Religious Values Motivating CSR: An Empirical Study from Corporate Leaders’ Perspective

Abstract

Using a panel data of 806 U.S. firms from 2006 to 2015, we find that in their ratings of corporate social responsibility (CSR) performance, firms with top managers who attended religiously affiliated schools outperform their peers with no such managers. The positive relationship between religious school attendance (RSA) and CSR performance is stronger among firms with lower level of community religiosity or less external monitoring (e.g., fewer analysts following or institutional investors). Our findings lend support to early theoretical work that suggests managerial CSR-oriented values (e.g., religious values) can be key motivating factors for CSR initiatives.

Introduction

A primary objective of business ethics research is to understand the drivers of ethical/unethical decisions a firm makes. Recently, as corporate social responsibility (CSR) has become a mainstream business activity, much academic interest has been devoted to understanding the determinants of firms’ CSR decisions. Broadly defined, CSR is a voluntary mechanism by which companies hold themselves to a set of ethical, social, and ecological standards. It concerns the relationship between business practices and the greater society. Most previous studies base their argument on the theory of the firm and treat CSR decision as a strategic choice from firms’ perspectives. For instance, CSR activities help maintain a positive corporate image and integrate with the local community, a strategic move consistent with the corporate objective of maximizing shareholder wealth.

In their theoretical work, Hemingway and Maclagan (2004) suggest that commercial imperative may not be the sole driver of CSR decision-making. According to upper echelons theory of Hambrick and Mason (1984), organizational decisions can be driven by the values and cognitive bases of their top managers. Ergo, when analyzing motives for engaging in CSR activities, one should not simply view managers as agents of corporate policy, rather, as the ultimate decision-makers. For instance, managers can directly set CSR policies, change specific projects to address their personal moral concerns, or informally inspire and promote CSR culture. Therefore, as formally modeled in Hemingway and Maclagan (2004), corporate managers’ values and interests in CSR initiatives (CSR-oriented values) can be an important motivating factor for a firm’s CSR decisions.

One example of such CSR-oriented values is religious value (Hemingway and Maclagan 2004).¹ Religion has long been regarded as a potential source of ethical norms to affect individual behavior. For example, compared to nonbelievers, religious individuals are more likely to restrain from unethical and immoral behavior. Religion is also shown to promote caring values as it stresses “giving back to the community” and “being socially responsible”. One thus would expect a positive association between managers’ religious values and firms’ CSR initiatives.

While the well-grounded theoretical work (Hambrick and Mason 1984; Hemingway and Maclagan 2004) has enlightened us with clear empirical predictions, we know of no studies that empirically examine the association between managers’ religious values and firms’ CSR performance. The challenge lies in the fact that religious values are unobservable to the public, and therefore difficult to measure for a large sample of corporate managers. In this paper, we aim to fill in the research gap by employing a proxy variable for managers’ religious value and investigate its relationship with firms’ CSR performance.

In the context of CSR activities, Wood (1991, p. 700) claims that “ethical training, cultural background, preferences, values, and life experiences all play a role in establishing the principles that motivate human behavior”. Motivated by Wood’s argument, we select a proxy variable which is based on managers’ past educational experience. We identify managers who attended religiously affiliated school (RSA managers) and perceive them to have more agreement on religious values than those without such experience. Our rationales for selecting this proxy variable are as follows. First, evidence suggest that individuals with religious adherence or those raised in religious families are more likely to attend church-affiliated colleges. In this regard, religious school attendance (RSA) is likely to be highly correlated with managers’ actual religious values. In addition, managers who attended church-affiliated schools, both religious and nonreligious, are likely to have exposure to religious education that, in turn, has a positive moderating influence on their “doing good” behaviors. Lastly, prior research has documented strong evidence that church-affiliated colleges provide better business ethics curriculum (Rutherford et al. 2012). Pascarella and Terenzini (1991) find that students from religiously affiliated schools obtain better moral development during college life. To the extent that educational experience in church-affiliated schools can reflect or shape managers’ religious values and/or positive attitude towards business ethics, we hypothesize that there is a positive link between managers’ RSA experience and firms’ CSR performance.

Our study contributes to the extensive literature on firms’ CSR decisions. Different from previous studies which treat CSR as a strategic decision from the firms’ perspective, our paper focuses on corporate leaders’ perspective and studies how their values and interests in CSR initiatives matter. From a broader perspective, our paper also adds to the growing literature that links college experience to an individual’s professional and personal life after college. Different from earlier studies which focus on college and labor market success, our work aims to explore the role of college education beyond disseminating knowledge.

The remainder of the paper is organized as follows. The next section reviews related studies and develops the hypotheses, followed by a section describing the data and sample. “Empirical Results” section examines the empirical relationship between RSA experience and CSR performance. “Discussion” section provides summary of main results, contribution, implications, limitations, and potential avenues for future research. The last section concludes.

Literature Review and Hypotheses Development

Our rationale for connecting managers’ RSA experience and firms’ CSR performance is based on the theoretical framework of Hambrick and Mason (1984) and Hemingway and Maclagan (2004). At the core of the upper echelons theory of Hambrick and Mason (1984) is the view that organizational strategies are reflections of the values and cognitive bases of their top managers. In this framework, the top management team is not an agent of corporate policy, rather the team that sets the policy. In the context of CSR, for instance, top managers can directly set CSR policies, change specific projects to address their personal moral concerns, or informally inspire and promote CSR culture. Therefore, as formally modeled in Hemingway and Maclagan (2004), corporate managers’ personal values and interests in a particular social cause can be an important motivating factor for a firm’s CSR decisions. In this school of thought, CSR is not so much a reflection of strategic business actions as in most previous literature, but a result of managers’ values and behaviors. One example of such CSR-oriented values is religious value.

Religion has long been regarded as a potential source of ethical norms to affect individual behavior. For instance, compared to nonbelievers, religious individuals are more likely to restrain from using alcohol, gambling, or committing crimes (Cochran and Akers 1989; Diaz 2000; Evans et al. 1995). In addition to suppressing unethical behaviors, religion is also shown to promote socially responsible behaviors. For example, using survey data from seven universities, Ibrahim et al. (2008) show that religion strongly influences students’ attitudes toward the economic, ethical, and philanthropic responsibilities of business. Recently, a growing body of evidence supports that religious norms in a geographic area can significantly influence managers in conducting ethical business practices Du et al. 2014; El Ghoul et al. 2012). Cui et al. (2015, 2016) find that senior managers, believers or nonbelievers, can be influenced by local religious moral values surrounding them: strong local religiosity is associated with greater CSR engagement.

While the theoretical work has enlightened us with clear empirical predictions, the challenge is to identify the right proxy for religious values. In Hambrick and Mason (1984, p. 196) and a follow-up work (Hambrick 2007), the authors suggest using observable demographic characteristics as proxies of managers’ values and cognitive frames. Such characteristics include “age, tenure in the organization, functional background, education, socioeconomic roots, and financial position”. Related, in the context of CSR, Wood (1991, p. 700) claims that “ethical training, cultural background, preferences, values, and life experiences all play a role in establishing the principles that motivate human behavior”. Therefore, even though it is difficult to measure managers’ actual religious values, we can identify an observable factor that have a close correlation with managers’ religious values. The proxy variable we select is based on managers’ past educational experience. We identify RSA managers and perceive them to have more agreement on religious values than those without such educational background. We believe that RSA can serve as a good proxy for religious values for the following three reasons.

First, evidence suggests that individuals with religious adherence or those raised in religious families are more likely to attend church-affiliated colleges. For instance, according to schools’ official websites, over 98% of the students are active members of the LDS Church at Brigham Young University (BYU), founded, supported, and guided by The Church of Jesus Christ of Latter-day Saints; and 66% of students have served in missions of the Church. In 2017, 81% of The University of Notre Dame’s incoming freshmen are identified as Catholic. Boston College and Loyola University reported this number as 70% and 59%, respectively. In this regard, RSA is likely to be highly correlated with managers’ actual religious values.

Second, managers who attended church-affiliated schools, both religious and nonreligious, are likely to have exposure to religious education that, in turn, has a positive moderating influence on their “doing good” behaviors. In the mission statements of most U.S. religious schools, there are such words as “Faith”, “God”, “Holy”, and the like. Many of these schools require taking courses on religion or attending certain kind of regular convocation or event. For example, at BYU, all students must take 14 credit hours of religious courses to graduate. At George Fox University in Oregon, even international students are required to take two Bible classes, attend a weekly chapel, and sign a student lifestyle agreement to “actively pursuing the highest call of God”. At Dordt College, a Christian school in Iowa, all classes in all fields of study are taught from a Christian perspective.² Hall et al. (2016, p. 208) assert that religiously affiliated schools provide an arena where young adults wrestle with their spiritual exploration and identity by performing chapel services and attending classes in theology or religious studies.

Hall et al. (2016, p. 206) argue that emerging adulthood is a critical phase of life because it is during these years that patterns of behaviors, values, relationships, etc. are forged that will shape the course of an individual’s entire life. We thus regard RSA as an essential life experience that has the potential to inculcating ethical behavior in future corporate leaders. To illustrate, in his book, “The Soul of a Business: Managing for Profit and the Common Good”, Tom Chappell, founder and CEO of Tom’s of Maine, discusses how his professor at Harvard Divinity School advised him to treat his business like a ministry and how this recommendation works its way to Tom’s mission statement which says, “to help create a better world by exchanging our faith, experience, and hope”.

Lastly, Rutherford et al. (2012) document strong evidence that church-affiliated colleges provide better business ethics curriculum than non-church-affiliated colleges. Pascarella and Terenzini (1991) find that students from religiously affiliated schools obtain better moral development during college life. Comegys (2010) provides evidence that completing courses in religious studies impacts college students’ attitudes towards business ethics. These findings suggest that not only RSA can be connected to religious value (albeit indirectly), but also can be linked to positive attitude towards business ethics. That is, even for a manager that was not religious before college, and was not influenced by religious education during college, by attending a church-affiliated college, she is more likely to have a better business ethics education and moral development than her non-attending peers. If we assume that religiously solidified and disciplined business ethics are persistent in guiding individuals’ behaviors in the long run (Hall et al. 2016), we would expect RSA managers to demonstrate better CSR performance.

Acknowledging these discussions above, we formally hypothesize that:

Hypothesis 1A

Firms with RSA managers have higher ratings in their CSR performance than those without.

Hypothesis 1B

Firms with more RSA managers have higher ratings in their CSR performance than those with less RSA managers.

Data and Sample

The sample data are from various sources: CSR performance ratings are from MSCI’s KLD database; schools’ religious affiliations are hand-collected and doublechecked from multiple sources including school websites, the U.S. Department of Education, and several other websites³; local (community) religiosity data is from the Association of Religion Data Archives (ARDA); firm fundamentals are from Compustat; stock prices are from CRSP; board-level information on directors’ education background and board independence are from BoardEx and ISS respectively; the mapping between county codes (FIPS) and zip codes are from the Department of Housing and Urban Development (HUD); the mapping between Federal Information Processing Standard (FIPS) and Metropolitan Statistical Area (MSA) codes are from NBER; county-level data including population, age, education, income, and race are from U.S. Census Bureau (USCB); presidential voters information from Guardian Newspaper and Townhall.com.

We start compiling our sample with top managers’ education background, i.e. main explanatory variable of interest, from BoardEx.⁴ Next, we intersect this education background dataset with firm fundamentals from Compustat and get a full list of firms. For each firm, we retrieve its managers and directors, and notably, the schools that each of them attends. “High-ranking” or “top” managers are then selected, only if the following strings are in the “role” column in BoardEx will the corresponding manager be included in the sample: CEO, CFO, COO, Chief, President, Chairman, Chairwoman, and VP. Managers with transitory roles are excluded.⁵ Using this rule, we make sure that only managers with enough authority in corporate-level decision-making are selected into the final sample.

After obtaining the full list of “top” managers and their respective alma mater, we hand-collect and doublecheck data with respect to which schools are religiously affiliated and their specific affiliations, using various sources specified above.⁶ Once we have gathered all sample religious schools’ affiliations, we aggregate the dataset from manager level to firm level and intersect it with the KLD database for CSR performance ratings, i.e. the dependent variable. This gives us an initial sample of 1484 firms. We then merge in the other seven databases for all other control variables, for example, stock prices from CRSP, and board independence data from ISS. We account for a 1-year lag for some of the control variables to account for serial correlation and require all variables in the baseline regression to be non-missing, obtaining a final sample of 3993 firm-year observations from 806 unique firms, across 10 years from 2006 to 2015.⁷

CSR Performance

The KLD database contains ratings of CSR performance in seven main categories: corporate governance, community, diversity, employee relations, environment, human rights, and products. Under each category, there are some rating factors representing strengths, and others representing concerns. Each factor is given a score of (zero) one for a firm in a year, if in this given year this given firm (does not meet) meets the assessment criteria set for this given factor. For example, in 2007, Costco Wholesale received a one in the factor of positive union relations, which is a strength factor, under the category of employee relations, and a zero in the factor of health and safety concern under the same category. Our main CSR measure, CSR SCORE, is the sum of scores for all strength (positive) indicators minus the sum of concern (negative) indicators from all seven dimensions of CSR performance rating, also known as a raw CSR score and in previous studies (e.g. Cai et al. 2011) called “Net CSR”. Consider Costco in 2007. It received a total score of 83 from all factors of strengths and a total score of 74 from all factors of concerns, across all seven CSR performance rating categories. Therefore, for this specific firm-year observation, Costco’s raw CSR score, CSR, is 9.

Hong et al. (2012) have shown that the corporate governance aspect in KLD data is heterogeneous with other aspects. Also, Krüger (2015) mentions that there are doubts about whether KLD measures corporate governance in the traditional understanding. Therefore, besides the CSR variable set CSR (All 7 Dimensions), we construct an alternative set of CSR measures, CSR (Excl. Corp. Gov.), in which the corporate governance dimension is excluded.

For additional robustness tests and CSR component analysis, we construct weighted CSR scores, raw CSR scores for strengths, and raw CSR scores for concerns. Weighted CSR scores (CSRw) are the raw CSR scores divided by the total maximum possible number of all KLD strengths and concerns items in a given year for a given firm. Raw CSR scores for strengths (CSRstr) and for concerns (CSRcon) are the sum of scores across all factors of strengths and the sum across all factors of concerns, respectively.

Panel A of Table 1 displays summary statistics for all CSR variables described above and used in this study. The mean (median) value of CSR, including all 7 dimensions is − 3.45 (− 1), and the mean (median) value of the corresponding weighted CSR score is − 0.99 (− 0.27). After excluding the corporate governance dimension, the mean (median) value of CSR increases to 0.5 (2), and the mean (median) value of the corresponding weighted CSR score increases to − 0.39 (0.37), indicating that the concerns are dominant in the ratings of corporate governance.

Table 1 Summary statistics

Religious School Attendance (RSA) and Community Religiosity

Individual managers’ RSA is indicated by the religious affiliation of the universities or colleges from which they graduated. Our main variable of interests, LEADERSHIP RSA, is an indicator variable that equals to 1, if there is at least one high-ranking manager who graduated from a church-affiliated school, and equal to 0 if there is no such manager. To ensure that our results are not sensitive to different measures, we construct the following variables of RSA and use them in alternative regression specifications: LOG LEADERSHIP RSA PERCENTAGE, which is the natural logarithm of one plus the number of top RSA managers divided by total number of top managers; LOG LEADERSHIP RSA PERCENTAGE SQUARED,⁸ which is the natural logarithm of one plus the squared term of the number of top RSA managers divided by total number of top managers and is used along with LOG LEADERSHIP RSA PERCENTAGE to address possible non-linear relation between LOG LEADERSHIP RSA PERCENTAGE and the dependent variable; LOG NUMBER OF RSA MANAGERS, which is the natural logarithm of one plus the number of high-ranking RSA managers. Normalization, i.e. adding one and taking the logarithm, is to address the high skewness in the raw measure of leadership RSA.

Following the literature (e.g. Hilary and Hui 2009; Dyreng et al. 2012; El Ghoul et al. 2012) to measure the strength of each county’s religious social norms, a key control variable is the level of religiosity where a firm is headquartered (COMMUNITY RELIGIOSITY), which is calculated as the number of religious adherents in the county divided by the population of the county. To identify the county where a specific firm is located, we employ the same methodology as previous studies, which use corporate headquarters’ location as its location (e.g. Loughran and Schultz, 2005; Pirinsky and Wang 2006). Specifically, Pirinsky and Wang (2006, p. 1994) point out that this approach seems “reasonable given that corporate headquarters are close to corporate core business activities”.

Panel B of Table 1 shows that the managerial RSA indicator has a mean value of 0.68, meaning that on average, about two-thirds of firms have at least one RSA manager in the leadership suite. From the community perspective, the overall religiosity on average is 0.51, indicating that half of the population in the sample counties attach to some religion. To gain a sense of the degree of religiosity of different regions in the U.S., Panel C of Table 1 presents the top and bottom 15 states ranked by average county religiosity aggregated at the state level. Although rankings for some individual states shift due to different dependent variables, compared with Hilary and Hui (2009), the ranges of average religiosity are quite consistent. For example, in the list of the top (bottom) 15 states, the range of average religiosity is between 73.4% for Utah (42.9% for Colorado) and 57.0% for Connecticut (31.2% for Maine), while in Hilary and Hui (2009), the range is from 74.57% for Utah (42.27% for Maryland) to 57.37% for Pennsylvania (31.16% for Oregon). Diving deeper into county-level religiosity, it is observed that there is a vast geographic difference between the Catholic and Protestant religions in the U.S. Most counties with Protestant adherents are concentrated in Utah, followed by Alabama and Arkansas, while most Catholic adherents live in the counties of Northeast region, including New York, Massachusetts, and Pennsylvania.⁹

Other Control Variables

The construction of other control variables follows Cui et al. (2016). Specifically, firm size (SIZE) is measured as the logarithm of total assets; growth opportunities (MARKET-TO-BOOK) is calculated as market value of equity divided by book value of equity; CAPITAL EXPENDITURE is capital expenditure expense over total sales; SALES GROWTH is the sales growth rate from previous year to current year; σ_RETURNS is the standard deviation of monthly stock returns of the previous year; LONG-TERM DEBT, ADVERTISING EXPENSES, and R&D EXPENSES, are long-term debt, advertising expenses, and research and development expenses, respectively, divided by lagged total assets. Also included are firm internationalization (e.g. Attig et al. 2016) as measured by FOREIGN EXCHANGE EARNINGS from Compustat, and BOARD INDEPENDENCE (e.g. Jo and Harjoto 2011, 2012) calculated as the number of independent outside directors divided by the number of total directors. For brevity, definitions of county-level variables such population, income, etc. are provided in the Appendix together with those of above-mentioned variables. The summary statistics for these variables are displayed in Panel D of Table 1.

Panel E of Table 1 shows the Fama–French 12-industry composition of the sample, by number of firms and number of observations. We see that the biggest group, about one-fourth of the sample (25.93% by number of firms and 25.69% by number of observations), is the business equipment industry, followed by the manufacturing (14.89% and 17.05%) and the finance industries (15.51% and 14.73%). There is no observation from utility companies after merging datasets from all sources. In total, the sample spans across nearly all industries. In terms of the scope of coverage, it is a fair representation of the actual market.

Table 2 presents the correlations among the raw CSR score, leadership RSA, community religiosity, and all other major control variables. First, between CSR score and leadership RSA, the Spearman correlation is 0.05, showing a positive relation, which is consistent with our hypotheses 1A and 1B. Second, the Spearman correlation between leadership RSA and community religiosity is 0.04, which suggests that they capture quite different aspects of religiosity internal and external to a firm. Third, except for the pairs between firm size and some other variables (e.g. stock return volatility, R&D expenses, and long-term debt), the vast majority of all correlations are generally and reasonably low. Last, most correlation coefficients are statistically significant at the 0.1 percent level.

Table 2 Correlation matrix

Empirical Results

Baseline Regressions

Table 3 presents the baseline regressions of CSR on leadership RSA, community religiosity, and other control variables. The standard errors in all specifications are double clustered by firm and year. The R-squared is calculated so that it is mostly be driven by the regressor of interest instead of the multiple fixed effects.¹⁰,¹¹ Specification (1) does not include industry, year, and state fixed effects; represents a univariate comparison, and shows a positive and significant relation between managerial RSA and CSR. Using various measures of leadership RSA and whether or not to control for community religiosity produces seven different specifications from (2) through (8). However, all six regressions in which measure of leadership RSA is included show the same statistically significant and positive association between CSR and RSA of corporate leadership at the 95% confidence level. Above all, the results are economically significant. For instance, in specification (1), the rating on how a firm behaves socially is 1.5 (1.494 to be exact) points higher for a firm having at least one RSA manager than a firm without any RSA manager. Given that the sample average (median) is − 3.45 (− 1), the effect is substantial. Worth mentioning is that in specifications (5) and (6) the coefficient of the squared term of leadership RSA percentage is statistically significant and negative at the 1% significance level, which indicates that the RSA of higher management is positively affecting firms’ social performance at a decreasing rate. That is, the promoting effect of RSA on CSR is the strongest when a management team recruits its first RSA manager; but as more and more RSA managers join, the marginal positive effect of RSA on CSR decreases. In specifications (7) and (8), we use number of RSA managers as an alternative measure of leadership RSA and see a positive and significant link between managerial RSA and CSR. To further check the robustness and account for unobserved heterogeneity of the baseline results, we employ random effects and show consistent results in specification (9). Pivotally, in specification (3) where community religiosity is included while leadership RSA is not, community religiosity has a significantly positive effect on CSR. However, the impact of community religiosity, which is shown to strongly influence the economic outcomes in previous studies and is also displayed in specification (3), is wholly muted in all specifications where it is included together with managerial RSA. We contend that while community or local religiosity is important by itself, but when leadership RSA is simultaneously considered, community religiosity becomes less critical, or its effect is substituted by the effect of leadership RSA. Managers’ intrinsic religious values proxied by RSA do significantly influence how they want their companies to behave toward the public good. Overall, the findings are consistent with hypotheses 1A and 1B.

Table 3 Leadership RSA and CSR: baseline results

Interaction Effects

While our paper focuses on the role of corporate insiders in setting the CSR choice, recent studies identify several external factors that determine the extent of CSR engagement in companies. Two important external factors are corporate governance (Harjoto and Jo 2011) and societal preferences (Cui et al. 2015, 2016). Harjoto and Jo (2011) contend that governance and monitoring mechanisms can serve as effective resolution for potential conflicts among various stakeholders. Thus, stronger governance and monitoring lead to more responsible behaviors of companies. Jo and Harjoto (2011) confirm that CSR is positively associated with the internal and external monitoring mechanisms, including analyst coverage (Frankel and Li 2004; Chen et al. 2016) and institutional ownership (Liu 2014). Another material factor that drives CSR, as discussed earlier, is the religious norms in a geographic area (Cui et al. 2015, 2016). They demonstrate that an external religious milieu, specifically the religiosity of community where a company is located, has predicting power over corporate decisions towards its shareholders and other stakeholders.

The prior work suggests that the CSR decisions we observe in practice not only reflect managers’ CSR-oriented values, but also the external pressures the firm faces. It is reasonable to assume that when the effect of external monitoring or social pressure is at work, the role of managers’ values and preferences becomes less weighty. We thus expect that the positive link between RSA and CSR performance we hypothesized earlier to be more pronounced among firms with less external monitoring or social pressure. In other words, we expect there is substitution effect between RSA and other driving forces of CSR, i.e. the effect of leadership RSA on CSR performance is stronger among firms with lower level of community religiosity or less external monitoring (e.g. fewer analyst following or institutional investors). Alternatively, one could also argue that there is complementary effect between managerial RSA and other driving factors of CSR which could reinforce the role of RSA in affecting the CSR performance. Therefore, whether complementary or substitution effect dominants is an empirical question.

Table 4 shows the empirical test results. The positive and significant effect of leadership RSA by itself confirms our baseline findings in Table 3. First and foremost, results show that the coefficients of all interaction terms are significant and negative at the 95% confidence level, meaning that leadership RSA and all the proxies of external monitoring or external pressure of doing good are substitutional in affecting CSR performance. In the first specification, the negative and significant interaction term between leadership RSA and community religiosity indicates that as the religiosity in the community decreases, i.e. as the external pressure of being socially responsible lessens, managerial RSA exerts a stronger impact on CSR. The second column looks at the interaction effect between managerial RSA and analyst coverage, which is significant and negative, indicating that when the external monitoring from analysts is weak, managerial RSA plays a bigger role in guiding firms to do good. Moreover, institutional investors act as another external monitoring mechanism for a firm. The third column shows that, when these investors account for a lower percentage of a company’s total investors, the firm needs stronger leadership RSA effect to act responsibly. The negative interaction effects are also economically significant. An interquartile range decrease in community religiosity, number of analysts following, and percentage of institutional investors for a firm with (without) RSA managers aboard would increase (decrease) its CSR performance score by 1 (0.5), 0.2 (1), and 0.7 (0.6) point, respectively. Given that the sample mean (median) of CSR performance score is − 3.45 (− 1), the effect is substantial. In sum, all three interaction tests, taken as a whole provide strong evidence that when firms face less external pressure of doing good or have weaker monitoring mechanisms, leadership RSA becomes more important in promoting CSR performance.

Table 4 Leadership RSA and CSR: interaction tests

To more intuitively show the interaction effects discussed above, we plot them in Fig. 1. Figure 1a–c illustrate the interaction effects between leadership RSA and different measures of external pressure of doing good or monitoring mechanism, i.e. community religiosity, analyst coverage, and institutional investors (the moderator variables), respectively. It shows that, across all three figures, as the moderator variable decreases from two standard deviations above the mean of itself (dashed line) to two standard deviations below the mean of itself (solid line), the slope of the line, i.e. the relation between leadership RSA and CSR, changes from slightly negative to positive, indicating a stronger supporting effect of leadership RSA on firms’ CSR performance, which is consistent with the results of regression analysis.

Fig. 1

Leadership RSA and CSR: Interaction Tests. These figures present the interaction effects between Leadership RSA and measures of external pressure of doing good or monitoring mechanism (the moderator variables), i.e. community religiosity, number of analysts following, and percentage of institutional investors, in Figs. a, b, and c, respectively. The Y axis denotes the predicted value of CSR rating. The X axis denotes Leadership RSA, ticked at 0 and 1. Solid (dashed) line is for moderator variable valued at two standard deviations below (above) the mean. Confidence intervals are at 1 percent significance level

Endogeneity and Propensity Score Matching (PSM) Methodology

One potential concern in our study is endogeneity. An RSA manager is likely to join a firm with better social performance; and, likewise, a company putting a lot of efforts in its social responsibilities might be more interested in hiring an RSA manager. Moreover, there might be omitted firm characteristics influencing both the hiring of RSA managers and corporate social responsibility. Therefore, to establish causality and strengthen our findings between managerial RSA and firms’ CSR performance, we perform PSM methodology. Specifically, in each sample year and for each firm with no top RSA managers, we match it with another firm from the same 3-digit SIC industry, in the same year, and with at least one top RSA manager based on the propensity score generated by community religiosity and all other control variables.¹² Next, we statistically compare the means of each control variable and also the outcome variable, CSR, to see whether CSR is significantly different, while all control variables are not statistically different between the two groups.

Table 5 presents the results of the PSM test. It shows that the CSR scores are significantly different between firms with RSA top managers and those without, and the mean of CSR performance rating for the firms with RSA top managers is higher by 2.19 with a p-value of 0.00. For control variables, 14 out of all 18 variables are not statistically different between the two groups. Although four of them are significantly different, the mean differences are all lower or equal to 0.04 in absolute value. The only exception is firm size, principally due to the fact that the matching is restricted to the same 3-digit SIC industry and year, and thus greatly reduces the pool of firms of similar size to select from and match to. Crucially, the mean difference of community religiosity is statistically insignificant, indicating that while the external religious environment for the two groups of firms is very similar, managerial RSA remains influential in affecting firms’ CSR performance. We perform a subsample test to address the concern that our results are not driven by firm size while observing the comparatively significant difference in absolute value for firm size between the control group and the treatment group. We split the sample into large firms and small firms using median firm size. Results show that for both groups of firms, the positive and significant relation between managerial RSA and CSR hold, while the effect is stronger in smaller firms.¹³

Table 5 Leadership RSA and CSR: propensity score matching

CSR Decomposition

There are many aspects in defining whether a firm acts socially responsible according to KLD’s CSR performance rating system. All the CSR performance rating factors can be categorized into strengths and concerns. Diving deeper, there are many elements in both strengths and concerns under each of the seven main categories. For instance, in terms of environment, using clean energy is a plus, while making a controversial investment is a minus from the corporate governance perspective. In regard to human rights, a lack of freedom of expression is considered a concern, while having a cash profit sharing policy in place is deemed a strength from an employee relations perspective.

Panel A of Table 6 shows the regressions of raw and weighted CSR with different components and respective strengths and concerns separately on leadership RSA and all other control variables. The positive effect of leadership RSA on CSR measures in both raw and weighted forms is statistically significant. Essentially, the positive impact is from strengths instead of the concerns. This is consonant with previous studies showing that for the most part religion positively affects personal and corporate behaviors (e.g. McGuire et al. 2012; Boone et al. 2012; Du et al. 2014). Next, besides the overall rating, each of the seven main rating categories in CSR is broken down into strengths and concerns and are regressed on managerial RSA. Results are provided in Panel B of Table 6. It is clearly seen that several factors contribute to the positive link between leadership RSA and firms’ CSR performance. Specifically, managerial RSA promotes overall diversity and strengths of environment and product. The only counterintuitive finding is that leadership RSA increases community concerns. Taking a thorough look into the detailed rating factors in this category, we find that there are only four rating criteria, i.e. investment controversies, negative economic impact, tax disputes, and others, which is the lowest compared to strengths and concerns under other main CSR performance rating categories, most of which have at least eight rating criteria. The low power of test is one possible explanation. When talking about community in the CSR context, we usually refer to charitable giving, housing support, volunteer programs, etc. instead of the above-mentioned economic viewpoints, thus making lack of relevance another explanation.

Table 6 Leadership RSA and CSR: CSR decomposition

To conclude, leadership RSA promotes firms’ overall CSR performance, particularly the positive aspects and the engagement with diversity, environment, and product.

Discussion

Summary of Main Results and Contribution

The empirical results of our study, based on a panel data of 806 firms from 2006 to 2015, suggest that there is a positive link between managers’ RSA experience and firms’ CSR performance. Specifically, in their ratings of CSR performance, firms with top managers who attended religiously affiliated school outperform their peers with no such managers. Furthermore, firms with more RSA managers receive higher CSR performance ratings than firms with fewer RSA managers, although the relationship between RSA and CSR is increasing at a decreasing rate. Our results are more pronounced among firms with lower level of community religiosity or less external monitoring (e.g., fewer analysts following or institutional investors). We interpret our results as empirical evidence that supports the theory of Hemingway and Maclagan (2004).

Our paper contributes to the extensive literature on firms’ CSR decisions. There is no argument that corporations bear responsibilities to society; but there is considerable debate as to how CSR is associated with the concept of shareholder primacy. Although some scholars regard CSR as an investment that comes at the expense of shareholders (Friedman 1970), extant literature largely acknowledges the value-enhancing role of CSR (Malik 2015). While most previous studies treat CSR as a strategic decision from the firms’ perspective, more recent work identifies external factors that determine the extent of CSR in companies. Such factors include corporate governance (Harjoto and Jo 2011), country legal regime (Liang and Renneboog 2017), and societal preferences (Cui et al. 2015, 2016), among others. In their theoretical work, Hemingway and Maclagan (2004) suggest that corporate managers’ values and interests in CSR initiatives can be an important motivating factor for a firm’s CSR decisions. Our paper adds to the literature by documenting that corporate managers who attended religiously affiliated school outperform their peers without such educational experience in their CSR initiatives. Our empirical findings lend support to early theoretical work that suggests managerial CSR-oriented values (i.e., religious values) can be key motivating factors for CSR initiatives.

From a broader perspective, our paper also adds to the growing literature that links college experience to an individual’s professional and personal life after college. Education, particularly post-secondary education, plays a vital role in disseminating knowledge and developing human capital. Earlier studies, therefore, use education as a proxy for skill (e.g., Bartel and Sicherman 1998; Chevalier and Ellison 1999) and examine the relationship between college degree attainment and labor market success (Bowen 1977; Feldman and Newcomb 1969; Pascarella and Terenzini 1991, 2005). Education, however, is not only associated with disseminating knowledge, but also with internalizing value. Different from earlier studies, our paper uses education as a proxy for value and documents a positive link between managers’ RSA experience and firms’ CSR performance. Our work thus contributes to this strand of literature by exploring the role of college education beyond disseminating knowledge.

RSA vs. Community Religiosity

In closely related work, Cui et al. (2015, 2016) find that senior managers, believers or nonbelievers, can be influenced by local religious and moral values surrounding them. Their findings raise a natural question: how does RSA measure compare with community religiosity? In the field of educational psychology, students’ learning behaviors are triggered by, in general, two types of motivations, intrinsic and extrinsic (e.g. de Charms 1968; Deci 1971, 1975). Intrinsic motivation generates enjoyment and satisfaction arising from voluntarily doing something (e.g. one reads a book because she enjoys reading) while extrinsic motivation refers to performing some task because it leads to a separate outcome (e.g. one reads a book because it is a homework assignment). Literature (e.g. Vallerand and Blssonnette 1992; Ryan and Deci 2000; Lepper et al. 2005; Corpus et al. 2009; Lemos and Veríssimo 2014) shows that it is intrinsic motivation that better promotes students’ academic achievement. However, intrinsic and extrinsic motivations are not opposites, as they may cooperate with each other.

The above discussion serves as a meaningful analogy to our view of the roles that RSA (intrinsic motivation, i.e. feels right or satisfied) and community religiosity (extrinsic motivation, e.g. to maintain corporate image) play in influencing CSR performance. As shown in the results of Table 3, where community religiosity is included, together with leadership RSA, its effect is muted in all tests. We do not argue that community religiosity is not salient in guiding or at least exerting pressure on firms to behave socially responsible. For sure, literature provides strong evidence that companies located in more religious areas tend to incur less unethical behaviors (e.g. McGuire et al. 2012; Boone et al. 2012). However, we contend that, corporate leaders are those who set policies, make decisions, and execute actions of how companies they lead should behave. When it comes to determine the pattern of expression of individuals’ decisions and actions, intrinsic values and motives can be more important than external pressure. Even if managers’ decisions are made under certain external pressure, for example, a company is located in an area where CSR is particularly valued by the residents, it is still at the managers’ discretion whether they solely satisfy shareholders’ benefits or voluntarily expand their responsibilities to the economy, environment and society, and if the latter, in what aspect(s) and by what degree.

Interestingly, RSA and community religiosity are not highly correlated (see Table 2), suggesting limited overlap between the two measures. However, the interaction term of RSA and community religiosity is negative and significant (see Table 4). We interpret this finding as consistent with the idea that when external pressure of being socially responsible lessens, managerial RSA exerts a stronger impact on CSR performance.

Limitations and Future Research

A primary limitation of our study is that we cannot observe managers’ actual religious values. Our measure of RSA serves merely as a proxy for religious value. Therefore, we cannot test the impact of attending religious colleges on shaping or reflecting religious values, nor can we demonstrate the effect of religious values on CSR performance. We acknowledge that while theoretical argument of Hemingway and Maclagan (2004) supports our hypothesis about the RSA-CSR link, we cannot directly test or demonstrate the channel in the model. For instance, our measure of managers’ RSA may capture two associated but different effects: managers’ religious belief before college and the religious education they receive in college. We are not unable to disentangle the two effects under the current empirical setting.

We also realize undoubtedly that RSA may not be a perfect proxy variable. There might be students who are nonreligious but went to church-affiliated schools. There could also be cases where students attended religious college but were not influenced by religious education at all, in terms of becoming a better citizen or an ethical manager in their future career. After all, some religious-affiliated schools have evolved greatly over time and the religious culture has faded. Not all church-affiliated schools require the student to attend religious classes or chapel services. However, we believe our proxy is valid and appropriate for at least the following two reasons. First, for a variable to be a good proxy, it needs to have a close correlation, not perfect correlation with the variable of interest. So long as the “measurement error” in RSA is not systematically associated with the outcome variable CSR performance, the noise in the proxy works against finding a statistically significant impact of RSA. Second, ultimately, we use RSA to proxy for CSR-oriented values (Hemingway and Maclagan 2004). Prior work has documented strong evidence that religious colleges provide better business ethics education. We believe this evidence further help justify RSA as a valid proxy for CSR-related values.

The above two limitations can be solved by using survey method to obtain managers’ true religious values and other relevant information. A common drawback of survey study is limited sample size and representativeness which greatly reduce the generality of findings. For example, Mazereeuw-van der Duijn Schouten et al. (2014) use 473 Dutch executives and Jamali and Sdiani (2013) survey only 149 Lebanese managers, while our final sample incorporates nearly 6000 top managers. What is more, the survey is usually done once, thus lacking more profound inferences that panel data could provide. Moving forward, it would be fruitful to conduct a similar study with large sample survey data spanning several years.

The current paper only considers attending (or not) the religious-affiliated school. The degree of actual religious education exposure that varies across different colleges and its effect on firms’ CSR behavior can be a natural further extension of the study. The current study investigates the effect of attending religiously affiliated schools on firm CSR performance. Further studies are worth conducting in respect to the effect of RSA on other firm decisions such as financial misreporting, earnings management, and other agency-related misbehaviors. Our prior is that RSA managers, due to higher moral standard, are much less likely to engage in these misbehaviors. Additionally, in an unreported test, we find that the effect of leadership RSA is stronger for small firms that for large firms, split by median firm size, which might indicate that managers of smaller firms are less affected by stakeholders, and are abler to instill their personal values into corporate culture. This is also a potential interesting point for further research.

Brammer et al. (2007) examine the association between religious denominations and individual attitudes towards CSR activities by conducting a large sample study, finding that religion may exert significant influence on shaping individuals’ perceptions of CSR. They show that there are variations across different religions in attitudes towards various aspects of CSR activities. For example, most Christians, Muslims, and Hindus think that the economic responsibility of companies is very substantial, while Jews and Buddhists see a firm’s material well-being significantly less meaningful. Although the impact of non-Protestant and non-Catholic religions in our sample is negligible due to data limitations, it is worth expanding our study in an international context to bring more insights on the relationship between managers’ religious value and firm’s CSR performance.

Conclusion

Religion has in the past been seen as a private matter with little or no place in corporate America. In this paper, based on theoretical work of Hemingway and Maclagan (2004), we assess the empirical association between our proxy of managers’ religious values and firms’ CSR performance. Using a panel data of 806 U.S. firms from 2006 to 2015, we find that in their CSR performance ratings, firms with top managers who attended religiously affiliated school (RSA managers) outperform their peers with no such managers. The positive relationship between RSA and CSR is stronger among firms with lower level of community religiosity or less external monitoring (e.g., fewer analyst following or institutional investors). Our findings lend support to early theoretical work that suggests managerial CSR-oriented values (e.g., religious values) can be key motivating factors for CSR initiatives.

Our paper contributes to the extensive literature on firms’ CSR decisions. Different from previous studies which treat CSR as a strategic decision from the firms’ perspective, our paper focuses on corporate leaders’ perspective and studies how their values and interests in CSR initiatives matter. From a broader perspective, our paper also adds to the growing literature that links college experience to an individual’s professional and personal life after college. Different from earlier studies which focus on college and labor market success, our work aims to explore the role of college education beyond disseminating knowledge. By doing so, we wish to highlight the importance of higher education in instilling students with good values.

Appendices

Appendix

Variables Definitions

CSR Score Raw CSR score, sum of scores of all CSR strengths factors minus the sum of scores of all CSR concerns factors for a firm-year.

CSRw Weighted CSR score, which is raw CSR score divided by the total maximum possible number of all KLD strengths factors and concerns factors for a firm-year.

CSRstr Sum of raw scores of all CSR strengths factors for a firm-year.

CSRcon Sum of raw scores of all CSR concerns factors for a firm-year.

Leadership RSA An indicator variable which equals to one if there is at least one high-ranking manager who graduated from a religiously affiliated school, for a firm-year, and zero otherwise.

Log Leadership RSA Percentage Natural logarithm of one plus number of high-ranking RSA managers divided by total number of high-ranking managers.

Log Leadership RSA Percentage Squared Natural logarithm of one plus the squared term of number of high-ranking RSA managers divided by total number of high-ranking managers.

Log Number of RSA Managers Natural logarithm of one plus the number of high-ranking RSA managers.

Community Religiosity Number of religious adherents in the county divided by the population of the county.

Size: Logarithm of total assets.

Market-To-Book Market value of equity divided by book value of equity.

Capital Expenditure Capital expenditure expense over total sales.

Advertising Expenses Advertising expenses divided by lagged total assets.

R&D Expenses Research and development expenses divided by lagged total assets.

Long-Term Debt Long-term debt divided by lagged total assets.

Sales Growth Sales growth rate from previous year to current year.

Stock Return Volatility Standard deviation of monthly stock returns of the previous year.

Foreign Exchange Earnings Foreign Exchange Earnings.

Board Independence Number of independent directors divided by total number of directors.

Age Average age of residents in each county, from U.S. Census Bureau’s estimates.

Education Percentage of the adult population in each county with a college education, from U.S. Census Bureau’s estimates.

Income Median household income in each county (in ten thousand), from U.S. Census Bureau’s estimates.

Rural An indicator variable that equals to 1 if the firm is headquartered outside the MSAs, and 0 otherwise.

Minority Percentage of racial minorities in each county, from U.S. Census Bureau’s estimates.

Population Natural logarithm of the population for each county in missions, from U.S. Census Bureau’s estimates.

Politics Percentage of the adult population in each county who is affiliated with the Republican party, from Guardian and Townhall.com.

Analyst Coverage Natural logarithm of number of analysts following.

Institutional Investors Percentage of total institutional ownership.