community social capital, social capital that arises from geographically bounded local communities, could also affect corporate behaviors such as CSR practices. Second, our research adds to a better understanding of community-based influences on organizations. Marquis and Battilana (2009, p. 297) argued that “the extensive focus on globalization and isomorphism … [has] led to a neglect of the particularities associated with local communities.” Our findings identify community civic norms and social networks as some of these particularities that affect corporate behaviors related to CSR, confirming their argument that institutions in local communities are significant antecedents of corporate behaviors.

Lastly, our research provides relevant implications for business practices and government agencies. CSR activities are important corporate decisions that affect constituents in the society, either directly or indirectly. A better understanding of the social antecedents that enable or constrain these corporate activities should promise a larger tool kit from which non-government organizations, interest groups, and policy makers can rely upon when seeking to influence the respective corporate activities so as to achieve desirable societal objectives. For instance, non-governmental organizations and interest groups pushing for CSR could prioritize their efforts and focus on those underperforming corporations that are situated in geographical regions with high social capital, as these firms could be “low-hanging fruits.” In contrast, since CSR is negatively related to corporate tax avoidance (Hoi et al. 2013) and earnings management (Kim et al. 2012), our findings imply that government agencies such as the Internal Revenue Service (IRS) and the Securities and Exchange Commission (SEC) should deploy more resources to scrutinize the tax planning and financial reporting activities of firms with headquarters that are located in geographical regions with low social capital. Additionally, both IRS and SEC could incorporate local community factors such as social capital into their monitoring and fraud-prediction models to improve the precision of the models. Further, given that social capital has a positive moderating effect on the CFP–CSP relationship, those investors who are oriented to socially responsible investment strategies should consider incorporating social capital, and perhaps other local factors, in their portfolio selection models.

Limitations and Future Research Directions

Marquis et al. (2007) argued that communities with different “frames of reference,” “city traditions,” or “cognitive templates” could have different “focus” or “form” in CSR such that companies in Minneapolis-St. Paul emphasize corporate philanthropy and firms in Atlanta focus on activities that are directed toward “local boosterism.” Although our results show a positive (negative) relation between community social capital and the level of positive CSR (negative CSR), they are silent on the extent to which community social capital affects the specific “focus and form” of CSR.

Another limitation of our study is that it uses an intra-country setting. In this study, we examine how variations in county-level social capital within a single nation, namely the US, affect CSR of resident firms with headquarters located in the local communities, namely US counties. As a consequence, it is debatable whether the tenet of our results would apply in a setting involving multiple countries. The intra-country empirical design is intentional because we intend to examine the effects of social institutions in small, geographically bounded local communities (Marquis and Battilana 2009). It helps to isolate the effects arising from differences in nation-level institutions on CSR. However, social capital could differ across nations (Fukuyama 1995), and such nation-level differences could affect CSR. We view this as a fruitful area of future research because there are significant variations in CSR/CSP across firms operating in different nations, but only a handful of studies have attempted to explain these variations.

The main insights we offer in this study are that community social capital constrains norm-deviant corporate behaviors and promotes norm-conforming behaviors. These insights complement recent development in the CSR literature that examines the influences of nation-level institutional factors on CSR (Jackson and Apostolakou 2010; Ioannou and Serafeim 2012). Additionally, they confirm the predictions of recent theoretical works such as Campbell (2007) and Marquis et al. (2007), which focus on the influences of local community factors on CSR. Given corporations’ increasing focus and resources expended on CSR activities, all these results point to a fruitful direction for future research development in the CSR literature, that it is opportune to explore how institutions, whether they are at the local community level or at the nation level, affect CSR. With respect to the local community level in particular, future research activities can be directed toward political, legal, religious, cultural and other social dimensions that neither this nor other extant studies have explored.

Evidence suggests that community social capital produces beneficial effects on societies, communities, and individuals (Fukuyama 1995; Knack and Keefer 1997; Buonanno et al. 2009; Guiso et al. 2004). However, it is debatable whether and how it influences the behaviors of local corporations. We use CSR activities as a context to analyze how community social capital affects corporate practices, and our findings reveal that there are strong links between community social capital and CSR practices. By the same token, it seems plausible that community social capital should affect other corporate practices, particularly those that are also perceived by people in the society as incongruent with the prescribed values associated with civic norms. These practices might include aggressive financial reporting practices, aggressive tax avoidance, excessive CEO compensation, etc. Future research should explore the effects of community social capital on these other corporate policies.

Lastly, there is significant evidence that CSR could potentially benefit shareholders through a range of channels. For instance, Kim et al. (2012) find that CSR reduces earnings management. Deng et al. (2013) find that CSR creates value for acquiring firms’ shareholders in mergers and acquisitions. On the other hand, our findings suggest that community social capital promotes CSR and it moderates the CFP-CSR relationship. Taken together, all these findings suggest that community social capital could also have indirect effects on corporate policies through its influences on CSR. These other corporate policies, such as earnings management and mergers and acquisitions, arguably have more pronounced effects on shareholders. As such, it would be interesting to explore how community social capital affects shareholders via its multi-faceted effects on CSR and these other corporate policies. These kinds of in-depth analysis could enrich our understanding of how community social capital affects shareholders of the firm.