From driver to enabler: the moderating effect of corporate social responsibility on firm performance

Abstract

Corporate social responsibility (CSR) is assumed to have a direct influence on firm performance. However, the existing literature provides a mixed depiction of the relationship between CSR and firm performance. In this study, CSR is considered as an enabler for firm performance, rather than a direct driving force. Using a sample of U.S. firms, we test the enabler hypothesis and find that CSR positively moderates the relationship between marketing investments and firm financial performance, i.e., the enabling hypothesis is supported. The moderating effect of CSR is further moderated by how firms treat their employees. Mistreating employees weakens CSR’s moderating effect because it may make customers to perceive CSR activities as self-interested makeups rather than purely charitable actions in nature. Overall, our study suggests a logic shift from considering CSR as a driver for firm performance to an enabler and provides implications for both future research and practices.

Keywords:

• corporate social responsibility (CSR)
• marketing investments
• employee mistreatment
• financial performance

JEL CLASSIFICATIONS:

• M21; M14; G34

1. Introduction

Corporate social responsibility (CSR) has become a strategic necessity for firms (Chen et al., 2020). Scholars have long been interested in examining whether and how firms can benefit from their investments in CSR (Cosans, 2009; Friedman, 1970). However, there is no theoretical and empirical consensus about the role of CSR (Cochran & Wood, 1984; Hou, 2019; Lu et al., 2020; Story & Neves, 2015; Waddock & Graves, 1997). Empirical evidence on CSR’s effects is mixed. On the one hand, some studies reported that CSR could benefit firms and their stakeholders (Godfrey, 2005; Kumar et al., 2018; Margolis & Walsh, 2001; Zhou & Ki, 2018); on the other hand, there are studies showing that CSR might be costly and only a few firms could benefit from it (Margolis & Walsh, 2001; Öberseder et al., 2011; Skarmeas & Leonidou, 2013). There is not much consensus regarding the relationship between CSR and firm performance, and some previous studies also reported insignificant, U-shape or inverted U-shape, which have been reviewed in the literature (Beurden & Goessling, 2008; Orlitzky et al., 2003; Saha et al., 2020). Also, the literature is inconclusive regarding the causality between CSR and firm performance; it is likely that CSR influences firm performance, but it is possible that also firm performance influences CSR (Galant & Cadez, 2017; McWilliams & Siegel, 2000).

Scholars have tried to reconcile these inconsistent findings of CSR’s effects by testing the nonlinear relationship between CSR and firm performance (Gao et al., 2019), introducing moderating or mediating factors (Hasan et al., 2018; Peloza & Papania, 2008; Story & Neves, 2015), and changing data analysis methods (Flammer, 2015; Godfrey, 2005). These efforts advanced our understanding of the effects of CSR on firm performance. Kemper et al. (2013) document that most of these studies have implicitly treated CSR investments as the direct inputs of firm performance (generally measured by financial indicators). In other words, the majority of scholars have taken CSR-related investments or actions as a driver that contributes directly to firm performance (Cochran & Wood, 1984; Lenssen et al., 2005; McWilliams & Siegel, 2000).

However, CSR investments can serve as an enabler that supports firms’ performance drivers. For example, CSR investments can enable value-added strategic and operational activities, such as marketing, which benefit firm performance (Kemper et al., 2013; Lee, 2019; Vlachos et al., 2009). The enabler logic means that CSR plays an indirect role in improving firm performance. The effects of CSR on firm performance are channeled by firm’s performance drivers that have direct effects on firm operational efficiency and final performance (Barnett & Salomon, 2006; Wang et al., 2008). Thus, there may be no direct relationship between CSR and firm financial performance (Surroca et al., 2010). In other words, CSR could moderate the effects of performance drivers. The remaining question is: what strategic or managerial activities can better benefit from the CSR investments?

The enabler logic models CSR as a moderating factor of the relationship between performance drivers and performance outcome variables. An emerging body of studies tend to associate CSR investments with firms’ specific actions, such as research and development (R&D), finance access, differentiation strategy and strategic marketing (Cheng et al., 2014; Gao & Hafsi, 2019; Luo & Bhattacharya, 2009). However, very few studies highlight CSR’s role as an enabler (for example, Kemper et al., 2013). In this article, we model CSR as a moderating factor for marketing, through which CSR contributes to firm performance. We model CSR as the enabler of marketing for several reasons. First, marketing directly targets customers who are the key drivers of firms’ social initiatives (Wang et al., 2016). Second, among firm stakeholders, customers who contribute most to a firm’s performance (Freeman, 1984; Maignan & Ferrell, 2004) appear to be particularly susceptible to a firm’s CSR initiatives (Bhattacharya & Sen, 2004). Third, social marketing theory suggests that marketing can be more effective if CSR is integrated into marketing strategies (Andreasen, 1994, 2006; Kotler & Zaltman, 1971; Lee & Kotler, 2011).

Therefore, we argue that CSR activities can facilitate marketing investments to improve firm performance. In particular, we argue that CSR could increase a firm’s good reputation because it provides a positive cue (a halo-effect) for customers to judge the validity and reliability of the firm’s marketing activities, such as advertising, promotion, and customer relation maintenance. However, whether CSR brings a halo-effect for the firm depends on whether customers perceive it as a true benefit for social welfare (Godfrey, 2005). For instance, if the firm treats its employees poorly, customers will doubt the motivation of firm’s CSR activities, which will then reduce the halo-effect of CSR. Thus, we argue that the moderating effect of CSR will be further moderated by customers’ perception of CSR legitimacy. In particular, we argue that mistreating employees will reduce the moderating effect of CSR. We use a sample of U.S. firms and find robust evidence to support our arguments.

The contribution of this study is threefold. First, it provides a novel explanation of the role of CSR as a moderating factor from the enabler’s perspective and showcases how CSR improves firm performance by facilitating marketing. Second, our study enriches the discussion of the association between CSR and marketing investments. Third, our study provides avenues for future studies to examine the relationships between CSR as an enabler and firm performance drivers.

The remainder is organized as follows. The next section reviews the literature and develops hypotheses. We then describe our methodology and present results right after. In the last section, we discuss and conclude.

2. Theory and hypotheses

A large number of studies have investigated the impact of marketing, which is typically proxied as the expenditures in advertising and promotional activities on firm performance (Andras & Srinivasan, 2003; Kim & Joo, 2013). These studies have primarily confirmed a positive relationship between marketing activities and firm performance (Brik et al., 2011). Marketing contributes to firm performance by reducing the degree of information asymmetry between firms and customers. Corporate managers usually know more about the characteristics and products of their firms than their customers (Montiel et al., 2012). Customers cannot search and identify qualified suppliers without bearing any costs. Marketing, such as intensive promotion and advertising, could inform customers about the existence of firms’ products and mitigate the asymmetric information issue (Kim & Joo, 2013; Nath et al., 2010). More importantly, marketing could increase firm reputation and sales by customer retention, acquisition, and coproduction (Andras & Srinivasan, 2003; Balmer et al., 2011). These increased sales could be subsequently transferred to better firm financial performance (Maignan & Ferrell, 2004; Walsh & Bartikowski, 2013).

However, according to the customer perception theory (Carbone & Haeckel, 1994; Iglesias et al., 2019; Sierra & McQuitty, 2005), customers do not passively receive information from firms’ marketing activities. Instead, consumers may be skeptical to persuasion through marketing (e.g. advertising) (Andras & Srinivasan, 2003; Yi & Jai, 2020). The effect of marketing or detailed promotional and advertising activities depends on to what extent customers will believe and accept the information of such activities. We argue that CSR could help firms to strengthen customers’ beliefs and acceptance of marketing and thus further improve firms' financial performance.

Scholars considered CSR as firms’ obligations to stakeholders and society at large (Graves & Waddock, 1994; Story & Neves, 2015; Wang et al., 2016). CSR manifests itself in a wide range of organizational activities, ranging from cause-related or social marketing such as green marketing, fair trade, corporate philanthropy and volunteering, and other practices intending to protect and improve social welfare (Andreasen, 2006; Lee & Kotler, 2011; Luo & Bhattacharya, 2006). CSR shows firms’ willingness to take responsibility beyond their organizational boundary and produces positive reputational capital among communities and stakeholders (Godfrey, 2005). Firms’ philanthropic activities could build a good corporate reputation (Chahal & Sharma, 2006; Cochran & Wood, 1984). This reputational capital could offer shareholders an ‘insurance-like’ protection for a firm’s idiosyncratic intangible assets and marketing activities (Peloza, 2006). As firms grow their influences, stakeholders, including customers, pay greater attention to not only their economic performance but also their social impact (Luo & Bhattacharya, 2006; Mohr & Webb, 2005).

CSR will also mitigate the information asymmetry between customers and firms and thus reduce the transaction costs and customers’ perceived risks (Godfrey, 2005). Similar to that public organizations benefit from communicating accountable information to citizens (Li, 2020), firms benefit from CSR activities because they deliver more positive and credible information, reduce the information asymmetry, and increase customer retention and acquisition (Balmer et al., 2011).

Customers perceive firms that engage in CSR as more responsible and thus increase their loyalty or switch their purchases away from other rivals. Therefore, compared to firms with fewer CSR activities, marketing made by firms with more CSR is more likely to be accepted and supported by customers because CSR reduces customers’ suspicion, mitigates information asymmetry, and increases marketing efficiency. These arguments lead to hypothesis 1.

Hypothesis 1:

The positive relationship between marketing investments on firm performance is strengthened in firms that invest more in CSR.

The motivation of firms’ CSR activities influences whether or not and how well CSR can be transferred to the firm’s reputational capital (Godfrey, 2005). Scholars have challenged the traditional wisdom of CSR increasing customers’ positive purchase willingness and intention and suggested that CSR activities may induce suspicion (Luo & Bhattacharya, 2006; Milne et al., 2008; Porter & Kramer, 2006). Recent anecdotal evidence also points out that consumers suspect firms’ CSR activities as a way of greenwashing (Mattila & Hanks, 2013; Vlachos et al., 2009). Corporate negative events, such as scandals, ethical lapses, and employee mistreating will lead customers to be skeptical of firms’ engagement in CSR (Vlachos et al., 2009). Firms’ motivations for CSR are often complex, ambiguous, and difficult to interpret because firms’ CSR might be driven by pure altruism, economic benefits, institutional pressures, or management opportunistic behaviors. Therefore, customers’ purchasing decisions vary because of their different interpretations of firm’s CSR motivations (Luo & Bhattacharya, 2006). Customers’ skeptical perceptions of firms’ CSR motivations will weaken the moderating effects of CSR.

Therefore, to ensure CSR to be a positive enabler, firms need to meet certain legitimacy conditions. According to the legitimacy theory (Suchman, 1995), the extent to which a firm’s CSR is valuable will rely on whether it has moral legitimacy in the eyes of stakeholders. Stakeholders judge a firm to be morally legitimate only if they perceive that the firm’s CSR motivation is sincerely to improve overall social welfare (Koh et al., 2014). However, if stakeholders perceive that the CSR motivation is to make up for misbehaving, they will judge the firm to be morally illegitimate (Godfrey, 2005). Scholars believe that CSR driven by appropriate charitable motivation will bring reputational capital (Gao, 2012; Godfrey, 2005). If the motivation of a firm’s CSR is believed to be impurely charitable, to gain certain private benefits, or to deliberately satisfy certain groups of people, this type of CSR will hurt the firm’s reputation and generate the backfiring effect (Yoon, 2003).

How firms treat their employees provides signals and basis for customers to evaluate the firms’ CSR motivations because treating employees well is a firm’s primary social responsibility (Story & Neves, 2015). Customers are more likely to perceive firms mistreating their employees as morally illegitimate (Campbell & Kirmani, 2000). This perceived illegitimacy leads to a negative judgement of firms’ CSR and further negatively influences the role of marketing in corporate financial performance (Campbell & Kirmani, 2000). If firms mistreat their employees, such as wrongfully dismissing employees, owing employee’s wages in arrears, and harming employees’ wellbeing, their CSR motivations will be negatively evaluated and their reputational capital gains of CSR will diminish (Vanhamme & Grobben, 2009; Yoon, 2003). These arguments lead to Hypotheses 2.

Hypothesis 2:

The positive moderation effect of CSR is weaker in firms with low levels of employee treatment compared to firms with high levels of employee treatment.

3. Research methodology

3.1. Sample selection

This study chooses the sample of listed firms. Stock prices and financial information of U.S. listed firms are acquired from the Center for Research in Security Prices (CRSP) and Compustat database. CSR information is collected from the MSCI ESG Stats database (formerly known as the KLD database). We acquire institutional ownership information from Thomson Institutional (13f) Holdings database and analyst following information from the Institutional Brokers' Estimate System (IBES) database. Due to data availability, our samples only include listed firms, which are the largest firms, and do not include small and medium enterprises. The number of covered companies in KLD database changed year by year. In the largest covering year, 5,244 firms were covered by the KLD database in 2014. However, only 660 firms were covered in the year of 2000. In total, KLD covers 13,760 firms from 1991 to 2016, with 53,154 observations. We choose the sample of almost all the U.S. listed firms from 2000 to 2016, and get a sample of 4,870 firms and 28,273 observations. The sample set does not go beyond 2016 because CSR information after 2016 is still not available in the KLD database. The number of the companies listed on the U.S. stock exchange in each year experiences a downward trend as in the last two decades, due to some businesses are getting bigger, and mergers and acquisitions have eliminated hundreds more. Next, we exclude the firms in the utility industry by identifying their standard industrial classification code (SIC code between 4000 to 4949), due to that public utility firms are often not profit-orientated and have strong linkage with governments, which are highly affected by the state. We also exclude the firms in the financial industry (SIC code between 6000 to 6999), due to that financial firms’ business model is highly different from other types of firms. We acquire a sample of 3,389 firms and 19,467 observations. Then, firms with missing variables are deleted. Finally, we acquire a sample with 2,892 firms and 17,528 observations. To reduce the impact of extreme outliers, all variables are winsorized at the 1% level by each fiscal year.

3.2. Variable operationalization

3.2.1. Measurement of corporate social responsibility

Many studies utilize the MSCI ESG Stats database (KLD database) to calculate the CSR score (Borisov et al., 2016;; Galant & Cadez, 2017; Krüger, 2015). The 1991 KLD database covers approximately 650 firms, including Domini 400 Social SM Index and Standard & Poor’s (S&P) 500. The KLD extended its coverage to the 1,000 largest firms from 2001 to 2002. The KLD further expands the firms to include the Russell 3000 index. KLD rates firms’ CSR behavior based on firms’ financial reports, websites, and corporate social responsibility reports.

This database covers CSR scores in 13 categories, including alcohol, community, corporate governance, diversity, employee relations, environment, human rights, product, gambling, firearms, military, nuclear, and tobacco. Following Lins et al. (2017), we select five stakeholder-oriented categories to measure firms’ CSR. The five categories include community, diversity, employee relationships, environment, and human rights. Each category and subcategories include the scores of strength and concern. Based on whether firms have a strength or concern, KLD defines the value as ‘0′ or ‘1′. Because some subcategories are only effective in a certain period, most CSR ratings are not available in the total sample period. To address this concern, we follow Lins et al. (2017) to utilize Equation (1)(1) $$\mathrm{CSR}=\sum \left(\frac{\sum {\mathit{Stren}\mathrm{g}th}_{i,j}}{{\mathit{Strength}\mathrm{}\mathit{count}}_i}-\frac{\sum {\mathit{Concern}}_{i,j}}{{\mathit{Concern}\mathrm{}\mathit{count}}_i}\right)$$(1) for the adjusted CSR score. (1) $$\mathrm{CSR}=\sum \left(\frac{\sum {\mathit{Stren}\mathrm{g}th}_{i,j}}{{\mathit{Strength}\mathrm{}\mathit{count}}_i}-\frac{\sum {\mathit{Concern}}_{i,j}}{{\mathit{Concern}\mathrm{}\mathit{count}}_i}\right)$$(1)

In Equation (1)(1) $$\mathrm{CSR}=\sum \left(\frac{\sum {\mathit{Stren}\mathrm{g}th}_{i,j}}{{\mathit{Strength}\mathrm{}\mathit{count}}_i}-\frac{\sum {\mathit{Concern}}_{i,j}}{{\mathit{Concern}\mathrm{}\mathit{count}}_i}\right)$$(1) , i denotes the category of CSR, and j represents the jth subcategory in category i.

In addition, we follow Davidson et al. (2019) to utilize the CSR net score as an alternative measure of CSR. (2) $$\text{CSR Net Score}=\sum {\mathit{Strength}}_{i,j}-\sum {\mathit{Concern}}_{i,j}$$(2)

3.2.2. Measurement of firm performance

We utilize the return on assets (ROA) as a proxy for firm performance because ROA is the most widely used measure to evaluate firms’ financial health (Hambrick & Quigley, 2014; Waddock & Graves, 1997). More specifically, we utilize the ratio of operating income before depreciation to total assets (ROA) as a measure for firm performance, and the ratio of earnings before interest and taxes to total assets (ROA2) as an alternative measure for ROA (Bennett et al., 2017).

3.2.3. Measurement of marketing expense

There is no specific marketing expense in the Compustat database. We follow Kurt and Hulland (2013) and utilize the ratio of SGA-R&D to total assets, which is the selling, general and administrative (SGA) expense minus R&D expenditure, as the proxy for marketing expense. Furthermore, we follow Engelberg and Gao (2011) and utilize the ratio of advertising expense to total sales as an alternative measure for marketing expenses (Baker & Wurgler, 2006; Dahlén et al., 2009).

3.2.4. Measurement of employee treatment

We follow C. Chen et al. (2016) and utilize the category of employee protection as the proxy for employee treatment. Specifically, we harness the similar way in Equation (1)(1) $$\mathrm{CSR}=\sum \left(\frac{\sum {\mathit{Stren}\mathrm{g}th}_{i,j}}{{\mathit{Strength}\mathrm{}\mathit{count}}_i}-\frac{\sum {\mathit{Concern}}_{i,j}}{{\mathit{Concern}\mathrm{}\mathit{count}}_i}\right)$$(1) to calculate the employee treatment score. Moreover, we divide our sample into two parts based on whether firms have a positive employee treatment score. A positive value for employees indicates that a firm’s employee strength exceeds employee concerns.

3.2.5. Control variables

To examine the moderating role of CSR on the relationship between marketing expense and firm performance, we follow prior studies to control firm-level characteristics, including firm size (Size), financial leverage (Leverage), firm age, capital expenditure (Capex), cash holding, sales growth, and Tobin Q. Moreover, in the robustness check, we also include R&D investment (R&D), analyst following (Analyst), product market competition (Competition), institutional shareholding (Institutional), corporate governance index (Governance), and financial constraint (WW). The control variables are expected to have an effect on the dependent variable, for example, (1) firm size would have a positive effect on ROA, as large firms acquire higher returns than small firms, and (2) leverage would have a negative effect on ROA, as firms with high leverage have weaker economic performance. Definitions of all variables are shown in Table 1.

Table 1. Variable definition.

VARIABLE	Definition
ROA	Return on assets, the ratio of operating income before depreciation to total assets
ROA2	Return on assets, the ratio of earnings before interest and taxes to total assets
CSR	CSR score of five stakeholder categories, including community, diversity, employee relationships, environment, and human rights. The calculation methodology is shown in equation 1(1) $$\mathrm{CSR}=\sum \left(\frac{\sum {\mathit{Stren}\mathrm{g}th}_{i,j}}{{\mathit{Strength}\mathrm{}\mathit{count}}_i}-\frac{\sum {\mathit{Concern}}_{i,j}}{{\mathit{Concern}\mathrm{}\mathit{count}}_i}\right)$$(1) .
CSR Net Score	An alternative measure for CSR and the calculation is shown in equation 2(2) $$\text{CSR Net Score}=\sum {\mathit{Strength}}_{i,j}-\sum {\mathit{Concern}}_{i,j}$$(2) .
Marketing	Marketing expense ratio, which is the ratio of selling, general and administrative expense excluding the R&D investment to total assets
Advertising	Advertising expense, which is the ratio of advertising expense to total sales. Advertising is set as zero if it is lacking
Size	Natural logarithm of total assets
Leverage	Firms financial leverage, which is the short-term debt plus long-term debt scaled by total assets
Firm age	The number of years since a firm first appeared in Compustat.
Capex	Firms capital expenditure, which is the ratio of capital expenditure to total assets
Cash Holding	Firms cash holding level, which is the ratio of cash and short-term investments to total assets
Sale Growth	Sales growth rate, which is the growth rate of sales from year t-1 to year t
Tobin Q	The ratio of market value plus debt to total assets
R&D	The ratio of R&D investment to total assets. R&D is set as zero if it is missing
Analyst	The natural logarithm of analyst following plus one
Competition	Product market competition, which is calculated as $1-\sum {\left(\frac{\mathit{Sales}}{\mathit{Total}\mathit{industry}\mathit{Sales}}\right)}^2.$ We utilize the first two digits of the SIC code to classify the industry.
Institution	Total institutional ownership, which is collected from Thomson Institutional (13f) Holdings
Own con	Ownership concentration, which is the Herfindahl index of largest institutional ownership
Governance	Corporate governance index, from MSCI ESG database
WW	WW financial constraint index, −0.091*FCF-0.062*Cash_Divident + 0.021*TLTD-0.044*size + 0.102*ISG-0.035*Sale Growth (Hennessy & Whited, 2007; Whited & Wu, 2006)

Source: The Authors.

3.3. Model specification

Our regression model is shown in Equation (3)(3) $$\mathit{ROA}=\alpha +{\mathrm{\beta }}_1\mathrm{Marketing}+{\mathrm{\beta }}_2\mathrm{C}\mathrm{SR}+{\mathrm{\beta }}_3\mathrm{C}\mathrm{SR}\times \mathrm{Marketing}+\mathrm{\gamma }\mathrm{Control}+\sum \mathit{Firm},\mathit{Industry},\mathit{Year}$$(3) . We utilize the time-fixed effect model to control for effects of time-invariable variables. We also include firm fixed effects and industry fixed effects in the model to control the unobservable factors related to firm and industry. Based on our hypothesis, if CSR positively moderates the relationship between marketing and firm performance, we would expect a positive and significant ${\mathrm{\beta }}_3.$ (3) $$\mathit{ROA}=\alpha +{\mathrm{\beta }}_1\mathrm{Marketing}+{\mathrm{\beta }}_2\mathrm{C}\mathrm{SR}+{\mathrm{\beta }}_3\mathrm{C}\mathrm{SR}\times \mathrm{Marketing}+\mathrm{\gamma }\mathrm{Control}+\sum \mathit{Firm},\mathit{Industry},\mathit{Year}$$(3)

To examine hypothesis 2, we separate our sample into two parts based on whether the employee's CSR score is positive. Next, we perform the regression in Equation (3)(3) $$\mathit{ROA}=\alpha +{\mathrm{\beta }}_1\mathrm{Marketing}+{\mathrm{\beta }}_2\mathrm{C}\mathrm{SR}+{\mathrm{\beta }}_3\mathrm{C}\mathrm{SR}\times \mathrm{Marketing}+\mathrm{\gamma }\mathrm{Control}+\sum \mathit{Firm},\mathit{Industry},\mathit{Year}$$(3) in the two subgroups to examine the impact of employee treatment.

4. Results

4.1. Descriptive results

Description statistics for the main variables are shown in Table 2. The mean value of ROA is 0.121, which indicates that every dollar that the sample firms invested in assets generated 12.1% of net income, measured by the operating income before depreciation. Next, CSR and CSR2 have negative mean values, indicating that sample firms had more CSR concerns than strength on average. In addition, firm marketing expense occupied approximately 22.7% of total assets in the sample firms.

Table 2. Description statistics.

VARIABLE	N	MEAN	SD	MIN	P25	P50	P75	MAX
ROA	17,528	0.121	0.121	−1.072	0.078	0.127	0.180	0.481
ROA2	17,528	0.077	0.124	−1.266	0.038	0.086	0.136	0.409
CSR	17,528	−0.124	0.485	−1.767	−0.400	−0.143	0.067	3.683
CSR Net	17,528	−0.026	2.038	−5.000	−1.000	0.000	1.000	11.000
Marketing	17,528	0.227	0.192	−0.009	0.089	0.170	0.304	1.324
Advertising	17,528	0.012	0.028	0.000	0.000	0.000	0.010	0.207
Size	17,528	7.032	1.570	3.209	5.872	6.881	8.028	12.757
Leverage	17,528	0.201	0.186	0.000	0.016	0.177	0.317	0.915
Firm age	17,528	22.580	16.390	3.000	10.000	16.000	34.000	66.000
Capex	17,528	0.052	0.058	0.001	0.018	0.033	0.061	0.440
Cash Holding	17,528	0.187	0.191	0.000	0.039	0.116	0.277	0.911
Sale Growth	17,528	0.126	0.280	−0.814	−0.005	0.087	0.207	2.060
Tobin Q	17,528	1.759	1.331	0.295	0.944	1.348	2.083	38.927
R&D	17,528	0.037	0.062	0.000	0.000	0.006	0.052	0.539
Analyst	17,528	2.028	0.776	0.000	1.609	2.079	2.639	3.784
Competition	17,528	0.803	0.180	0.000	0.746	0.866	0.926	0.964
Institution	14,437	0.729	0.220	0.014	0.601	0.766	0.886	1.322
Own con	14,439	0.062	0.046	0.018	0.036	0.049	0.070	0.614
Governance	17,411	−0.061	0.207	−0.667	−0.167	0.000	0.000	1.000
WW	17,528	0.476	0.329	−0.422	0.322	0.444	0.573	14.848

Source: The Authors.

Table 3. Pearson correlation analysis.

	1	2	3	4	5	6	7	8	9
1. ROA	1
2. CSR	0.071***	1
3. Marketing	0.026***	0.050***	1
4. Size	0.198***	0.265***	−0.317***	1
5. Leverage	−0.052***	0.003	−0.214***	0.331***	1
6. Firm Age	0.104***	0.152***	−0.058***	0.456***	0.106***	1
7. Capex	0.118***	−0.041***	−0.083***	0.041***	0.092***	−0.070***	1
8. Cash Holding	−0.239***	0.009	0.108***	−0.393***	−0.404***	−0.276***	−0.205***	1
9. Sale Growth	0.082***	−0.080***	−0.059***	−0.100***	−0.065***	−0.176***	0.116***	0.099***	1
10. Tobin Q	0.168***	0.081***	0.175***	−0.215***	−0.224***	−0.190***	0.022***	0.404***	0.253***

(1) *** p < 0.01; ** p < 0.05; * p < 0.1.

Source: The Authors.

Next, we conduct the Pearson correlation analysis to reveal the correlation relationships among the main variables (see Table 3). Marketing and CSR correlate positively with ROA, demonstrating that firms with higher marketing expenses and more CSR engagements are associated with a higher earning capacity. Besides, the largest correlation coefficient between firm age and firm size is 0.456, which is unlikely to trigger multicollinearity problems. We also calculate the variance inflation factor. The largest VIF is 2.3, and the mean VIF is 1.53. Both VIF values are less than 5, indicating that our regression would not suffer from multicollinearity problems.

4.2. Regression model

The main regression result is shown in Table 4. Model 1 is the baseline model, only including control variables. Independent variable—marketing expense, and the moderator—CSR, and its interaction with marketing expense are added into models 2–4. Model 5 is the full model. Year, industry, and firm fixed effects are controlled in all the models. All models pass the 1% level of the F test, proving the significance of the regression model. Adjusted R² values are greater than 0.66, demonstrating that our model has good explanatory power for firm performance.

Table 4. Main regression results.

						Treatment on employees
	Full Sample					Bad	Good
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
VARIABLES	ROA	ROA	ROA	ROA	ROA	ROA	ROA
CSR		0.156			0.112	0.111	0.121
		(1.502)			(1.366)	(1.135)	(1.412)
Marketing			0.184***	0.165***	0.157**	0.143***	0.131***
			(3.125)	(3.431)	(2.109)	(3.288)	(2.930)
CSR x Marketing				0.143***	0.132**	0.113	0.149**
				(2.839)	(2.197)	(1.558)	(2.202)
Size	0.011***	0.014***	0.011***	0.015***	0.013***	0.012*	0.012**
	(2.680)	(2.661)	(2.767)	(2.821)	(2.597)	(1.779)	(2.129)
Leverage	−0.158***	−0.154***	−0.163***	−0.098***	−0.114***	−0.081***	−0.091***
	(−3.674)	(−3.257)	(−3.114)	(−2.914)	(−3.903)	(−3.497)	(−3.564)
Firm Age	−0.014	−0.015	−0.013	−0.014	−0.017	−0.019*	−0.012*
	(−1.250)	(−1.352)	(−1.497)	(−1.511)	(−1.502)	(−1.750)	(−1.862)
Capex	0.050	0.101	0.074	0.063	0.089	0.105*	0.009
	(1.489)	(1.522)	(1.301)	(1.417)	(1.365)	(1.648)	(1.152)
Cash Holding	−0.014	−0.017	−0.023	−0.044	−0.014	−0.056	−0.052
	(−1.234)	(−1.005)	(−1.231)	(−1.047)	(−1.068)	(−1.406)	(−1.535)
Sale Growth	0.053***	0.058***	0.059***	0.054***	0.051***	0.053***	0.062***
	(10.919)	(10.919)	(10.919)	(10.919)	(10.919)	(8.137)	(9.392)
Tobin Q	0.016***	0.014***	0.017***	0.019***	0.015***	0.021***	0.012***
	(5.187)	(4.789)	(4.676)	(5.295)	(6.291)	(8.051)	(4.346)
Constant	0.134***	0.254***	0.175***	0.334***	0.416*	0.603**	0.580**
	(4.779)	(3.145)	(3.442)	(3.657)	(1.803)	(2.188)	(2.010)
Observations	17,528	17,528	17,528	17,528	17,528	8,581	8,947
Adjusted R^2	0.660	0.665	0.670	0.681	0.699	0.764	0.697
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
F	7.784***	10.035***	12.234***	15.647***	30.72***	19.04***	21.08***

(1) T statistics are adjusted by clustering at the firm level and shown in parentheses.

(2) *** p < 0.01, ** p < 0.05, * p < 0.1.

Source: The Authors.

In models 3–5, marketing expense has a positive and significant coefficient, demonstrating that marketing expense has a positive relationship with a firm’s ROA. These findings suggest that marketing expenses could improve a firm’s performance. The coefficient of CSR x Marketing in model 5 is positive and significant at the 5% level, demonstrating that CSR could enhance the relationship between marketing and ROA. To be more specific, the positive effect of marketing on ROA would be improved by CSR, as this positive relationship becomes stronger when CSR increases. Thus, CSR could improve the relationship between marketing expense and firm performance, which is consistent with Hypothesis 1.

To test Hypothesis 2, we split our sample into two groups: bad employee treatment group (below the mean value) and good employee treatment group (above the mean value). The results are presented in models 6 and 7 in Table 4. It is shown that CSR x Marketing has an insignificant coefficient in model 6 but a positive coefficient in model 7, illustrating that the moderating role of CSR is only substantial in the firms that treat employees well. In firms with poor employee treatment, there is no substantial change between the high CSR and low CSR groups. However, in the good employee treatment group, marketing has a negative slope when CSR is low, and the slope of the line becomes positive when CSR is high. Thus, the positive moderating role of CSR only exists for firms that treat employees well, which is consistent with Hypothesis 2.

Some controls are worth noting. Firm size has a positive and significant coefficient in models 1 to 7, demonstrating that large firms acquire higher returns than small firms. In addition, this effect is more pronounced in firms that treat employees well. Leverage has a negative and significant coefficient in models 1 to 7, indicating that firms with high leverage have weaker economic performance. The coefficient of sales growth is significantly positive in models 1 to 7, illustrating that firms with high sale growth also earn a higher return on assets. Finally, Tobin Q has a significantly positive coefficient in all models, demonstrating that firms with high investment opportunities would obtain more profits.

4.3. Robustness checks

To make our results robust, we first include more control variables to alleviate the endogenous concerns of omitted variables. Next, we utilize alternative measures for firm performance, CSR, and marketing expense to address possible measurement error problems. Finally, we conduct the instrument variable regression to address possible endogenous problems.

4.3.1. Adding more control variables

We include more control variables that may correlate with firm performance, including R&D investment, analyst following, product market competition, institutional ownership, ownership concentration, corporate governance, and financial constraint. The regression results are shown in Table 5. The coefficient of CSR x Marketing is positive and significant at the 5% level in models 1 and 3, and the coefficient of marketing is positive and significant too. These results are consistent with hypotheses 1 and hypothesis 2, verifying the robustness of our result.

Table 5. Robustness checks: More control variables.

	Full sample	Treatment on Employees
		Bad	Good
	Model 1	Model 2	Model 3
VARIABLES	ROA	ROA	ROA
CSR	0.115	0.112	0.127
	(1.430)	(1.233)	(1.380)
Marketing	0.155**	0.107**	0.153**
	(2.210)	(2.218)	(2.191)
CSR x Marketing	0.147**	0.094	0.172**
	(2.435)	(1.204)	(2.456)
Size	0.001	0.012*	0.015*
	(0.242)	(1.661)	(1.785)
Leverage	−0.134***	−0.063**	−0.152***
	(−8.257)	(−2.299)	(−7.448)
Firm Age	−0.020*	−0.016	−0.019
	(−1.868)	(−1.415)	(−1.267)
Capex	0.023	0.071	−0.033
	(0.481)	(1.029)	(−0.532)
Cash Holding	−0.032	−0.049	−0.039
	(−1.130)	(−1.112)	(−1.181)
Sale Growth	0.143***	0.155***	0.144***
	(7.358)	(3.481)	(6.628)
Tobin Q	0.014***	0.017***	0.012***
	(8.182)	(6.135)	(5.247)
R&D	0.500***	0.428***	0.415***
	(6.131)	(3.991)	(3.856)
Analyst	0.008***	0.012***	0.009***
	(3.153)	(2.964)	(2.628)
Competition	0.013	0.023	0.008
	(0.869)	(1.101)	(0.395)
Institution	0.018*	0.039***	0.028***
	(1.941)	(2.657)	(2.639)
Ownership Concentration	−0.146***	−0.163**	−0.157***
	(−3.654)	(−2.385)	(−3.304)
Governance	−0.004	−0.009	0.001
	(−0.645)	(−1.334)	(0.166)
Financial constraint	−0.089***	−0.105**	−0.080***
	(−4.945)	(−2.434)	(−4.109)
Constant	0.561**	0.567**	0.508
	(2.369)	(2.018)	(1.600)
Observations	14,332	6,736	7,586
Adjusted R-squared	0.704	0.766	0.799
Year FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
F	29.66***	11.23***	19.50***

(1) T statistics are adjusted by clustering at the firm level and shown in parentheses.

(2) *** p < 0.01, ** p < 0.05, * p < 0.1.

Source: The Authors.

4.3.2. Alternative measures for firm performance

Here, we utilize alternative measurements for firm performance to conduct robustness checks. We calculate the ROA using the earnings before interest and tax and present the regression result in Table A1 in the appendices. CSR x Marketing presents a significantly positive coefficient in model 1 and model 3, which is consistent with prior results. This result proves the positive moderating role of CSR on economic performance, and this impact is only pronounced in firms that treat employees well. Thus, this result further verifies the robustness of our results.

4.3.3. Alternative measures for corporate social responsibility

Moreover, we utilize the CSR net score to replace the original CSR score and present the result in table A2 in the appendices. CSR Net x Marketing shows a significantly positive coefficient in models 1 and 3, indicating that CSR improves the relationship between marketing and firm performance. This impact is only substantial in firms that treat employees well. These results are consistent with prior findings and indicate the robustness of the findings.

4.3.4. Alternative measure for marketing expense

Lastly, we utilize advertising expenses to replace marketing expenses and present the results in Table A3 in the appendices. CSR x Advertising is only positive and significant in firms that treat employees well. Although CSR x Advertising is not significant in model 1, the T value is 1.629, which is close to the 10% critical value of 1.65. A possible reason is that the CSR’s positive moderating role is only effective in firms that treat employees well, leading to an insignificant result in the full sample. However, we can still prove the significant moderating effect of CSR on the relationship between advertising and firm performance in the firms that treat employees well. Thus, our results remain robust.

4.4. Instrumental variable estimation

The issue of reverse endogeneity might affect our results. For example, firms with high earning capacity are more likely to engage in CSR activities and invest in marketing activities, leading to a biased regression result. To address this concern, we employ the instrumental variable (IV) regression. We follow Fisman and Svensson (2007), Lin et al. (2010), and Cheng et al. (2014) by using the industry-year average value, excluding the firm itself, as the instrumental variable. The industry level CSR and marketing expenses might affect a firm’s CSR and marketing expenses but cannot have a direct impact on the error term in the equation of interest. Thus, we utilize the industry-year average CSR and marketing expense as the instrumental variable for CSR and marketing expenses and conduct instrumental variable regression. IV regression results are shown in Table A4 in the appendices. CSR x Marketing has a positive and significant coefficient in models 1 and 2. The value of coefficient of CSR x Marketing of the full sample as shown in model 1 is 0.241, and when firms treat their employees better, as shown in the model 2, the value of coefficient of CSR x Marketing increases to 0.307. The result indicates that the positive moderation effect of CSR is stronger in firms with high levels of employee treatment compared to firms with low levels of employee treatment. This result is consistent with our hypothesis and verifies the robustness of our results.

5. Conclusion

In this study, we argue that the CSR is an enabler rather than a driver of firm financial performance and provide empirical evidence to back up our argument. We find that CSR moderates the effects of marketing, which is a direct driver of firm performance. The theoretical argument indicates that CSR signals customers to overcome the information asymmetry between them and firms, yields reputational capital gains of firms’ marketing efforts, and further improves firms’ financial performance. Consistent with the CSR literature, this study supports the assumption that CSR does not always positively influence firms’ performance. CSR negatively influences on a firm’s performance when stakeholders suspect the firm’s CSR motivation is not purely charitable. Firms may use CSR to make up employee mistreatment. Our results show that employee mistreatment reduces the reputational capital gains of CSR and thus weakens the moderating effect of CSR on the marketing-performance association.

Our findings have both theoretical and practical implications. First, we build on previous studies and contribute to the CSR literature by theoretically proposing and empirically testing CSR as an enabler rather than a direct driver of firms’ performance. In particular, our findings demonstrate that CSR yields reputational capital gains of firms’ marketing efforts that directly improve performance. CSR’s moderating effects on marketing is no longer salient when firms’ CSR motivation is not purely altruistic. For example, if customers perceive that firms are using CSR to make up employee mistreatment, CSR will backfire and weaken marketing’s positive effects on performance. Unlike studies examining the effects of CSR on firm performance and on marketing, R&D, and strategic differentiation, this study sheds light on the mechanism by which CSR plays its role in influencing firm performance, i.e., moderating the effects of marketing on firm performance.

Our results also provide practical implications. Our study suggests that firm managers and executives can utilize CSR to improve their firms’ strategic and managerial practices that benefit performance. In particular, marketing managers can utilize CSR to increase firms’ reputational returns and financial performance. However, if their firms mistreat employees, fixing the mistreatment might be a more urgent issue than rushing in CSR. When customers perceive a firm’s CSR as a makeup instead of a pure charitable action, CSR may backfire and could not play its positive moderating role in enhancing firm performance. In addition, when firms are treating employees well, firm leaders should effectively communicate such information to their stakeholders because then CSR would play a more effective role in moderating the marketing-performance association. There are multiple factors contributing to firm performance, and our study shows that one factor may have contradicting effects on firm performance. Therefore, we should pay closer attention to the context and environment that fosters one particular factor’s contributions to firm performance.

We identify several areas for further research. First, in this study, we only focus on the CSR’s moderating effect on marketing. However, there are many other direct performance drivers. Future studies could focus on CSR’s moderating roles in relationships between other drivers and firm performance to test whether CSR can enable other drivers. These studies would help firms design and execute their CSR strategies and optimize their CSR investments. Second, this study only focuses on how mistreating employees exposes the impurity of firms’ CSR motivation and finds that employee mistreatment weakens the positive effect of CSR in moderating the marketing-performance relationship. Effective management is far beyond treating employees. Future studies could discuss relationships between other managerial practices’ influences and CSR’s moderating effects, such as environmental management, in particular climate change mitigation (Cadez & Czerny, 2016; Lyon & Maxwell, 2008), the grand challenge of mankind. Third, previous studies document that CSR is contextually contingent and is largely influenced by external institutions. While we only use data from the U.S. firms, future studies could perform more examinations in settings other than the U.S. CSR is also contingent upon industry sector (Cadez & Guilding, 2017) and the magnitude of stakeholder pressures (Perez-Batres et al., 2012), which could be investigated in future studies.

Finally, we acknowledge some limitations of our study. Due to data availability, the use of MSCI database includes only the biggest companies from the U.S., which leads to the generalizability problem, as the situations for small and medium enterprises may be different. Meanwhile, there are some concerns on the causality of the relationships. Compared to data and measures from randomized experiments, the archival data and proxies used in this study for the main constructs of interest are relatively crude. The correlations between ROA, CSR, and marketing in our results are relatively low, and it may influence the validity of outcomes. The reverse causality of the relationships also could also be an issue that this study has not resolved.

Disclosure statement

The authors declare no potential conflict of interest.

Appendices:

Results of robustness checks

This study conducts a serise of robustness check to verify the results. We put the tables of additional robustness checks in the appendices.

Table A1. Robustness checks: Alternative measurement for firm performance (ROA2).

	Full sample	Treatment on employees
		Bad	Good
	Model 1	Model 2	Model 3
VARIABLES	ROA2	ROA2	ROA2
CSR	0.114	0.113	0.024
	(1.122)	(1.278)	(1.112)
Marketing	0.147***	0.263***	0.116***
	(3.970)	(3.328)	(3.180)
CSR x Marketing	0.136**	0.106	0.126**
	(2.260)	(1.289)	(2.335)
Size	0.009*	0.009*	0.013**
	(1.916)	(1.958)	(2.308)
Leverage	−0.147***	−0.068**	−0.161***
	(−3.430)	(−2.420)	(−3.736)
Firm Age	−0.017	−0.019	−0.012
	(−1.558)	(−1.544)	(−0.851)
Capex	0.054	0.098	0.019
	(1.197)	(1.449)	(0.326)
Cash Holding	0.015	0.087	0.076
	(1.021)	(1.350)	(1.366)
Sale Growth	0.057***	0.057***	0.064***
	(10.180)	(7.961)	(8.702)
Tobin Q	0.015***	0.022***	0.012***
	(6.262)	(7.789)	(4.336)
Constant	0.419*	0.626**	0.674**
	(1.660)	(1.975)	(1.986)
Observations	17,528	8,581	8,947
Adjusted R − squared	0.667	0.713	0.730
Year FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
F	31.51***	23.44***	24.79***

(1) T statistics are adjusted by clustering at the firm level and shown in parentheses.

(2) *** p < 0.01, ** p < 0.05, * p < 0.1.

Source: The Authors.

Table A2. Robustness checks: Alternative measurement for CSR.

	Full sample	Treatment on employees
		Bad	Good
	Model 1	Model 2	Model 3
VARIABLES	ROA	ROA	ROA
CSR Net	0.002	0.001	0.005
	(1.403)	(1.213)	(1.377)
Marketing	0.160**	0.143**	0.136**
	(2.245)	(2.319)	(2.065)
CSR Net x Marketing	0.007**	0.011	0.014**
	(1.979)	(1.158)	(2.076)
Size	0.010***	0.011**	0.012**
	(2.588)	(2.372)	(2.426)
Leverage	−0.141***	−0.081***	−0.153***
	(−9.893)	(−3.488)	(−8.538)
Firm Age	−0.016	−0.019*	−0.012
	(−1.628)	(−1.753)	(−0.896)
Capex	0.059	0.105*	0.097*
	(1.374)	(1.649)	(1.679)
Cash Holding	−0.014	−0.056	−0.042
	(1.015)	(−1.706)	(−1.352)
Sale Growth	0.055***	0.053***	0.062***
	(10.913)	(8.138)	(9.381)
Tobin Q	0.015***	0.021***	0.012***
	(6.311)	(5.057)	(4.346)
Constant	0.424*	0.605**	0.611**
	(1.839)	(2.190)	(2.054)
Observations	17,528	8,581	8,947
Adjusted R − squared	0.699	0.764	0.796
Year FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
F	30.61	18.95	20.88

(1) T statistics are adjusted by clustering at the firm level and shown in parentheses.

(2) *** p < 0.01, ** p < 0.05, * p < 0.1.

Source: The Authors.

Table A3. Robustness checks: Alternative measurement for marketing expense.

	Full sample	Treatment on employees
		Bad	Good
	Model 1	Model 2	Model 3
VARIABLES	ROA	ROA	ROA
CSR	0.017	0.009	0.013
	(1.418)	(1.081)	(1.012)
Advertising	0.727***	0.567***	0.868***
	(6.207)	(2.830)	(5.983)
CSR x Advertising	0.108	0.044	0.170*
	(1.629)	(0.395)	(1.853)
Size	0.015***	0.009*	0.015***
	(3.983)	(1.665)	(3.191)
Leverage	−0.142***	−0.080***	−0.152***
	(−3.984)	(−3.342)	(−3.610)
Firm Age	−0.016*	−0.018	−0.012
	(−1.664)	(−1.612)	(−0.928)
Capex	0.052	0.097	−0.003
	(1.213)	(1.528)	(−0.046)
Cash Holding	−0.010	−0.069	−0.021
	(−0.728)	(−1.392)	(−1.318)
Sale Growth	0.055***	0.054***	0.062***
	(11.047)	(8.261)	(9.528)
Tobin Q	0.015***	0.020***	0.012***
	(6.424)	(7.755)	(4.488)
Constant	0.391*	0.455*	0.398*
	(1.733)	(1.656)	(1.737)
Observations	17,528	8,581	8,947
Adjusted R − squared	0.699	0.760	0.798
Year FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
F	34.16	17.25	24.71

(1) T statistics are adjusted by clustering at the firm level and shown in parentheses.

(2) *** p < 0.01, ** p < 0.05, * p < 0.1.

Source: The Authors.

Table A4 Robustness checks: Instrumental variable regression.

	Full sample	Treatment on employees
		Good
	Model 1	Model 2
VARIABLES	ROA	ROA
CSR	0.118	0.151
	(1.494)	(1.333)
Marketing	0.270*	0.147*
	(1.782)	(1.845)
CSR x Marketing	0.241**	0.307**
	(2.526)	(2.071)
Size	0.036	0.027
	(1.549)	(0.766)
Leverage	−0.125***	−0.130***
	(−3.443)	(−3.749)
Firm Age	−0.013	−0.006
	(−1.265)	(−0.460)
Capex	0.014	−0.010
	(0.260)	(−0.112)
Cash Holding	−0.016	−0.114
	(−1.416)	(−1.194)
Sale Growth	0.058***	0.061***
	(10.139)	(8.204)
Tobin Q	0.013***	0.010***
	(5.651)	(3.659)
Observations	16,982	9,419
Year FE	Yes	Yes
Industry FE	Yes	Yes
Firm FE	Yes	Yes
F	29.68	21.52
Under identification test (p-value)	0.000	0.000

(1) T statistics are adjusted by clustering at the firm level and shown in parentheses.

(2) *** p < 0.01, ** p < 0.05, * p < 0.1.

Source: The Authors.