Abstract
According to surveys of companies, branding is one of the main objectives of their corporate social responsibility (CSR). With advantageous data from Brand Finance, we address three contextual factors that may condition the relationship between CSR and brand value. First, we hypothesize that the relationship between CSR and brand value obtains across major world regions and industrial sectors (“the convergence thesis”). Second, we hypothesize that the relationship has weakened with time, as companies have had increasing difficulty using CSR to differentiate their brands in a sea of CSR-espousing competitors (“the crowding out thesis”). Third, we hypothesize that the relationship between CSR and brand value is weaker where a brand’s identity is different from that of its corporate owner, which may make it difficult for observers to readily link (corporate-level) CSR with its potential (lower level) brand beneficiaries (“the identity-match thesis”). We support these hypotheses with random-effects, fixed-effects, and instrumental-variable regressions before ending with contributions, limitations, implications, and potential next steps.
Keywords
Corporate social responsibility (CSR) is the idea that companies have not only economic obligations toward their owners but also social obligations toward other stakeholders (Carroll, 1991). These obligations range from complying with laws to leaving a smaller environmental footprint, supporting local communities through charitable donations, and increasing gender and ethnic diversity on boards of directors. A major perceived benefit of CSR, according to surveys of companies themselves, is branding. Surveys with this finding have been conducted by academics (Lawrence et al., 2013, Figure 14), professional bodies (American Management Association, 2005, p. 2), consultancies (McKinsey & Company, 2008, Exhibit 2), intergovernmental groups (United Nations Development Programme, 2012, p. 41), business associations (Responsible Ireland, 2012, p. 3; World Economic Forum & International Business Leaders Forum, 2003, Figure 1), investment houses (Tshikululu Social Investments, 2013, Exhibit 14), and media organizations (The Economist Intelligence Unit, 2005, p. 28). While CSR may bring many other business benefits (Sprinkle & Maines, 2010), these surveys suggest that branding is among those that matter most.
Percent of global 250 companies that produce sustainability reports.
Scholarly work on whether CSR, in actuality, strengthens brands has overwhelmingly tended to predict consumer brand perceptions rather than the ultimate effectiveness indicator of brand financial performance. Reflecting this gap is the large literature on CSR and brand equity (D. H. M. Wang et al., 2015) compared with the small literature on CSR and brand value (Harjoto & Salas, 2017; Melo & Galan, 2011). This gap is significant as brand equity and brand value are very different in nature (Raggio & Leone, 2009; Tiwari, 2010), in measurement approach (Chandon, 2003), and in susceptibility to a problem that may plague CSR research—social desirability bias (Carrigan & Attalla, 2001; Eckhardt et al., 2010). In this latter regard, ample research finds that consumer brand perceptions (the underlying substance of most estimates of brand equity) often do not translate into actual brand purchases (the revenue streams upon which brand-value estimates are ultimately based; Kuokkanen, 2017; Kuokkanen & Sun, 2020). More generally, as observed by Raggio and Leone (2007), “a distinction [between brand equity and brand value] is important because, from a managerial perspective, the ultimate goal of brand management and brand equity research should be to understand how to leverage [brand] equity to create [brand] value” (p. 380).
Greatly limiting the ability to conduct research on CSR and brand value has been data availability. Researchers can somewhat easily use existing questions and scales to build indicators of brand equity by surveying consumers of target brands (Hur et al., 2014). By contrast, to estimate brand value, researchers need information on brand earnings over time, which is generally not public. Lack of data availability has provided an opportunity for several brand consultancies to carve a niche for themselves in brand-value estimation. Three of them now publish annual lists with estimates of the brand values of the largest global brands, namely, Interbrand, BrandZ, and Brand Finance. These organizations reflect another imbalance in the literature—The dataset of the first has appeared in a handful of studies (Harjoto & Salas, 2017; Melo & Galan, 2011; Torres et al., 2012; Yang & Basile, 2019; see Table 1 for a listing of previous panel studies of brand value on CSR), whereas the datasets of the latter two have been relatively underutilized.
Previous Panel Regressions of Brand Value (BV) on CSR.
Note. Torres and colleagues (2012) do not supply information on the number of non-U.S. observations under analysis. Moreover, this information cannot be precisely calculated from Interbrand’s lists because many brands there are owned by private companies (and would therefore be dropped from the analysis due to lack of public financial records), and many others would undoubtedly have missing values on a key variable of interest. Nonetheless, the estimate above assumes that 43% of the study’s sample comes from non-U.S. brands, because this is approximately the percentage of non-U.S. brands in the published lists in a given year. We exclude H.-M. D. Wang (2010) and Cowan and Guzman (2020) from above, because their studies are cross-sectional. CSR = corporate social responsibility; DJSI = Dow Jones Sustainability Index.
The study at hand, which makes the first significant use of the Brand Finance database in the CSR literature, seeks both methodological and theoretical contributions. Methodologically, we aim to improve upon existing panel regressions in six ways. Our study (a) adds robustness by debuting an alternative measure of brand value; (b) has increased statistical power, at least threefold larger than previous studies, allowing for more control variables to rule out the possibility of spurious statistical relationships (addressing a limitation of prior work pointed out by Bhattacharya et al., 2020, p. 2069); (c) strengthens causal inference with instrumental-variable (IV) modeling to address the potential endogeneity of CSR; (d) analyzes a highly international brand sample, representing brand-owning firms from 44 countries (responding to calls from Harjoto & Salas, 2017; Torres et al., 2012, p. 22; Yang & Basile, 2019, p. 14); (e) has greater generalizability to smaller brands by analyzing hundreds more outside the top 100 in brand value (responding to calls from Yang & Basile, 2019, p. 14); and (f) increases robustness with a CSR indicator based, not on a single ratings provider but on five different well-recognized global evaluation schemes (responding to calls from Bhattacharya et al., 2020, p. 2070, to move beyond the typical usage of the scores of KLD Analytics).
As for theoretical contributions, with respect to the influential rubric of Whetten (1989), we address the where, when, and who of CSR and brand value. Although, like previous studies, we document a general relationship between CSR and brand value, our hypotheses move beyond this established finding to raise contextual factors that might govern the relationship. This approach positions our study within a line of research showing that CSR’s effects on business outcomes are oftentimes mediated, moderated, and subject to boundary conditions such as time, place, and industry (Q. Wang et al., 2016). In particular, our hypotheses examine whether the CSR–brand-value relationship depends on the contextual factors of where, in terms of world regions and industrial sectors (first hypothesis); when, in terms of periods in CSR’s development (second hypothesis); and who, in terms of specific brand identities that allow stronger connections between (corporate-level) CSR and (lower level) brand value (third hypothesis).
While each hypothesis hangs together around contextual factors that may condition CSR’s effect on brand value, each also makes its own lower level theoretical contribution. In terms of the rubric of Business & Society (Crane et al., 2016), our first hypothesis tests a highly influential argument from Matten and Moon (2008) that companies are now using “explicit CSR” for branding purposes across major world regions. We follow Jamali and Neville (2011) in calling this the “convergence thesis,” which we test in answer to a recent call for regional-comparative research on CSR and brand value (Harjoto & Salas, 2017). Our second hypothesis refines the convergence thesis by attending to a corollary: As more companies have converged upon explicit CSR, its branding benefits may have diminished with time. To facilitate discussion, we label this the “crowding out thesis.” Our third hypothesis generates a theory about which brands are most likely to benefit from CSR. Here, we argue that CSR (by its very terminology, a corporate-level construct) has stronger brand benefits when the corporate identity is similar to that of the owned brand. Where these identities match up, audiences can more easily link CSR with its potential brand beneficiaries. We label this the “identity-match thesis,” and motivate it with insights from the brand architecture and brand identity literatures (Aaker & Joachimsthaler, 2000).
Core Concepts and Theoretical Underpinnings
CSR implies that companies have social responsibilities, which are often understood in contrast to their more traditional financial and economic obligations (e.g., to be efficient, well-governed, and profitable). 1 The term suggests accountability at the company level, that the onus for managing social impacts is vested with companies themselves and not just with governments that provide more compulsory forms of regulation. CSR can therefore be viewed as a type of decentralized social movement in which companies, sometimes with guidance or pressure from civil society, claim to address social and environmental issues through a range of beyond-compliance activities. The common understanding of CSR as voluntaristic action is closely linked to the observation that companies can use discretionary, individualistic, and charismatic CSR practices as the basis for branding campaigns. This may allow companies to differentiate their product offerings from those of competitors. Companies that brand themselves or their products as socially responsible may ultimately come to realize higher firm performance (Kerin & Sethuraman, 1998; Rao et al., 2004), as well as many intermediate benefits, such as enhanced credibility and greater customer loyalty and advocacy (Aaker, 2010; Keller, 2001; Taylor et al., 2004). Good branding, in other words, is one of many possible “business cases for CSR” (Carroll & Shabana, 2011; Hafenbradl & Waeger, 2017; Margolis & Walsh, 2003; Porter & Kramer, 2006).
Brands are definable as a set of “marketing-related intangible assets that may include names, terms, and logos that are intended to identify goods and create distinctive images and associations in the minds of stakeholders, thereby creating economic benefits for the owner” (Australian Marketing Institute & Brand Finance, 2011). 2 Brand value is “the monetary value of a brand to a company in a transaction, whether it is internal or external (as with an investment, purchase, sale or licensing agreement)” (ISO 20671, 2019). 3 As such, brand value often refers to the “sale or replacement value of a brand” (Raggio & Leone, 2007, p. 380). The term suggests that investments in advertising, public relations, or marketing to perform such activities as cultivating a brand personality or brand awareness can be assessed in terms of how much these activities ultimately increase a brand’s monetary value as might be realized in its sale (Tiwari, 2010).
A potential tool to improve brand value is CSR. Surveys of business leaders, at least, report that branding is one of CSR’s main benefits. A Boston Consulting Group and MIT Sloan Management Review (2009, Exhibit 3) survey of about 1,500 business leaders from all major world regions on “The greatest benefits to your organization from addressing sustainability issues” found that the top response category was “improved company or brand image.” This category was selected by 36% of respondents, much larger than the second-place “cost savings,” which garnered only 10%. Similarly, a McKinsey & Company (2008, Exhibit 2) survey found that the top pathway by which a global sample of Chief Financial Officers believe that CSR improves company financial performance is “maintaining a good reputation or brand equity.” 4 As a final example, Accenture’s (2010, Figures 1–3) global survey of over 1,000 executives on “the factors that have driven you, as a CEO, to take action on sustainability issues” found that the top one, by far, was “brand, trust, and reputation,” affirmed by 72% of respondents, with the second-place “potential for revenue growth/cost reduction” trailing well behind at 44%.
There are several lines of theory that shed light on why CSR might be good branding. Institutional theory is one (DiMaggio & Powell, 1983; Meyer & Rowan, 1977). It concerns the durable structures of organizational life, including the policies, practices, and initiatives by which the global CSR movement has spread (Kaplan & Kinderman, 2020; Lim & Pope, 2020; Strang & Meyer, 1993). Institutional theory suggests that organizations converge upon a similar set of structures to bolster their legitimacy in the eyes of their audiences (Etter et al., 2016). Conformance to institutional expectations makes an organization’s activities seem more understandable, appropriate, and necessary, and may thereby increase access to resources vital for organizational survival and success (Jeong & Kim, 2019). Viewed through an institutional theoretical lens, CSR branding may have emerged and spread because it allows organizations to position their identities in ways that resonate with the beliefs, values, and cognitions of their cultural environments (Athanasopoulou & Selsky, 2015). As preferences for CSR have become widespread (Cone Communications, 2015), CSR branding has been occurring on deeper levels. Many organizations have become isomorphic with the new expectations by revising even their mission, vision, and core-value statements to include CSR language (Pope et al., 2018).
Another insightful body of thought is stakeholder theory (Freeman, 1984). Its core claim is that organizational leaders, in the course of decision making, should (or actually do) consider how their choices affect or can be affected by social groups beyond shareholders, such as workers, suppliers, investors, unions, and local communities (Laplume et al., 2008). The variant that has been called instrumental stakeholder theory (Donaldson & Preston, 1995) proposes that businesses, even when pursuing their traditional objectives, should cultivate stakeholder relationships the same as any other organizational resource (Stoney & Winstanley, 2001). A strategy that may serve well here is CSR. It may deliver more purchases from consumers (Martínez & Rodríguez del Bosque, 2013), more capital from investors (Cheng et al., 2014), more productivity from employees (Gubler et al., 2018), less pressure from activists (Helmig et al., 2016), and a lighter regulatory touch from governments (Kinderman, 2011). Given the breadth of stakeholders found to respond favorably to CSR, business leaders may decide that CSR branding provides favorable positioning with respect to a widespread stakeholder preference. Business leaders that manage to cultivate strong CSR identities may ultimately come to realize higher firm value (Jo & Harjoto, 2011; McWilliams & Siegel, 2000).
Alongside institutional and stakeholder theories, which provide a broad backdrop for understanding the rise of CSR branding, is a complementary and highly empirical literature that investigates the factors that condition whether CSR branding is successful. Among those documented are the credibility of CSR claims (Hur et al., 2014) and observers’ perceptions of brand quality (Liu et al., 2014). Others are industry affiliation (Lai et al., 2010) and consumer awareness of CSR as obtained through courses of advertising (Du et al., 2010). Similarly, CSR may have a greater branding impact if it aligns with the preexisting brand identity (Deng & Xu, 2017) and if consumers perceive the CSR as stemming from altruistic rather than self-interested motivations (Groza et al., 2011). Below, we seek to further this literature on the factors that condition CSR’s relationship with brand value by building three hypotheses on where, when, and for whom the relationship holds.
Hypotheses
Where: The Convergence Thesis
This section answers calls for research that investigates the relationship between CSR and brand value across world regions (Harjoto & Salas, 2017, p. 22; Torres et al., 2012, p. 22; Yang & Basile, 2019, p. 14). These calls reflect the reliance of existing studies (Table 1) on predominantly American brands. The theoretical significance of answering these calls is that it can inform “rival theses” as to whether the nature and effects of CSR are the same or different by geography (Jamali & Neville, 2011, p. 599). In setting out the convergence and divergence theses below, we engage particularly with the convergence arguments of Matten and Moon (2008). This is not only because this study is a landmark in the CSR literature but also because it has temporal correspondence with our data, being published near the beginning of our observation window.
The core claim of the “divergence thesis” is that CSR differs substantially by geography, that it interacts with local familial, linguistic, economic, governmental, religious, and ethical systems (Jamali & Carroll, 2017). CSR may manifest differently, for example, in Muslim societies (Murphy & Smolarski, 2020) or vary by civil and common law contexts (Mehmet et al., 2017). A country’s developmental status may also influence the CSR topics that business leaders confront (Jamali et al., 2017). Differences in countries’ governance capacities, as a final example, may condition whether a CSR framework, constructed at the world level by an international nongovernmental group, diffuses to a domestic context to regulate business activities (Jackson & Apostolakou, 2010; Pope & Lim, 2020).
On the contrary is the “convergence thesis,” which proposes an increasing worldwide similarity in CSR understandings, values, principles, and practices (Chiu, 2013). Trappings of this thesis are the rise and spread of “common models” of CSR (Jamali & Neville, 2011, p. 599; Pope & Meyer, 2015). In the 1980s and 1990s, dozens of corporate-membered CSR initiatives diffused to standardize business practices in industries as diverse as chemical manufacturing (Responsible Care; est. 1985), logging (Rain Forest Alliance, 1987), agriculture (World Fair Trade Organization, 1989), manufacturing (ISO 14001, 1992), fishing (Marine Stewardship Council, 1996), apparel (Fair Labor Association, 1999), banking (Equator Principles, 2003), and mining (Extractive Industries Transparency Initiative, 2003). Around the turn of the millennium, a second wave of CSR initiatives sought to standardize CSR communications for companies in any industry, including the Global Reporting Initiative (1997), United Nations Global Compact (UNGC, 2000), and Carbon Disclosure Project (2000). At the same time, CSR ratings proliferated to commensurate and evaluate the CSR of companies across nations and industries, including the Dow Jones Sustainability Index (DJSI, 1999), FTSE4Good Index (2001), and “World’s Most Ethical Companies” (2005).
In taking up the convergence thesis, we follow Jamali and Neville (2011) in treating Matten and Moon (2008) as the most influential statement. This article, now among the top-10 most-cited articles on CSR in Web of Science, has particular relevance for us, because its core argument suggests that CSR is currently being used worldwide for branding. It also has empirical relevance because our observation window begins in 2007, the year before the article’s publication, undoubtedly when the authors were refining their ideas and making key observations. This temporal correspondence presents an opportunity for a history-matched test of an implication of an especially prominent CSR theory.
Matten and Moon (2008) contend that a global convergence is occurring toward a type of CSR that is highly conducive to branding. In the authors’ terminology, companies in world regions formerly associated with “implicit CSR” have begun to turn toward “explicit CSR.” Implicit CSR, in the authors’ conceptualization, reflects business systems that are national and regional in scope rather than international. Historically typified by European nations, implicit CSR tends to be mandatory, codified into law, and inconsistent with “individual corporations articulating their own versions” (Matten & Moon, 2008, p. 410). Explicit CSR, by contrast, historically associated with the United States, “normally consists of voluntary programs and strategies” (Matten & Moon, 2008, p. 409) that are expressly discussed in CSR language. Explicit CSR vests companies with “discretion and initiative” to form and externally advertise their own “firm-specific responsibility practices” (Matten & Moon, 2008, p. 405). As explicit CSR consists of “clearly articulated and communicated practices and policies” (Matten & Moon, 2008, p. 405) that emphasize voluntaristic, firm-specific social responsibilities, it is a type of CSR by which companies can distinguish themselves and thereby gain branding benefits.
The hypothesis below goes a step further than Matten and Moon (2008) to theorize that the shift toward explicit CSR is occurring, not only across world regions but also across industrial sectors. In other words, it recognizes that Matten and Moon (2008) are painting the picture of a broad cultural shift, where geography is one dimension of many. As for industrial sectors, here too, one could contrast the first hypothesis below with a large comparative literature that suggests divergence. Studies are abundant that portray CSR as a unique reflection of industry-specific institutional logics, competitive dynamics, and resource dependencies (Arminen et al., 2019). CSR may differ, for example, across retail and business-to-business organizations (Haddock-Fraser & Fraser, 2008) or by industries that sell search, experience, and credence goods (Tian & Wang, 2011). CSR may have limited branding effects in industries that are inherently deleterious to the natural environment or human health, such as cigarette manufacturers (Palazzo & Richter, 2005). Nevertheless, if the shift toward explicit CSR theorized by Matten and Moon (2008) is a diffuse cultural phenomenon, its spread should also be pan-industrial. We draw inspiration here from recent scholarship showing that some CSR practices have spread, not only among businesses but also throughout the nonprofit and public sectors (Pope et al., 2018).
When: The Crowding Out Thesis
This section draws out a corollary of Matten and Moon (2008): As CSR has diffused across geographies and industries to become a commonplace and routine activity, its branding benefits may have diminished with time. As a range of CSR practices have become standardized, mainstream, and even taken for granted, their capacity to provide competitive differentiation may have attenuated. We call this the “crowding out” thesis and further discuss it below.
The underlying logic of the crowding out thesis is marketing advice that branding should be based on company and product attributes that are “distinct and difficult-to-imitate” (Blankson, 2016, p. 162). To develop a brand positioning (Gwin & Gwin, 2003), marketers generally recommend that companies find unoccupied market niches, where they can sell products with perceived attributes that cannot be found elsewhere (Kotler, 2003). An indicator of a good brand differentiation strategy is the ability to claim a set of unique, concise, credible, specific, memorable, and valued propositions for a consumer to buy an offering over that of competitors (Anderson et al., 2006; Ghodeswar, 2008).
The problem for CSR is that many associated activities are now quite ordinary rather than cutting edge. The dramatic diffusion that brought this about occurred in the years just before and after Matten and Moon’s (2008) publication. It was then that many authors began to observe a marketplace in which CSR advertisements are “everywhere” (Alves, 2009). A striking example of the extent of this diffusion is CSR reporting. KPMG (2020, p. 10) has tracked the incidence of CSR reporting from 1999 to 2020 among the world’s largest 250 companies by revenue (Figure 1). In the first decade of the new millennium, figures climbed dramatically from 35% (1999) to reach almost full saturation by 2011. Throughout the most recent decade, figures appear to have topped out, fluctuating in the narrow range of 93% to 96%, with very little remaining upside to be realized.
Many have proceeded from the observation of a saturated marketplace to argue that companies need to do more and more CSR just to stand out. Perceiving an era of “ratcheting expectations,” Bertels and Peloza (2008) claimed that companies who use conventional CSR to burnish their brands may be “running just to stand still” (p. 62). Huemer (2010) noted that “globally shared principles” of CSR, termed “hypernorms,” have become “central and enduring attributes” for many companies and, as such, “do not provide the distinctiveness needed for organizational identities” (p. 273). Dowling and Moran (2012) noted that most CSR practices, such as giving money to worthy causes or joining third-party initiatives, are easy for competitors to imitate and thus cannot form the basis of a sustainable reputation strategy (p. 25). Peloza and colleagues (2012) noted that “sustainability messages have become ubiquitous . . . for most large firms,” and for those “that have not pursued a leadership position, it is difficult . . . to use sustainability to create meaningful differentiation from competitors” (p. 74). These observations tie into the broader “de-radicalization” (Shamir, 2004) and “mainstreaming” of CSR (Berger et al., 2007), whereby CSR has transitioned from a formerly “subversive doctrine” to a consensus movement (Friedman, 1970).
A possible outgrowth of this cultural transformation is the recent rise of brand activism (Sarkar & Kotler, 2020). First surfacing as an academic topic only around 2017, brand activism occurs when companies take progressive stances on social issues that are politicized (Kotler & Sarkar, 2017). Examples are Nike’s advertising in support of racial justice and Starbucks’ efforts to advance marriage equality (Avery & Pauwels, 2018). Brand activism seems to be a natural consequence of the mainstreaming of CSR, which has forced companies to do more to differentiate themselves. It is not without risk, however. A drawback of being among the first to champion a social cause is the possibility of courting controversy (Mukherjee & Althuizen, 2020). In 2019, Gillette ran a series of advertisements against “toxic masculinity” that were targeted to male users of its popular shaving razors. Many observers felt that the campaign was unwise. It appeared that Gillette was lecturing its own customer base. Indeed, some customers who perceived the advertisements as an affront to their character took to YouTube to post videos of themselves throwing Gillett’s products in the garbage while narrating their own views on the state of gender relations. Nonetheless, the vanguard positions that many companies are now taking on emergent social issues suggest that many consider CSR, as it has been practiced for the past two decades, as no longer having the power to set brands apart in a sea of CSR-espousing competitors.
For Whom: The Identity-Match Thesis
This section raises an issue seldom problematized in studies of the effects of CSR on brands. That is, the constructs exist at two different levels. CSR, as denoted by the “corporate” in the acronym, applies to the corporate level. Brands, however, exist at a lower level. They are owned by corporations and do not necessarily have a one-to-one relationship with them. A single corporation can own multiple brands, or more than 400 of them, as does Unilever. The question arises, then, for which brands does (corporate-level) CSR increase (lower level) brand value? We proffer an answer below by drawing insights from the brand architecture and brand identity literatures.
The brand architecture literature starts from the observation that many companies offer a range of products and services. As such, many grapple with whether to invest their limited resources in developing a singular brand (Balmer, 2001; Hatch & Schultz, 2003) or in developing multiple brands for their individual (or subsets of their) products and services (as in a “house of brands” approach; Aaker, 2004; Hsu et al., 2016; Lane Keller, 2015). To help companies decide, the brand architecture literature seeks to define the inherent nature of branding at different levels (Rajagopal & Romulo, 2004), identify the variety of marketing linkages that can be engineered among various owned brands (Aaker & Joachimsthaler, 2000; Petromilli et al., 2002), and offer prescriptions for when one branding approach is preferable (Strebinger, 2014). Branding at the corporate level, for example, may be conducive to highlighting organization-wide human resource capabilities and the overall company culture (Konopik, 1975). Meanwhile, product branding allows companies to target their offerings to specific demographic, psychographic, geographic, or behavioral consumer segments (Muzellec & Lambkin, 2009). Where there are both corporate and product brands, the brand architecture literature suggests that there are cases, variously, where the corporate brand should be “leveraged, separated, or distanced” from the product brands (Hsu et al., 2016). Companies also have the option to design their corporate and product brands to maximize complementarities, ensuring that all draw from and reinforce a common set of cognitive associations (Uggla, 2006).
These issues centrally concern the brand identities of corporations and their products. Brand identity has at least two distinct meanings. Presently, it does not refer to a contradistinction between brand identity (how a brand intends to be perceived with regard to attributes such as brand personality or brand culture) and brand image (how a brand is actually perceived; Janonis et al., 2007; Nandan, 2005). Rather, brand identity refers to the attributes that enable audiences to identify a corporation or product and tell it apart from the competition. This narrower usage applies to Investopedia’s definition of brand identity as “visible elements of a brand, such as colors, design, and logo, that identify and distinguish the brand in the consumers’ minds.” While this conceptualization of identity is narrower than the one that refers to the intended overall character, personality, culture, and meaning of a brand, the conceptualization provides the foundation for the broader one to develop. The name of a company, as one element of the narrower usage of identity, is perhaps the primary point of attachment in consumers’ minds for a litany of brand attributes. In other words, names provide “central, distinct, and enduring attributes that constitute an organization’s [or brand’s] essential character” (Kalaignanam & Bahadir, 2013, p. 458).
Although developing a separate corporate brand identity can be beneficial (Petromilli et al., 2002), we propose that one consequence is a reduced positive effect of (corporate-level) CSR on lower level brands. This because “it seems appropriate to assume that customers associate CSR activities to the corporate brand,” rather than its various product and service brands (Bhattacharya et al., 2020, p. 2070). Identity differences among these brands may make it difficult for audiences to link a corporation’s activities to its potential noncorporate brand beneficiaries. The idea that brand names, for example, are critical attachment points for positive information is one reason that marketers advise that they should appear early and often in TV commercials (Baker et al., 2004; Hutchinson et al., 1994; Romaniuk, 2009; Zaichkowsky, 2010).
Turning to CSR, in keeping with its corporate-level nature, much CSR communication centers around the corporate identity rather than the (potentially numerous) owned brand identities (Vallaster et al., 2012). It should be underscored here that CSR communication may be necessary for obtaining CSR benefits (Du et al., 2010), because many consumers have low awareness of all the prosocial activities that companies are doing (Servaes & Tamayo, 2013). Illustrating the corporate-centered nature of CSR communication, it is the corporate name that is added to the DJSI or the lists of the “World’s 100 Most Ethical Companies” of Ethisphere. It is the corporate name that receives a rating in CSR databases such as KLD Analytics, Bloomberg’s Environmental, Social and Governance (ESG) scores, and Newsweek’s Green Ratings. It is the corporate name that is associated with participation in the UNGC or Carbon Disclosure Project. It is the corporate name, finally, that is listed on websites to indicate membership in CSR-themed associations such as the World Business Council for Sustainable Development (WBCSD) and the Boston College Center for Corporate Citizenship.
The corporate-centered nature of CSR communication may limit CSR’s branding effects where consumers have limited awareness of the brands that corporations own (Uggla, 2006). An example is the American clothier, Gap Inc. It has received numerous CSR awards, such as inclusion in the prestigious list of “100 Best Corporate Citizens” of CRO Magazine. Many consumers, however, do not know that Gap Inc., in addition to its namesake brand (Gap), owns several other large ones. These include Old Navy, Banana Republic, and Athletica. Indeed, although Gap Inc. might want consumers to associate its CSR with its portfolio brands, it may otherwise have an incentive to distance these brands from one another. This is because they target different consumer segments, ranging from casual (Old Navy) to business (Banana Republic) and sportswear (Athletica).
Method
Sample
The sample consists of all brands in the annual “Global 500” brand-value rankings of Brand Finance in the years 2007–2013. 5 A London-based brand consultancy (est. 1996) affiliated with a network of 20 branding offices worldwide, 6 Brand Finance advises clients on brand management, with specialties in brand taxation, financing, analytics, and strategy. Positioning itself as a thought leader, Brand Finance engages in a host of epistemic activities such as authoring white papers, maintaining a newsletter, and arranging conferences, workshops, and award ceremonies. Brand Finance (2009) markets its annual brand-value rankings as relevant for the calendar year at publishing, but notes in promotional materials that they are current as of the last day of the prior year (p. 1), which is the coding treatment in our models and the basis of the observation window cited above.
To gather our corporate-level variables, we linked the brands in Brand Finance’s rankings to their owners. We identified brand owners primarily from the brands’ websites (e.g., the “about us” sections), but where this was not fruitful, from queries of Google for “who owns [brand name]?” careful to confirm the results from multiple sources (news media, the corporate owner’s home page, etc.). At the corporate level, the sample includes many of the world’s largest companies, with 55% of the 2010 7 Fortune 500 represented as brand owners. Compared with the 2010 Fortune 500, sample brand-owning companies are smaller in revenue (~US$46 billion vs. US$56 billion), more frequently in the Consumer Discretionary or Consumer Staples 8 sector (~35% vs. 23%), and more likely to be U.S.-headquartered (~34% vs. 27%). 9 The sample with nonmissing data on key variables consists of 697 brands owned by 618 companies from 44 headquarter countries. The panel dataset is moderately balanced, as brands appearing in the rankings do so for an average of 3.8 of 7.0 possible years. 10
Dependent Variable: Brand Value
Brand Finance (2008, 2009) details its brand-valuation methodology and provides commentary and supplemental analytics in online “Global 500” reports. 11 Brand Finance takes a royalty-relief approach that equates brand value with the open-market cost to license a brand in a hypothetical arm’s length transaction (for extended explanation, see Brand Finance, 2018, or ISO 10688, 2010; for a comparison with other approaches, see Salinas & Ambler, 2009). This cost is a calculation of the net present value of all estimated future brand-specific royalties; is based on public data from sources such as Bloomberg, investment analyst reports, and company financials; and draws upon industry-specific norms about licensing rates in addition to benchmarks from recently observed brand transactions. Brand Finance’s market-based methodology is beneficial due to concerns that commonly gathered consumer-attitudinal measures of brand strength may not ultimately engender brand-supporting behaviors (Deng & Xu, 2017). Brand Finance is well aware of the selling point: “The attraction of our method is that it is based on commercial practice in the real world.” 12
Supporting processual validity (Hayashi et al., 2019), Brand Finance’s (2011) methodology complies with ISO Standard 10688, which “provides a consistent framework for the valuation of local, national and international brands both large and small . . . to create an approach to brand valuation which is transparent, reconcilable and repeatable” (p. 48). As for convergent validity, Brand Finance’s valuations have substantial correlations with alternative global datasets. On average across years, about 89% of BrandZ’s “Top 100 Global Brands” 13 appear in Brand Finance’s “Global 500,” with about 60% also appearing within the top 100 positions (authors’ calculations). Corresponding figures for Interbrand’s “Top 100 Best Global Brands” are 90% and 44%. For brands appearing in Brand Finance’s lists and one of these alternatives, brand-value correlations are moderately high and statistically significant (with BrandZ: ρ = .60, p < .01; with Interbrand: ρ = .64, p < .01).
Brand Finance’s valuations have ample variation, ranging in the latest data year from $104 billion (“Apple” of Apple Inc.) to $3 billion (“Bouygues” of Bouygues Telecom). Due to a right skew—for example, Apple’s valuation is about as large as that of the fifth- (Verizon, $53 billion) and sixth-ranked (GE, $53 billion) valuations combined—we apply a log transformation to render the variable more Gaussian (justifiable also on the basis of statistical tests for normality, for example, Stata’s sktest). We standardize all nonbinary variables to facilitate comparison of coefficient magnitudes within and across models. Univariate statistics appear in Table 2, the highest and lowest brands with respect to CSR and brand value are shown in Table 3, correlation statistics appear in Table 4, and variable descriptions are shown in Appendix A.
Univariate statistics and sector and country representation for sample brands.
N = 2,237. Variables appear in the order introduced in the text and are unlogged and unstandardized to convey magnitudes in intuitive terms (standardized variables would be less informative, being constrained to a mean of zero and a standard deviation of one). bSector representation and country representation are by brand-years. CSR = corporate social responsibility; DJSI = Dow Jones Sustainability Index; IV = instrumental variable; WBCSD = World Business Council for Sustainable Development; SIC = Standard Industrial Classification.
Highest and Lowest Brands With Respect to Key Variables.
Note. Brand value is in $ (in millions). As there is considerable year-to-year turnover at the very bottom of Brand Finance’s rankings and to render the data more representative of sample brand-years, the analysis above is based only on brands appearing in the sample for a majority of observation years (i.e., four or more). Similarly, for representativeness, brand value and CSR figures above are averages across all available observation years. The list of companies that do not appear in a CSR index (and therefore whose CSR indicator score is 0) is not complete; companies are presented there in alphabetical order. CSR = corporate social responsibility.
Correlation Matrix.
Note. The correlation of the CSR indicator with the scores from KLD Analytics (the number of strengths in the areas of the environment, the community, employee relations, diversity, products, corporate governance, and human rights) is .49. CSR = corporate social responsibility; DJSI = Dow Jones Sustainability Index; IV = instrumental variable; WBCSD = World Business Council for Sustainable Development.
p < .05.
Key Independent Variables
All hypotheses contained the construct, CSR, which we indicated with professional third-party ratings. While this is a very common approach, we note that these ratings have been implicated in debates about CSR’s meaning and measurement, and scrutinized for having limited agreement and external validity (Chatterji et al., 2009). Some ratings have also been criticized for their lack of granularity (many are published as binary indices rather than continuous scales), for opacity in their construction methodologies, for exorbitant costs of access to the full set of subindicators on which the published ratings are based, and for heavy reliance on quantifiable, commensurable, and public information on company policies rather than in-depth, prospective, and outcomes-based research on actual social practices (Conway, 2019). Aware of these concerns, we heeded recent advice to increase robustness by indicating CSR with multiple rather than single ratings (Chatterji et al., 2016, p. 1612). We sought, more specifically, to gather all available, global, overall CSR ratings that covered our entire observation window. We methodically applied these criteria (Appendix B) to a compilation of 18 global CSR ratings gathered in the final year of our observation window by sustainability consultant, Globescan (2013). This yielded five qualifying CSR ratings—Asset4 (ASSET4, launched 2002, headquartered in Toronto, Ontario, produced by Thomson Reuters), the DJSI (1999, New York City, S&P Dow Jones with RobecoSAM), the FTSE4Good Index (FTSE4GOOD, 2001, London, UK, FTSE Group with Ethical Investment Research Services), the “Global 100 Most Sustainable Corporations” (G100, 2005, Toronto, Ontario, Corporate Knights), and the “World’s Most Ethical Companies” (ETHICAL, 2006, New York City, Ethisphere Institute). These include three of the five ratings that Globescan (2013) finds to have the most practitioner familiarity, with the two others addressing only the dimension of environmentalism (the Carbon Disclosure Leadership Index) or not being a strict measure of CSR (Fortune’s “World’s Most Admired Companies) (p. 8). As indicated below, each of these ratings gives some assurance of little sample selection bias, as each has a target population of companies among the largest several thousand worldwide, constituting a much more comprehensive set than Brand Finance’s lists of the largest 500 brands worldwide.
Lacking space for a full review of each rating’s methodology, we point readers seeking complete background to available published materials (Corporate Knights, 2019; Ethisphere, 2019; FTSE, 2012; RobecoSAM Sustainability Investing, 2013; Thomson Reuters ASSET4, 2013). ASSET4, DJSI, FTSE4GOOD, and G100 are similar in being constructed by private professionals who conceive CSR as a reflection of a company’s overall performance across numerous constituent CSR dimensions such as the environment, society, employees, and supply chains. Within these dimensions, professionals assess CSR by dozens of annually updated quantitative indicators coded from disparate data sources such as company filings, sustainability reports, lawsuits, and media coverage. As a general summary, these ratings could each be described as “designed to reflect a broad consensus on what constitutes good corporate social responsibility globally” by taking into account a variety of objective, quantifiable corporate policies and practices (FTSE4Good, 2012, p. 2). G100, DJSI, and FTSE4GOOD are publicly released as lists of companies whose CSR is assessed, respectively, to be in the top 100, top 10%, or top 30% within large corporate samples such as all companies with revenues exceeding $1 billion (G100), the Dow Jones Global Total Stock Market Index (DJSI), or the FTSE All-Share Index (FTSE4Good). ASSET4’s ratings are percentile ranks of companies against industry peers in the areas of society, environment, and corporate governance, 14 as assessed within a global population of 3,500 15 publicly traded companies in major stock indices such as the Russell 1000 and MSCI World. As a short summary of its methodology, ETHICAL makes available an online proprietary survey that employees at several thousand companies worldwide fill out annually with self-reported data on CSR subindicators in the domains of ethics, citizenship, and governance. ETHICAL then systematically analyzes these data to calculate lists of winners that, since inception, have grown from about 90 to 140 companies.
Our CSR indicator is a count variable from 0 to 5 constructed by adding annual binary appearances in FTSE4GOOD, DJSI, G100, and ETHICAL. We then add ASSET4 after coding it as a binary variable to approximate the “best-of” lists of FTSE4GOOD, DJSI, G100, and ETHICAL by assigning a 1 to companies with scores in the top 10% of the database. 16 We use unconverted ASSET4 scores in supplemental, stand-alone regressions in Appendix C, which similarly include regressions with the other CSR ratings as sole CSR indicators. In our main models, we add together the five CSR ratings assuming that companies appearing in many have better CSR than companies in few. Adding them together also recognizes the modest inter-rating correlations (averaging .19 across statistically significant dyadic matchups), which suggests that, while the ratings measure common aspects of CSR, they also measure unique aspects that can be captured through aggregation. We note that our results are highly stable to the arbitrary deletion of any single underlying CSR rating from our count variable, which we performed as an unreported robustness check.
For H1, we code brands by owners’ headquarter regions, running separate models for the ones in our sample with enough cases to ensure adequate statistical power. These include Asia (China, Hong Kong, Japan, South Korea, and Taiwan), Europe (European Union plus Switzerland, Norway, and the United Kingdom), and North America (the United States and Canada). 17 For H1, we also code brands by the Standard Industrial Classification (SIC) sectors with enough cases to permit adequate statistical power, 18 which are D (“Manufacturing”), E (“Transportation, Communications, Electric, Gas, and Sanitary Services”), and H (“Finance, Insurance, and Real Estate”).
For H2, that CSR’s effect on brand value has diminished with time, we created a count variable that increases from one (in the first year of observation) to seven (in the final year). 19 We assume a linear trend as a baseline expectation, while discussing in our results a somewhat more complex trend that better fits our data. Finally, for H3, we indicated the identities of brands and their owners by their respective names, recognizing that names are perhaps the most common cues for audience recognition and recall of business entities (Kara et al., 2020). In other words, names are “infused with meaning to clearly communicate the identity of the corporation [or brand] to external stakeholders” (Kalaignanam & Bahadir, 2013, p. 457). We created a binary variable that defines identity dissimilarity as 1 where owners and brands have different names (e.g., “Delhaize Group” clearly differs from its sample brand, “Food Lion,” just as “Champion” clearly differs from “Hanesbrands”). We considered names to be similar if the brand name was embedded in the company name (e.g., “Exxon” and “ExxonMobil”), because these cases permit ready linguistic associations between companies and brands. This treatment also allowed us to ignore common legal suffixes such as Inc., Corp., and Ltd. We note that in cases where a company owns multiple brands, it is possible that one of those brands would still have a coding of 0 if its name is similar to the owner’s name. For example, Gap is the namesake brand of Gap Inc. (= 0), a corporation that also owns brands with very dissimilar names, that is, Old Navy, Banana Republic, and Athletica (= 1).
Control Variables
We drew controls from existing studies and from research linking the controls to CSR and brand value. For example, company size appears in similar research (Melo & Galan, 2011; Torres et al., 2012; H.-M. D. Wang, 2010) and has plausible relationships with both brand value (Janoskova & Kliestikova, 2018) and CSR (Udayasankar, 2008). We include two size indicators, revenues and assets, both in logged end-of-year U.S. dollars. Profitability is return on assets (ROA) as earnings before interest and taxes divided by total assets. Resource slackness is free cash flow, as well as leverage as total debt divided by stockholders’ equity (Campbell, 2007, p. 952; Riantani & Nurzamzam, 2015). 20 In addition to research and development (Torres et al., 2012), we controlled for advertising as annual appearances (= 1) in Advertising Age’s lists of the “World’s Top 100 Advertisers” (following Melo & Garrido-Morgado, 2012). 21 As several values of our ratio variables (i.e., ROA and debt/equity) were extreme due to a few instances of negative numerators or small denominators, we Winsorized them at cut points of 1% and 99%. Where possible, 22 Wald tests of combined joint significance (p < .05) separately justified the inclusion of three groups of dummy variables—for year, recognizing the generally upward trend in brand valuations; for SIC sector 23 ; and for the headquarter regions of Sub-Sahara Africa; Asia; Southeast Asia, Australia, and New Zealand; Europe; Central, Latin America, and the Caribbean; the Middle East and North Africa; North America; and the former Soviet countries. 24
Regression Models
The Breusch and Pagan Lagrangian multiplier test (χ2 p < .001) suggested panel rather than pooled regressions. Hausman tests (χ2 p < .01) recommended fixed-effects regressions where possible for models lacking time-invariant predictors. We implemented all models in Stata 13 with the functionalities of the xtreg and xtivreg commands. We employed cluster-robust standard errors to guard against cross-sectional heteroscedasticity and within-panel serial correlation. Right-hand-side variables received a 1-year lag to guard against simultaneity. 25 Power analyses suggested that all models have an adequate sample size. 26 Diagnostics revealed little concern of multicollinearity (variance inflation factors for noncontrol variables 27 are always much lower than the common rule of thumb of 10). Diagnostics suggested no unusual data or data with undue influence. 28
Endogeneity: IV Regressions
Endogeneity is a concern because it is plausible, not only that CSR affects brand value but also that brand value affects CSR. Strong brands may generate resources for CSR investment, and whether CSR has the capacity to improve brands may depend on prior brand characteristics such as credibility and authenticity (Alcañiz et al., 2010). The potential for reverse causality is also suggested in the “brand-insurance” literature (Minor & Morgan, 2011; Werther & Chandler, 2005), which argues that some companies use CSR to protect already established brands from adverse events. Here, a valuable brand leads to CSR, not the other way around.
We therefore estimated IV regressions to isolate exogenous variation in our CSR indicator (Stock & Trebbi, 2003), because it is plausible that this indicator correlates with the error term in our regressions, leading to biased coefficient estimates. Suitable IVs need to satisfy three requirements: sensibility (the relationship between the IVs and CSR is theoretically justifiable), relevance (the relationship is sufficiently large), and validity (the IVs affect the dependent variable (DV) primarily through the potentially endogenous regressor, as evidenced by low covariance between the IVs and the error term). Qualifying IVs were annual company membership in two international business associations, the UNGC (for extensive background on it, see Rasche et al., 2013) and the WBCSD (Pope & Lim, 2017). 29 Below, we describe these variables and systematically discuss them in relation to the three requirements for IVs.
Perhaps the flagship global CSR initiative, the Global Compact (GC), boasts over 10,000 corporate members from more than 150 countries. Members must report annually against a slate of 10 CSR principles that distill prior UN declarations in the areas of human rights, labor, natural environment, and corruption. 30 We gathered GC membership from the online GC database. The WBCSD is a CEO-led coalition of nearly 200 major multinational firms (e.g., 3M, Apple, BASF) whose mission is to “accelerate the transition to a sustainable world by making more sustainable businesses more successful.” Among its many operations, the WBCSD partners with nongovernmental groups, convenes global conferences and events around emergent social issues, works with governments on policy formation, and oversees a network of about 60 national sustainability councils. 31 We obtained WBCSD membership lists for past years by consulting archived versions of the WBCSD website in the Wayback Machine.
As for sensibility and relevance, empirical work suggests that business associations may improve members’ CSR (Ortas et al., 2015). As for the GC, Ruggie (2001) has argued influentially that it should be understood as a “learning network” through which companies, in collaboration with representatives of civil society, come to construct and ultimately adhere to CSR compliance norms. As for the WBCSD, Pope and Lim (2017) have linked company membership in it to higher performance in several CSR ratings herein (i.e., DJSI, FTSE4GOOD, and G100). As for validity, the concern here is whether the GC and WBCSD have an independent effect on our outcome variable, brand value. We expect (and Sargan tests support) that this is unlikely to be the case. The WBCSD, for example, has limited ability to generate direct branding benefits as it has a low public profile, as an elite organization focused on high-level policy engagement. The GC, for its part, expressly limits the use of its logo for branding purposes, which might help to explain its low awareness even among MBA students (Haski-Leventhal, 2013, p. 15).
Results
We first discuss the results for a general relationship between CSR and brand value before turning to the results for the contextual factors in our hypotheses. Zero-order correlations among CSR’s subindicators in Table 4 are moderately positive and generally significant (.04–.44, p < .05), suggesting a degree of commonality in what the subindicators assess. Descriptive analyses in Figure 2A show that members of the CSR subindicators always have higher average estimated brand values than nonmembers. A generally upward trend in average brand values as the overall CSR indicator increases is apparent in Figure 2B. Table 5’s random-effects (Models 1, 3, 4, and 5) and fixed-effects regressions (Model 2) report positive, significant associations between the indicators of CSR and brand value (p < .01). Controls here are always in the expected directions and are frequently statistically significant, while the R2 statistics suggest that a considerable amount of brand-value variation is explained (.21–.42). Coefficient magnitudes for random-effects regressions are interpretable as the average effect of X on Y when X changes across time and between companies by one unit. 32 This average effect in Model 1 as the CSR indicator increases by 1 is about 0.07 standard deviations in logged brand value (we note here that the unlogged standard deviation of brand value is about 7,505, per Table 2). Notably, DJSI, FTSE4GOOD, and G100, as sole CSR indicators (Appendix C), each have positive, significant associations with brand value (p < .05) that are similar in magnitude (β = .09–.12). This is not the case for ETHICAL (p < .14) and ASSET4 (p < .11). ASSET4 remains insignificant even when the more granular raw percentile ranks are sole CSR indicators (Model C1A). We expound on this heterogeneity in the “Discussion” section in relation to ongoing research on CSR ratings’ convergent validity.
Brand value by membership in included CSR ratings: (A) brand value by individual memberships in CSR ratings and (B) brand value by count of all memberships in CSR ratings.
Panel Regressions of Brand Value on the CSR Indicator.
Note. Cluster robust standard errors in parentheses; all nonbinary variables standardized. STATA does not report Wald χ2 for fixed-effects models but F statistics. CSR = corporate social responsibility; ROA = return on assets; FE = fixed effects.
p < .10. *p < .05. **p < .01. ***p < .001.
To address the potential endogeneity of our CSR indicator, Table 6 presents two-stage least-squares IV regressions. Supporting their relevance, each first-stage IV (Model 6) is a significantly positive CSR predictor. The Cragg–Donald F statistic (19.99, p < .05) is larger than not only a common rule of thumb of 10 but also the most restrictive Stock–Yogo weak identification value. Supporting their validity, the second-stage regression (Model 7) reports an insignificant Sargan statistic (0.01, p = .96). This regression also reports a significant, positive association between instrumented CSR and brand value (β = .28, p < .01). We note here that obtaining a larger β in IV regressions than in uninstrumented ones is common, occurring in about 80% of studies in a recent IV-regression meta-analysis (Jiang, 2017). Overall, the bivariate analyses as well as fixed-effects, random-effects, and IV regressions are consistent that CSR associates with higher brand value, although there is some heterogeneity in the results for specific CSR subindicators. These findings, as a general matter, serve to strengthen the argument that a major objective of CSR that has been affirmed in company surveys is generally attainable in practice.
2SLS Instrumental-Variable Panel Regressions of Brand Value on CSR Indicator.
Note. Cluster robust standard errors in parentheses; all nonbinary variables standardized. 2SLS = two-stage least squares; CSR = corporate social responsibility; DV = dependent variable; ROA = return on assets; WBCSD = World Business Council for Sustainable Development; FE = fixed effects.
p < .10. *p < .05. **p < .01. ***p < .001.
For H1, Table 7 displays subsampled regressions for the headquarter world regions and industrial sectors that are sufficiently represented in our sample. Here, there is a positive association between CSR and brand value in North America (p = .009), Europe (p = .006), and Asia (p = .054). There is also a positive association between CSR and brand value in the sectors of manufacturing (p = .018); transportation, communications, electric, gas, and sanitary services (p = .010); and finance, insurance, and real estate (p = .007). Notably, the coefficients with the smallest magnitudes are North America (β = .05) and manufacturing (β = .06). This is somewhat surprising given that many large American firms that are classified in the manufacturing sector (e.g., Nike, Coca-Cola) come to mind as being outspoken about their CSR. These results have some resonance with the “crowding out” thesis, given that Matten and Moon (2008) observe that American companies were the first to adopt “explicit CSR,” but perhaps now are experiencing dwindling results from it in an American marketplace increasingly saturated with CSR claims. Notwithstanding some differences in coefficient magnitudes, these results comport with the “convergence thesis” of Matten and Moon (2008) that companies worldwide have shifted to “explicit CSR,” a type that is conducive to branding. Furthermore, in showing consistent effects of CSR on brand value across, not only regions but also sectors, the results suggest that the shift in CSR theorized by Matten and Moon (2008) was a multidimensional one.
Panel Regressions of Brand Value on CSR Ratings by Geographic Region and Industrial Sectors.
Note. Cluster robust standard errors in parentheses; all nonbinary variables standardized. Too few sample companies in the finance industry reported R&D expenditures during our window of observation to calculate a coefficient. CSR = corporate social responsibility; ROA = return on assets; FE = fixed effects.
p < .10. *p < .05. **p < .01. ***p < .001.
Compatible with H2, the intermediate (Model 3) and full model (Model 5) in Table 5 report a significant interaction (p < .01) between CSR and time. As one might expect, due to inflation alone, the models show a positive main effect of time, suggesting that brand values have generally increased each year by about 0.11 standard deviations (p < 1). The positive main effect of CSR from Models 1 and 2 remains in Model 3 (but is larger at 0.14 standard deviations; p < 1). This suggests that CSR is generally associated with an increase in brand value irrespective of observation year. Finally, the positive and significant interaction term suggests that, in each additional year of observation, the effect of CSR on brand value decreased by about 0.02 standard deviations (p < .01). Further probing these temporal effects, unreported analyses suggest that the decline in CSR’s effects on brand value was not strictly linear. Rather, the relationship weakened most dramatically in the final years of the 2000s (2007–2010) and stabilized as the 2010s began (2011–2013). These results are consistent with the crowding out thesis that, as more and more companies adopted CSR, it became increasingly difficult for them to use CSR to obtain branding differentiation.
Corroborating H3, the interaction of CSR with dissimilar name is negatively significant in Models 4 and 5. Although the coefficient for the main effect of CSR in Model 4 is 0.08 for each unit increase in the CSR indicator, there is an offsetting coefficient for the interaction term of negative 0.11 when the names are different. This finding is consistent with the identity-match thesis proposed earlier, suggesting that there may be limitations in using CSR to bolster brands at a lower level than the corporate brand.
Discussion
In this study, we used advantageous data from Brand Finance to document several contextual factors that may condition CSR’s effects on brand value. In doing so, we strove for contributions that were both methodological and theoretical. Methodologically, our study answered calls from prior researchers to explore CSR’s branding effects with greater rigor and empirical scope. For example, we increased sample size, expanded beyond American brands, put smaller brands under analysis, gathered alternative indicators of CSR than the typical usage of the scores of KLD Analytics, included more control variables, and used IV modeling to strengthen causality. We believe that these methodological contributions, altogether, are substantial.
As for theoretical contributions, while we found evidence of a general relationship between CSR and brand value, we also sought to move beyond this established finding. In particular, we investigated where, when, and for whom this relationship might hold. As such, we contributed to a growing literature on the moderating, mediating, and boundary conditions that govern CSR’s business benefits (Q. Wang et al., 2016). In terms of where, employing a dataset whose first observation year began around the time that Matten and Moon (2008) were forming their highly influential “convergence thesis,” we provided corroborating evidence that CSR’s branding effects appear to be cross-geographic and pan-industrial. As for when, we supported an implication of this line of thinking. CSR’s effects on brand value may have diminished with time as the market became increasingly saturated with CSR claims and activities. Here, we found that the most dramatic fall-off occurred in the final years of the first decades of the 2000s. Finally, in terms of for whom, our findings supported the “identity-match thesis.” Here, we proposed that the branding effects of CSR, which is corporate level in its very terminology, do not transfer as well onto lower level brands when those brands have a dissimilar identity as that of the owning corporation, which makes it difficult for audiences to readily associate the CSR with the potential lower level brand beneficiaries.
Managerial Implications
Our findings provide some encouragement to the many businesses who report in surveys that good branding is a major objective of their CSR. Our subsample analyses suggested that this encouragement may be taken by businesses in world regions as far apart as North America and Europe and in industrial sectors as different as manufacturing and finance.
That said, our findings suggest that CSR’s branding benefits may not be as strong today as in the early and mid-2000s, when CSR was much less ubiquitous. In putting forward the “crowding out thesis,” we proposed that the diminishing effects may be because it is difficult for companies these days to use CSR to stand out in a sea of competitors all claiming to be socially responsible. A managerial implication is that, if companies want to use CSR for brand differentiation, they may consider avoiding conventional, standard, routine, or commonplace activities. Rather, they may wish to develop ones that are especially bold or cutting edge. More specifically, companies may decide to engage in brand activism (Kotler & Sarkar, 2017), be the first to address an emergent social problem, design initiatives that have an inimitable resonance with their preexisting brand and corporate identities, advertise in ways that accentuate points of difference, or practice CSR in an especially sincere and thoroughgoing manner.
Our findings also suggested that there is a potential case where the expected increase in brand values from CSR may be weaker. This was when a focal brand cannot be readily associated with its corporate owner by having a similar identity. In such cases, managers may need to devise compensatory strategies. As a simple example, companies that own many brands that produce CSR reports through such platforms as the GC or Global Reporting Initiative may take care to ensure that those reports highlight their brand offerings. They may, for example, place the brand logos on the title page or include a brand architecture diagram in the preface.
Implications for CSR Ratings
Our analyses are relevant for ongoing research on CSR ratings’ validity. We systematically gathered multiple CSR indicators fitting several basic criteria in recognition of recent questions about CSR ratings’ convergent validity and predictive power (Chatterji et al., 2016). Although our scale behaved as hypothesized, it concealed underlying heterogeneity. This heterogeneity is consistent with observations that leading CSR ratings “have common dimensions, but on aggregate, they do not converge” (Semenova & Hassel, 2015, p. 249) and that there is “a surprising lack of agreement across social ratings” (Chatterji et al., 2016, p. 1597).
Underlying CSR ratings with relatively weak relationships with brand value were ETHICAL and ASSET4, whose idiosyncrasies may be telling. Unlike the other subindicators, ETHICAL does not assign scores to each company in a target population but makes available an online survey that employees at any company can self-submit for consideration for inclusion in the annual list. The survey processing fee was $3,000 last year. ETHICAL is also unique in being created by Ethisphere, a for-profit consultancy. These features may create selection bias into ETHICAL in favor of resource-equipped companies or those that have awareness of the survey through a consulting relationship with Ethisphere. As for ASSET4, its methodology, at least during our observation window, took a “lowest common denominator approach” to CSR evaluation (Thomson Reuters ASSET4, 2013, p. 3). Scores were based only on public, quantifiable, and commensurable company information. As ASSET4 disallowed analysts from using “subjective assessments or overrides” (Thomson Reuters ASSET4, 2013, p. 3), the scores oftentimes appear impervious to corporate scandals. Thomson Reuters notes, for example, that “companies that have been involved in environmental, social or governance controversies will only find their scores affected within the pillar where the controversy occurs and even there according to objective metrics that are applied uniformly” (Thomson Reuters ASSET4, 2013, p. 3). This approach can have unseemly results. For instance, in the year after the Deepwater Horizon oil spill in 2010, after BP plc instituted a raft of new policies and charitable programs, the company actually saw its ASSET4 environmental score improve from 88.8 to 93.5. Meanwhile, BP was immediately dropped from DJSI and FTSE4GOOD, never appeared again in the G100, and saw its brand value in the Brand Finance database plunge by 28%. Such results are, perhaps, behind Thomson Reuters’ recent methodological change to “discount scores based on negative media stories” that are “significant” and “material” (Thomson Reuters ASSET4, 2018, p. 7). In pointing out these idiosyncrasies, our intention is not to present ASSET4 and ETHICAL as “worse” than other CSR ratings but simply to note several clear methodological differences that future researchers may wish to keep in mind as they continue to explore CSR ratings’ convergent validity.
Limitations and Future Research
As for limitations, we employed only a single brand-value indicator, the only one to the best of our knowledge that is publicly available, and offered the statistical power to test our hypotheses. Even this indicator, however, although advantageous in many ways compared with alternatives (BrandZ and Interbrand), limited our power to investigate CSR’s branding impacts by geography. The indicator had a small representation from world regions such as Africa and Central America, from sectors such as agriculture and mining, and from small brands outside the largest 500 in the world in value. As another limitation, our dataset was available only for the years 2007–2013. While we took advantage of this observation window to test a contemporaneous theory (and one of its corollaries) of Matten and Moon (2008), we understand that it would be helpful to confirm our hypotheses with newer data.
We hope that future researchers will use qualitative, survey, ethnographic, and experimental methods to observe consumers more directly to investigate whether the theories that we have tested in this study with relatively high-level regressions are robust to other empirical approaches. We also call for research that further examines the mechanisms, pathways, contexts, and boundary conditions that govern CSR’s branding effects. For example, does CSR improve brand value by enhancing credibility or trustworthiness, or does it strengthen broader company attributes such as favorability, status, legitimacy, and reputation? Another avenue of future inquiry is to examine whether our proposed relationships are found comparatively. With the benefit of more statistical power, future researchers can explore these relationships across industries and nations. Do industries with higher concentrations of CSR experienced diminished returns from CSR as a result of crowding out? Unfortunately, we were unable to explore such patterns because many of our subsamples had too few observations (e.g., only 340 in Asia). However, we encourage future researchers, when possible, to move the analysis from sectors to industries, and from world regions to countries.
Future researchers might also alight from here to examine more closely how and why the effects of CSR may differ for corporate-level and lower level branding approaches. How might companies with highly evaluated CSR that operate as a “house of brands” develop communication and positioning strategies such that their corporate-level social programs and policies ultimately do benefit the brands of their variously named product offerings?
We also see a large opening for future researchers to examine the CSR branding’s costs. Are the brand-value gains greater than the needed expenditures? Noteworthy here is that companies, when asked in surveys why they do not do more CSR, often cite costs as an impediment (Accenture & United Nations Global Compact, 2013, Figure 9). A cost analysis, although beyond the scope of the present article, can assess whether CSR is the best branding strategy. Costs include the opportunity costs of more traditional branding approaches that emphasize product quality or performance. Although imperfect in this and undoubtedly other respects, the present effort has done much to strengthen the existing literature by documenting that companies with highly evaluated CSR are associated with higher branding values, a finding that may serve to further the CSR movement, which holds out promise for benefiting the commonweal.
Footnotes
Appendix A
Variables Used in the Analysis.
| Variable | Source | Type | Notes |
|---|---|---|---|
| Brand value | The Brand Directory of Brand Finance | Continuous | Calendar year-end U.S. dollars, in millions, logged, lagged, standardized |
| CSR indicator | Various (see below) | Count | Count of ADMIRED, DJSI, ETHICAL, FTSE, and G100 |
| ASSET4 | Thomson Reuters Datastream | Binary | 1 for companies with scores in top 10%, 0 otherwise |
| DJSI | S&P Dow Jones | Binary | 1 = member of list |
| ETHICAL | Ethisphere | Binary | 1 = member of list |
| FTSE4GOOD | FTSE Group | Binary | 1 = member of list |
| G100 | Corporate Knights | Binary | 1 = member of list |
| Time effect | Authors’ codings | Count | = 1 in 2007, and growing by one each subsequent year |
| Dissimilar name | Authors’ codings | Binary | 1 = brand and owner do not share the same name; 0 otherwise |
| World regions | Compustat | Dummies | Dummies for the headquarter regions of North America, Asia, and Europe of the brand-owning company according to the country–region codings described in the text |
| Industrial sectors | Compustat | Dummies | Dummies for Sector D (manufacturing), E (transportation, communications, electric, gas, and sanitary services), and H (finance, insurance, and real estate) of the Standard Industrial Classification Scheme |
| IV: UNGC | UN member database | Binary | 1 = UNGC signatory |
| IV: WBCSD | Current and archived WBCSD websites | Binary | 1 = Member of World Business Council for Sustainability |
| Revenues | Compustat | Continuous | Calendar-year-end U.S. dollars, in millions, logged, lagged, standardized |
| Assets | Compustat | Continuous | Calendar-year-end U.S. dollars, in millions, logged, lagged, standardized |
| ROA | Compustat | Ratio | EBIT/total assets; Winsorized at 1% and 99%, lagged |
| Free cash flow | Compustat | Continuous | Calendar-year-end U.S. dollars, in millions, Winsorized at 1% and 99%, lagged |
| Debt/equity | Compustat | Continuous | Total debt/shareholder equity, logged, lagged, standardized |
| R&D | Compustat | Continuous | Calendar-year-end U.S. dollars, in millions, logged, lagged, standardized |
| Top global advertiser | Advertising Age | Binary | Appearance in Top 100 Global Advertiser lists |
| Sector effects (as controls) | Compustat | Dummies | The 10 sectors of the Standard Industrial Classification scheme |
| Region effects (as controls) | Compustat | Dummies | Sub-Sahara Africa; Asia; Southeast Asia, Australia, and New Zealand; Europe; Central, Latin America, and the Caribbean; the Middle East and North Africa; North America; and the former Soviet countries |
Note. CSR = corporate social responsibility; DJSI = Dow Jones Sustainability Index; UNGC = United Nations Global Compact; WBCSD = World Business Council for Sustainable Development; ROA = return on assets; EBIT = Earnings before interest and taxes.
Appendix B
Characteristics of Leading CSR Evaluation Schemes.
| CSR evaluation scheme | Available a | Globalb | Overall CSR ratingc | Covers 2007–2013d | Strictly CSR-relatede |
|---|---|---|---|---|---|
| Access to Medicine Index | Yes | Yes | Medicine only | No | Yes |
| Asset4 ESG scores | Yes | Yes | Yes | Yes | Yes |
| Bloomberg ESG scores | No | Yes | Yes | No | Yes |
| Carbon Disclosure Project | Yes | Yes | Environment only | Yes | Yes |
| Climate Counts | Yes | Yes | Environment only | Yes | Yes |
| CR Magazine’s Best Corporate Citizens | Yes | No | Yes | Yes | Yes |
| DJSI | Yes | Yes | Yes | Yes | Yes |
| Fortune’s Most Admired | Yes | Yes | Yes | Yes | No |
| FTSE4Good Index | Yes | Yes | Yes | Yes | Yes |
| INrate Sustainability Assessment | No | Yes | Yes | Yes | Yes |
| Global 100 Most Sustainable Corporations | Yes | Yes | Yes | Yes | Yes |
| Goodguide | No | Yes | Yes | No | Yes |
| GS SUSTAIN | No | Yes | Yes | No | No |
| MSCI ESG Data | No | Yes | Yes | Yes | Yes |
| Oekom Corporate Ratings | No | Yes | Yes | Yes | Yes |
| Sustainalytics Co. Profiles | No | Yes | Yes | No | Yes |
| World’s Most Ethical Companies | Yes | Yes | Yes | Yes | Yes |
| Vigeo Ratings | No | Yes | Yes | Yes | Yes |
Note. CSR = corporate social responsibility; DJSI = Dow Jones Sustainability Index.
Meaning freely available to us through our university libraries (e.g., ASSET4), online (e.g., G100, ETHICAL), or from the ratings providers (e.g., FTSE4GOOD and DJSI), rather than requiring very large access fees. bNecessary because our dataset is global, excluding, for example, CRO Magazine’s “100 Best Corporate Citizens,” which covers only American companies. cThe ratings do not concern only single CSR issue areas (e.g., the Access to Medicine Index concerns only inexpensive pharmaceutical availability). dExcluding, for example, Fortune’s “World’s Most Admired Companies,” commonly considered a measure of corporate reputation and based, at any rate, on many indicators that are not CSR-related, such as the quality of products and services. eExcluding, for example, Sustainalytics, whose historical data are available on Bloomberg only from 2009.
Appendix C
Panel Regressions of Brand Value on Individual CSR Ratings.
| Model C1A | Model C1B | Model C2 | Model C3 | Model C4 | Model C5 | |
|---|---|---|---|---|---|---|
| ASSET4 (raw score) | 0.00 (0.02) | |||||
| ASSET4 (binary coded) | 0.04 (0.02) | |||||
| DJSI | 0.09* (0.04) | |||||
| ETHICAL | 0.11 (0.08) | |||||
| FTSE4GOOD | 0.12* (0.05) | |||||
| G100 | 0.10** (0.03) | |||||
| Revenues ($ in millions, log) | 0.34*** (0.08) | 0.19* (0.08) | 0.18* (0.08) | 0.19* (0.08) | 0.08 (0.11) | 0.19* (0.08) |
| Assets ($ in millions, log) | 0.45*** (0.09) | 0.30*** (0.09) | 0.31*** (0.09) | 0.31*** (0.09) | 0.32*** (0.12) | 0.30*** (0.09) |
| ROA (Winsor) | 0.11** (0.03) | 0.08** (0.03) | 0.08** (0.03) | 0.08** (0.03) | 0.14** (0.04) | 0.08** (0.03) |
| Free cash flow | 0.02 (0.03) | 0.03 (0.02) | 0.03 (0.02) | 0.03 (0.02) | 0.08* (0.04) | 0.03 (0.02) |
| ($ in millions, log) | ||||||
| Debt/equity (Winsor) | –0.05* (0.02) | –0.03 † (0.01) | –0.03 (0.01) | –0.03 † (0.01) | 0.00 (0.03) | –0.03 † (0.01) |
| R&D ($ in millions, log) | 0.08 (0.05) | 0.10 † (0.05) | 0.09 † (0.04) | 0.10* (0.05) | 0.15* (0.06) | 0.09 † (0.05) |
| Top global advertiser | 0.07 (0.06) | 0.11 (0.06) | 0.11 † (0.06) | 0.11 † (0.06) | 0.12 (0.10) | 0.11 † (0.06) |
| Year FE | Yes | Yes | Yes | Yes | Yes | Yes |
| Industry FE | Yes | Yes | Yes | Yes | Yes | Yes |
| Region FE | Yes | Yes | Yes | Yes | Yes | Yes |
| Constant | –1.35*** (0.25) | –0.38 (0.95) | –0.38 (0.95) | –0.37 (0.96) | 1.05** (0.34) | –0.37 (0.95) |
| n | 1,933 | 2,237 | 2,237 | 2,237 | 985 | 2,237 |
| Overall R2 | .39 | .30 | .30 | .30 | .29 | .30 |
| Within R2 | .40 | .41 | .41 | .41 | .38 | .41 |
| Wald χ2 | 856.94 | 968.88 | 938.66 | 952.98 | 536.48 | 944.06 |
Note. Cluster robust standard errors in parentheses; all nonbinary variables standardized. CSR = corporate social responsibility; DJSI = Dow Jones Sustainability Index; ROA = return on assets; FE = fixed effects.
p < .10. *p < .05. **p < .01. ***p < .001.
Authors’ Note
This article received feedback from John W. Meyer, Mark Granovetter, and Xueguang Zhou of Stanford University. The article incorporates comments from Alwyn Lim and participants at several conferences, including the “2018 Summer Workshop” of Organization Studies in Samos, Greece, and an autumn 2018 research colloquium at the School of Business and Economics at the Norwegian University of Life Sciences.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
