What happens abroad, stays abroad? Exploring how corporate social irresponsibility in domestic and international markets influences corporate reputation

,


A R T I C L E I N F O
Corporate social irresponsibility (CSI) can occur in the multinational enterprise's (MNE) domestic and international markets, thereby risking corporate reputation. However, are corporate reputations differentially influenced by the location of CSI events? Drawing on the ethnocentric bias perspective, we examine how CSI affects corporate reputations according to whether CSI emerges in the MNE's home or international markets. We theorize that, when CSI occurs in an international host market, the negative relationship between CSI and corporate reputation is generally weaker. Conversely, when CSI arises within the home location, home countrylocated CSI has the strongest negative relationship to corporate reputation. Our findings generally reflect the core argument of the paper: home-country based CSI incidents may be more consequential to an MNE's corporate reputation compared to those CSI incidents which unfold in certain host countries. Our longitudinal analysis, comprising of 2,401 CSI events, involving 465 MNEs, confirms our theorizing. Among our principal contributions, this study adds to the growing and important literature on the dark side of international business (IB).
While instances of irresponsible conduct are presumed to have significant consequences for the firm and its reputation (Wang & Li, 2019), practical examples of CSI, such as the Rana Plaza disaster involving companies such as Primark and H&M, or Apple's numerous human rights abuses in China, illustrate that MNEs can be associated with CSI and still, their corporate reputations may appear only modestly influenced (e.g., Nardella, Brammer & Surdu, 2020). Accordingly, scholars have increasingly begun to question whether CSI presents consistent and significant risks for organizations (e.g., Barnett, 2014;Brammer et al., 2021;Jackson et al., 2012), leading to calls for renewed interest in the mechanisms that penalize and deter CSI (c.f. Buckley, 2021;Nardella, Brammer & Surdu, 2022;Whittington & Yakis-Douglas, 2020).
In parallel, IB research continues to hold the idea that the 'location matters' when considering sources of market and non-market risk (see for instance, Oh et al., 2020), although studies which specifically explore the impact of where CSI events are located remain scarce. Exploring the location-effects of CSI remains important because decision-making can often be biased toward the home market (Bohas, Morley & Kinra, 2021;Michailova et al., 2017), which implies that internationally located CSI events may be penalized less severely than CSI events located at home, or not penalized at all. Indeed, despite continued interest in CSI (Bu & Wagner, 2016;Cuervo-Cazurra et al., 2021;Nardella & Brammer, 2021) and corporate reputation in IB (Newburry, Deephouse & Gardberg, 2019;Sun & Ding, 2020;Wettstein et al., 2019), our understanding of the complex and (potentially) location-contingent relationship between CSI behavior and MNE reputation remains underdeveloped.
In order to advance our understanding, this study focuses on how and when the location of CSI influences corporate reputations. Our study primarily draws on extant IB research which explicates how decisions are often ethnocentrically biased in their orientation (Bohas et al., 2021;Michailova et al., 2017). Here, studies have illustrated that MNE stakeholders, such as investors (Sialm, Sun & Zheng, 2020), employees (Templer, 2010), and managers (Birkinshaw, Bouquet & Ambos, 2007) can often preference their home market and de-emphasize the importance of 'foreign' markets. Accordingly, in this paper, we explore the linkage between 'where' CSI events are located and corporate reputation. MNE stakeholders may be likely biased to assess CSI more readily and knowledgably when it occurs in the home location compared to more 'foreign', international locations. Readiness to negatively assess CSI, we theorize, stems from the behavioral tendency to respond more strongly towards threats that concern the ingroup (e.g., 'home' market) versus an outgroup (e.g., 'foreign' market). Knowledgably assessing CSI, a key form of moral reasoning, requires a set of rules, norms, and values upon which to draw. These rules are often shared within groups and provide decision-makers with instructions around how to process information about what isand what is not -legitimate and acceptable behavior in a given environment. The ingroup may provide a framework to help guide responses to CSI. Ethnocentric biases may therefore exist and strongly influence the relationship between CSI and MNE reputation.
This paper attends to calls for research on the 'dark side' of IB, specifically on MNE CSI behavior (Cuervo-Cazurra et al., 2021), its consequences, and the mechanisms that penalize and may deter irresponsibility (Whittington & Yakis-Douglas, 2020). We focus on the media disclosure of CSI as a suitable setting to examine the nature of the relationship between CSI and MNE reputation. We test our hypotheses on a sample of 2,401 CSI events involving 465 U.S.-based MNEs between 2005 and 2012. Overall, our study unearths a strong ethnocentric relationship between CSI and corporate reputation.
Consequently, our study makes three main contributions. First, we contribute to the emergingyet growingbody of research on the dark side of IB (Ghauri, Strange & Cooke, 2021;Rygh et al., 2021), and in particular, that which examines the CSI behavior of MNEs (e.g., Cuervo-Cazurra et al., 2021;Fiaschi, Giuliani & Nieri, 2017;Surroca, Tribó & Zahra, 2013). We enrich these literatures by illustrating that the location of MNE CSI matters when it comes to the social regulation of their behavior. In doing so, we unearth a relationship between CSI and corporate reputation that is generally weaker when CSI emerges in an international or 'foreign' market. Second, we extend theory concerning the ethnocentric bias (Perlmutter, 1969; see also Fischer et al., 2022) with insights from behavioral theory research. In doing so, we specifically provide IB and management scholars with a more nuanced conceptualization of ethnocentric biases which assists with overcoming previous over-reliance on simplistic moral reasoning rationales (Michailova et al., 2017). Our theorizing explicates how greater divergence in the norms, values, and legitimacy pressures of home and host international markets can contribute to ethnocentrism, by influencing the difficulty with which stakeholders are able to assess internationally located (CSI) events. Third, we contribute to advancing global perspectives on corporate reputation (Deephouse et al., 2016;Newburry et al., 2019) by providing a more holistic theorization of how the location of MNE (CSI) behavior is related to the reputation assessments generated in response to those (CSI) behaviors. Policy and practice implications concerning MNE CSI, and its potential consequences are discussed to further elaborate on the mechanisms that penalize and may deter irresponsibility, including when corporate reputation may become a key social regulation mechanism.

Corporate reputation
To become an MNE, a firm must possess firm-specific resources and capabilities which can be successfully exploited abroad (Narula & Verbeke, 2015). Intangible resources, such as corporate reputationi.e., "a perceptual representation of a company's past actions and future prospects that describes the firm's overall appeal to its key constituents" (Fombrun, 1996: 72) are hence considered to be of critical strategic value (Barney, 1991), as they can enhance the international competitiveness of the firm (Bell, Filatotchev & Rasheed, 2011). This is because a 'good' corporate reputation has been shown to influence a range of stakeholder behaviors, including those of investors, employees, customers, among many others (Antonetti & Maklan, 2016;Kölbel, Busch & Jancso, 2017;McDonnell & King, 2013;Turban & Greening, 1997;Zavyalova et al., 2016). Accordingly, prior reputation research has paid particular attention to questions concerning how, when, and to what extent corporate reputations are an outcome of specific organizational behaviors (Love & Kraatz, 2017;Mishina, Block & Mannor, 2012;Shiu & Yang, 2017;Zavyalova et al., 2016). Over time, two principal branches of corporate reputation research have emerged. The first strand of studies examined the reputational implications of evolving institutional norms against which corporate conduct and performance tend to be evaluated (Doh et al., 2010;Dorobantu, Henisz & Nartey, 2017;McDonnell, 2016;McDonnell & King, 2013). A second branch of research has examined the psychological foundations of corporate reputation. Here, studies have emphasized how stakeholder perceptions emerge in response to new information (Antonetti & Maklan, 2016;Ravasi et al., 2018). Scholars have shown that expectations, framing, reference points and cognitive biases have the potential to influence the formation of corporate reputations (Bitektine, 2011;Mishina et al., 2012;Ravasi et al., 2018).
In IB, there remain unresolved theoretical and empirical questions about the relationship between CSI and MNE corporate reputations. In fact, given the richness and complexity of the reputation construct, IB scholarship has remained largely uncertain about the extent to which organizational behavior is related to reputation outcomes (Rabbiosi & Santangelo, 2019). Some scholars have asserted that CSI carries a significant risk to reputation. Thus, MNEs are often expected to divest from markets where CSI emerges, in order to avoid reputation damage (see for instance, the study by Wang & Li, 2019). In contrast, more pragmatic views have emerged, whereby scholars have illustrated that MNEs may transfer their 'irresponsible' practices from the home market to their subsidiaries located abroad, particularly when international host markets are characterized by reduced institutional development (e.g., Bu, Xu & Tang, 2022;Surroca et al., 2013). This implies that the negative consequences of CSI for the MNE may diminish when CSI is located away from home, in a 'foreign' environment with differently enforced (generally weaker) rules and norms. A smallyet growingbody of research has specifically considered how corporate reputations may be associated with cross-national differences (Borda et al., 2017;Deephouse et al., 2016;Gardberg, 2006), reinforcing the idea that stakeholder perceptions may differ -to some extent -across international markets. In an increasingly 'complex' IB environment, where MNEs are expected to effectively fulfill their obligations towards a variety of stakeholders (Brammer et al., 2021), more research is necessary to explore when and how factors such as CSI location influence the reputations of MNEs.

Ethnocentric biases
In parallel, a considerable body of IB literature has examined the effects of 'location' on decision-making, behavior, and organizational outcomes (e.g., Li, Hernandez & Gwon, 2019;Oh et al., 2020;Xu & Shenkar, 2002). Notably, IB scholars have observed the tendency of company managers, as well as a broader range of other MNE stakeholders, to preference their ingroup(s) -i.e., individuals with similar value systems, experiences, beliefs, and goals - (Perlmutter, 1969; see also Bohas et al., 2021) when making decisions. Should the parameters of ingroup preferences center (often loosely) around aspects related to ethnicity, this tendency is considered to be the result of the ethnocentric bias.
Unsurprisingly, scholars emphasized that ethnocentrism is a largely negative attribute (for a review, see Michailova et al., 2017). Early studies mainly defined the ethnocentric bias as an attitude where people assume that their "particular vision of what is and what should be is the best and that all other systems of knowledge and belief are not only different from but inferior" (Wortzel & Wortzel, 1985: 412). Implicit in most (if not all) conjectures of ethnocentrism is the existence of an "ingroup" that is prioritized and an "outgroup" that is not (Sumner, 1906). By focusing on the ingroup, studies suggest that anti-outgroup sentiments tend to emerge. Behaviors such as discrimination (Templer, 2010), a reduced motivation to build positive relationships with stakeholders from other cultures (Wu & Bodigerel-Koehler, 2003), as well as the poor/unfair management of global resources by MNE managers (Caligiuri & Tarique, 2012) have all been theorized to result from ethnocentric biases.
However, there are studies which also question whether ethnocentrism does, in fact, emerge mainly out of an attitude of superiority (see: Michailova et al., 2017). To further this line of enquiry, we put forward an alternative perspective of ethnocentrism which draws on behavioral theory. Specifically, we propose that behavioral theory can help us understand how the complexity of the external, institutional environment exposes decision-makers to various risks. In circumstances where significant risks tend to emerge, individuals are often found to respond by applying the same cognitive strategies which have evolved to assist them in navigating their environments (Fiske & Taylor, 1984; see also Marewski, Gaissmaier & Gigerenzer, 2010;Nardella, Narula & Surdu, 2021;. Studies have proposed that circumstances where individuals are presented with risks to themselves and what they consider as their ingroup are prioritized, thereby responding more readily, urgently, and critically to such circumstances than contexts which do not (Epley & Caruso, 2004). From this view, we propose that the ethnocentric bias may emerge, in part, from an adaptive psychological tendency (Tversky & Kahneman, 1973), one which has developed to be protective to the individual as well as to the (in)group.
In this context, we acknowledge the fact that MNEs do not only operate in a home, domestic market, but they also typically operate in multiple 'foreign' (international) contexts, characterized by different institutional rules, norms, and values. In each location, institutions, including the state, can assist in propagating a dominant set of rules, norms, and values which become held widely in that society and which dictate what is (il)legitimate and (in)appropriate MNE conduct (Brammer et al., 2021; see also Aguilera & Grøgaard, 2019;Fainshmidt et al., 2018). Institutions act with the dual purpose of affording societal participants a set of core instructions and guidelines to guide their perceptions within a given context, as well as producing a structure that can further regulate (organizational) behavior. Achieving legitimacy -"a generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions" (Suchman, 1995: 574) is important for an MNE's reputation. As such, institutions can legitimize the behavior of firms by rewarding behavior that reinforces the tenets of a particular set of rules, norms, and values, and punishing or delegitimizing behavior which violates them (Meyer & Rowan, 1977; see also Hannan, 2010). Aside from the state, there are other institutions that propagate shared values; from non-governmental institutions (Besharov & Smith, 2014), religious institutions (Gümüsay, Smets & Morris, 2020), and professional bodies (Roulet, 2019). Being part of an (in)group, allows MNE stakeholders to share the same (or a similar) set of core institutional instructions and guidelines which aids processing and evaluating information.
We continue below to discuss why, when MNE stakeholders are presented with the social harm caused by CSI, their assessments are likely to be strongly influenced by the similarity or divergence between markets, i.e., markets perceived as more similar ('ingroup') and 'foreign' markets ('outgroup'). What constitutes '(in)appropriate', '(il)legitimate' or 'severe' corporate conduct, such as CSI, can vary significantly at the level of location due to ethnocentric tendencies.

The relationship between home country CSI and corporate reputation
CSI is most commonly identified when a firm is associated with (sometimes severe) harm to its stakeholders (c.f., Clarke, Riera & Iborra, 2021). Hence, we recognize that, overall, CSI presents risks (e.g., physical, emotional and/or economic) (Nardella et al., 2020). For the MNE, stakeholders' responses to the risks produced by CSI may have a significant impact on its reputation. At the same time, stakeholders' responses to risk may not be homogenous. Behavioral insights further suggest that risks which are situated within people's immediate environment produce heightened responses (Gineikiene & Diamantopoulos, 2017;Rozin & Royzman, 2001). The tendency to emphasize and respond to the immediate, local environment is mainly attributed to the salience of risks to the 'self' and the 'ingroup' (Epley & Caruso, 2004). Ethnocentric responses, therefore, may be more likely to arise when the risks associated with CSI emerge in the immediate, home market, and less likely to arise when CSI risks emerge elsewhere, in international host markets.
In addition, international markets themselves, and the legitimacy pressures that characterize them, are diverse. Some markets may hold significant similarities to the home market, particularly when rules, values and norms are mostly shared (e.g., principles around freedom of speech, freedom of association, ideas around individual liberty, rule of law enforcement, etc.). Contrastingly, other international markets may hold rules, norms, and values with far greater divergence, whereupon substantive difficulties in understanding local expectations and sources of legitimacy for these contexts may emerge. When assessing information referring to a 'foreign' location, a source of uncertainty faced by informal regulators, such as non-governmental MNE stakeholders, may largely stem from the comparative divergence around what constitutes (ill)legitimate behavior across locations.
We therefore propose that, in addition to the bias to respond to the (immediate) risks of CSI more readily in the home market, internationally located CSI may produce different responses, depending on the (di)similarity between locations. Should institutional rules, norms, and values between locations be largely similar/shared, CSI events, we argue, are likely far less challenging to (morally) comprehend. Conversely, considerable divergence between markets may mean that some internationally located CSI events are substantially more difficult to assess, thereby mitigating any subsequent reputation effects.
In Fig. 1, we propose a theoretical framework which maps out how CSI influences MNE reputations in terms of whether CSI risks impact; 1) the nuclear ingroup (the home country) also associated with a core 'shared' set of institutions, values, and thus, ideas around what constitutes (il)legitimate or (in)acceptable behavior; 2) the extended ingroup (located in a Western host country) which is also associated with 'similar' institutions, and values; and 3) an outgroup (located in a non-Western host country) which, we propose, is associated with more 'divergent' institutions, values, and expectations around how organizations should behave in a given environment. In the remainder of this section, we hypothesize how the location of CSI may influence MNE reputations.

CSI: impacting the MNE's 'nuclear ingroup' and MNE reputation
Since the majority of MNE stakeholders do not experience CSI or its impacts first-hand, they tend to rely on intermediaries, such as the news media, to become informed (Deephouse & Heugens, 2009;Pollock & Rindova, 2003). Upon receiving information about CSI, stakeholders then interpret this information. As CSI events can differ in their effects, studies have shown that of principal importance to stakeholders is the attribution of CSI's severity (Lange & Washburn, 2012). Since the severity of CSI is important to subsequent reputational effects, scholars have noted that the victims of CSI are fundamental to stakeholder judgements of a CSI event's severity (Nardella et al., 2020;Valor, Antoneti & Zasuwa, 2022). Namely, as the primary group impacted, victims of CSI are expected to experience the risks and harm(s) associated with irresponsible corporate conduct. Given the primacy afforded to the core ingroup (Perlmutter, 1969) and the tendency to respond with urgency when threats to the ingroup emerge (Epley & Caruso, 2004), we, overall, theorize that stakeholders are more likely to respond negatively to CSI behavior when the nuclear ingroup is impacted by CSI in the home country.
Further, we theorize that attributing the relative level of severity to CSI, is enhanced by a shared set of rules, norms and values held within a home location (Chan & Ananthram, 2020;Greenwood et al., 2008;Greenwood et al., 2012). In addition to the tendency to prioritize 'risk' to the ingroup, shared principles among the ingroup also strengthen stakeholders' ability to assess the severity of CSI located in a home institutional environment. Specifically, shared principles regarding what is legitimate MNE behavior (or not) in a given society (e.g., expectations around individual liberty, freedom of speech, enforcement of the rule of law, rights of employees) will enhance stakeholders' understanding of CSI events, as well as their assessments of CSI severity, thus amplifying the ethnocentric bias.
Conversely, when assessors of CSI lack a set of shared rules and values, efficient judgements about the inappropriateness or illegitimacy of MNE behavior are reduced. Hence, understanding CSI and its severity is likely made more difficult because applying shared principles and approaches for evaluating CSI become unavailable (Brammer et al., 2021). In such circumstances, assessors may encounter more ambiguity associated with evaluating the causes, outcomes, and severity of CSI incidents. When difficulties in appreciating CSI incidents emerge, we propose that there is a relatively weaker relationship between MNE CSI behavior and corporate reputations.
Overall, in line with established theory on CSI and corporate reputation (Lange & Washburn, 2012;Mishina et al., 2012), we recognize that MNE stakeholders generally disapprove of irresponsible corporate conduct. However, our rationale departs from prior studies, as we theorize that stakeholders are ethnocentrically biased to respond most strongly to those CSI events which occur in the home market location. This process is mainly facilitated by perceived urgency of home market located events as well as a shared set of rules, norms, and values that can be readily accessed to understand and evaluate CSI and its risks. Accordingly, our first hypothesis states the following: Hypothesis 1: There will be a negative relationship between CSI and corporate reputation; and the negative effect will be strongest when CSI is associated with the nuclear ingroup (i.e., home market).

CSI: impacting the MNE's 'extended ingroup' and MNE reputation
Though ethnocentric biases can help explain the significance of home market located CSI, CSI events can be globally dispersed. Internationally located CSI may therefore be more nuanced in its effects -and as such -it does not pose the same level of perceived risk to all MNE stakeholders. Internationally located CSI, we propose, is less likely to produce a magnified response to the MNE's 'nuclear ingroup,' as this group is likely not exposed to the same level of risk from internationally located CSI. Therefore, the process of assessing internationally located CSI is less urgently responded to by stakeholders.
Furthermore, ethnocentric responses may also emerge as the consequence of "limited knowledge" (c.f. Newberry, Gardberg & Blekin, 2006: 668). Responding to internationally located CSI is made more difficult because of a lack of shared rules, norms, and values by which to understand the moral significance of events. Since stakeholders are predisposed to respond more strongly to events that pose risks to ingroups located principally within the home market, by contrast, we explain that they may respond less strongly, overall, to events that influence other groups located in 'foreign' host markets. This propensity is then magnified by the relative divergence in institutions and legitimacy pressures between home and host, international locations. The process of moral reasoning, particularly in terms of the severity of international CSI can be made less challenging by greater convergence in what constitutes appropriate and legitimate behavior between groups in different locations.
Broadly, we propose that CSI events that emerge across an international host market holding a relatively similar set of institutions may be (somewhat) less challenging for stakeholders to evaluate. This is because perceived relatedness between rules and values among members of what we term an extended ingroup, will lead to biases manifesting toward those affected by CSI and those assessing CSI (rather than merely the nuclear 'ingroup'). The presence of relatively similar institutions, with similar core beliefs (e.g., individual liberty, freedom of speech, and the rule of law) between those affected and those assessing CSI can make it easier for individuals assessing CSI to appreciate its severity and significance. Therefore, these 'extended ingroup' membersi.e., in our study, the potentially harmed stakeholders located in Western host marketsmay be perceived to share similar views and expectations around what is deemed as (il)legitimate corporate behavior (Thornton & Ocasio, 2008) as the home market. Accordingly, a polycentric tendency (Perlmutter, 1969) is likely to emerge when CSI is associated with harming an extended ingroup i.e., stakeholders may stillalthough to a lesser extent -penalize firms for irresponsible behavior associated with a Western host market location.
Hypothesis 2: There will be a negative relationship between CSI and corporate reputation; whilst the negative effect will be weaker when CSI is associated with the extended ingroup (i.e., Western host markets) compared to the nuclear ingroup (i.e., home market).

CSI: impacting the MNE's 'outgroup' and MNE reputation
CSI associated with, and perceived to influence, the outgroup (here, CSI in non-Western host locations) -where fewer institutional rules, values and legitimacy pressures are shared -will be associated with much weaker responses. As MNE behavior "is socially irresponsible only to the extent that observers perceive it as such" (Lange & Washburn, 2012: 301), CSI associated with the global value chains of MNEs in many non-Western countries, including instances of use of discriminatory business practices, child labor and undermining of employees' rights, are all relevant examples of CSI that repeatedly impacts 'outgroups' with potentially limited reputational consequences for MNEs (Antonetti & Maklan, 2018;Lee et al., 2020). Such instances of CSI further serve to illustrate the subjective nature of stakeholder perceptions.
As such, we emphasize that it may be considerably more difficult for stakeholders to interpret information about a CSI event (including information about its causes, severity, and outcomes) when the rules, norms, and values between locations differ significantly. What constitutes a source of legitimacy in the home market, may not constitute a source of legitimacy in a distant host market, leading to more ambiguity around the severity and undesirability of a CSI event. Applying the same explicit and codified 'rules of the game' may largely become unavailable, which oftentimes results in inertia, where no strong views or perceptions about the CSI event become widely formed. We expect that CSI located outside the nuclear or extended ingroup will represent significantly reduced risks for MNE reputations, as greater divergence in home-host pressures around what constitutes (il)legitimate firm behavior presents a barrier to judging the severity (or even presence) of CSI. Hence: Hypothesis 3: The negative relationship between CSI and corporate reputation will be weakest when CSI is associated with the MNE's outgroup (i.e., non-Western host markets).

Sample and data collection
We gathered our data from two main sources: Fortune Magazine's "World's Most Admired Companies" (WMAC) survey (formerly America's Most Admired Companies -AMAC during our sample period) 1 , which has been conducted since 1983 and provides a comprehensive dataset on corporate reputation for MNEs (dependent variable); and the ASSET4 dataset -compiled by Thomson Reuters -to provide information on media disclosed CSI events (independent variables). The WMAC survey measures perceptions of firms and their behaviors among top executives and directors of MNEs, together with financial analysts who cover these organizations to calculate reputation scores. To feature in the WMAC survey, an organization must have annual revenue exceeding $10 billion and rank among the largest in its industry. Respondents are invited to name the leading firms in their own industry and then asked to rate each company on nine key performance attributes: (1) ability to attract and retain talent; (2) quality of management; (3) social responsibility to the community and the environment; (4) innovativeness; 1 Over the time period sampled, the main focus of our reputation data was "America's Most Admired Companies". In this period, representation was focused on the U.S. context. In 2014, the survey was adjusted, gaining greater inclusivity in terms of international firms and respondents surveyed, thereupon rebranded to Fortune Magazine's "World's Most Admired Companies" ranking.
Our second key data source -ASSET4 (Thomson Reuters) -collects data from media reports and classifies CSI events into twenty categories including, but not limited to, "intellectual property", "anti-competition", "taxation", "management compensation", "wages and working conditions", "child labor", and "public health". As noted by Pollock and Rindova (2003: 632), "in performing its functions of informing, highlighting, and framing, the media presents market participants with information that affects impression formation and the legitimation of firms." In order to extend the ASSET4 dataset (and collect information about CSI event characteristics such as its location), we conducted our own media searches (Chatterji et al., 2016;Fiaschi et al., 2020). We used the LexisNexis search directory, which draws data from a wide range of reliable media sources (e.g., The Wall Street Journal, Financial Times). We followed a systematic process (see : Flammer, 2013) to collect and code the CSI-related information available from the media sources. We proceeded as follows. First, we restricted the focus of the analysis to CSI events disclosed between 2004 and 2011 (one year prior to the last survey which measured the reputation score). Second, we searched for CSI events implicating U.S.-based MNEs present in the WMAC(AMAC) survey. Third, we searched for key terms related to the underlying CSI events (following the event categories provided and defined by ASSET4), as well as search terms constructed to capture broader media reporting of CSI. 3 Fourth, we eliminated duplicates and media articles that did not refer to CSI implicating the MNEs in our sample. This left us with over 250,000 media articles to code. Fifth, in order to make the coding process transparent and replicable, we developed a pro-forma, and recoded each of the disclosed CSI events into one of the twenty categories of CSI event types originally described in ASSET4 (see Table 1 for the full list of CSI event classifications and their definitions). We also added, under the relevant categories, the CSI events that had not been originally identified in the ASSET4 dataset. We read each media article to extract relevant CSI-related information in particular around event characteristics, namely the location of the CSI event. 4 Lastly, we merged the extended ASSET4 dataset with our reputation scores  (Fombrun & Rindova, 1996;Mishina et al., 2012). Engelen and van Essen (2012: 56) already explain in detail that "stock prices are simply an aggregate of a huge amount of buy-and-sell decisions", reinforcing the idea that stockholders' core concern represented in this evaluation is the increase in their own financial risk and/or decrease in return on investments, rather than their broader perceptions about the reputation of the firm. 3 Supplementary material is available in Appendix 1 where we disclose the full list of search terms. 4 We ensured inter-coder reliability (Gaur & Kumar, 2018;Lombard, Snyder-Duch, & Bracken, 2002) by starting the coding process with a pilot test based on a small number of CSI events (20 CSI events, 700 media documents). All coders examined the media documents. We refined the pro-forma and coding instructions until the informal, pilot study suggested an adequate level of agreement, namely over 90% (e.g., agreement about the category in which a CSI event would be placed, the most appropriate geographic location classifications, how to manage "ghost" events i.e., events which re-emerge in subsequent years etc.). Next, we formally assessed reliability using the percent agreement technique based on a sample of 100 CSI events; inter-coder agreement was high, as coders agreed on their coding results over 95% of instances.
(WMAC/AMAC) and the data on firm-specific financial characteristics which were employed as our main control variables (DataStream 5 ). Our final sample comprises 465 U.S. MNEs implicated in 2,401 disclosed CSI events in their home market (i.e., the U.S.), other Western markets, and/or non-Western markets. Our sample is formatted as a panel dataset defined by company-years -we have a total of 1,856 company-year observations. 6

Dependent variable: MNE corporate reputation
Our reputation measure is the WMAC/AMAC corporate reputation score following CSI disclosure -as mentioned earlier, we use the overall reputation score (calculated in the WMAC/AMAC survey as the arithmetic mean of the nine performance scorese.g., Love & Kraatz, 2017). "REPT" is a continuous dependent variable which measures the overall reputational score for a given company, in a given year.

Independent variables: CSI
ASSET4 classifies events into twenty CSI categories (Engelen and van Essen, 2012). These are: Financial CSI ("Shareholder rights CSI", "Earnings CSI", "Insider trading CSI", "Accounting CSI"); Consumer CSI ("Consumer harm CSI", "Product and service quality CSI", "Product recalls CSI"); Labor CSI ("Diversity and opportunity CSI"; "Freedom of association CSI"; "Employee health and safety CSI"; "Wages and working conditions CSI"); Human rights CSI ("Human rights CSI", "Child labor CSI"); Ethics CSI; Competitive CSI ("Intellectual property CSI", "Anti-competition CSI"); Environmental CSI ("Spills and pollution CSI"); Management CSI ("Management compensation CSI"); Public health CSI; and Taxation CSI. For each category of CSI, we added the total number of events noted in that category in a company-year; and then totaled all CSI events disclosed for each company-year to construct the independent variable, "Total CSI events" 7 . We considered that a CSI event at time t− 1 can be expected to be related to stakeholder reputation assessments at time t. 8 This measure enables us to explore the main relationship between CSI and reputation assessments.

Moderating variable
To examine the effects of CSI location, we split the sample into three categories: nuclear ingroup (i.e., U.S. home market), extended ingroup (i. e., Western host markets); and outgroup (i.e., non-Western host markets) CSI. In line with previous studies (notably, Epley & Caruso, 2004), we assume that stakeholder interpretations of CSI are influenced by biases, which is why individuals are often predisposed to categorize even the most complex situations into somewhat crude classifications, such as 'me'/ 'we'/ 'them'. The specific classification of host country locations into Western and non-Western markets is common, although what is defined as a typically "Western" market can differ based on the research purpose. In our sample, we use "Western" to refer to the "rich West," which generally includes the following host countries: Australia, Austria, Belgium, Canada, Finland, France, Germany, Iceland, Ireland, Italy, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom. We classify all other countries as "non-Western." Our main rationale here is that institutional rules, norms, and sources of legitimacy rooted in a given society are widely recognized to be values-based (Dahlmann & Grosvold, 2017). Hence, these countries share significant overlapping similarities in reference points regarding values, and therefore, many shared tenets within their overarching, location-based ideas around what constitutes (il)legitimate firm behavior. We use indicator variables to measure whether a CSI event occurred in the U.S. (1,0); a Western market (1,0), or a non-Western market (1,0).

Control variables
Prior studies have identified a potential 'home bias' regarding the financial performance of the MNE (e.g., Ke, Ng & Wang, 2010;Wolf, 2000). Namely, the often-greater financial importance associated with the home market may influence stakeholder perceptions and thus, the relationship between CSI and reputation. We therefore thoroughly control for aspects related to financial performance. Since large MNEs may be more visible and/or salient to our stakeholder groups of interest, we control for "Firm size" by utilizing a proxy of market capitalization; the variable is measured as the stock price multiplied by the total number of shares outstanding. Further, since high RandD intensity has been associated with more reputable organizations (Fombrun & Shanley, 1990); the variable "RandD intensity" is measured as the ratio of RandD expenditures to total sales. Higher(lower) levels of return on assets (ROA) and sales growth(decline) are indicative of positive(negative) MNE strategy, as well as favorable(unfavorable) organizational behavior, all of which may be associated with reputation assessments (e. g., Brammer & Pavelin, 2006;Brammer, Jackson & Matten, 2012;Roberts & Dowling, 2002). We therefore also control for "ROA," measured as the ratio of pre-tax profits to total assets, and "Sales growth, " measured as year-on-year growth in the firm's average sales volume. In turn, the leverage ratio, reflecting the degree of financial flexibility, may be negatively related to reputation as it could be perceived to reduce the firm's ability to strategize and even prioritize the management of reputation risks against other activities; accordingly, we control for "Leverage", measured by the ratio of total debt to total assets. The variables pertaining to firm-specific characteristics were collected from DataStream for each company-year at t− 1.
We further controlled for how well our sampled organizations score in areas associated with ESG (environmental, social, and governance) performance using data from Thomson Reuters' ASSET4. "Environmental performance" measures the overall degree to which a company uses best management practices to avoid environmental harm and risks (the category scores are: resource use, emissions, and innovation). "Social performance" (Nardella et al., 2020;Walker et al., 2010), is calculated by ASSET4 from four category scores associated with a company's commitment towards being a good citizen (the category scores are: human rights, workforce, community, and product responsibility). Lastly, "Corporate governance performance" measures the proportion of equity held by long-term institutional investor groups (i.e., pension funds, insurance companies, and life assurors) (Brammer & Pavelin, 2006;Ryan & Schneider, 2002); a strong presence of institutional investors signals that firm activity is well monitored, thus increasing confidence in firm behavior (the category scores here are: management, shareholders, and CSR strategy) 9 . ASSET4 measures each of these three variables on a scale from 0 to 100, with a higher value indicating higher performance scores in each E, S, G area. The data associated with ESG performance was collected for each company-year 5 In 2018, Thomson Reuters sold part of the data services to Blackstone Group LP, and the business was later formed as Refinitiv (previously known as Data-Stream and briefly, as EIKON). The dataset is now called Refinitiv Datastreamsee https://www.refinitiv.com/en. 6 In a given company year, a firm may have been associated with CSI in two or three location categoriesin this case, the dummy variables constructed for the location categories noted "1" in the relevant category. 7 Table 1 provides an overview of CSI categories included. 8 We note that a CSI event is coded as "1" in the dataset according to the year in which the event has emerged in the media and has been disclosed to the public. The only exception to this coding rule, are events which re-emerge as a result of a new lawsuit initiated against the company (in such instances, we, again, code the event as "1" for that respective company-year). 9 Full details on the ESG methodology as developed by Thompson Reuters are available here: https://www.esade.edu/itemsweb/biblioteca/bbdd/inbbdd/ archivos/Thomson_Reuters_ESG_Scores.pdf.
at t− 1. 10 Importantly, we also control for the effect of negative media valence, i.e., the total amount of critical media received by a firm in a given year. Prior studies have suggested that the nature and frequency of media reporting may strongly influence stakeholder perceptions (e.g., Kölbel et al., 2017). Some firms receive more media attention than others (e.g., tobacco companies, oil and gas companies, and technology companies), and international variation in media attention to CSI cannot always be captured by industry dummies. News media may also be biased to unfavorably depict instances of CSI, in particular, home market CSI. When doing so, the media may deploy negative lexical choices in their reporting as well as a greater frequency of news reports (Deephouse & Heugens, 2009). The content and frequency of negative press may then influence stakeholder perceptions. To measure negative media valence, we used the Linguistic Inquiry and Word Count software (LIWC). LIWC calculates the frequency and extent to which a body of text contains specific key words; the software codes words and phrases using underlying dictionaries developed in psychology and linguistics research and it is used to understand how word use leads to perception and behavior formation (see Tausczik & Pennebaker, 2010). We constructed separate Microsoft Word documents for each CSI event in a given year including all related media articles (after eliminating duplicate articles) and introduced these documents into LIWC. LIWC searched for terms that accurately identified negative emotions in the language used in the media reports 11 ; namely, the program searched the content of media articles in the English language and automatically calculated the percentage of words that pertained to 'negative affect' (see Choi & Lin, 2009 for details); Using LIWC, we therefore calculated the two main components of our negative media valence variable: (i) the cumulative percentage of words expressing negative affect, and (ii) the volume of media coverage. "Negative media valence" is therefore a continuous variable computed by multiplying the cumulative percentage of words in the media articles expressing negative affect by the overall volume of media coverage (i.e., the total word count of articles). Lastly, we included year (2005-2012) and industry (2-digit SIC codes) dummies in all the models.

Estimation methods
Before testing the hypotheses, we estimated a series of econometric models to confirm the validity of our results. All continuous variables are censored at the 1% and 99% tails to remove the impact of potential outliers. To detect autocorrelation in the residuals, we used the Durbin-Watson test: the statistical value of the test was 1.810, confirming that the assumption of instance independence has not been violated (per the rule of thumb, values between 1.5 and 2.5 are "normal"). The variance inflation factors (VIFs) range between 1.032 and 3.615, suggesting no serious problems of multicollinearity. As confirmed in Table 2, the independent sample t-test test shows statistically significant differences in firm characteristics and reputation scores amongst the three CSI location categories. As also shown in Table 2, the firms in our sample are large, with average sales growth between $110 and $120 million, and leverage ratios between around 16-19%, an average RandD intensity of around 2%, and a mean return on assets of 6-10%. These MNEs also score fairly high on social performance (64-74/100), environmental performance (61-77/100), and corporate governance performance (80-84/100).
Our dataset is an unbalanced panel, within which we have between one and eight observations (one per year) for each of 465 sampled companies. Repeated observation of particular units of analysis violates the independence assumptions of ordinary least squares estimation, necessitating the use of panel data techniques. Following prior research (e.g., Ballinger, 2004;Ren, Eisingerich & Tsai, 2015), we estimated the effects of CSI and corporate reputation using a Generalized estimating equations (GEE) approach. The GEE model allows for the modeling of correlated observations within firms resulting from the repeated measures across years. This estimation strategy assumes that coefficients can vary for each unit of observation i.e., each firm, recognizing therefore, that firm-specific characteristics are important. Consequently, reputational assessments attributed to the same firm are not considered 10 The inclusion of ESG ratings as control variables raises potential endogeneity concerns because a firm's ESG score might be reduced where social, environmental, or governance controversies are encountered. While ASSET4 scores, in contrast to other ESG ratings (e.g., the KLD), reflect irresponsible conduct to only a relatively minor extent (c. 25 indicators out of 250+), we also ran an additional set of models in which ASSET4 ESG control variables were removed. We found that our findings were robust to this change, indicating that endogeneity concerns do not bias our research findings. 11 An overview of LIWK, its language categories and how to use the software is presented in Tausczik and Pennebaker (2010). One of the key language categories in LIWK is "emotionality: positive and negative emotions"; the degree to which individuals express emotion, and the valence of that emotion can help understand reactions to important events. The emotions of interest in this paper were "sadness" and "anger"; hence, the percentage of words in each media article that pertained to "sadness" and "anger" were extracted for analysis. As explained in Tausczik & Pennebaker (2010:41), there are 184 words pertaining to "anger" (e.g., hate, kill, annoyed) and 101 words pertaining to "sadness" (e. g., crying, grief, sad) that the programme searched.
independent observations. In GEE models, the regression coefficients represent the mean change in the dependent variable for every unit of change in the independent, while other predictors are held constant. 12 Table 3 presents the regression results for CSI effects on MNE corporate reputation. Model 1 in Table 3 is the baseline model, including the moderator and control variables. Models 2 and 4 report results for the nuclear ingroup or the home market (i.e., U.S.) CSI sample, while Models 3 and 4 show regression results for the host market extended ingroup (i.e., Host Western) and outgroup (i.e., Host Non-Western) CSI samples. To start with, firm characteristics such as size, performance (e. g., ROA, sales growth) and low leverage are positively related to

Appendix 1
List of search terms.
Abuse* OR Accuse* OR Accusa* OR Accident* OR Anger* OR Anticompetit* OR Anticompet* OR Antitrust OR Attack* OR Axe* OR "Backdating stock options" OR Bad* OR Benevolen* OR Blame* OR Boycott* OR Breach* OR Break OR Bribe* OR Broke* OR Catastroph* OR "CEO compensation" OR Cheat* OR "child labour" OR "child labor" OR "compensation package" OR Complain* OR Contaminat* OR Controvers* OR "cooking the books" OR Corrupt* OR "creative accounting" OR Crime OR Criminal* OR Crisis OR Crises OR Cut* OR Damag* OR Danger* OR Deceive* OR Decept* OR Defect* OR Deforest* OR Detriment* OR Devastat* OR Died OR Dies OR Disaster* OR Discredit* OR Discriminat* OR Dishonest* OR Disput* OR Distort* OR Distrust* OR Endanger* OR Evad* OR Evasion OR "executive compensation" OR "executive pay" OR "executive salaries" OR "executive bonuses" OR Explod* OR Exploit* OR Explos* OR Expose* OR False* OR Fatal* OR Fault* OR Feud* OR Foul* OR Fraud* OR Harm* OR Hike* OR Hospitali* OR "human rights" OR Hurt OR Illegal* OR Ill-treat* OR Immoral* OR Inequal* OR Infring* OR Injure* OR "insider trading" OR "insider trades" OR "insider dealing" OR "insider deals" OR "insider dealings" OR Irresponsib* OR Litigat* OR Malic* OR Malevolen* OR Malfeasan* OR "management compensation" OR Manipula* OR Mend* OR Misconduct* OR Misdeed* OR Mislead* OR Misrepresent* OR Mistreat* OR Mistrust* OR Misuse* OR Neglect* OR Negligen* OR Negative* OR Offend* OR Outcry* OR Outrage* OR Overcharg* OR Penalt* OR Prevent* OR "price fixing" OR "price-fixing" OR Prosecut* OR Protest* OR Recall* OR Redundan* OR Repair* OR Repercussion* OR Restate* OR Revelation* OR Revolt* OR Risk* OR Ruin* OR Sabotage* OR Scandal* OR Settle* OR Shame* OR Shatter* OR Shock* OR Spill* OR Spoil* OR Strike* OR "stock options backdating" OR Sued OR Suing OR Suit OR Suspici* OR Tamper* OR Tarnish* OR Threat* OR Trust* OR Unequal* OR Uneth* OR Underage* OR Undermine* OR Unfair* OR Unlawful* OR Unsafe OR Uproar* OR Urgen* OR Victim* OR Violat* OR Violen* OR Weak* OR Wound* OR Wrongdoing 12 On the basis of a priori theoretical or statistical considerations, the specification of the model to be estimated requires the clarification of three main issues: (i) the statistical properties of the dependent variable, (ii) the form of the link function (a transformation of the dependent variable that allows estimation of the model) being used, and (iii) assumptions regarding the nature of the within-subject correlations. In our case, tests indicate that the assumption of normally distributed dependent variables cannot be rejected for our data, and therefore we assume our dependent variable is normally distributed. Further, regarding the link function, we deploy the identity link function in which the dependent variable is not transformed, in part, because this link function is robust to assumptions regarding the distribution of dependent variable and within-subject correlations. Lastly, with concerns to the within-subject correlation matrix, we adopt the default choice which adjusts the correlation estimates by the number of nonredundant parameters.
corporate reputation. Where significant, the negative effect of RandD intensity may mean that RandD investments take time to come to fruition and may come at the expense of exploiting immediate market opportunities. Consistent with Hypothesis 1, Model 2 in Table 3 shows that the relationship between CSI and reputation is negative and statistically significant (Model 2, β = − 0.039, p = 0.000) for the nuclear ingroup i.e., the U.S., home market. In Model 3, we show that CSI media disclosure also has a negative and statistically significant relationship with MNE corporate reputation when CSI threatens what we classified as stakeholders' extended ingroup, i.e., CSI occurring in a Western host market (Model 3, β = − 0.027, p = 0.000), and the effect is evidently weaker as per Hypothesis 2. As per Hypothesis 3, when the media associates CSI with an MNE's operations in a non-Western market, we did not find significant effects on reputation (Model 4, β = − 0.011, p = 0.233). This finding contrasts previous CSI research (e.g., Wang & Li, 2019) which theorizes that CSI represents an internationally consistent risk to corporate reputation. This result suggests that − insteadrisks to MNE reputation may oftentimes be location bound.
Overall, in interpreting our findings, we highlight that the relationship between CSI location and corporate reputation is certainly more nuanced than prior studies have implied over time (Hall, 1993;Wang & Li, 2019). In line with research on CSI in an international context, which suggests that CSI located away from the home market can be associated with reduced risks for MNEs (Surroca et al., 2013), we generally find that CSI has the strongest negative relationship on corporate reputation when it is home-based. With regards to international CSI, we further find strong evidence that CSI affecting a related ingroup of stakeholders (i.e., in Western host markets) has substantively greater reputational effects than CSI affecting outgroups (i.e., in non-Western host markets). At this point, our results provide strong support to extant theory concerning the reduced influence of 'outgroups' on stakeholder perception (c.f. Antonetti & Maklan, 2018;Michailova et al., 2017), and theory regarding 'polycentric' favoritism between similar populations (c.f. Perlmutter, 1969). Also, our findings have implications for the literature concerned with the application of ethnocentric bias to real world phenomena (Michailova et al., 2017) and the social regulation of CSI (Brammer et al., 2021;Buckley, 2018;Devinney, 2011). We point directly to the significant heterogeneity in stakeholder responses to CSI, which oftentimes, appears insufficient in penalizing MNEs when they are accused of behaving irresponsibly.

Robustness checks
We subjected our results to a series of robustness checks. First, we recognize that our categorization of "U.S.", "Western" and "non-Western" markets represent one way in which location could be categorized. Therefore, we conducted robustness checks by sub-sampling according to institutional quality/distance scores, which are expected to quantify the (differences in) quality of institutions for MNE home and host markets. We collected data from the Economic Freedom of the World

Appendix 2a
Robustness checks -GEE results for the institutional quality sub-samples. Industry and year dummies are included but not shown. a Variable is a logarithm.

Appendix 2b
Robustness checks -GEE results for the institutional distance sub-samples. Industry and year dummies are included but not shown. a Variable is a logarithm.

Appendix 3
Robustness checks -GEE results for the relationship between specific types of CSI events and corporate reputation, moderated by CSI location. (2) legal structure (including property rights enforcement); (3) access to sound money; (4) freedom to engage in international trade; and (5) regulation (i.e., of credit, labor, and business). We used EFW's overall score aggregating the five components (due to the five dimensions being highly correlated) and sub-sampled according to (1) the institutional quality of the locations associated with CSI (Appendix 2a), and (2) the institutional distance between home (U.S.) and host locations (Appendix 2b). As shown in Appendix 2a, our results hold, even when we classify location as "High quality" and "Low quality" institutional environments i.e., we observed a weaker effect of CSI on reputation, when CSI occurs in the "low institutional quality" location. These resultswhich generally show relatively weaker negative effects or positive reputation effects for the "low institutional quality" sub-samplesupport our findings in that, a strong breach of stakeholder values tends to occur for events that take place in higher quality institutional environments. In turn, CSI may be less questioned in locations where, for instance, poverty is a more significant threat to wellbeing, and events such as child labour, diversity, opportunity or working conditions violations, have become more commonplace, part of the 'way' of doing business in those markets, or, at the very least, ambiguous. With regards to our institutional distance subsamples, we found that moving from lower to higher quality locations (referred to as "positive distance") has a stronger moderating effect on the negative relationship between CSI and reputation. In future studies, scholars could further explore different thresholds of distance; overall, objective measures of distance alone may not capture how MNE stakeholders, in practice, perceive and categorize the social world (see also Bargh, 1994). Second, because our dataset is unbalanced, with various years for the unit of observation (i.e., the MNE), we re-ran our models for a balanced panel of firms for which we had complete data. Findings remained consistent, suggesting that results are not skewed by the numbers of observations available.
Third, previous studies (e.g., Clarke et al., 2021;Lange & Washburn, 2012) have suggested that there may be heterogeneity between how the different categories of CSI themselves are perceived by organizational assessors, and thus penalized. We also unpack in our robustness analysis between the categories of CSI events that have been disclosed to the public by the media (Appendix 3). We observed that many 'types' of CSI negatively affect reputation, whilst there are other CSI types which show no, or even beneficial effects. Most CSI events were negatively linked to reputation assessments for the nuclear and extended ingroups. Worth noting is that we identified some Financial CSI types to pose a risk to MNE reputation, even when associated with outgroups i.e., non-Western markets (e.g., "Insider trading": β = − 0.769, p = 0.000; "Accounting": β = − 0.565, p = 0.026). We also uncovered positive relationships to corporate reputation following some CSI events (e.g., Child Labor and Freedom of Association CSI, respectively). One interpretation for these findings is that firms may be motivated to better respond to accusations of child labor CSI and freedom of association controversies due to the topicality and scrutiny that typically follows these events (Caruana et al., 2021). Effective management of CSI may then lead to reputation betterment (Coombs, 2007). An alternative interpretation of these results might be that some instances of CSI are diagnostic of efficiency management practices (e.g., Anderson, 2000;De Schutter, 2010;Massarani, Drakos & Pajkowska, 2007), which may be perceived positively among stakeholder groups, such as those included in our sample.
Fourth, CSI characteristics such as whether the firm was found culpable for the associated event may also influence corporate reputation (see Lange & Washburn, 2012). We control for culpability using an indicator variable that equals "1" when the firm was found legally culpable for CSI, and "0" otherwise. Our results suggest that legally determined culpability has an overall negative, but insignificant effect on reputation (e.g., for U.S. sub-sample: β = − 0.084, p = 0.204). Large, prominent firms such as MNEs face increased public scrutiny, which may mean that being accused of CSI weighs significantly to stakeholders, irrespective of whether culpability is established by formal regulators.

Discussion
By behaving irresponsibly, organizations are assumed to place their corporate reputations at risk (Fiaschi et al., 2017;Wettstein et al., 2019;Wang & Li, 2019). For MNEs, the complexity of their international operations exposes them to accusations of CSI in domestic and international markets. Even so, IB research had yet to explore whether 'location matters' for CSI and its organizational outcomes.
This study makes several theoretical and practical contributions towards better understanding MNE CSI. First, our analysis reveals that CSI location may trigger ethnocentric biases, resulting in home market CSI risking corporate reputation most significantly. Therefore, we reveal that the location of CSI does, indeed, matter. By exploring the location of CSI, we also contribute to theory on what we may refer to as 'international CSI' i.e., CSI in an international context (Cuervo-Cazurra et al., 2021;Strike, Gao & Bansal, 2006). We extend theory concerning CSI by adding a behavioral lens to our understanding of how stakeholders perceive and respond to MNE CSI differently, depending on the location of CSI behaviors observed. The notion that CSI represents consistent risks to MNE reputations and performance had mostly neglected the observation that MNE stakeholders are still, in fact, largely bound to (and therefore biased by) a home location, as well as informed by the overarching rules, values and ideas around legitimate behavior propagated in the home institutional environment.
Second, and relatedly, our study enriches prior conversations in the IB literature (Rabbiosi & Santangelo, 2019;Surroca et al., 2013;Wang & Li, 2019)  Industry and year dummies are included but not shown. a Variable is a logarithm. b Since a firm may have multiple events in different locations in a given year, the "home U.S." sub-sample only includes events identified in the U.S. and not in other locations.
home environment, which is why MNE stakeholders tend to be biased to remain most vigilant to, and assess most knowledgably, CSI which occurs in that home environment. Home-based CSI is likely better understood, and thus, tends to be perceived as more severe and deserving of penalties. Importantly therefore, by providing the first large-scale empirical study to view CSI as an ethnocentrically biased phenomenon, we make a strong case that corporate reputations may not always be significantly influenced by internationally located CSI. In particular, stakeholders appear unable/unwilling to respond to CSI arising in international markets characterized by relatively divergent institutional rules, norms, and legitimacy pressures. Third, this study extends our conceptual understanding of ethnocentric biases in IB and management (Bohas et al., 2021). Here, we developed and furthered an alternative explanation for the effect of ethnocentrism that goes beyond simplistic moral reasoning rationales and the idea that ethnocentric biases mainly emerge out of an expression of superiority. We proposed that increased stakeholder knowledge and shared ideas around what is (il)legitimate and (in)acceptable may undergird and intensify ethnocentric perception. The evidence we provide largely supports our conceptualization that biases can emerge because of the heightened tendency to assess more knowledgably, and respond to, individual and ingroup sources of risk (Epley & Caruso, 2004). As CSI can expose decision-makers -as well as their ingroup -to various threats/harms, ethnocentric biases arise when CSI influences the home market, and comparatively less so when CSI is associated with a distant and oftentimes 'foreign' international market. By nuancing CSI location, we also find evidence which suggests that convergence between institutional values strengthens (negative) perception of MNE conduct when compared to MNE (CSI) behavior that occurs in markets where values diverge. Therefore, we provide evidence of a 'polycentric' (Perlmutter, 1969) tendency to respond strongly to CSI that affects a related ingroup of stakeholders, particularly when institutional rules, norms, and values between regions are similar/shared. In sum, we attended to calls for a more contemporary conceptualization of ethnocentric biases for IB and management (Michailova et al., 2017) by proposing that the alignment/divergence between location-based sources of legitimacy may, at least in part, explain ethnocentric behavior.

Key managerial and policy implications
From a managerial perspective, our findings have practical implications for the global strategic management of socially constructed resources (Buckley et al., 2022), particularly corporate reputation. As managers have become increasingly concerned with perception management challenges, it is especially pertinent for them to better understand the risks posed by CSI. Our findings imply that perceptual biases towards the home market location and those international locations with significant similarities may expose MNEs to serious perception management challenges in some contexts, but not others. For practicing managers, our results suggest that CSI which emerges in a relatively distant host country location may not necessarily be negatively related to corporate reputation. Hence, divesting or disassociating the firm from CSI in host markets may (potentially) not always be the most appropriate strategic response. At the same time, when CSI is more likely to go unpenalized, the firm should still consider self-regulation from a CSR perspective . This is particularly relevant for CSI that is known to arise in locations which lack a clear instruction from formal and informal regulators around what constitutes irresponsible behavior and how it should be addressed. Extant research has shown the many benefits associated with foreign firms pursuing socially responsible actions in foreign locations to avoid the negative spillovers of their activities (c.f. Marano, Tashman & Kostova, 2017;Parente et al., 2019). Future investigators can -with increased accuracy -map the (likely) complex and multi-domestic nature of socially constructed intangible resources, such as corporate reputation, to guide appropriate corporate actions.
From a policy perspective, our research provides important insights concerning the social regulation of international enterprise behavior (Buckley, 2021;Devinney, 2011;Nardella et al., 2022). In both academic works and the media, it is expected that corporate reputation "serves as an informal enforcement mechanism" (Atanasov, Ivanov & Litvak 2012: 2215 to penalize, and hence, deter MNEs from behaving irresponsibly (Aguilera et al., 2015;. However, our findings illustrate that outsourcing responsibility to informal, societal regulators to penalize CSI may not always be an appropriate strategy for policymakers. In fact, in certain conditions -such as when CSI occurs 'away' from home -social regulation may fail to deter irresponsible MNE behavior. As such, questions emerge around the appropriateness of continuing to outsource the regulation of CSI to local, host market institutions, particularly when such institutions are characterized by significant voids. Organizational actors may engage in irresponsible or questionable business practices, oftentimes because of the perceived futility of opposing institutional voids. We propose that more nuanced (home market) policies will help to deter MNEs from engaging in CSI where differences between institutional legitimacy expectations and (potentially) voids can minimize the downside risks for MNEs.

Limitations and future research
Naturally, our research is not without limitations, some of which could be the focus of future studies. For instance, our research focused on a distinct set of MNE stakeholders, i.e., managers and market analysts. Whilst we include various controls to account for the financial bias which could be associated with our study, more research is needed to uncover the perceptions of other stakeholders, such as customers and employees, amongst many other MNE stakeholders. Furthermore, we note, as recent research has highlighted (Mariconda, Zamparini & Lurati, 2021), that it would be of particular value to tease out the temporal dimension more broadly in MNE reputation research. Time and the MNE's (potentially extensive) history of CSI may influence when, and to what extent, stakeholder perceptions, such as negative reputational assessments, are formed. In this study, our dataset takes into consideration the temporal dimension of reputation by assuming that CSI in year t− 1 will influence stakeholders' reputational assessments in year t; we do not, however, consider the specific point in the year t− 1 when the CSI event is disclosed. A longer delay between CSI and the measurement of corporate reputation may influence the strength of the impact, leading to potentially weaker reputation effects in the presence of stakeholder forgetfulness (Mena et al., 2016), or stronger effects for highly recent CSI, for which stakeholder memory is likely stronger. Hence, exploring the temporal dimension of MNE reputation, we propose, may be a valuable future undertaking.
Whilst our study placed emphasis on better understanding the effects of CSI location on MNE reputation, stakeholders from different geographies may be exposed to multiple sources of information salient to their assessments. Although it is beyond the scope of this study to explore the effect of multiple sources of information on corporate reputation, it may be valuable to further explore the information generated by bodies such as regulators, NGOs, and social media, as well as the implicated organizations' communicative and remedial responses to CSI.
Another potential limitation of our study arises from CSI complexity. The robustness analysis we present suggests some heterogeneity in how stakeholders respond to CSI types. We focused on the heterogeneity arising from CSI location; however, more work is needed to enhance the theoretical clarity and predictive value of an ethnocentric bias perspective. It may be that a given stakeholder can belong to a variety of 'ingroups' at a given point in time, each with their own rules, norms, and values (Besharov & Smith, 2014). Future research, we believe, should specifically focus on whether the notion of the 'ingroup' exists simultaneously at the stakeholder, national and/or supranational levels, potentially leading to varying stakeholder responses to CSI behaviors.
Finally, we have theorized and shown that it is necessary to explain how the location of MNE behavior influences responses to that behavior. Alternative avenues might include the exploration of other CSI measures, such as RepRisk and disaggregated measures of corporate reputation, such as RepQuotient. Further contributions might be generated by exploring the concurrent influence of location and stakeholder perception of socially 'responsible' MNE behavior. In our view, CSR, ESG, and non-market strategy scholars (e.g., Jamali et al., 2017;Lee et al., 2020;Sun et al., 2021) may find value in examining the effects of where philanthropy, sustainability, and social responsibility activities of MNEs are located. In providing novel theorizing to extend the ethnocentric bias perspective, we have endeavored to provide scholars with a framework to explore the location of MNE (mis)conduct. The abovementioned limitations represent, in our view, valuable future research opportunities. By considering the location of MNE behavior as a key moderator of the relationship between CSI and corporate reputation, we hope to inspire further debate on the dark side of international business.

Conclusions
Corporate social irresponsibility and its outcomes remains a topical and growing area of IB research. Previously, it was assumed that MNEs significantly risk their reputations when behaving irresponsibly in their global operations. In this study, we examined how the location CSI events are related to corporate reputation. We found that MNE stakeholders are ethnocentrically biased in their responses to CSI. A key highlight of our study is that CSI's outcomes are contingent on location, with CSI presenting the most severe reputation risks when it emerges in the home location and significantly less so when CSI occurs abroad. Indeed, vis-à-vis the reputational outcomes of CSI, our study suggests that what happens abroad, can largely stay abroad.

Declaration of Competing Interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.