Carrots or sticks? Inferring motives of corporate CSR Engagement from empirical data

What drives Corporate Social Responsibility (CSR) engagement of firms? Most empirical literature on firms’ CSR engagement focuses on benefits resulting from CSR engagement which are granted by the public. Other studies argue that firms are driven by fear of negative public reactions. This raises, firstly, the question, whether rewards or punishments drive CSR engagement of firms. Secondly, how can we find out, whether benefits or punishments are more relevant? This study starts out from the observation that CSR is defined by going beyond statutory prescriptions monitored and enforced by legal institutions. Existing literature argues that firms orient themselves at public expectations regarding CSR, which are presumed to increase over time. But the specific drivers and mechanisms of this increase are yet unclear. Based on simple behavioral assumptions, we derive developments of CSR engagement for the overall population of firms and sub-populations. These developments differ in crucial aspects, depending on whether firms engage in CSR because they expect benefits or punishments. Comparing empirical data on CSR engagement at the population and sub-group level from different economic regions with implications derived from the behavioral assumptions indicates that in the US, the public exercises pressure by imposing punishments rather than by offering rewards, while this is not the case in other economic regions.


Introduction
To retain their "license to operate", firms are ever more in need of public goodwill.What started out as business activity in compliance with the law increasingly expanded into a more active role of firms as corporate citizens from which pro-active behavior in the domain of corporate social responsibility, CSR, is expected (Carroll 1999;Hess et al. 2002;Orlitzky et al. 2003;Bertels & Peloza 2008;Soppe et al. 2011).Substantial research efforts are concerned with the causes and mechanisms driving this long term development.This study supplements existing research by inquiring into whether the CSR engagement of firms is driven by their hopes for benefits or their fear of punishment imposed on them by the public.The crucial question is, what happens, if the firm does nothing?Does a firm's CSR engagement lead to benefits, which are simply absent, if the firm does nothing?Or, is the firm punished for doing nothing?The dominating view in the literature is that firms engage in CSR and are rewarded -by the public at large, investors, and customers -for doing so.Implicit to this view is that if firms do not engage, the status quo persists -in the dominating view, there are no "punishments" for non-engagement.This dominating view is criticized the part of the literature and moreover not consistent with the perceptions of practitioners from firms' management.Existing research does not address this question, moreover, existing methods are not able to answer this question.
In terms of the motivation for CSR engagement, some studies see an intrinsic motivation at work, which expresses itself in a strong role for the management staff and its ideological orientations regarding CSR as drivers (Hemingway & Maclagan 2014;Hafenbrädl & Waeger 2017), the moral motivation in the classification of Aguilera et al. (2007).This is even more the case in firms where directors are politically appointed (de Andres et al. 2022).Managers, like the society as a whole, are assumed to undergo a value change with sustainability and CSR becoming more relevant (Davis 1960), and as a consequence, they act upon these values.This intrinsic view does not refer to short-or long-term economic advantages or trade-offs but sees managers as engaging in CSR for personal reasons, at times against the interests of the firm's owners (Cheng et al. 2013).
Most studies, however, argue that CSR engagement is beneficial for firms, allowing them to reap manifold benefits (Kurucz et al. 2008), the instrumental and relational motivation in the classification by Aguilera et al. (2007).Implicit to this instrumental view is the assumption assumes that CSR engagement pays off in the end: firms engage in CSR in order to achieve, ultimately, higher financial performance (Aupperle/Van Pham 1989; Plewnia/Günther 2017, Huang 2021), or better performance in the stock market (Eccles et al. 2014;Pfister et al. 2020;Serafeim 2020).These advantages are also central elements in firms' justification of their -often costly -CSR engagement (Peifer/Newman 2020).As for how exactly these benefits arise, the business case is more indirect.By engaging in CSR, firms achieve a better relationship with stakeholders, who are, in the end, also potential investors, employees, business-partners, and customers (e.g.Berman et al. 1999;Brammer/Pavelin 2006;Tang et al. 2012, Chen et al. 2015;Burke et al. 2019;Michelon et al. 2019).Benefits can arise within the firm, e.g., it is argued that the organizational changes involved in dealing successfully with CSR issues also improve the firm's overall organizational abilities, which increases financial performance (Eccles et al. 2014;Flammer 2015;Schwens/ Wagner 2019), or help the firm to cope with crises and shocks (Epure 2022), as firms with a strong focus on sustainability weather shocks better.But most benefits originate in the firm's environment: shareholders by showing a preference for investing in firms with a strong CSR record (Starr 2008;Cheah et al. 2011;Kordsachia et al. 2021) but also stakeholders outside the domain of actual economic relationships are seen as driving CSR engagement of firms (Dawkins/Lewis 2003;Aguilera et al. 2007).Regarding the firm's public standing, CSR engagement is seen as a means to increase public reputation (Doh et al. 2010;Du et al. 2010;Fernandez Sanchez et al. 2015;Zavyalova et al. 2016) which will ultimately pay off in some form or other, be it in relations to consumers (Sen/Bhattacharya 2001), who are more actively calling for CSR engagement and favor firms with a strong CSR engagement (Yang et al. 2021) or by being able to recruit CSR-minded employees, (Turban/Greening 1997).Central to the instrumental view is that firms invest in CSR to realize benefits and basically make a trade-off which is fully equivalent to other investments.While the socio-political environment -which differs among countries and economic models -may affect the role and relevance of stakeholders for CSR engagement, (Steurer 2010; Ioannou/Serafeim 2012, Brown/Steen Knudsen 2015; Desender/Epure 2021), the explanatory mechanism of stakeholders driving the firm remains in place.Despite this strong focus, the issue of whether stakeholders use benefits or punishments as a means to drive firms' CSR engagement is not addressed.
The state of the art on the drivers of CSR engagement can be summarized as follows: Many mechanisms driving CSR engagement are discussed.But despite long standing research efforts, issues of causality in the relationship between CSR engagement and economic performance (McWilliams/Siegel 2000;Flammer 2015;Lin et al. 2019) persist, as CSR engagement may also be an expression and signal of financial strength (Lys et al. 2015).There might also be a publication bias for studies finding positive relationships of CSR with various economic indicators (Rost/Ehrmann 2017).In particular, the actual micro-level mechanisms by which CSR engagement translates into economic performance are not fully understood (Aguinis/Glavas 2012, Barnett/Salomon 2012;Dixon-Fowler et al. 2013;Kang et al. 2016).However, there is a widely shared assumption that CSR engagement is driven by the rewards society at large, specific stakeholders, or the marketplace bestow on firms engaging in CSR (Margolis/Walsh 2003;Orlitzky et al. 2003;Aguilera et al. 2007;Doh et al. 2010), with most of the research sharing the basic assumption that society rewards CSR engagement.

An alternative conception of the mechanisms driving CSR engagement
These highly diverse approaches to CSR engagement share some methodological and ontological features.Methodologically, the factors driving CSR engagement of specific firms are seen as exogenous in the explanations proposed.Regardless of whether the drivers considered are stakeholders or properties of the firm, whether objective features or perceptions of features are considered, there is a cut between the driv-ers and the firm's reaction in terms of CSR engagement.There are neither feedback loops nor is a system-level dynamic included in the analysis, where the current level of CSR engagement in the firm population affects future engagement of firms in the population.Ontologically, the focus is on advantages of CSR for firms: What firms do, is presumed to be driven by their expectations of rewards for doing something, not by fear of punishments for doing nothing.These assumptions can be questioned.From doing so, a different methodology follows.
Regarding the system-level dynamic of CSR engagement, observers and practitioners alike acknowledge the evolution of societal demands imposed on firms (Davis 1960;Hess et al. 2002).In the instrumental view of CSR engagement, this would result in higher benefits for CSR engagement, increasing the (monetary) payoff of CSR engagement and thus higher levels of CSR engagement.By investing, e.g., in a sustainable supply chain, certain costs arise, but these lead to benefits, which are either directly of monetary nature or can be valued in monetary terms, see Karpoff et al. (2005) or Kurucz et al. (2008).The return -in terms of customers, reputation etc. -for having a sustainable supply chain increases as societal demand for sustainability increases.This trade off holds true regardless of what the public expects firms to do or what other firms do.The decision is seen as a unilateral trade-off individual firms make.However, considerations on the relationship between the level of CSR engagement a firm chooses and the benefits arising show a disconnect.The general public and specific stakeholders are expecting evermore CSR engagement from firms and there is substantial evidence that these expectations matter for firms' choices (Aguilera et al. 2007).True, the higher demand increases the returns of CSR investments, which is in line with the predominating instrumental view.But the raising expectations concern also the status quo.Studies dealing with CSR reporting show the relevance of meeting stakeholder expectations (Nason et al. 2018).It is not sufficient to have, in objective terms, a strong CSR performance (Rindova et al. 2005).As Brammer/Pavelin (2006) point out, a firm's CSR activities have to match the public's expectations and concerns regarding the firm's activities.The returns of CSR investment not only depends on the firm's investment per se, as the instrumental view implicitly assumes, but on public expectations.Blatant violations of public expectations typically come in the form of a CSR-related crises, their impact on the firm's reputation and how firms can respond, have been widely discussed already (e.g. for many, Karpoff et al. 2005;Cho 2009 andHahn/Lülfs 2014).But the importance and consequences of standing out or to under-perform in terms of "everyday" CSR are studied in much less detail and the evaluation differs.For instance, Barnett/Salomon (2012) conclude from their studies that firms should either engage only little in CSR, which increases their financial performance, as low CSR engagement saves money, or very much, as outperforming other firms incurs substantial advantages which offset the costs of CSR.In this view, doing as others do does not pay off.Just as in other economic areas, the returns of an investment, here, CSR engagement, depend on what other firms do.This inter-firm interaction was also seen by other studies: firms orient themselves at their peer groups when deciding about CSR engagement (Marquis/ Tilcsik 2016).CSR activists' pressure not only affect the targeted firms, but other firms as well (Briscoe et al. 2015) and firms are evaluated in relative terms to all other firms/their peer group, in particular in the case of negative events (Zavyalova et al. 2012).Regarding public scrutiny for CSR issues, existing research investigated into the effects of public exposure (Brammer/Millington 2006) or the media (El Ghoul et al. 2019) on CSR engagement, finding that firms respond to higher scrutiny with more CSR engagement, also indicating that public expectations matter.To conclude, there is a strong case that firms are oriented at meeting expectations.What remains open is, what consequences follow from this.
Regarding the question of whether CSR engagement is driven by hopes for benefits or fears of punishment the public imposes, the former view clearly dominates, often without considering the alternative (see the review in Rost/Ehrmann 2017).Basically the argument, made in the literature reviewed in the introduction, is that since there is a strong societal demand for CSR, producing this output confers, ultimately financial, advantages on the firms.But there is also the view that firms engage in CSR out of fear of negative consequences imposed on them by the public.Already McGuire et al. (1988) pointed out, that CSR engagement is used as a means to avoid costly legal consequences of CSR-related problems and as an insurance in the case of acute problems, relating to CSR or other issues (see Minor/Morgan 2011 and Epure 2022 for similar arguments).CSR engagement is also done to avoid negative consequences, like public campaigns or consumer reactions (Russell et al. 2016).These are not only reactions to acute CSR-related crises, but also to a constant under-performing of public expectations.Two developments also support this view: on the one hand, the public is ever more watchful regarding CSR issues (Bansal 2005;Blevins/Ragozzino 2019;El Ghoul et al. 2019).This scrutiny is exerted by traditional media, business databases, which include CSR reporting in their coverage, and from individuals and NGOs active in social media.Firms can be compared in terms of CSR engagement more easily.On the other hand, the vulnerability of firms to activities of social and individual actors, for instance by social media campaigns, is higher than ever, a fact already stated by Dawkins/Lewis (2003).The managerial relevance of the benefits vs. punishment issue arises from the fact that benefits as opposed to punishments define different baselines for the situation that the firm does nothing.
Considering CSR engagement as driven by interdependent expectations and a mix of rewards and punishments results in several issues and different mechanisms.
First, on what basis are rewards and punishments determined?We argue that what matters is no longer compliance with legal rules, but compliance with expectations: compliance regarding CSR differs from compliance in other themes, e.g., financial mis-conduct, in that while there are legal rules on CSR-related conduct of firms, mere law-abidance is insufficient for avoiding negative consequences.Going beyond what the law stipulates as an obligation is the defining criterion of CSR (Carroll 1998;Carroll/Brown 2018;European Commission 2011).Firms not just have to be lawabiding, but good corporate citizens.Further, CSR compliance differs from compliance in a narrower sense in that, e.g., while compliance with financial regulations is monitored by public institutions, like the SEC or ESMA, CSR in a broader sense is not monitored by public institutions but by the public itself: while there are also institutions in charge of enforcing CSR-related laws, say, the EPA in the environmental domain, firms can run into major and even existential CSR-related problems without violating any law.Utilities from the energy sector may serve as a case in point.The business model of burning coal to produce electricity is legal, but in some societies no longer seen as socially responsible and legitimate behavior.Just as CSR is essentially defined as engagement beyond and above legal prescriptions (basically the core of the European Commission's definition), it is also essentially judged beyond and above legal rules.Which is so say, CSR compliance is related to meeting or not meeting the public's expectations (Brammer/Pavelin 2006).We argue that the defining characteristic of firms' CSR standing is to be perceived by the public as (not) meeting the public's expectations.Basically, a CSR-related compliance problem concerning a firm arises from the interplay between the expectations of the public, the public's sensitivity regarding CSR, and firm's factual CSR engagement.
Second, how do public expectations come about?Here, we argue that the public's expectations reflect the CSR engagement all firms show.Just as with the establishment of social institutions (Berger/Luckmann 1966), common and typical behavior constitutes behavioral expectations, which in turn drive actors' behavior, perpetuating and institutionalizing the expectations.What the public expects, depends on what firms do.What firms do, depends on what the public expects.Instead of focusing on exogenous drivers of firms' CSR engagement, like the demands imposed on them by specific stakeholders, we argue that firms themselves, by their choices, affect public expectations about the appropriate level of CSR engagement and thus drive their own future behavior.
Third, how do public expectations drive CSR engagement of individual firms?Public expectations become relevant for firms, when the public detects divergences between firms' behavior and its expectations and reacts by imposing rewards or punishments on the firms.We argue that based on its expectations, the public detects and sanctions CSR under-performers as well as CSR out-performers, which causes firms to adapt their CSR engagement in response the public's reactions.CSR engagement is not driven by one side, be it firms or the public, but by the interaction of both sides.Thus, as for the behavioral foundations of population-level CSR engagement, firms are, on the one hand, in an interaction with the public, which monitors and sanctions their CSR engagement.On the other hand, firms are in a prototypical red queen situation (Barnett/Hansen 1996;Barnett 2008) with all other firms, where benefits and costs are determined by their relative position in the population of firms in terms of CSR engagement.
This conception of CSR engagement has several implications.For a firm to run into a major CSR problem, two preconditions have to be met: first, there must be a divergence between the firm's CSR engagement and the public's expectations regarding the appropriate level of CSR engagement for this firm.Second, the public must be able to detect this difference.Once the mismatch between expectations and factual CSR engagement is detected, the public responds, depending on whether the expectations are missed, met, or exceeded.Depending on the public's dealings with CSR, rewards are bestowed or punishments imposed, both of which also reflect the degree of mismatch.This interactive perspective puts the focus on specific elements of the relationship between firms and the public: whether the public responds more by rewards or more by punishments, and the role of the public's ability to detect firms whose CSR engagement diverges from expectations.
Using simple behavioral assumptions, we derive how CSR engagement in a population of firms and the CSR engagement of individual firms evolve under various assumptions about how the public detects and responds to CSR engagement of individual firms relative to the public's expectations, and how firms with various levels of costs for CSR engagement behave in such a situation.Comparing the implications of different constellations with the empirical development of CSR engagement in a firm population allows to infer whether rewards or punishments predominate in the population considered.

The firm-public-interaction
Regarding the firm-public-interaction, we use different assumptions about the conditions of interest: the predominance of public rewards or punishments and the public's ability to detect CSR out-/under-performance.Assuming rational behavior on the side of the firms we derive, how a public, which punishes CSR under-performance, rewards CSR over-performance, or is not interested in CSR at all, drives firms' behavior and thus the development of CSR engagement of firms and the firm population.

The firms: costs and benefits
We do not include an intrinsic motivation for CSR engagement (as opposed to Hemingway & Maclagan 2014 or Cheng et al. 2013), but assume that firms decide about their level of CSR engagement rationally, based on costs and benefits.Firms observe the public's reaction to their current CSR engagement and adapt in response to these reactions but also based on firm-specific considerations of CSR-related costs and benefits.As we explicitly consider benefits and punishments, we allow for costs of engagement and non-engagement alike.
1) Costs of CSR engagement arise in many forms.For instance, from choosing suppliers, which are more environmentally-oriented but more expensive, from social sponsoring, from setting up a monitoring of compliance in the supply chain etc. Depending, notably, on the firm's sector, engagement in CSR is more or less costly.Take as an example firms providing internet-based services, which have a much more limited domain in which they have to deal with CSR-related problems.As they are not producing physical objects, they have no supply chain from which CSR related problems might arise like problematic labor conditions in the apparel industry or problematic origins of raw materials in electronics.
2) Costs of CSR non-engagement arise due to being perceived to be a CSR underperformer and the public critique or reputational costs associated with this (Karpoff et al. 2005;Rindova et al. 2005, Rothenhoefer 2019).This public critique can come in many forms, ranging from activists' campaigns against a specific product or against the firm's overall business model.Reputational costs of a negative image among customers and business partners arise in terms of losing existing business, but also in business, which is not obtained.
Benefits of CSR engagement are frequently mentioned in the literature.They may be very direct, as in the case when investing in being more sustainable actually saves costs in the future, or indirect, in form of intangible benefits like legitimacy (Fombrun et al. 2000), or, more concrete, a positive public image and the even more tangible benefits associated with it (Peloza et al. 2012), which range from being able to sell products at a premium, to attract additional customers who value CSR engagement (Sen/Bhattacharya 2001;Russel et al. 2016), to recruiting employees (Turban/Greening 1997) to being able to wheather acute crises better because a high CSR reputation shields the firm from severe consequences like loosing market share, reputational damages or even litigation by NGOs active in the domain concerned (Minor/Morgan 2011, Fauser/Utz 2021, Epure 2022).While the mechanisms are more complex, involving reputation, the investment in CSR is functionally comparable to an insurance fee, where relatively small costs are borne to limit potential damages, often of unknown magnitude.
Given the variety of costs and benefits, modeling each type of costs and benefits is beyond the scope of the paper, but not actually required either: while costs of CSR engagement are quantifiable in practice, the diffuse costs of CSR non-engagement are much harder to quantify, especially from an outsider's perspective.The same holds true for benefits, which are in empirical studies often listed qualitatively but rarely quantified.As for how firms determine and trade-off costs and benefits, we have only to presume that firms themselves are able to do so when setting their level of CSR engagement.

The public: capabilities and motives
The notion of "the public" refers, on the one hand, to the aggregate of individuals in a population, in their various stakeholder roles as customers, citizens, investors and consumers of products and users of natural resources, and, on the other hand, to the traditional and social media and the public opinion reflected in these.We abstract from the fine grained nature of the public and simplify it for the sake of the model into a single actor, conceptualized as the aggregate of the actors and groups of actors, like CSR-related NGOs, expressing expectations using various media.At the behavioral level, we assume the public to be able perceive the average of CSR engagement in the population of firms, which constitutes its expectations, to be able to compare the CSR engagement of a specific firm with this average, and to sanction firms diverging from the expectations.This is equivalent to the existence of an established and expected standard for CSR engagement, the ability of the public to detect under-and out-performers, and the public's willingness to respond to this with rewards or punishments.As for the real-world detection mechanisms, several are possible: individual CSR activists and NGOs might survey the CSR engagement of firms and publish the information, in particular on under-performers.Detection may also be a result of CSR surveillance exercised by the general media or by professional information providers of data on non-financial performance.However, the public's sensitivity for CSR engagement is to some degree fuzzy, as the public may be unable recognize the CSR engagement level perfectly.Based on anecdotal evidence, one might argue that the scrutiny of watchdog NGOs and the public is primarily concerned with CSR underperformance.However, conceptually, there are two distinct detection capabilities: first, the probability to receive a reward, given that the firm outperforms, second, the probability to receive a punishment, given that the firm under-performs.The public's reaction depends on the degree, to which the firm diverges from the public's expecta-tion, but also on some weighting factor, capturing the importance the public assigns to the type deviation (out-vs.under-performance).This allows to explore the effects of a public, which primarily rewards CSR out-performance as opposed to a situation where the public primarily punishes CSR under-performance.
Analytically, the magnitude of a punishment/reward can be differentiated: it depends on the degree to which the firm out/under-performs the public's expectations, the probability that the public detects and reacts to the out/under-performance, and the degree to which the public weights the gap between expected and the factual engagement when imposing a reward/punishment.For instance, an out-performing firm may receive a large benefit, either, because it is actually out-performing public expectations to a large degree, because the public is very sensitive to out-performance, or because the public attaches high weight to out-performance.A firm may receive a very small benefit, because the firm is out-performing only to a small degree, the public is insensitive to out-performance, or does attach little weight to the degree of out-performance.Thus, benefits of out-performance and punishments of under-performance are factually expected values, which depend on the three variables: probability, distance, and weight.For this reason, it is for applying the model sufficient to differentiate between the expected value of under-performance and the expected value of out-performance, i.e., the four situations in which the public, only rewards, only punishes, does both or neither.

CSR engagement: actions, reactions, and Developments
The presumed interactive nature of CSR engagement implies that firms set their CSR engagement, to which the public responds, then firms adapt in response to the public's reactions towards their engagement, which changes the public's expectations and so on.Regarding the behavior of firms, and the behavior of the public, we make the following assumptions, ordered by the chronology of the interaction.

Assumption 1
Firms chose their (initial) level of CSR engagement, which we assume to correspond with their costs of CSR engagement: the higher the CSR costs, the lower the initial engagement.The costs of CSR engagement are assumed to be stable over time, as they are closely linked to the firm's business model.

Assumption 2
Based on the CSR engagement levels chosen by all firms, the public forms an expectation as to the appropriate CSR engagement level.This expectation is the average CSR engagement of all firms.
The public detects with a certain probability under-/out-performance of a firm relative to the average CSR engagement of all firms and imposes rewards and/or punishments on the firm.The reward/punishment depends on the degree of out-/underperformance, the probability that the public detects the out-/under-performance and the weight, the public attaches to out-/under-performance.All three factors, which 1 3 together constitute a specific reward/punishment, are driving the predominance of rewards and or punishments in the public's dealing with the CSR engagement of firms.Regarding the costs and benefits, we assume the following.

Assumption 3a
Firms, which are neither out-nor under-performing the public's expectations or are not detected by the public to do so receive neither rewards nor punishments.They only have to bear the costs of their CSR engagement, which depends to the level of CSR engagement chosen and their firm-specific costs capturing how expensive CSR engagement is for them.

Assumption 3b
Firms, which under-perform the public's expectations and are detected to do so by the public receive a punishment, which depends on a weighting factor and the magnitude to which the firm is below the average CSR engagement.In addition, the firm bears the costs of its CSR engagement, resulting from the level and how expensive CSR engagement is for the firm.For instance, the costs of a firm arising from the actual CSR engagement may be low, because the firm decided not to engage to a substantial degree in CSR, but the costs of CSR as an issue are high nevertheless, as reputational costs arise.

Assumption 3c
Firms, which out-perform the public's expectations and are recognized to be outperforming receive a reward, which equalizes or even overcompensates their costs arising from CSR engagement.Again, the reward depends on the degree by which the firm in question exceeds the population average CSR engagement, weighted by a factor indicating how highly the population values CSR out-performance.
The total costs arising for a firm in a round are constituted by the costs of CSR engagement plus the rewards and punishments arising from being below or above the average of CSR engagement.In terms of managerial decision-making, this implies that CSR as an issue cannot be ignored: engaging in CSR is costly, but not engaging in CSR is costly, too.We assume firms to observe their payoff and adapt their CSR engagement in line with the following behavioral assumptions.

Assumption 4a
By default, firms maintain their level of CSR engagement.This holds in particular for firms which receive neither a reward nor a punishment relating to their CSR engagement.We do not presume that cases of actively reducing the CSR engagement occur, nor do we assume that the CSR engagement is increased for its own sake.

Assumption 4b
If the firm's costs in terms of punishment due to CSR under-performance are higher than the costs of the current CSR engagement, the firm will increase its CSR engagement.The cheaper CSR engagement is for the firm, the more it will increase its CSR engagement.

Assumption 4c
If the firm's benefits from CSR engagement are higher than the costs of its CSR engagement, the firm will also increase its CSR engagement further.Again: the less costly CSR engagement is for the firm, the more it will increase its CSR engagement.

Explorations
While firm properties -basically their costs of CSR engagement, how expensive engagement is for them -are assumed to be fixed, outcomes of the public-firm interaction depend the relevance of rewards and punishments for out-and underperforming.Exploring the behavioral model allows statements about how the population-level of CSR engagement develops in a specific situation, e.g., high relevance of punishments, and how firms with different levels of CSR-related costs behave.
The public has four options respond to firms not matching the public's expectations: only rewarding, only punishing, doing neither, or doing both.As we argued above, the three components -degree of out-/under-performance of the firm, the public's sensitivity to mismatches, and the magnitude, by which the public weights deviations from expectations -all result in high or low relevance of rewards or punishments for the firms, thus there is no need to consider these components in isolation.Each of the public's four options leads to very different developments of CSR engagement of firms and at the level of the firm population.
1) If the public imposes neither rewards nor punishments, firms receive neither rewards nor punishments.The same holds true for the hypothetical situation where the public is completely unable to recognize the CSR engagement of firms, which also makes the expected values of out-/under-performance irrelevant for firms.As a consequence, firm-level behavior remains unchanged, i.e., no firm changes its engagement, and the population-level CSR engagement also remains unchanged in level and dispersion, understood as the range of CSR engagement levels of individual firms occurring in the population.
2) If the public reacts to the CSR engagement of firms by (predominantly) bestowing rewards for out-performance, but no punishments, the CSR engagement in the population is driven upwards and persistently so.Firms for which CSR engagement comes at very low costs and which had therefore the highest initial level of CSR engagement constantly increase their CSR engagement, as they receive substantial benefits at low costs, dragging the population-level CSR engagement along.Firms, which are either at the average of CSR engagement of or under-perform the average, retain their initial CSR engagement as they face no negative consequences for doing 1 3 so.As there are no punishments for firms which are increasingly under-performing, the dispersion of CSR engagement will increase: under-performers retain their level, out-performers increase it.
3) If the public responds to CSR engagement by (predominantly) imposing punishments for under-performance, the population-level CSR engagement basically stagnates after a phase of adaptation by the under-performers, which increase their engagement till they no longer receive punishments for under-performance to a degree which outweighs their costs.As there is no reward for firms out-performing the average, firms which initially out-perform the average will not increase their CSR engagement.The dispersion of the CSR engagement in the population will decrease, as the under-performers catch up with the rest of the population, whose members do not change their engagement levels.
4) If the public uses both, rewards and punishments, the resulting development of population-level CSR engagement is similar to the situation with only rewards: it increases, but does so with a stronger dynamic.Not only are out-performers driven to increase their CSR engagement, but so are under-performers.The increase in engagement by the under-performers drives the out-performers to further increases in their engagement, in order to keep their distance and the associated benefits, the increase of the out-performers forces the under-performers to increase their engagement, in order to limit their distance to the average and the associated punishments.Due to the constant out-pacing of the out-performers and the catching-up of the under-performers, the dispersion of the CSR engagement in the population will remain basically unchanged.Depending on the severity of the punishment and the magnitude of the rewards, this process is faster or slower.

Confronting reality
While the behavioral model is simplified and stylized, it allows interesting insights into the development of CSR engagement, for the firm population and also subgroups in the firm population.The above arguments show that there are highly different developments expectable for constellations where the public predominantly relies on punishments, rewards, or both in dealing with firms' CSR engagement.In order to evaluate the factual relevance of rewards and punishments, we compare patterns in CSR levels and developments found in reality with the implications of the above arguments in order to identify those constellations which can account for the empirical observations.In particular the development of the CSR engagement at the firm population level, among firms which under-/out-perform at the beginning of the observation period, but also the development of the range of observable CSR engagement in the firm population.The implications of the behavioral model and empirical data to not have to match perfectly.Rather, the question is, Which constellation is closest to the empirical data?
Empirical data on CSR engagement of firms can be obtained from several sources, such as csrsmonitor.org,MSCI ESG Rating, csrhub.com,or sustainalytics.com.We chose the Refinitiv ESG database (Refinitiv 2019) as a source as it offers the largest and longest coverage of ESG engagement.As to what firms to sample, the model hinges on how the public deals with CSR, specifically, if the public tends to handing out rewards as opposed to handing out punishments.This is a genuinely cultural, and, consequentially, country-specific feature (see Steurer 2010 and Brown/Steen Knudsen 2015 for effects of the political system on CSR and Desender/Epure 2021 for the interaction between economic and political features in the varieties of capitalism framework).Thus, the relevance of rewards, punishments and also the public's capability to detect deviations from the expected CSR engagement may differ among cultural regions and countries.Despite the internationalization of firms, we treat firms as belonging to the country in which they are based as firms, aside from multinationals, typically have their strongest footprint and media presence in their home countries.As for the countries selected, we commence with the US, as the largest set of firms, and compare this sample with another country from the same region, Canada, and another region, continental Europe.
As for measuring the ESG engagement, the ESG scoring provided by Refinitiv is based on about 400 ESG-related items of information, some qualitative, indicating the presence of institutional features relating to CSR, some quantitative, like energy consumption or emissions.Typically, the ESG scoring is used as a variable, e.g., Desender/Epure (2021), which however is not appropriate for the given research interest of this study, as the overall range of Refinitiv's ESG scores stays constant, covering a potential range from 0 to 100 regardless of the absolute development of the firms' CSR engagement.For instance, the firm with the highest performance in a year is scored 100.If this firm increases its engagement in the next year, but is still the best in class, it will also score 100, to indicate it to be the best firm in the sample/ group.Thus, using Refinitiv's ESG score to demonstrate an absolute increase in CSR engagement, or to compare firms over time and across countries, is by construction impossible.
For this reason, we went back to the raw data on qualitative binary ESG related items, about 200 features, coded as present or absent.Refinitiv provides raw information on the individual items of ESG engagement, like having a diversity management in place.These features concern institutional aspects, for instance, having a certain policy in place, rather than quantitative features, like a firm's CO 2 emissions.Doing so has several advantages: first, institutional aspects, like having a diversity management policy, apply to all firms, regardless of their business model, as opposed to, say, the quantity of CO 2 emitted, which depends very much on the business model.Second, computing a proxy measure of absolute CSR engagement by counting the institutional CSR features present for a firm in a given year allows to track the firm's absolute CSR development over time and thus to test, e.g., whether population-level CSR engagement increases over time.The list of items going into our absolute CSR engagement measure is listed in the Appendix.
Figure 1 shows the development of CSR engagement of US firms sampled from the Refinitiv database over the period of 2008 to 2018 for three sub-samples of firms: (1) for firms, which under-performed in terms of CSR engagement in 2008, i.e. the bottom quartile of the firm population, (2) for out-performers, the top quartile of firms in terms of CSR engagement in 2008, and (3) the middle 50% of the population.
Looking at all available cases (firm-years), we see that the absolute CSR engagement increases only very little over time.For the whole sample (ca.7,730 firm-years), 1 3 the increase per year (b-coefficient) is 0.41 points (t=4.42), on a scale which ranges from a minimum of 13 to a maximum of 137.
Looking only at the cohort of 2008, i.e., the information given in Fig. 1, developments of the groups differ.The out-performers of 2008 show, of course, substantially higher initial CSR levels than the other two groups, but do not increase their CSR engagement to a substantial degree (b=0.92t=5.37), and not in a way which is much different from the "mid-fielders" (b=1.67 t=12.36).This supports the interpretation that they are not progressively increasing their CSR engagement, which would be indicative of a situation where strong rewards are offered by the public for CSR outperformance.Instead, the CSR development of the out-performers matches a situation were benefits granted by the public are very limited.
The picture for the under-performers of 2008 differs substantially: they do increase their CSR engagement substantially (b=2.93 t=17.13),till, at the end of the observation period, they have caught up with the mid-fielders.This behavior fits with a situation where the public imposes strong punishments on under-performance.This process occurs in eight years time.While it is difficult to say, whether this development is fast or slow, there is good reason to argue that it is fast: under-performers achieve an increase of CSR engagement of 24 points in the period from 2008 to 2016, the out-performers only 13 points, whereas the mid-fielders of 2008 increase their CSR engagement by 19 points during this period.The behavioral model also allows statements about the conditions where the range of observable of CSR engagement increase or decrease over time.Refinitiv expands its coverage of firms and countries over time, a development which, by construction, affects the range and heterogeneity of CSR engagement occurring in the database, thus, comparing two years in terms of range is misleading.For this reason, we only track the cohort of firms which were already part of the sample in 2008.Looking at Fig. 1, we see that in line with a constellation with high punishments and low rewards imposed by the public on firms, the range of the observable CSR engagement reduces over time.
Thus, the overall development of the CSR engagement in the population of US firms is characterized by under-performers quickly closing up to the average.The average CSR engagement in the population increases, but only slightly.The topquartile of the firm population does not increase its engagement any more than all other firms do.The constellation most in line with the empirical pattern observable is one in which there are strong punishments associated with high public awareness for under-performance, but no or very low rewards for out-performance.
There are two possible alternative explanations, which might account for the observations made.
First, it may be the case CSR engagement and the probability to survive as a firm are correlated.One might argue that firms, which are doing poorly in terms of CSR, are doing poorly in other aspects as well and drop out of the sample, in particular by going bankrupt.If this were the case, firms with low CSR engagement would drop out and this might seemingly increase the CSR engagement in the remaining sample.However, the sample composition remains stable over time: initially, the group of under-performers in 2008, which are tracked over time in the above figure, consists of 93 cases in 2008, in 2015, 84 of them remain, in 2016 77, which might also be due to a delay in data availability.Similar, the out-performer group is also stable.It is not the case that low performers in this subgroup drop out of the sample.In 2008, the out-performer group consists of 95 cases, in 2016, there are still 87 firms covered.
Second, while we consider only the cohort of firms covered in 2008, one might presume that most of the development of CSR is driven by firms entering the database which feature higher levels of CSR engagement right from the start.To test this argument, we compared the CSR engagement scores of firms, which enter the database and are covered for the first time, with the scores of existing ones.For each year from 2009 to 2017, we computed a t-test, comparing the mean CSR engagement of the old cases with the CSR engagement of the new cases.We found, that the new entries feature lower CSR engagement than the old cases in each of these years, the difference range from 30 points in 2012 to 10 points in 2011, the differences are statistically significant in each year apart from 2017, for which only one new entry is added to the database.
In order to get an insight into how public dealings with CSR engagement of firms differ across countries, we compare the development of CSR engagement in the US with Canada and Europe.Due to differences in data coverage, we were unable to use the same time-frame as in the case of the US.
Comparing the situation in Canada with the US, we see that the CSR development in Canada is more uniform.All three groups increase their CSR engagement over the course of the period under consideration.No group, e.g., the under-performers, increases its engagement much more than the other two groups, which would be indicative of strong punishments.Nor is it the case that the out-performers go ahead and progressively increase their engagement, as would be indicative of high rewards.Despite some increase, there is but little convergence among the three groups.In terms of the constellation of rewards and punishments, the situation of uniform and moderate increase is indicating a development driven by a combination of rewards and punishments.
Figure 3 gives the situation in continental Europe (France, Germany, Norway, Denmark, and Sweden).All three groups, out-performers, mid-fielders, and underperformers, increase their CSR engagement.The out-performers increase their engagement much more than out-performers in the US, which is indicating a higher relevance of rewards.The under-performers increase their engagement much more than the other two groups and are catching up with the mid-fielders, which indicates strong punishments, too.As a consequence, the range of the CSR engagement decreases.This situation is in line with rewards for out-performance and but stronger punishments for under-performers, which are catching up to the mid-fielders.
Comparing the three populations indicates notable differences in how CSR engagement develops.In the US, the CSR engagement in the full population increases only very slowly by 0.22 points per year (b=0.22;t=2.13), basically stagnating.Only in the sub-sample of under-performers, the increase is substantial.We conclude that CSR development in the US is driven by strong punishments but little or no rewards.
In Canada, the CSR engagement of the overall population increases by 3.6 points per year.There is basically no difference in the increase between the under-performers (b=2.89) and the out-performers (b=3.06), and only a small difference to the midfielders (b=4.20).We conclude, that both, rewards and punishments are in place.
In Europe, all groups increase their CSR engagement substantially.The overall population CSR engagement increases by about 5 points per year (b=4.89),among the mid-fielders of 2000.the increases is even higher (b= 5.20).The out-performers of 2000 increase their engagement less (b=4.06),while the under-performers increase it much more (b=6.14).The fact, that the out-performers increase their engagement indicates rewards for out-performance and but the stronger increase of the underperformers indicates even stronger punishments for under-performance.

Synthesis of results
Starting out from the conception of CSR engagement as an interactive phenomenon, this theoretical approach was reconstructed using an actor-based model, where firms react to public reactions to their choices and the public's reactions reflect expectations formed by what firms do.Of interest was, 1) how different sets of assumptions about the predominance of rewards and punishments imposed by the public affect the development of CSR engagement.2), which of the four possible constellations -the public imposes rewards only, punishments only, does both or neither -fits closest with empirical CSR development in the firm population of a country.The constellations considered result in very different patterns of CSR engagement of the firm population and sub-groups in the population.These expected patterns were compared with the empirical patterns gained from data on ESG scores from firm populations in the US, Canada, and continental Europe.The comparison indicates that empirically in the US, CSR engagement is driven by firms avoiding public punishment rather than by going for a reward granted by the public for outstanding CSR engagement.The picture in Europe and Canada differs, indicating that how the public deals with CSR engagement is indeed a cultural feature.In particular, it has to be noted that fear of punishment is a much stronger motive for CSR engagement than the majority of research indicates.

Theoretical contributions and implications
The approach of conceptualizing CSR compliance as a socially constructed phenomenon is complementary in its aims and method to existing approaches which focus on explaining CSR engagement of individual firms based on firm properties, like industry membership or public exposure (Brammer/Millington 2006).The papers first theoretical contribution consists of putting the focus on the system-level dynamics of CSR, which can, but rarely are, captured in empirical data on CSR engagement of firm populations.This approach allows an analysis without having to resort to any information about specific firms beyond their CSR engagement.This system-level perspective, the degree to which developments of CSR engagement by some firms affect the CSR engagement of other firms was not yet considered in a systematic way, and the behavioral model suggested allows to gain substantial insights on factors driving CSR engagement which are absent in other analyses.It also allows to derive implications of changes in structural features in the firms societal and business environment, like changes in the costs of CSR or changes in the public's dealings with CSR under-/out-performance.By allowing for country specific effects, like the level of public attention for CSR, the paper expands existing literature on the effects of country-level regulation (Steurer 2010 and Brown/Steen Knudsen 2015) and the variety of capitalism (Desender/Epure 2021) on CSR, which were studied already.The present study offers a complementary approach which allows for variation among economically similar but culturally different systems (US vs. Canada).

Implications for practice
In terms of a practical contribution, the public's tendency to reward or punish CSR engagement turned out to be a crucial factor.Acknowledging the importance of this factor allows to derive predictions about how features like public attention impact on CSR development.Attention to CSR per se, which corresponds to the attention conventional and social media pay to CSR out-/under-performance, speeds up processes, but does not change their direction.Central is the public's reaction towards out-and under-performance.A society, which deals with CSR only in terms of punishing under-performers will achieve no increase in CSR engagement but stagnation.Negative events dominate the public discourse about CSR and our model would imply that this dealing is counterproductive if the aim is to increase CSR engagement.To achieve this, rewards should play a larger role.

Limitations and future research directions
In terms of potential limitations, assumptions were made on the costs of CSR and the nature of the public.We can, however, discuss the consequences of changes in the firm's environment, both societal (concerning the public) and economic (concerning the costs of CSR) on the model and the insights gained from it.
The behavioral model to some degree assumed that the costs of CSR engagement are stable for the firms.We can loosen this assumption, and, in terms of a comparative statics analysis, inquire how changes in the costs of CSR engagement affect the implications of the model for the development of the CSR engagement in the population of firms.We consider two cases.
First, the costs of CSR engagement for all firms decreases / increases.The model hinges on the relative positions of firms: Behavior of firms is determined by their relative position in the population of all firms in particular their position relative to the average constituted by all firms.Now, if costs of CSR engagement in general and for all firms decreases / increases, all firms increase / decrease their CSR engagement, as it became cheaper / more expensive relative to their perceived benefits of CSR engagement.Still, the relative positions remain unchanged, the development of the CSR engagement of specific firms still depends on their relative position relative to the public's expectations and in particular, on whether the public rewards outperformance or punishes under-performance.For instance, even if CSR engagement becomes generally cheaper, the absence of benefits granted by the public will result in out-performers increasing their engagement less than under-performers do, the observable CSR engagement in the population converges, as out-performers stagnate while under-performers catch up.
Second, consider the case that costs of CSR engagement for a specific firm, which features high initial costs of CSR and thus, in our model, an initially low level of CSR engagement, decrease.The firm shows more CSR engagement, moves up in the relative ranking of CSR engagement of the population and increases the average of CSR engagement in the population to some degree.However, regardless of the individual firm which changes its behavior and relative position, there are still firms, which are below the average and firms which are above the average.The behavioral model hinges on the relative positions and how the public reacts to deviations from the public's expectations.Thus, if the public primarily sanctions firms showing to little CSR engagement relative to its expectations, there are still firms, which are below average and will receive sanctions.For those firms above the level of CSR engagement the public expects, there are still no benefits granted by the public -thus, their CSR engagement does not change.For a firm, for which the CSR costs increase, the same applies: it moves downwards in the relative positioning (by actively reducing its CSR engagement or by not increasing it) but there is still the relative ranking of the firms, notably, whether they are above of below the expectations, and depending on the public's dealing with CSR, the under-performers still receive punishments and /or the out-performers still receive benefits.
The model also treated the public more or less as a single actor.Given that societies become more heterogeneous, this assumption can be loosened and the implications of increasing heterogeneity in terms of CSR attitudes can be discussed.The primary diverge can be seen between shareholders and stakeholders.For shareholders, the costs of CSR engagement are relevant.As they ultimately pay for it, they can be presumed to trade off costs and benefits and call for the firm's management to do so as well.For other stakeholders this is not the case -for them, the benefits arising through CSR engagement for the (global) society dominate.In part, differences in the stakeholder composition may account for the differences between societies, e.g., also the observation that in some societies, punishments for under-performers are so much more relevant than rewards.
In terms of the model, more heterogeneous stakeholders and in particular more stakeholders with a higher attention and a more critical view on firm's CSR engagement result, in the end, in higher costs for CSR non-engagement, thus even higher punishments for under-performers.This implies, if the model holds true, that underperformers catch up with the mid-fielders even more quickly.But still, lacking rewards for out-performers, the stakeholders will not achieve their aim of constantly more CSR engagement.
In terms of extensions, future research might consider the possibility that for different CSR domains, i.e., the three domains of environment, social and governance, the public's reactions in terms of granting benefits or imposing punishments may differ, with for instance the developments in the domain of environment being more driven by rewards imposed, while in the social domain, punishments for not meeting the expectations dominate.

Conclusion
Many theoretical and empirical studies highlight the benefits of CSR engagement, but we argued that we simply do not know whether overall firms are actually motivated by the benefits offered or by the punishments threatened by the public.This paper started out from a different view on the nature of CSR and of how CSR engagement in the population of firms comes about, modeling it as an interactive, socially constructed phenomenon: Actions of firms constitute public expectations and are, in turn, affected by these expectations.Contrary to studies focusing on drivers of CSR engagement of individual firms which are conceptually exogenous from the explanatory approach (e.g., Ioannou/Serafeim 2012 or Desender/Epure 2021), our model focuses on the firm population, system-level dynamics and features an interactive feedback component: the stronger some firms invest in CSR, the higher the pressure on other firms to follow, in order to avoid negative consequences imposed on them by the public.This red queen-type interaction was discussed in other domains, like organizational innovation (Barnett/Hansen 1996), but is also applicable to the domain of CSR.This conception allowed to focus on empirical questions for which no estab-lished methods exist.Most importantly, whether the public drives CSR engagement by rewarding out-performance or by punishing under-performance relative to what is expected.It turns out that the observable patterns in CSR engagement of populations and sub-populations are not consistent with the dominating view of CSR engagement being driven by rewards granted by the public.

Fig. 1
Fig. 1 Development of CSR Engagement among US-Firms.(Note: Trends in the development of the CSR Engagement Score for US firms computed based on information from the Refinitiv database for the years up to 2017.Sub-samples for middle 50% ("Mid-Fielders"), and the 25% with the lowest ("Under-Performers") and the 25% with the highest ("Out-Performers") CSR engagement in 2008.Shaded areas indicate the 95% confidence intervals.)

Fig. 2
Fig. 2 Development of CSR Engagement in Canada.(Note: Trends in the development of the CSR Engagement Score for Canadian firms computed based on information from the Refinitiv database for the years 2008 up to 2012.Sub-samples were created for the middle 50%, and the quartiles with the lowest / highest CSR engagement in 2008.Shaded areas indicate the 95% confidence intervals.)

Fig. 3
Fig. 3 Development of CSR Engagement in Europe.(Note: Trends in the development of the CSR Engagement Score for European firms computed based on information from the Refinitiv database for the years 2000 up to 2010.Sub-samples were created for the middle 50%, and the quartiles with the lowest / highest CSR engagement in 2000.Shaded areas indicate the 95% confidence intervals.)