Introduction

Firms’ corporate social responsibility (CSR) investments have reached unprecedented levels; the top 500 global companies spend a combined $20 billion annually on CSR-related activities (Thompson, 2020). Such vast investments appear warranted, considering industry reports that suggest 66% of global consumers are willing to pay a premium for sustainable brands (NielsenIQ, 2015), and 90% would switch to a brand supporting a worthy cause (Cone Communication, 2015; Havas Media Group Report, 2021).

Although academic literature on CSR initiatives similarly has expanded, its findings are less consistent than those produced by industry analysts (Acquier et al., 2017). In empirical studies of the impact of CSR initiatives that provide estimates of CSR treatment effects, we find widely divergent results. Some studies claim CSR initiatives improve consumers’ appreciation of firms or brands (e.g., Muniz et al., 2019; Sen & Bhattacharya, 2001), but others indicate mixed (e.g., Simmons & Becker-Olsen, 2006; Yoon et al., 2006) or even adverse (e.g., Dubé et al., 2017; Ginder et al., 2021) effects.

Notably, we also find conflicting results pertaining to a specific mechanism related to how CSR influences consumers, namely, the altruistic inferences that consumers make in response to a CSR initiative. In research that predicts how such inferences drive the effects of CSR initiatives, results indicate both positive (e.g., Romani et al., 2013; Vlachos et al., 2009) and negative (e.g., Lichtenstein et al., 2004; Yoon et al., 2006) effects. Yet we know of no studies that explicitly address these discrepancies, a gap that hinders theoretical progress (Ji et al., 2022; Kuokkanen & Sun, 2020; Vishwanathan et al., 2020) and creates a risk that CSR initiative research will remain marked by collections of disparate empirical results, rather than establishing a cohesive body of knowledge.

By taking an integrated view on these gaps, we propose that consumers’ inferences about firms’ altruistic motives drive the effects of CSR initiatives differently, depending on several contingency factors. Consumers tend to adopt a virtue ethics view of CSR, such that firms appear virtuous if their CSR initiatives are properly motivated (van de Ven, 2008), driven by sincere concern for the welfare of others, rather than self-interest (Grigore et al., 2021). Considering the complex effects of such altruistic inferences can provide a more cohesive view of CSR initiative literature, in that we predict they might lead to positive or negative CSR effects, depending on the situation (Parguel et al., 2011; Romani et al., 2013; Zasuwa, 2017). That is, to reconcile inconsistent study findings and consolidate extant literature, we investigate whether altruistic motive inferences represent a core consideration for research focused on consumer responses to CSR initiatives.

To answer this question, we need both a comprehensive perspective on the CSR initiative paradigm and a dialogical inquiry to confront assumptions about the role of altruistic inferences with specific cases that reflect the diverse characteristics (contingencies) of prior empirical studies. Therefore, we perform a meta-analysis of research into how CSR initiatives affect consumer responses (i.e., attitudes and behaviors directed toward the firm/brand) and seek to identify contingencies at the firm (e.g., whether the initiative is business related, philanthropy, or cause marketing), cause (e.g., environmental vs. health vs. social domains), and contextual (e.g., cultural influences) levels. We gather 337 effect sizes from 118 articles that include 162 individual studies and 584,990 unique respondents, published between 1996 and 2021. Beyond offering comprehensiveness and a summary of the research field, a meta-analysis enables us to test whether contingencies systematically affect the impact of the independent variable (CSR initiatives) on dependent variables (consumer responses), using a moderator analysis (Chan & Arvey, 2012) at the scale of the entire research domain.

Our empirical strategy is twofold. First, we test whether CSR initiatives affect consumer responses indirectly through altruistic inferences, as well as whether this indirect effect varies across firm-, cause-, and contextual-level contingencies. Second, we determine whether the effects of individual contingency factors on the CSR–consumer response link follow the pattern predicted by an altruistic inference mechanism, such that contingencies that facilitate (inhibit) altruistic inferences should strengthen (weaken) the positive impact of CSR initiatives. To ensure causal identification, we focus exclusively on studies that investigate the effects of CSR initiatives on consumer responses in comparison with a control condition, rather than correlations of CSR perceptions and consumer responses. Only such comparative investigation can isolate the impacts of actual firm CSR initiatives on consumer responses (Dekkers et al., 2019) and provide actionable findings.

In pursuing these insights, we build on prior attempts to address mixed CSR findings with conceptual reviews (e.g., Fatima & Elbanna, 2023; Fatma & Rahman, 2015; Peloza & Shang, 2011; Thomas et al., 2020; Yuan et al., 2011; Zerbini, 2017), as well as meta-analyses that focus specifically on one type of CSR initiative, such as cause marketing (Fan et al., 2022; Schamp et al., 2023), or one type of contingency factor, such as cause–company fit (Zasuwa, 2017).Footnote 1 In an effort to deal with the mixed findings at a more fundamental level, we investigate a unifying mechanism that is rooted in consumers’ perceptions of CSR ethicality.

Therefore, this study makes several theoretical and practical contributions. First, we establish altruistic inferences as a core paradigmatic belief that can bring together disparate findings across studies and provide a coherent structure for understanding consumer responses to CSR initiatives. This insight has important ethical implications regarding the role of virtue-based ethics, compared with consequentialist-based ethics, for CSR. Second, in addition to building consensus about the current paradigm (Chan & Arvey, 2012), our dialogical inquiry challenges it. Some results cannot be fully accounted for by altruistic inferences, so consumers’ expectations of how firms’ situations might influence their CSR initiatives—which have not been studied formally thus far—likely should be conceptualized as part of the underlying mechanism. Third, our findings reveal which CSR initiatives consumers appreciate most, across a range of business conditions, which can help practitioners define their firms’ engagement in CSR and foster greater societal benefits.

Theoretical Foundations and Hypotheses

Altruistic Inferences as a Paradigmatic CSR Belief

It is important to note that CSR is related but distinct from, corporate sustainability (Husted & Allen, 2007). Although many characteristics have been argued to differentiate CSR and corporate sustainability, the time dimension is the most relevant in the context of our research because firms achieve corporate sustainability through their continued long-term commitment to CSR initiatives (Hockerts & Searcy, 2023). Additionally, over time, CSR has progressively broadened to include a vast range of stakeholders, and its capacity to elicit appreciation from consumers has become especially prominent (Luo & Bhattacharya, 2006; Vishwanathan et al., 2020). The current CSR concept reflects the expectations of all stakeholders (Yuan et al., 2020), such that it encompasses co-created value (Agudelo et al., 2020). Accordingly, we adopt Chandler’s (2016, p. 248) description of CSR as “a holistic approach whereby strategic planning and core operations are managed in the interests of a broad set of stakeholders to optimize value over the medium to long term.”

Altruism also has been defined in different ways, though two common principles emerge from popular conceptualizations. First, an altruistic action must be disinterested; the actor expects no benefit, even indirectly, in return. This principle aligns with a definition from Joo et al. (2016) that altruism implies a genuine and supportive motive, rather than profit-seeking or self-serving claims. It also fits with Perera and Chaminda’s (2013, p. 245) definition of altruistic motives as “a business firm’s selfless engagements in social welfare activities.” Second, an action is altruistic if it is performed voluntarily, such that no external factors determine it (Muniz et al., 2019). For a firm’s altruistic behavior specifically, the definition by Simmons (1991, p. 3), which Guinot et al. (2016) apply in an organizational context, captures these principles: “altruism seeks to increase another’s welfare, not one’s own; is voluntary; is intentional, meant to help someone else, and expects no external reward.” Thus, we predict that voluntary CSR initiatives that appear driven by genuine concern for the interests of others likely trigger altruistic inferences among consumers (Vlachos et al., 2009).

This assertion—placing altruistic inferences as central to the CSR paradigm—also is consistent with research predictions that consumers make virtue-based ethical judgments of CSR. They favor displays of sincere rather than self-interested CSR (van De Ven, 2008). Altruism also is a dominant theme in CSR research (Ferrero & Sison, 2014; Romani et al., 2013; Zasuwa, 2017); more than 60% of the articles eligible for inclusion in our current meta-analysis offer conceptual arguments rooted in altruistic inferences to explain the impact of CSR initiatives. A main consequence of virtue-based ethics is that consumers regard CSR initiatives as less ethical if they are not altruistically motivated (Acquier et al., 2017), regardless of their societal outcomes (Ferrero & Sison, 2014). This theoretical research approach reflects consumers’ sense that firms are more accountable to “do good” than to “do well” (Wall, 2021). For example, more US consumers assert that the primary responsibility of firms is social (49%) rather than economic (37%) (Stobierski, 2021).

Meta-Analysis as a Dialogical Ethical Approach

In proposing that research on consumer responses to CSR initiatives reflects a paradigmatic belief in altruistic inferences that drive the impact of CSR across a range of contingency factors, we define a paradigm as “a framework of beliefs, values, and techniques that informs researchers in their generation, development, and application of knowledge” (Chan & Arvey, 2012, p. 82). Because it integrates empirical literature in a field, a meta-analysis can determine if important questions have been sufficiently addressed by a paradigm and thus foster normal science (Chan & Arvey, 2012).

No methodology is ethically neutral, so we also consider the ethics of a meta-analytical method to help clarify its potential contribution to business ethics. For example, a meta-analysis devoted to stocktaking or documenting overall effect sizes in a field could be characterized as a casuistic approach. Casuistry is the “process by which one can deliberate over moral decisions without needing to employ explicit moral principles and theories” (Lurie & Albin, 2007, p. 204). A purely casuistic meta-analysis could exclude paradigms. In Derry and Green’s (1989) typology of ethical theory inquiries, a casuistic view of meta-analysis represents a minimalist perspective, emphasizing mainly analytical techniques while overlooking beliefs and values. But meta-analysis also can yield contributions beyond stocktaking, such as by identifying contingencies of the main effect of interest (Steel et al., 2021), which can increase the precision of a paradigm, broaden its scope, or build consensus about it (Chan & Arvey, 2012). Therefore, the ethical theory inquiry underpinning our meta-analysis more closely reflects what Derry and Green (1989) qualify as dialogical:

Refusing to accept a simple distinction between theoretical or applied ethics, it proposes a unitary inquiry marked by an active dialogue between theoretical investigations and the analysis of concrete issues or cases. The ultimate aim of this inquiry is to test and enhance our theoretical sophistication. (p. 531)

With such a dialogical approach, we develop and test the meta-analytical framework in Fig. 1, reflecting a theory of CSR ethics from consumers’ viewpoint. In this framework, altruistic inferences constitute the fundamental paradigmatic belief that consumers leverage to react to firms’ CSR initiatives, according to multiple contingency factors. In keeping with a dialogical inquiry, we confront this theory with specific cases of CSR initiatives in prior literature, and we estimate the potential mediating effect of altruistic inferences across the contingency factors.

Fig. 1
figure 1

Meta-analytical framework

Full size image

Thereafter, we test whether the individual effects of contingency factors are consistent with altruistic inferences. By focusing on distinct CSR initiatives, this approach explicitly addresses the adequacy of altruistic inferences as a theoretical principle and their application to specific ethical situations. Our predictions about the directions of the moderating effects are predicated on whether contingency factors facilitate altruistic inferences, in such a way that stronger altruistic inferences generally improve consumer responses. Effects inconsistent with this posited theoretical ethical principle require reconsideration or refinement, in line with a dialogical approach (Derry & Green, 1989). In contrast, a lack of effect would indicate that a casuistic approach to the ethics of CSR initiatives is more valid, because no generalizable ethical principles emerge from particular cases (Lurie & Albin, 2007).

Mediation by Altruistic Inferences

Attribution theory guides our predictions of how altruistic inferences drive consumer responses to CSR initiatives. It provides a strong conceptual basis for making predictions about the conditions that lead to stronger or weaker altruistic inferences, because its fundamental premise is that observers (i.e., consumers) try to understand why actors (i.e., firms) behave as they do by inferring the nature of their motivations, depending on the circumstances (Heider, 1958). Furthermore, attribution theory is widespread and well-accepted as a theoretical foundation for the mechanisms underlying the effect of CSR initiatives on consumer responses. It has appeared in many prior studies, as demonstrated in the bibliometric analysis of CSR literature by Ji et al. (2022).

By default, observers tend to infer that actors’ behaviors are motivated by their inner dispositions and reject this default assumption only if evidence signals that actors behave in reaction to external circumstances, in which case the observers infer situational motives (Kelley, 1973). Conceptually, this evidence should reflect one of three general principles: Behaviors that are consensual (i.e., observed among other actors as well), lack consistency (i.e., not consistently observed over time), or are distinct (i.e., only observed in specific contexts) are more conducive to situational attributions, because they suggest compliance with normative expectations, weak commitment, or opportunism, respectively (Kelley, 1973). In the absence of sufficient alternative evidence that the situation influences the behavior, dispositional motives appear more likely.

When they encounter CSR, consumers thus draw causal inferences by making either dispositional attributions (the CSR initiative is aligned with the firm’s true values) or situational attributions if there is evidence that warrants doing so (the CSR is motivated by the firm’s self-interest) (Ginder et al., 2021). Dispositional attributions prompt inferences consistent with defining components of altruistic behaviors, which in turn should evoke more favorable consumer responses (Ellen et al., 2006; Vlachos et al., 2009). That is, consumers likely perceive that the firm’s CSR is not driven by the expectations of receiving something in return, and it results from a voluntary decision (Guinot et al., 2016). In contrast, situational attributions lead consumers to infer that the firm expects to reap economic benefits (Gautier & Pache, 2015), which garners less favorable reactions (Choi et al., 2016). In turn, we predict that the mediating effect of altruistic inferences is contingent on whether conditions are conducive to dispositional or situational CSR motive attributions. Formally:

Hypothesis 1

The mediating effect of altruistic inferences in driving the impact of CSR initiatives on consumer responses is stronger (weaker) in conditions favoring dispositional (situational) attributions.

Moderating Effects of Contingency Factors

To integrate ethical theory and specific ethical cases of CSR initiatives (Derry & Green, 1989), we also investigate the moderating effect of firm-, cause-, and contextual-level contingencies (Fig. 1) that may influence altruistic inferences, according to attribution theory. Although it does not provide direct evidence of mediation, a process-in-moderation approach establishes more valid causal claims about the role of altruistic inferences in the CSR–consumer response link (Spencer et al., 2005) while also providing substantive insights about each contingency factor. It leverages a large set of effect sizes too. Consistent with recent business ethics meta-analyses (Fan et al., 2022; Schamp et al., 2023; Smith et al., 2023), we use this approach to control for methodological characteristics and publication biases.

Firm-Level Contingencies

Contingencies at the firm level include characteristics related to the types of CSR initiatives, their consistency, the hedonic versus utilitarian nature of the firm offering, and whether the firm is controversial and engages in corporate social irresponsibility (CSiR).

Types of initiatives. Firms engage in various CSR initiatives, such as business practices, philanthropy, and cause marketing (Ellen et al., 2006; Peloza & Shang, 2011). These types of CSR initiatives generally have been examined independently (Öberseder et al., 2013), despite their varying potential influences on altruistic inferences. Business initiatives relate directly to the firm’s operations, such as sustainability in the production process (Choi & Ng, 2011), fair trade (Carlsson et al., 2010), or employee welfare (Peloza et al., 2015). Thus, they are associated with ethical causes but also with the fulfillment of economically relevant objectives—like increased profits from efficiencies, competitive advantages, and greater market shares (van Rekom et al., 2014)—and with mandated compliance with regulations or industry norms (Perez-Batres et al., 2012). Thus, firms’ motivation for business initiatives may appear mostly self-interested and constrained, which would limit altruistic inferences. In contrast, philanthropy and cause marketing likely trigger altruistic inferences. Philanthropy refers to contributions to causes purely for the betterment of society, without being linked to sales targets (Lii & Lee, 2012). That is, philanthropic initiatives do not aim to benefit the firm (Ginder et al., 2021), and there is no external pressure to engage in them, likely making them appear disinterested and voluntary. Cause marketing refers to short-term sales promotion mechanisms, during which the firm donates to a cause for each product sold (Fan et al., 2022). Although the sale proceeds benefit the firm, engaging in cause marketing is more likely to result from the firm’s volition, whereas regulations and norms (i.e., situational motives) are more influential on business initiatives (Christensen et al., 2021). In summary, business initiatives should be the least favorably evaluated by consumers, because they do not demonstrate a lack of concern about the firm’s own interests and genuine care for the interests of others, nor do they represent an unconstrained decision compared with philanthropy and cause marketing.

Hypothesis 2

The impact of CSR initiatives on consumer responses is stronger for cause marketing and philanthropy than business-related types of CSR.

Consistency: Ongoing Versus One-Off

Firms may implement recurring (ongoing) CSR initiatives or opt for one-off actions (Yuan et al., 2011). In contrast to one-off actions, ongoing CSR actions indicate long-term support for a cause (Porter & Kramer, 2006). According to Kelley (1973), such consistency signals that the action is motivated by disposition rather than situational influences, leading to altruistic inferences (Choi et al., 2016; Parguel et al., 2011). As a case in point, Youn and Kim (2018) find that donation duration is positively linked with altruistic inferences. Thus:

Hypothesis 3

The impact of CSR initiatives on consumer responses is stronger for ongoing compared with one-off initiatives.

Hedonic Versus Utilitarian Offering

Utilitarian consumption is predominantly cognitively driven, instrumental/functional, and goal-oriented, whereas hedonic consumption is focused on attaining multisensory, emotional experiences (Dhar & Wertenbroch, 2000). Consumers evaluating utilitarian offerings are more likely to make dispositional attributions (Sen & Lerman, 2007). Because of the congruence between the cognitive nature of the attribution mechanism and the cognitive mode of evaluation of utilitarian offerings (Zhao et al., 2011), consumers in need of utilitarian offerings are less likely to engage in cognitive elaboration to find alternative evidence that supports situational motives (Kelley, 1973). Therefore, CSR initiatives may be less favorably received if the firm’s offering is principally hedonic rather than utilitarian.

Hypothesis 4

The impact of CSR initiatives on consumer responses is weaker for hedonic versus utilitarian offerings.

Firm Controversy

Allegations of corruption, breaches of privacy, unfit products, or misleading marketing are examples of CSiR that stir controversy around firms. These controversial, unethical practices can hurt various stakeholders (Swaen et al., 2021). Although it may still be beneficial for such firms to engage in CSR, rather than remain passive, to seek or facilitate forgiveness (Valor et al., 2022), these benefits are likely mitigated, compared with those that might be gained by uncontroversial firms. Multiple studies show that consumers interpret CSR actions by controversial firms as insincere, self-serving, and constrained (Swaen et al., 2021; Yoon et al., 2006). They thus might appear as distinctive actions (Kelley, 1973), conditional on the firm’s specific situation. Consequently, we predict that altruistic inferences are limited for controversial firms, whose CSR seems situationally rather than dispositionally motivated:

Hypothesis 5

The impact of CSR initiatives on consumer responses is weaker for controversial than for non-controversial firms.

Cause-Level Contingencies

Empirical investigations of CSR initiatives span different domains and varying levels of fit with the firm’s products.

Cause Domain

Causes pertain to various domains, such as environmental efforts to improve the natural environment, health-focused attempts to facilitate a healthy society, and human services or social causes that advance social structures (Zasuwa, 2017). Consumers may view non-human causes as less self-serving than human causes (Ginder et al., 2021), because they more readily assume the latter are being exploited for firms’ benefits (Sabri, 2018), to the detriment of altruistic inferences. Thus:

Hypothesis 6

The impact of CSR initiatives on consumer responses is stronger for non-human (e.g., environmental) than human (e.g., health, social) causes.

Cause Fit

The notion of perceived fit refers to the similarity between a cause and the firm’s product category or brand image/positioning (Berens et al., 2005). Cause fit is high if congruence exists between the cause supported by the CSR initiatives and the attributes of the firm’s products (e.g., Simmons & Becker-Olsen, 2006), as when Starbucks gives away free reusable cups to consumers to limit the waste associated with disposable cups. Although the effects of cause fit on consumer responses to CSR initiatives remain contested (Fan et al., 2022; Kuokkanen & Sun, 2020), from an attribution theory perspective, a close alignment between the cause and the firm should make the benefits more salient (Yoon et al. 2006) and heighten skepticism about the sincerity of the firm's motives (Ellen et al., 2000). Formally:

Hypothesis 7

The impact of CSR initiatives on consumer responses is weaker for high than low cause fit.

Contextual-Level Contingencies

Cappelli and Sherer (1991, p. 56) define the context as “factors associated with units of analysis above those expressly under investigation.” In our case, the contexts in which CSR initiatives are embedded include the firm’s industry, the cultural orientation of the country where they occur, and the time period during which they occur.

Stigmatized Industry

Companies operating in stigmatized industries, defined as those “whose products or manufacturing processes negatively influence society and generate public concerns” (Lee & Cho, 2022, p. 1), might seek to improve their image by engaging in CSR. For example, British American Tobacco (2018) invests £60 million annually in sustainable farming practices. Consumer scrutiny of companies from such stigmatized industries tends to be greater than that for firms in non-stigmatized industries (Lee & Cho, 2022) and may lead to allegations of CSR hypocrisy (Wagner et al., 2009). When their suspicion is evoked regarding a firm’s CSR motives, consumers tend to make situational motive attributions, impeding altruistic inferences (Ginder et al., 2021).

Hypothesis 8

The impact of CSR initiatives on consumer responses is weaker for companies in stigmatized compared with non-stigmatized industries.

Culture

Prior CSR initiative studies have been performed in dozens of countries, representing various cultures. To capture how altruistic inferences vary across cultures, we use Inglehart and Baker’s (2000) framework, which stipulates that a country’s cultural orientation is organized around two main dimensions: traditional versus secular-rational and survival versus self-expressive. We consider Inglehart and Baker’s framework particularly well adapted to our study, because it accounts for the evolving nature of culture (Tung & Verbeke, 2010), as it progressively moves toward more self-expressive and secular-rational values (Beugelsdijk & Welzel, 2018). This feature is essential to examine cultures across the four decades spanned by extant CSR literature. Fittingly, these cultural values get updated regularly (i.e., seven waves of research in the past 40 years). A crux of Inglehart and Baker’s (2000) thesis is that cultures evolve toward more democracy because economic development increases autonomy, rationality, and self-affirmation, which in turn elevate free choice as an essential value. Cultures that value free choice—leaning toward the secular-rational and self-expressive ends of the two dimensions—likely favor dispositional rather than situational attributions. Pervasive free choice in a society implies that behaviors reflect the true inclination of actors because external constraints do not determine them, consistent with the democratic notion that individuals are the main causal agents in their lives (Nisbett, 2014). Therefore, rational-secular and self-expressive cultures should be more conducive to altruistic inferences, because they encourage observers to attribute the cause of actors’ behaviors to dispositional rather than situational motivations. Formally:

Hypothesis 9

The impact of CSR initiatives on consumer responses is stronger among consumers from (a) secular-rational than traditional and (b) self-expressive than survival cultures.

Time

Firms have embraced CSR more extensively in the past decade; only 20% of S&P 500 companies published sustainability/responsibility reports in 2011, whereas 92% did so in 2020 (Governance & Accountability Institute, 2021). Considering these rising norms surrounding social responsibility (Bertels & Peloza, 2008), CSR initiatives likely appear more and more common, and according to the consensus principle of dispositional attributions (Kelley, 1973), this trend should hinder altruistic inferences.

Hypothesis 10

The impact of CSR initiatives on consumer responses weakens over time.

Methodology

Data Collection

We present a flow diagram of the data collection process in Web Appendix 1. We employed several resources to identify relevant papers. First, we performed extensive searches on journal aggregators, research platforms, and academic databases—Google Scholar, EBSCOhost, ProQuest, and EBSCO Open Dissertations—using keywords such as “CSR,” “corporate social responsibility,” “cause marketing,” “cause-related marketing,” “philanthropy,” “sustainability,” “environmentally friendly,” and “green marketing.” The references of the initial set of papers obtained through these searches were checked to identify other likely sources of data. Second, we retrieved papers included in systematic CSR reviews (e.g., Peloza & Shang, 2011). Third, to complement these search efforts, we conducted manual searches of leading journals that regularly publish about CSR (European Journal of Marketing, International Journal of Research in Marketing, Journal of the Academy of Marketing Science, Journal of Business Ethics, Journal of Business Research, Journal of Consumer Psychology, Journal of Consumer Research, Journal of Marketing, Journal of Marketing Research, Journal of Retailing, Management Science, and Marketing Science). Fourth, we sent a request for published and unpublished manuscripts through the electronic distribution list of the American Marketing Association. Fifth, we sent individual emails to 80 scholars in the CSR domain who had recently published in leading journals, requesting their unpublished manuscripts. These procedures yielded 1001 papers for potential inclusion.

Inclusion Criteria

To determine the final sample, we applied several criteria to these 1001 papers (see Web Appendix 1 for details). They must investigate the effect of CSR initiatives on consumer responses directed toward the firm or related entities (e.g., brands, products) that can be clearly linked with the initiative. We thus included studies that measured consumer responses such as brand attitudes, attitudes toward the firm, behavioral intentions, or behaviors. To test for the mediating effect of altruistic inferences, we also included investigations of the impact of CSR initiatives on altruistic inferences or the relationship between altruistic inferences and consumer responses (Web Appendix 2 details the indicators of altruistic inferences).

However, we excluded any studies that featured tests of outcomes other than consumer responses, such as financial performance or felt emotions (e.g., pride, guilt) (Kim & Johnson, 2013), which are not directly targeted toward the firm initiating the CSR. Other reasons for exclusions included papers without quantitative data, that developed a scale, that were not about CSR, or that did not provide the information needed to calculate the effect sizes or determine the country of the respondents. In addition, we only included studies that adopted experimental or quasi-experimental designs and contrasted a treatment condition (e.g., presence of a CSR initiative, high-intensity CSR initiative such as zero carbon emissions, high sustainability signage) with a control condition (e.g., absence of CSR initiative, low-intensity CSR initiative such as low carbon emission, low sustainability signage). These studies provide more precise tests of the causal link between CSR initiatives and consumer responses than cross-sectional studies can (Metelli & Chaimani, 2020).Footnote 2 Moreover, cross-sectional research generally offers insufficient quality to be included in meta-analyses when experimental trials are available, because their design does not reliably establish the impact of interventions (Dekkers et al., 2019). Thus, we did not include cross-sectional studies.

Finally, no exclusion criterion related to the publication year, but only papers published in English were included. We removed two outliers with standardized residuals greater than 2.57 (Viechtbauer & Cheung, 2010). In total, 118 articles, reporting on 162 individual studies, spanning four decades from 1996 to 2021 and representing a total of 584,990 unique respondents,met our criteria for inclusion. We identify 315 effect sizes pertaining to the impact of CSR initiatives on consumer responses, including 40 from unpublished papers (13%). The mediation analysis also includes the effect sizes for the relationships between CSR initiatives and altruistic inferences and between altruistic inferences and consumer responses, bringing the total number to 337.

Effect Size Calculation

Because we focus on experimental and quasi-experimental studies, we use Hedges’ g (g)—the standardized mean difference between treatment and control groups—as the primary effect size measure. It removes the potential bias in small sample studies and provides a more precise estimate than d (Borenstein et al., 2021). We also account for measurement error in the dependent variables when appropriate (Hunter & Schmidt, 2004). (We performed the main analyses with the error-corrected effect sizes and confirmed their robustness using raw effect sizes.) The effect sizes were coded according to a positive frame so that we could integrate positive and negative variables (e.g., the negative impact of a CSR initiative on intentions to switch brands is coded as a positive effect). Web Appendix 3 details the effect size calculations and formulas.

Moderator Coding

The coding scheme (definitions and operationalizations) for the substantive moderators, as well as for our methodological and publication bias controls, is in Web Appendix 4. Binary moderators are contrast-coded (Abraham & Hamilton, 2018), multicategory moderators are dummy-coded, and continuous moderators are mean-centered. The moderators and the effect sizes were double coded independently, and the intercoder reliability of 87.5 to 100% indicates strong agreement. Any disagreements were resolved through discussion.

The methodological controls include the type of consumer responses, student sample, randomization of participants to conditions, between- or within-subject design, control group without CSR or with lower CSR intensity, real firm, and scenario. Publication bias arises because only studies with statistically significant results tend to be published and are thus more likely to be included in the meta-analysis (Sterne & Egger, 2001). To mitigate this potential issue, our meta-regression model includes three controls. First, we consider whether the study was published (DeSimone et al., 2021). Second, we note journal quality, according to whether the article appears in a journal listed in the Financial Times 50 list; top-tier journals often require stronger effects (Borenstein et al., 2021). In a robustness check, we also control for journals assigned a score of 4 or above by the Chartered Association of Business Schools. Third, we include effect size precision, defined as the inverse of the standard error of g (Stanley & Doucouliagos, 2012). In the presence of publication bias, less precise studies, with smaller sample sizes, may yield stronger effects (DeSimone et al., 2021).

Results

Distribution of Effect Sizes and Assessment of Multicollinearity

The 315 individual effect sizes of CSR initiatives’ impact on consumer responses range from g = −1.486 to 1.900 (M = 0.276; median = 0.302; SD = 0.484). More than 50% of them lie in the 0.10–0.50 interval, representing effects of small to medium strength. We depict the frequency distribution of these effect sizes in Web Appendix 5. The grand mean effect size shows that, on average, CSR initiatives have a small positive impact on consumer responses (Hedges’ g = 0.288, 95% confidence interval = 0.227; 0.348). The mean is not informative, however, because of the substantial heterogeneity of effect sizes (Borenstein et al., 2021), reflecting fragmentation in the literature and the need to identify contingency factors (i.e., moderators). We reject the null hypothesis of a common effect shared among all studies (Cochran’s Q = 3,397.21, df = 314, p < 0.01). The between effect size variance stemming from moderators (i.e., τ2 = 0.119) represents almost all the observed variance (I2 = 96.92%). Furthermore, the 95% prediction interval is very wide (PredInt95% [−0.344; 0.919]), indicating that effect sizes from the same populations as those studied can fall anywhere between g = −0.344 and 0.919.

Multicollinearity is not at play in the data. First, all the moderators have low variance inflation factors; the highest is 2.05, far below the cut-off value of 10. Second, as the correlation matrix in Web Appendix 6 indicates, the highest pairwise correlation is −0.62 (between brand attitude and behavioral intentions). Although this finding does not indicate multicollinearity, because it involves two categories of the same moderator (Allison, 2012; Sinapuelas et al., 2015), we show subsequently that the results remain robust when we exclude behavioral intentions from the model. No other correlations exceed 0.6.

Moderated Mediation

With a moderated mediation analysis, we test whether the role of altruistic inferences, as the underlying mechanism, is a paradigmatic belief in CSR research. We estimate a model in which the mediating effect of altruistic inferences, enabling the impact of CSR initiatives on consumer responses, is moderated by whether the contingencies (i.e., moderators) are favorable to dispositional or situational attributions. If our dialogical approach to CSR ethics is valid, there should be some reliable contingency factors in the mediation, such that the structural path from CSR initiatives to altruistic inferences should be stronger (weaker) in conditions favorable to dispositional (situational) motive attributions.

We use the meta-analytic structural path approach suggested by Bergh et al. (2016) and estimate the mean pairwise correlations between constructs with the meta-analysis to form a correlation matrix. A meta-analytic correlation matrix, built for each set of conditions (i.e., favorable to dispositional or situational motive attributions), estimates the mediation model. The dispositional attributions correlation matrix includes the effect sizes of CSR’s impact in conditions that we predict are conducive to altruistic inferences (e.g., utilitarian product offering, non-stigmatized industry, support of non-human causes such as the environment). The situational attributions correlation matrix instead includes effect sizes in conditions less conducive to altruistic inferences (e.g., hedonic products, stigmatized industry, culture leaning toward traditional). We test Hypothesis 1 in a multigroup analysis, by assessing if the structural path from CSR to altruistic inferences is stronger for the dispositional mediation model than for the situational mediation model (see Web Appendix 7).

The mediation model fits the data very well in both conditions (dispositional: χ2 = 7.45, df = 6, p = 0.282, confirmatory fit index [CFI] = 0.99, normed fit index [NFI] = 0.96, adjusted goodness-of-fit index [AGFI] = 0.99, root mean residual [RMR] = 0.043, root mean square error of approximation [RMSEA] = 0.013; situational: 4.92, df = 3, p = 0.178, confirmatory fit index [CFI] = 0.98, normed fit index [NFI] = 0.95, adjusted goodness-of-fit index [AGFI] = 0.99, root mean residual [RMR] = 0.052, root mean square error of approximation [RMSEA] = 0.032). The CSR–altruistic inferences path is positive and significant in the dispositional condition (b = 0.191, p < 0.001) but negative and significant in the situational condition (b = −0.258, p < 0.001), suggesting a positive indirect effect in the former case and a negative indirect effect in the latter. When we constrain the CSR–altruistic inferences path to be equal in both conditions, we confirm that they significantly differ. That is, the fit with the data worsens when the path is constrained rather than unconstrained (χ2 = 7.45, df = 6, p > 0.281 versus χ2 = 78.93, df = 7, p < 0.001; CFI = 0.99 versus 0.61, NFI > 0.96 versus 0.60, AGFI = 0.99 versus 0.94, RMR = 0.043 versus 0.106, RMSEA = 0.013 versus 0.084). Therefore we can reject the null hypothesis that both models fit the data equally well (Δχ2 = 71.49, p < 0.001). In support of Hypothesis 1, the mediating effect of altruistic inferences on the impact of CSR initiatives on consumer responses is stronger in conditions favorable to dispositional rather than situational attributions.

Meta-Regression

A meta-regression model must satisfy two main requirements. First, we need to model any systematic variance between effect sizes due to moderators. Second, we must account for the possibility that the 315 effect sizes are not independent, because they span 156 individual studies from 113 papers. With a multilevel mixed-effect model (Van den Noortgate et al., 2015), we can assume that the retrieved effect sizes are randomly drawn from a population of “true” effect sizes that vary depending on the moderators (i.e., fixed effects). This model also includes random effects at the study and paper levels, so the variance is partitioned according to nesting within papers. We apply the Metafor package in R for our analyses (Viechtbauer, 2010), with restricted maximum likelihood as an estimator, because of its greater efficiency compared with alternatives (Langan et al., 2019). We thus estimate the effect size from study j in paper z using the following model:

Level 1: \({\text{ES}}_{{\text{jz}}} = \beta_{0{\text{jz}}} + {\text{e}}_{{\text{jz}}},\)

Level 2: \(\beta_{0jz} = \theta_{00z} + \sum_{(k = 1)}^K {\beta_k X_{(k,jz)} + vjz},\) and

Level 3: \(\theta 00z = \gamma 000 + \sum_{(l = 1)}^L {\beta_{(K + l)} Z_{(l,lz)} + \psi z}\), where ejz is the sampling variance, υjz \(\sim \) N(0, \({\upsigma }_{\upupsilon }^{2}\)) is a random-effect estimate of the variance between studies nested within papers, and ψz \(\sim \) N(0, \({\upsigma }_{\uppsi }^{2}\)) is a random-effect estimate of the variance between papers. The study-level explanatory variables, denoted as \({X}_{k,jz}\), include cause marketing, philanthropy, other initiatives, strategic consistency, hedonic–utilitarian, controversial firm, general domain, health domain, social domain, other domain, cause fit, stigmatized industry, behavior, firm attitude, behavioral intentions, student sample, and effect size precision. The paper-level explanatory variables, denoted as \({Z}_{l,lz}\), include traditional–secular-rational, survival–self-expressive, time, between-subject design, control group, randomized experiment, real firm, scenario, journal quality, and publication status.

Model 1 in Table 1 includes the effects of the moderators in Fig. 1; Model 3 adds interaction terms to identify second-order contingencies of the cause fit moderator (we provide the rationale for these analyses subsequently). We rely on standard exploratory meta-analytic procedures to test interactions between cause fit and each of the other theoretical moderators, one at a time, then add those that are significant (Bijmolt et al., 2005). Models 2 and 4 are reduced versions of Models 1 and 3, respectively, from which we exclude the methodological and publication bias controls. These models provide more degrees of freedom per parameter to ensure that overfitting is not an issue. We find no evidence that reduced and full models fit the data differently; the null hypothesis of equal fit cannot be rejected, according to Akaike’s information criterion (main effect models: p = 0.147; interaction models: p = 0.268). Thus, overfitting is unlikely. We also discuss the predicted mean effect sizes for each category of the significant moderators, obtained while we held the other moderators in the model at their average values across all the papers in the meta-analysis (Bijmolt et al., 2005).

Table 1 Meta-regression results for CSR initiatives and consumer response
Full size table

Firm-Level Contingencies

We set business initiatives as the reference category to test the effects of the type of initiative. In support of Hypothesis 2, business initiatives are not as appreciated by consumers (predicted \(\overline{{\text{g}} }\) = 0.211) as cause marketing (predicted \(\overline{{\text{g}} }\) = 0.397, b = 0.186, p = 0.042) or philanthropy (predicted \(\overline{{\text{g}} }\) = 0.356, b = 0.145, p = 0.055). Consistently, the “other” types of initiatives generate better consumer responses than business initiatives, though without reaching statistical significance (predicted \(\overline{{\text{g}} }\) = 0.471, b = 0.260, p = 0.111). Models with alternative reference categories indicate that cause marketing and philanthropy are not significantly different, nor are their comparisons with other types (ps ≥ 0.457). In terms of consistency, consumer responses tend to be more positive when CSR initiatives are ongoing rather than a one-off effort, though this effect is not significant (b = 0.094, p = 0.128), so we cannot confirm Hypothesis 3. Consumers respond less favorably to CSR initiatives from firms whose offering is principally hedonic (predicted \(\overline{{\text{g}} }\) = 0.208) rather than utilitarian (predicted \(\overline{{\text{g}} }\) = 0.387, b = -0.089, p = 0.003), in support of Hypothesis 4. There is no significant difference between non-controversial and controversial firms (b = 0.111, p = 0.115), in contrast with our prediction in Hypothesis 5.

Cause-Level Contingencies

With “environment” as the reference category for the cause domain moderator, we find that environmental initiatives trigger more favorable consumer responses (predicted \(\overline{{\text{g}} }\) = 0.405) than social initiatives (predicted \(\overline{{\text{g}} }\) = 0.260, b = −0.144, p = 0.039). Similar trends arise for the health (predicted \(\overline{{\text{g}} }\) = 0.259, b = −0.145, p = 0.105) and other (predicted \(\overline{{\text{g}} }\) = 0.259, b = −0.145, p = 0.211) domains, though they do not reach significance. No other comparisons of domains yield significant differences (ps ≥ 0.295). These results provide partial support for Hypothesis 6. However, the lack of significant difference between high and low cause fit (b = −0.021, p = 0.486) does not support hypothesis 7. Considering the ongoing debate about this moderator, we subsequently explore some moderating factors of this effect through interactions.

Contextual-Level Contingencies

In support of Hypothesis 8, CSR initiatives performed by companies in stigmatized industries garner less favorable consumer responses (predicted \(\overline{{\text{g}} }\) = −0.411) than those in non-stigmatized industries (predicted \(\overline{{\text{g}} }\) = 0.352, b = −0.382, p < 0.001). In contrast with Hypothesis 9a, consumers in cultures leaning toward the secular-rational end of the traditional–secular-rational dimension do not react better to CSR initiatives than consumers in cultures leaning toward the traditional end (b = 0.100, p = 0.194). As we predicted in Hypothesis 9b, consumers in cultures that value self-expression are more favorable to CSR initiatives than consumers in cultures that value survival (b = 0.116, p = 0.004). Finally, over time, CSR initiatives meet with less favorable consumer reactions (b = −0.019, p = 0.014), in support of Hypothesis 10.

Moderators of the Cause–Fit Contingency

The effect of cause fit is contingent on two other moderators (Table 1, Model 3). First, whether companies operate in a stigmatized industry alters the impact of fit (b = 0.148, p = 0.025). As we show in Fig. 2, Panel A, in non-stigmatized industries, CSR initiatives are well-received regardless of fit (high fit: predicted \(\overline{{\text{g}} }\) = 0.326; low fit: predicted \(\overline{{\text{g}} }\) = 0.350). In stigmatized industries though, low fit leads to negative consumer reactions (predicted \(\overline{{\text{g}} }\) = −0.634), and CSR has no impact when the fit is high (predicted \(\overline{{\text{g}} }\) = −0.065). Second, whether the firm is controversial provides another moderator (b = −0.092, p = 0.009), as depicted in Fig. 2, Panel B. Fit does not make a difference for non-controversial firms (high fit: predicted \(\overline{{\text{g}} }\) = 0.311; low fit: predicted \(\overline{{\text{g}} }\) = 0.260), whereas for controversial firms, low fit is superior (predicted \(\overline{{\text{g}} }\) = 0.599) to high fit (predicted \(\overline{{\text{g}} }\) = 0.281).Footnote 3

Fig. 2
figure 2

Interactions. Panel A Industry stigmatization and cause fit. Panel B Firm controversy and cause fit

Full size image

Control Variables

We find that CSR initiatives affect behavioral intentions and firm attitudes more positively than they do brand attitudes (p < 0.001 and p < 0.01, respectively). Control groups that do not feature a CSR initiative yield marginally stronger effects (p = 0.096) than control groups featuring a lower intensity initiative. No other methodological characteristics are significant (ps ≥ 0.123). Similarly, the publication bias controls are not significant (ps ≥ 0.201), indicating that this bias is unlikely.

Robustness Checks

We conducted several additional analyses to assess the reliability of the results. First, we ran several diagnostic tests of publication bias, including the fail-safe N (Rosenthal, 1979), Egger’s regression (Egger et al., 1997), contour-enhanced funnel plots (Peters et al., 2008), and trim-and-fill analysis (Duval & Tweedie, 2000). In keeping with the non-significant publication bias controls in the meta-regression, these diagnostics indicate that publication bias is unlikely (see Web Appendix 8).

Second, we estimated several models to test for robustness to alternative specifications, in which we (1) leave the two outliers in the sample, (2) use the raw effect sizes and their variance instead of correcting for measurement error, (3) control for journal quality by considering top journals from ABS category 4 and above, (4) add GDP growth as a country-level control, (5) include the percentage of female respondents, (6) include the mean age of the participants, and (7) exclude behavioral intentions from the type of consumer response due to its −0.62 correlation with brand attitude. All the results are analogous to the main models (Web Appendix 9 provides the full details).

Discussion

Our contention that the CSR research field forms a paradigm of inquiry in which altruistic inferences have a central role is supported by the tests of our hypotheses, which we summarize in Table 2. Two different analyses support this conclusion. First, with a meta-analytic structural path model, we confirm the mediating role of altruistic inferences and the nature of the attributions that lead to such inferences. According to the moderated mediation analysis, CSR initiatives strengthen altruistic inferences in conditions that facilitate dispositional attributions, but in conditions that facilitate situational attributions, CSR initiatives reduce altruistic inferences, contributing negatively to consumer responses.

Table 2 Summary of findings
Full size table

Second, with a process-in-moderation approach, we find largely convergent evidence about the role of altruistic inferences. As expected, in most cases, contingencies corresponding to dispositional (vs. situational) attributions—cause marketing and philanthropy (vs. business) initiatives, firms with utilitarian (vs. hedonic) offerings, environmental (vs. social) initiatives, self-expression (vs. survival) cultures, and earlier (vs. later) periods—yield more positive impacts for CSR initiatives. Although the positive impact of initiatives that are ongoing (vs. one-off) or that take place in cultures with stronger secular-rational (vs. traditional) values do not reach statistical significance (p = 0.128 and p = 0.194, respectively), their effects are directionally consistent with our predictions.

Ethical Implications

Our research demonstrates that altruistic inferences are a core belief of the CSR initiative research paradigm, which implies that consumers adopt virtue lenses to judge the ethicality of firms engaging in CSR. This insight has profound implications for CSR ethical theory and research on the societal impact of CSR.

Virtue Ethics as a Basis for Judging CSR Initiatives

Our study provides important insights into the field of virtue ethics in consumer studies of CSR. First, we demonstrate that a more virtuous approach to CSR initiatives can benefit firms, in terms of consumer responses. These findings align with burgeoning evidence of the positive consequences of virtue-based ethics CSR for firms. For example, Chakrabarty and Bass (2015) find that institutions adopting a virtue-based ethics approach to microfinance mitigate their risk to a greater extent than firms adopting a consequentialist (i.e., maximizing benefits for society as a whole) or a deontological (i.e., conforming to societal rules, norms, or regulations) approach to ethics. Although few empirical investigations focus on the effects of a virtue-based ethical approach, and most studies remain conceptual, the comprehensiveness of the meta-analysis supports estimates of cross-cultural effects that are critically underresearched in virtue ethics domains (Karimova et al., 2020).

Second, we advance debates about the plausibility of whether ethical principles can drive “effective altruism.” According to Singer (2015), effective altruists believe that the wealthy ought to do more to prevent suffering, and they choose to support socially responsible organizations that deliver the most surplus benefits possible for society (beneficence principle), by applying reason and evidence. On the one hand, the mediating role of altruistic inferences, as predicted by attribution theory (Kelley, 1973), is strongly aligned with a view of consumers as effective altruists who behave as rational agents and observe firms to determine which ones are most beneficial for society. On the other hand, the paradigmatic belief status of altruistic inferences challenges Singer’s (2015) notion that consumers must support organizations that deliver the most surplus good. At the very least, it is necessary to recognize that the purity of firms’ intentions to engage in CSR is important for consumers, which does not support Singer’s (2015) claim that beneficence is the only ethical principle.

Third, CSR literature tends to focus on the business outcomes of CSR initiatives, at the expense of social outcomes (Du et al., 2023; Islam & Greenwood, 2021), perhaps due to an outdated but persistent conception that business is strictly an economic activity. Yet as growing recognition realizes, business is an important societal institution that can determine well-being (Greenwood & Freeman, 2017). We thus call for research that examines the societal consequences of CSR more systematically (Du et al., 2023) as well as how CSR initiatives that are best received by consumers—as evidenced in the current research—can best be leveraged by firms to manage the triple bottom line, which consists of the social, environmental, and economic dimensions (Carmine & De Marchi, 2023). We also provide some guidance for such efforts, in revealing that consumers take a virtue ethics view of CSR rather than a purely consequentialist view, whereby CSR ethicality is strictly a function of the magnitude of its positive consequences for society (Kuokkanen & Sun, 2020; McWilliams & Siegel, 2001; Sharp & Zaidman, 2010; van de Ven, 2008). That is, a virtue ethics perspective should not be overlooked by researchers investigating the societal consequences of CSR and the role of CSR in transitioning toward corporate sustainability (Hockerts & Searcy, 2023), who may be tempted to adopt a purely consequentialist ethical perspective.

Consolidation of Paradigmatic Ethical Beliefs

Discrepant findings about the impact of CSR on consumer responses have jeopardized the emergence of a cohesive research field (Fan et al., 2022; Kuokkanen & Sun, 2020; Vishwanathan et al., 2020). Our meta-analysis addresses this apparent fragmentation by showing that altruistic inferences can unify prior findings, by acknowledging firm-, cause-, and contextual-level contingencies that facilitate dispositional or situational attributions (Heider, 1958). Building consensus about the paradigm of CSR initiatives also should facilitate more programmatic research (Chan & Arvey, 2012) focused on altruistic inferences. Existing frameworks based on altruistic inferences rarely apply the theoretical underpinnings of attribution theory either. Formalizing the central role of altruistic inferences might guide research efforts toward normal science (Kuhn, 1962). Furthermore, attributions consist of a vast theoretical apparatus—including fundamental principles of consensus, consistency, and distinctiveness (Kelley, 1973)—which might account for an even broader set of contingency factors than those examined in current research, as we propose in our research agenda.

Refinement of Paradigmatic Beliefs

Several unexpected findings emerge, suggesting the need to refine extant theoretical explanations with a dialogical approach (Derry & Green, 1989). We propose integrating, into an attributional account of altruistic inferences, a key construct that has not been theorized by CSR literature, namely, consumer expectations.

Our results do not support predictions that high-fit CSR is better received than low-fit CSR and that CSR is more effective for non-controversial firms. We posit that expectations can shed light on these results, in that our exploratory interactions reveal that the impact of fit varies depending on whether the industry is stigmatized and whether the firm is controversial. Cause fit does not influence consumer responses to CSR initiatives in non-stigmatized industries or by non-controversial firms (Fig. 2). Still, in stigmatized industries, a low fit produces adverse consumer reactions, whereas high fit is neither beneficial nor harmful. Stakeholders might expect firms from stigmatized industries to engage in CSR initiatives linked to their industry, if their motivations are sincere (Bertels & Peloza, 2008), so low-fit CSR initiatives might seem situationally motivated, because they do not match this expectation, which weakens altruistic inferences. In addition, low fit is more beneficial than high fit for controversial firms, which also might reflect expectations. Stakeholders often expect controversial firms to engage in CSR self-interestedly, to compensate for their irresponsibility (Valor et al., 2022). If their CSR initiatives feature poor-fitting causes—which are less likely to be interpreted as a strategic compensation for CSiR, because these causes are not within the firm’s primary domain—controversial firms behave more ethically than their situation seemingly dictates, which could improve altruistic inferences.

Therefore, we propose that altruistic inferences should be treated as a paradigmatic belief and also that researchers examining CSR initiatives should entrench consumer expectations in relation to firms’ situations into their theories. For example, expectations are key antecedents of motive attributions (Gilbert & Malone, 1995). To be clear, we do not merely claim that firms should exceed consumer expectations to garner more favorable opinions. Rather, we suggest integrating a key tenet of attribution theory into the CSR paradigm, according to which consumers’ assessment of how much a given situation shapes the behavior of firms depends on their expectations about the extent to which the situation influences their behavior. We anticipate that expectations relate to the power of situations rather than to normative firm behavior.

Practical Implications

This meta-analysis also suggests new ways of thinking about CSR for different stakeholders. Noting the paradigmatic belief status of altruistic inferences and the importance of accounting for consumer expectations, firms can benefit more from their CSR if consumers do not anticipate that the environment pressures them into engaging in CSR. This finding has profound consequences: If consumers are aware of regulations from governments or pressures from nongovernmental entities that constrain firms to be more sustainable or more pro-social (Carroll, 1979; Ferrero & Sison, 2014; Muniz et al., 2019), they may discount the ethical value of CSR initiatives. For example, consumers may be more receptive to cause marketing and philanthropy than business-related initiatives because the latter appear driven mainly by industry norms or regulations (Christensen et al., 2021). Generally, our findings indicate that firms should investigate how their customers expect different situational circumstances to influence their CSR.

Regarding specific contingency factors that determine CSR effectiveness, we find that contextual-level moderators are particularly strong. Although CSR impacts weaken over time, the dynamic nature of Inglehart and Baker’s (2000) cultural framework indicates that as societies progress toward more self-expressive values, the effectiveness of CSR should also improve. A region witnessing strong growth in self-expressive values (e.g., increasing democratic values, autonomy, expressions of uniqueness) should become more receptive to CSR too. A meta-analytic approach is an effective method for uncovering time and cultural impacts, whereas it is difficult to test, in a primary study, for the effect of these factors at the same scale as our research, which spans 25 years of empirical literature from 28 countries.

Because the effect of cause fit depends on firm controversy (CSiR) and industry stigmatization, we also illustrate some practical consequences of these two-way interactions between moderators with simulations, to clarify the net effects, using the Model 3 results (Table 1). We simulate the impact of CSR initiatives on consumer responses for eight scenarios represented by the combinations of the three moderators. To yield relevant predictions, we set the year to 2021 (which accounts for 43 effect sizes in our dataset) and define consumer responses as behavioral. The other moderators are set to their average sample values (Bijmolt et al., 2005). As the full results in Web Appendix 10 reveal, when the firm is not controversial and is not from a stigmatized industry, high (predicted \(\overline{{\text{g}} }\) = 0.242) and low (predicted \(\overline{{\text{g}} }\) = 0.220) fit yield equivalent positive behavioral responses, of a small magnitude. In stigmatized industries, high fit insulates the firm from negative reactions, whether or not it is controversial, such that the effect is null in both cases. But the effect of low fit can differ radically, depending on firm controversy and the industry. In a stigmatized industry, a low fit for non-controversial firms yields medium-sized negative reactions (predicted \(\overline{{\text{g}} }\) = −0.765) but is inconsequential if the firm is controversial. In a non-stigmatized industry, a non-controversial firm can obtain significantly higher positive responses with low fit (predicted \(\overline{{\text{g}} }\) = 0.559).

Limitations, Research Agenda, and Conclusion

Although we contribute to a better definition of the normal science paradigm for CSR initiatives, the current research has several limitations. Due to insufficient data, we could not delve into the relative effects of different initiatives in the business category, such as corporate volunteering, fair trade, diversity, or employee relations (Peloza & Shang, 2011). When more primary studies address specific business initiatives, finer-grained integration efforts will be possible. Cause fit generally refers to the level of congruence between the CSR issue and the attributes of the company’s products (Simmons & Becker-Olsen, 2006), but it also might arise from congruence between the CSR initiative and the company’s brand image, mission, or target market (e.g., Lee et al., 2012). Insufficient data also prevented us from assessing the impacts of cause fit beyond its standard definition, but continued studies might explore how different types of fit affect responses to CSR. In addition, we focused on altruistic inferences, which is the most widely cited explanatory principle among researchers in the CSR–consumer response paradigm (about 60% of studies) and also is a well-defined form of altruism (André et al., 2017). A given paradigm can accommodate several theoretical explanations though (Kuhn, 1962). Other cognitive drivers such as self-serving motives (Kim & Choi, 2018) or affective drivers such as gratitude (Romani et al., 2013) might enable CSR effects. Although we could not test whether other drivers of CSR effects operate alongside altruistic inferences, due to insufficient data, we hope further research examines this possibility as more mediation evidence accumulates.

Beyond addressing these limitations, we propose a research agenda, dedicated to continued investigations of contingency factors at the firm, cause, and contextual levels (Table 3). We also propose ideas for advancing toward a theory of CSR effects. Critics claim that because it operates from within the market system, CSR stabilizes rather than addresses the root causes of issues. It would be helpful to investigate how consumers react to transformative CSR that aims to eliminate the causes altogether (Nickel & Eikenberry, 2009).

Table 3 Agenda for future research directions
Full size table

Finally, we recommend research into the role of consumer expectations about the capacity of the situation to influence firm behavior. If consumers expect industry norms to pressure firms into engaging in CSR, does this expectation affect their appreciation of individual firms’ CSR? Do they systematically attribute situational motives to situational demands (Gilbert & Malone, 1995)? We hope our meta-analytic efforts to build a comprehensive conceptual framework centered on altruistic inferences and suggestions for continued research inspire researchers to refine the current paradigm and thereby improve managerial practice regarding CSR initiatives.