Signaling legitimacy across institutional contexts-The intermediary role of corporate social responsibility rating agencies

: Research Summary Good corporate social responsibility (CSR) ratings can increase a firm’s legitimacy and reduce its default risk. Yet, the interpretation of CSR varies between different countries. We investigate whether CSR ratings have a risk‐mitigating effect across different institutional contexts. We find that good CSR ratings have a general risk‐mitigating effect. Yet, we also find that the effect decreases when the rating agency is embedded in the institutional context of its home country and the rated firm operates in a country with a different culture or regulatory system. This suggests that a rating agency’s country of origin and its embeddedness in that country’s context play an important role in the relationship between CSR ratings and default risk. Managerial Summary In this article, we investigate to what extent CSR rating agencies signal organizational legitimacy across institutional contexts. Specifically, we evaluate how institutional distance between the rating agency and the rated firm moderates the negative effect of the CSR rating on the firm’s default risk. We compare two CSR rating agencies, OEKOM, which is strongly embedded in its institutional context, and ASSET4, which operates across diverse institutional contexts. Based on a panel of 604 firms in 13 countries from 2011 to 2016, we show that CSR ratings from both rating agencies have a risk‐mitigating main effect. Yet, only in the case of OEKOM is the effect negatively moderated by regulatory and cultural distance. Thus, the legitimating effect of institutionally embedded CSR rating agencies decreases with institutional distance. Abstract Research Summary : Good corporate social responsibility (CSR) ratings can increase a firm's legitimacy and reduce its default risk. Yet, the interpretation of CSR varies between different countries. We investigate whether CSR ratings have a risk-mitigating effect across different institutional contexts. We find that good CSR ratings have a general risk-mitigating effect. Yet, we also find that the effect decreases when the rating agency is embedded in the institutional context of its home country and the rated firm operates in a country with a different culture or regulatory system. This suggests that a rating agency's country of origin and its embeddedness in that country's context play an important role in the relationship between CSR ratings and default risk. Managerial Summary : In this article, we investigate to what extent CSR rating agencies signal organizational legitimacy across institutional contexts. Specifically, we evaluate how institutional distance between the rating agency and the rated firm moderates the negative effect of the CSR rating on the firm's default risk. We compare two CSR rating agencies, OEKOM, which is strongly embedded in its institutional context, and ASSET4, which operates across diverse institutional contexts. Based on a panel of 604 firms in 13 countries from 2011 to 2016, we show that CSR ratings from both rating agencies have a

negative effect of CSR ratings on default risk is most pronounced when the rated firm originates in the same country as the rating agency, and, conversely, that the negative effect decreases as the institutional distance between the rating agency and the rated firm increases.
Our empirical analysis relies on a panel data set of 604 firms from 13 countries between 2011 and 2016. We use data from two different CSR rating agencies. The first agency is OEKOM, which originates in, and has been firmly embedded in, the German institutional context. The second is ASSET4, which originates in Switzerland, but has a more international character and is not embedded in any particular institutional context. The results show that high CSR ratings from either of the two agencies reduce default risk for the international sample of rated firms. However, in the case of OEKOM, this effect is moderated by differences in the institutional context between Germany-the rating agency's home country-and the home country of the rated firm. Specifically, the risk-reducing effect is decreased by regulatory and cultural distance from Germany. For the ASSET4 data, we do not find a similar moderating effect.
This study makes two interrelated contributions to the international business literature. First, we demonstrate that the CSR insurance effect is partially, but not entirely, moderated by context-specific conceptualizations of CSR. On the one hand, the relationship between CSR ratings and default risk seems to be driven by widely accepted CSR practices, which is in line with prior studies on the insurance effect (Godfrey et al., 2009;Koh et al., 2013;Shiu & Yang, 2017) and with the view that CSR is increasingly being standardized (Christmann, 2004;Fortanier, Kolk, & Pinkse, 2011). On the other hand, the relationship is moderated by regulatory and cultural distance between the rating agency and the rated firm, when the rating agency is embedded in a particular institutional context. This lends support to the view that CSR is shaped by local norms and pressures (Campbell et al., 2012;Crilly et al., 2016;Ioannou & Serafeim, 2012). This study reconciles these two literatures by suggesting that the insurance effect of CSR ratings is greatest within one homogeneous institutional context. But since there are widely accepted CSR practices that are consistently reflected in different CSR ratings, the insurance effect of a particular CSR rating also persists in cross-national settings.
Second, we highlight the role of CSR rating agencies as institutional intermediaries (Doh et al., 2010;Slager et al., 2012). Prior work emphasizes the importance of institutional intermediaries in representing and standardizing institutional norms (Deephouse & Suchman, 2008;Durand & McGuire, 2005;Rao, 1994). While institutional pressures normally favor a convergence of performance criteria, Chatterji et al. (2016) highlight the problematic fact that CSR ratings greatly and persistently diverge in their definitions of CSR. We address this problem by examining institutional intermediaries facing cross-national heterogeneity of norms. Our results suggest that CSR rating agencies cannot develop one CSR performance standard that reflects all institutional contexts equally well. Instead, it seems that agencies must choose between representing one specific context, attempting to average across different contexts, or using multiple context-specific standards. The fact that these various choices exist helps to explain why the assessment results of CSR rating agencies currently continue to diverge.

| CSR, LEGITIMACY, AND DEFAULT RISK
Corporate social responsibility has become an important and diverse field of management research (Aguinis & Glavas, 2012;Garriga & Melé, 2004). A growing stream within this field takes an institutional theory perspective on CSR (Brammer et al., 2012;Campbell, 2007;Matten & Moon, 2008). In this article, we adopt this institutional theory perspective to investigate the relationship between CSR, legitimacy, and default risk, with a particular emphasis on the role of CSR rating agencies. By framing CSR rating agencies as institutional intermediaries, we synthesize a body of literature that studies how the interpretation of CSR varies in cross-national settings and a body of literature that empirically demonstrates the risk-mitigating effects of CSR performance standards.
From an institutional theory perspective, CSR is a construct that means different things in different places since it is shaped by regulatory, normative, and cognitive pressures (Brammer et al., 2012;Campbell, 2007). As institutional pressures vary across institutional contexts, CSR practices differ from context to context (Ioannou & Serafeim, 2012;Jamali & Mirshak, 2007;Miska et al., 2016). This leads to a diversity in the practice of CSR that is particularly salient in cross-national settings. For example, CSR practices differ between the United States and Europe (Doh & Guay, 2006;Matten & Moon, 2008). Also, multinational corporations adapt their CSR practices when venturing to other countries (Campbell et al., 2012;Rathert, 2016), and stakeholders in different countries respond in heterogeneous ways to the same CSR practices (Crilly et al., 2016).
We follow Ioannou and Serafeim in adopting a broad definition of CSR as "a business organization's configuration of principles of social responsibility, processes of social responsiveness, and policies, programs, and observable outcomes as they relate to the firm's societal relationships" (Wood, 1991, p. 693). This definition is ideal for investigating CSR in a cross-national setting because to the phrase "societal relationships" implies that CSR may comprise different sets of practices for different societies. In that regard, the definition reflects the context-specific nature of CSR, which is central to a comparative institutional analysis of CSR practices.
Gaining and maintaining legitimacy is a key reason firms practice CSR (Chiu & Sharfman, 2011;Godfrey, 2005;Kostova & Zaheer, 1999). Legitimacy is 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, p. 574). By legitimacy, we specifically mean sociopolitical legitimacy, a type of legitimacy that refers to the compliance with legal rules and moral norms that govern a firm's relationship with society (Bitektine, 2011). This includes both the social and the ecological domains, for example, compliance with labor and environmental regulations, and adhering to what is perceived as socially and environmentally acceptable by customers, employees, and other stakeholders.
Legitimacy, in turn, increases a firm's probability of survival (Oliver, 1991;Ruef & Scott, 1998;Suchman, 1995). Organizations that satisfy culturally approved criteria and enjoy the support of surrounding institutions have a higher likelihood of surviving, sometimes irrespective of their performance and stability in other dimensions (Meyer & Rowan, 1977;Scott, 2008). Conversely, organizations that lack legitimacy are fundamentally at risk. Organizations that lack legitimacythose that violate widely accepted norms and understandings and are discredited by respected authorities-have a decreased chance of survival (Dowling & Pfeffer, 1975;Zuckerman, 1999). Thus, the institutional theory perspective on CSR suggests that CSR enhances a firm's chances of survival by increasing its legitimacy.
There is a body of empirical literature that supports these theoretical arguments by demonstrating that CSR has risk-mitigating effects. Often referring to the "insurance hypothesis" (Godfrey, 2005), a string of studies has provided evidence that CSR reduces a variety of financial risks (Chava, 2014;Cheng et al., 2014;El Ghoul, Guedhami, Kwok, & Mishra, 2011;Godfrey et al., 2009;Goss & Roberts, 2011;Koh et al., 2013;Shiu & Yang, 2017). These studies provide strong and consistent evidence that firms that practice CSR are perceived as less risky investments by investors.
The empirical studies most directly related to the concept of firm survival are those that investigate the effect of CSR on default risk (Drago, Carnevale, & Gallo, 2019;Chava, 2014;Goss & Roberts, 2011). Default risk is defined as the risk that a firm "will fail to meet its obligations in accordance with agreed terms" (Basel Committee on Banking Supervision, 2000). It reflects investors' assessments of the probability that a firm will survive and remain solvent until funds it has borrowed are due to be paid back. Given that fixed income investors commonly lend to firms with a time horizon of several years, default risk inherently reflects a firm's probability of surviving in the long term. High default risk indicates that investors have doubts about whether a firm will survive, or whether it might instead file for bankruptcy before its debt investment matures. Bankruptcy is a form of organizational death (Sheppard, 1994), that is, the opposite of firm survival. Low default risk indicates that investors are fairly certain that a firm will survive until its debt investment matures. Default risk, thus, reflects a firm's likelihood of survival, as it is perceived by its creditors.
However, there is a yet unsolved tension in the literature regarding the intermediary role of CSR rating agencies. On the one hand, the empirical studies testing the insurance hypothesis are consistent with an institutional theory perspective on CSR. The fact that CSR is associated with lower default risk is consistent with the view that CSR enhances a firm's legitimacy and, consequently, its probability of survival. On the other hand, empirical studies on CSR rely on CSR ratings to test the authors' hypotheses, implicitly assuming that there is one standard of CSR against which all firms can be evaluated. This contrasts with the institutional theory perspective on CSR, which emphasizes that CSR is a fluid construct that is shaped by institutional pressures. Several empirical studies that study international samples have already highlighted that the effects of CSR ratings on financial performance are contingent on institutional context (Ghoul, Guedhami, & Kim, 2017;Hoepner, Oikonomou, Scholtens, & Schröder, 2016). However, the reliance on a particular CSR rating in cross-national settings implicitly assumes that CSR can be measured against a single standard, despite the strong evidence that CSR is a context-dependent construct (Gond & Crane, 2010).
To ease this tension, we generate hypotheses that include CSR rating agencies in the relationship, and compare two different CSR rating agencies. We argue that CSR rating agencies do not simply measure the construct of CSR, but also shape its meaning in response to the rating agencies' institutional contexts. In the following hypotheses development, we refine the theoretical link between CSR, legitimacy, and default risk, by including CSR rating agencies as institutional intermediaries.

| CSR ratings and default risk
Corporate social responsibility rating agencies are institutional intermediaries (Doh et al., 2010;Rao, 1994), which formalize norms and expectations vis-à-vis CSR into a performance standard, and provide assessments in the form of CSR ratings. The main audience for CSR ratings are investors, who commonly pay for the rating agencies' services (SustainAbility, 2017). However, CSR rating agencies also cater to the rated firms and to the wider public (Scalet & Kelly, 2010).
As institutional intermediaries, CSR rating agencies perform three related functions. First, CSR rating agencies reduce information asymmetry between firms and investors by revealing the latent quality of a firm's CSR practices. In the absence of rating agencies, it would be difficult for outsiders to assess whether and how well a firm is practicing CSR because firms hold superior information about their CSR and may choose whether to disclose it (Lyon & Maxwell, 2008). CSR rating agencies develop screening methodologies to evaluate and communicate firms' CSR practices.
Second, CSR rating agencies create CSR performance standards, which are implicit in the rating methodology. The basis of any assessment of CSR is a definitive interpretation of what CSR means (Chatterji et al., 2016). This interpretation is formalized in the rating methodology, which determines which specific CSR practices are included in the assessment, how they are evaluated, and how they are weighted and aggregated into a rating score. Given that CSR is a broad construct, rating agencies have considerable latitude on how to interpret it (Delmas & Blass, 2010;Delmas, Etzion, & Nairn-Birch, 2013). For example, there are large differences in the weights that CSR rating agencies assign to social and environmental aspects of CSR (Capelle-Blancard & Petit, 2017). Chatterji et al. (2016) show that CSR ratings from different providers do not always converge because each rating agency defines CSR in its own particular way. Consequently, a CSR rating score reflects both a measure of CSR, as well as the rating agency's interpretation of what CSR means.
Third, a CSR rating agency signals the legitimacy of rated firms, thereby playing the role of a "legitimating agency" (Deephouse & Suchman, 2008;Durand & McGuire, 2005;Durand, Rao, & Monin, 2007). By creating a CSR performance standard that reflects societal norms and expectations, CSR rating agencies spell out which concrete CSR practices are deemed legitimate. Accordingly, firms that perform well against this CSR performance standard are legitimate because they conform to societal norms and expectations. The same principle of third-party legitimation has been observed in other domains: Michelin stars signal to what extent chefs and restaurants conform to the norms of French haute cuisine (Durand et al., 2007), and the Association to Advance Collegiate Schools of Business (AACSB) signals to what extent business schools meet the expectations of students, employers, and faculty (Durand & McGuire, 2005). In each case, a high score suggests that the rated organization is behaving in ways that are legitimate in the eyes of its key constituencies and stakeholders. In the case of CSR ratings, a high score signals that a firm's CSR practices are considered legitimate by its employees, customers, and investors, as well as by regulators, NGOs, and society at large.

| Signal fit
The fact that CSR ratings signal legitimacy forms the basis of the relationship between CSR ratings and default risk. CSR ratings signal legitimacy to investors, and investors prefer legitimate organizations because they are more likely to be able to survive until borrowed funds are due to be paid back. Legitimate firms are more resilient for two main reasons. First, legitimate firms have preferred access to resources (Meyer & Rowan, 1977;Oliver, 1991). For instance, firms with high CSR ratings find it easier to attract talent (Bhattacharya, Sen, & Korschun, 2008). Also, other organizations are more willing to provide resources to legitimate firms, as this allows them to be associated with high legitimacy actors (Ruef & Scott, 1998). Second, legitimate firms are more likely to be trusted (Zucker, 1987), which is an important asset in times of crisis. For example, firms that are involved in accidents are penalized less when they are considered legitimate (Godfrey et al., 2009). Firms that had accumulated trust through CSR also emerged from the financial crisis of 2008 with less damage than firms that had not accumulated trust (Lins, Servaes, & Tamayo, 2017). For these reasons, high CSR ratings are associated with low default risk.
However, there is likely to be variation in the extent to which CSR ratings signal legitimacy. This is because CSR rating agencies develop standards of CSR performance, even though different CSR practices are associated with legitimacy in different institutional contexts (Crilly et al., 2016). To explore this variation, we define the extent to which CSR ratings are correlated with a firm's underlying legitimacy as "signal fit" (Connelly, Certo, Ireland, & Reutzel, 2010). Signal fit describes the correlation between a firm's CSR rating and a firm's effective legitimacy in the institutional context in which the firm is operating. Signal fit is high when a rating agency's CSR performance standard represents the rated firm's stakeholder norms and expectations vis-à-vis CSR. In this case, a strong CSR rating score signals legitimacy. Signal fit is low when a rating agency's CSR performance standard does not represent the rated firm's stakeholder norms and expectations vis-à-vis CSR. In this case, neither a high nor a low CSR rating score signals organizational legitimacy. The concept of signal fit allows us to formulate hypotheses about the relationship between specific CSR ratings (rather than a general concept of CSR) and default risk. In the following hypotheses, we explore the conditions that are likely to increase or decrease this signal fit.

| Widely accepted CSR practices
Signal fit is increased by the presence of widely accepted CSR practices. Assuming that CSR has completely opposite meanings in two different countries, a CSR rating score that has a high signal fit for one of the countries would have a signal fit of zero for the other one. However, to the extent that there are certain CSR practices that are consistent with each of the two country's specific interpretations of CSR, then the CSR rating score that has a good signal fit for one of the countries will also have a minimal level of signal fit for the other country.
Corporate social responsibility practices have increasingly converged at the international level (Christmann, 2004;Fortanier et al., 2011;Slager et al., 2012). Fortanier et al. (2011) show that the efforts of supranational institutions, such as the United Nations, the International Labour Organization, and the International Organization for Standardization (ISO), increasingly harmonize the CSR activities of firms from different countries. Even more directly, organizations such as the sustainability accounting standards board seek to standardize and formalize CSR reporting globally, focusing on the aspects that are consistently relevant to stakeholders (Khan, Serafeim, & Yoon, 2016). These standardization efforts are likely to drive diverse and context-specific CSR practices toward more uniform and widely accepted practices.
Provided that there are widely accepted CSR practices, CSR ratings are likely to reflect them. First, CSR rating agencies themselves are active participants in the dialogue about the standardization of CSR reporting (Slager et al., 2012). Second, CSR rating agencies source a considerable amount of their data from firms' own CSR reporting (SustainAbility, 2017), meaning that CSR practices that are consistently reported are likely to be part of the CSR rating score. Therefore, in the presence of widely accepted CSR practices, we expect that CSR ratings have high signal fit and signal a firm's legitimacy. Based on the outlined relationship between legitimacy and default risk, this suggests a general negative effect of CSR ratings on default risk. H1: CSR ratings have a negative effect on default risk.

| Context-dependent interpretations of CSR
In contrast to the increasing global standardization of CSR practices, signal fit is reduced by contextdependent interpretations of CSR. Many firms practice CSR in a context-specific way that depends on the country where they originate and where they establish new operations (Aguilera & Jackson, 2010;Campbell et al., 2012;Miska et al., 2016;Rathert, 2016). The explanation for this diversity is that different firms are shaped by different institutional pressures.
This line of reasoning can be applied to CSR rating agencies. Rating agencies, just like any other organization, are shaped by the institutional context in which they originate and in which they are embedded. The institutional context is especially influential for institutional intermediaries because their role is to formalize the institutional pressures that surround them. Thus, the institutional context that a rating agency is embedded in is likely to leave an imprint on the way in which CSR is interpreted and formalized in its rating methodology (Durand & McGuire, 2005).
To the extent that a CSR rating agency is embedded in a particular institutional context, the signal fit of its ratings is high-within that institutional context. However, the signal fit deteriorates if the rated firm operates in a country other than where the rating agency is based.

| Regulatory distance as a moderator
The regulatory environment has an important influence on the way in which CSR is practiced (Campbell et al., 2012;Reimann, Rauer, & Kaufmann, 2015). A country's regulations are a prominent feature of the institutional context and a common differentiator in management studies (Ioannou & Serafeim, 2012). Faced with similar or identical environmental and social issues, different countries take different regulatory approaches to address these issues. Accordingly, we define regulatory distance as the degree to which regulations in the environmental and social domains are different between two institutional contexts.
As Campbell et al. (2012) observe, CSR commonly comprises practices that go beyond legal requirements and fill existing institutional voids. Therefore, different laws and regulations immediately change the set of activities that are recognized as CSR activities. Consider, for example, the provision of healthcare for employees by firms. In countries with mandatory health insurance, offering basic healthcare to employees will be seen as less valuable as compared to countries where workers do not have health insurance by default (Matten & Moon, 2008). Many other types of regulation, such as pollution control or subsidies for environmentally friendly technologies, can turn what is considered a specific CSR practice in one country into a practice that is merely an act of compliance in another country.
The regulatory environment of a rating agency's home country is thus likely to be reflected in the agency's rating methodology. On this basis, we propose that regulatory distance between the home country of a CSR rating agency and the home country of a rated firm decreases the signal fit of CSR ratings. The greater the regulatory distance, the greater the mismatch between the regulatory environment that is embedded in the CSR rating and the regulatory environment in which the firm is operating will be. For instance, a firm might receive a high rating score for using energy saving equipment. When the use of this equipment is mandated by local law, it is questionable whether the firm's legitimacy has increased. Consequently, as regulatory distance increases, CSR ratings increasingly fail to signal a firm's legitimacy. We therefore propose that when a rating agency is firmly embedded in a national regulatory context, regulatory distance to the rated firm weakens the relationship between the CSR rating score and default risk.
H2: The effect of CSR ratings on default risk will be negatively moderated by regulatory distance between the rating agency's home country and the rated firm's home country, such that the negative effect of CSR ratings on default risk is weaker when regulatory distance is larger.

| Cultural distance as a moderator
National culture also influences the interpretation of CSR in different countries (Scholtens & Dam, 2007). National culture refers to common experiences and shared meanings that are common within a country and are valid despite the fact that they are not formally stated or enforced (Tsui, Nifadkar, & Ou, 2007). National culture is a primary component of a country's institutional context and has an important influence on the way in which societies and organizations function. It also has an important influence on the way in which abstract concepts, such as "fairness," are interpreted in different ways in different countries (Tsui et al., 2007). Wang and Juslin (2009) show that the Chinese interpretation of CSR is distinctly shaped by Taoist ideas about harmony, which are central to Chinese culture. Consequently, Chinese firms emphasize contributions to social stability and employment. However, ensuring equality between different ethnic groups, which is important in Western societies' interpretations of CSR, is less emphasized in China. In a more general vein, Campbell et al. (2012) show that with increasing cultural distance, the variations in firms' CSR activities and related stakeholder interpretations also increase. This provides evidence that different national cultures generally result in different interpretations of CSR.
Corporate social responsibility rating agencies that are embedded within a specific national culture are likely to reflect this cultural background in their rating methodologies. For example, a CSR rating agency that originates in a Christian country might reflect Christian ideas about honoring god's creation and thus specifically address protecting the environment in its rating methodology (Teck Hui, 2008). Consequently, CSR ratings that reflect one single cultural context might be incongruent with other cultural backgrounds. On this basis, we propose that cultural distance between the home country of a CSR rating agency and the home country of a rated firm decreases the signal fit of CSR ratings. The greater the cultural distance, the greater the mismatch between the cultural background that is embedded in the CSR rating and the cultural background in which the firm operates will be. We therefore predict that when a rating agency is firmly embedded in a national cultural context, cultural distance to the rated firm weakens the relationship between the CSR rating score and default risk.
H3: The effect of CSR ratings on default risk will be negatively moderated by cultural distance between the rating agency's home country and the rated firm's home country, such that the negative effect of CSR ratings on default risk is weaker when cultural distance is larger.

| DATA AND METHODOLOGY
Our sample consists of listed firms from 13 different countries. For the independent variable, we used two alternative CSR ratings-one from OEKOM research AG (henceforth OEKOM), and one from ASSET4. Data on default risk and financial control variables were obtained via Thomson Reuters Eikon. Data on cultural distances were obtained from the GLOBE project (globeproject.com/studies), and data on regulatory distances were obtained from the Organisation for Economic Co-operation and Development (OECD). Our final sample consists of 604 firms at the intersection of these data sets. Tables 1 and 2 provide an overview of the sample across countries and time periods.

| OEKOM ratings
OEKOM utilizes a total of 95 indicators on 22 themes equally distributed over the social and environmental domains. The indicators are aggregated into a single continuous rating score, which can take any value between one (lowest grade) and four (highest grade) in decimal increments. OEKOM assesses all companies, regardless of geographic origin, against the same CSR definition. The rating methodology is adapted to fit specific industries, but it is not adapted for a company's geographic origin. Thus, OEKOM assesses all companies from the same industry against the same universal definition of CSR. OEKOM ratings are revised on a quarterly basis. A complete description of the indicators and the rating methodology is available in detail online (ARISTA, 2012).
As an organization, OEKOM is deeply embedded in the institutional context of Germany. Founded in 1993 in Munich, OEKOM pioneered CSR ratings in Germany and was among the earliest services of this kind worldwide. OEKOM's offices have been in Munich since its foundation, and almost all of OEKOM's roughly 70 employees are located in Munich. The vast majority of OEKOM's employees speak German, and most were born in and have been educated in Germany. Furthermore, the founders are German and the advisory board is comprised of native Germans. OEKOM was an independent company until 2018, 2 years after the end of our study period, when it was acquired by Institutional Shareholder Services, a financial service provider based in London.
OEKOM's rating methodology is based on prior academic work, the Frankfurt-Hohenheimer guidelines for the ethical assessment of firms, developed by three German professors (Hoffmann, Ott, & Scherhorn, 1997). These guidelines were written with an explicit ethical objective, namely to make business more compatible with cultural development, social equity, and environmental sustainability. Although they encompass a global perspective, some elements of the Frankfurt-Hohenheimer guidelines hint at their German origin. For instance, the criteria in the social domain include the importance of labor unions and employee participation-aspects that are firmly institutionalized in Germany, and more so than in other countries such as the United States. Notwithstanding the scientific quality of the guidelines, the fact that they were written exclusively by Germans suggests that the interpretation of CSR contained within them is shaped by the institutional context of Germany. Combined with the embeddedness of OEKOM as an organization in Germany, these are strong reasons to believe that CSR ratings from OEKOM reflect a uniquely German perspective on CSR.

| ASSET4 ratings
ASSET4 in many ways provides the same service as OEKOM. However, there are important differences regarding assessment methodology, organizational history, and institutional embeddedness of the organization. ASSET4 utilizes 250 individual indicators across 15 subcategories of CSR, which are aggregated into three main pillars: environmental, social, and corporate governance. Thus, a first difference is that ASSET4 explicitly includes corporate governance in its CSR ratings. A fairly detailed description of the company's rating methodology is available upon request at Thompson Reuters. Similar to the OEKOM method, ASSET4 aggregates the indicators into a single rating score, which ranges from 0 to 100. Also like OEKOM, the scores are calculated in an industry specific way. ASSET4 ratings are updated continuously. However, in a striking departure from how OEKOM calculates its scores, ASSET4 scores are calculated differently for different regions, in order to reflect the different importance assigned to certain aspects of CSR in different regions.
As an organization, ASSET4 is similar in size to OEKOM with approximately 100 employees, but the employees come from diverse international backgrounds and are not associated with any particular geographical context. ASSET4 was founded as a private company in 2003 in Switzerland. However, as an organization it was never deeply embedded in the Swiss context, and it has had an international perspective from the start. In 2006, the company received a private equity investment from Goldman Sachs (2006), and as early as 2007 the company had offices in London, New York, Mauritius, and India. In 2009, ASSET4 was acquired by Thomson Reuters (Reuters, 2009), an internationally active financial information broker based in Canada, and became a part of its extensive data service offerings for investors.
The rating methodology of ASSET4 is primarily driven by the needs of investors. It emphasizes the objectivity and comparability of the ratings, as well as the possibility that investors can create their own ratings based on ASSET4 data. In the documentation of its ratings, ASSET4 acknowledges that there is no perfect way to construct CSR ratings. ASSET4 has a ratings committee that meets on a quarterly basis to decide on potential changes to the methodology, but information about who is on that ratings committee is not publicly disclosed. The information that is available does not reveal any details about how ASSET4 interprets the notion of CSR in its methodology. In contrast to OEKOM, the methodology does not seem to be based on ethical considerations. Taken together, ASSET4 as an organization is not firmly embedded in any particular organizational context, but tends to operate in a very cross-national setting. Consequently ASSET4 ratings represent a cross-national perspective on CSR.

| Default risk
We measure default risk in terms of the spreads on firm-specific credit default swaps (CDSs). A CDS is a derivative that is linked to a borrower's outstanding debt and represents, in principle, an insurance against default. CDSs are internationally standardized financial assets that are completely independent of a firm's location and tax status (Fabozzi & Mann, 2012). CDS spreads, thus, represent an internationally comparable measure of default risk. These spreads are popular in financial studies because they offer an accurate and timely measure of default risk (O'Kane & Sen, 2005). CDSs are traded on a large and liquid financial marketplace, which ensures efficient pricing. We chose the 5-year CDS because it is the most common and most liquid derivative on the market. For each firm, we observe the last quoted CDS spread within a quarter as our measure of default risk. The spread is expressed in basis points (1/10,000), comparable to the level of interest paid on a loan. Given that the distribution of CDS spreads is skewed, we take the natural logarithm, which yields a dependent variable with a normal distribution.

| Institutional context
We compute the variables regulatory distance and cultural distance to measure differences between the institutional contexts of a firm's country of incorporation and the home country of the rating agency. To compute this variable, we apply the squared Mahalanobis distance measure across several dimensions of institutional distance. In the following equation, I denotes vectors for country i and reference country j, containing country-specific observations of institutional context across n dimensions. S is the covariance matrix of the data set with i countries and n dimensions.
Mahalanobis distance is recommended as an improvement upon earlier measures of cross-national distance, such as the one introduced by Kogut and Singh (1988). This is because by including the covariance matrix, Mahalanobis distance takes into account that different dimensions of institutional distance are usually correlated and have heterogeneous scales and variances (Berry, Guillén, & Zhou, 2010;Kandogan, 2012). This avoids, for example, overweighting differences in dimensions with high variance, as well as overweighting differences in separate dimensions which are highly correlated. Since OEKOM is based in Germany and ASSET4 is based in Switzerland, we compute institutional difference twice: first against the reference country of Germany, and a second time against Switzerland.
Regulatory distance is based on 14 dimensions: five dimensions stem from the environmental policy domain, and nine dimensions stem from the social policy domain. The environmental policy dimensions are taken from the environmental policy stringency index, provided by OECD. The dimensions of the index are (a) environmentally related taxes, (b) emission trading schemes, (c) feedin tariffs for renewable energy, (d) emission limit values, and (e) government expenditure on research and development in environmentally friendly technologies. The social policy dimensions are taken from the social expenditure index, also provided by the OECD. The dimensions are expenditures in the categories (a) active labor market programs, (b) family, (c) health, (d) housing, (e) incapacity related, (f) old age, (g) survivors, (h) unemployment, and (i) other social expenditures. Regulatory distance describes how different a country's regulatory context is relative to Germany or Switzerland, respectively, based on these 14 regulatory dimensions. To allow for calculation of the covariance matrix, data from all OECD countries is used.
Cultural distance has been widely applied in management studies. Cultural distance calculations are based on the nine cultural value dimensions established by House (2004), namely (a) assertiveness, (b) institutional collectivism, (c) in-group collectivism, (d) future orientation, (e) gender egalitarianism, (f) humane orientation, (g) performance orientation, (h) power distance, and (i) uncertainty avoidance. Cultural distance measures how different a country's cultural context is across these dimensions, relative to Germany or Switzerland, respectively.

| Control variables
Our model specification follows the empirical approach of Chava (2014). First, we control for firm size because large firms generally experience lower default risk. Firm size is measured as the natural logarithm of total assets in billions of U.S. dollars (USD) and labeled Firm size. Second, we control for return on assets (ROA), since the ability to generate income from existing assets is a riskmitigating factor. ROA is measured in USD billions and labeled ROA. Third, we control for leverage, since higher leverage makes it increasingly uncertain whether the firm can fulfill all of its debt obligations. Leverage is defined as the ratio of outstanding liabilities over total assets and labeled Leverage.
Our analysis differs from Chava (2014) as it is based on a global sample of firms. Therefore, adhering to the recommendations of Tsui et al. (2007), we also control for country-specific factors that exist in addition to institutional aspects and are likely to have an effect on the dependent variable default risk. First, we control for the baseline default risk within a country, operationalized as the average of the dependent variable per country-year, and labeled country risk. Second, we control for the sovereign credit risk of the country as an indicator of the overall risk perception of a country's economy. It is operationalized as the logarithm of the 5-year CDS spread for the respective country's sovereign debt, and labeled sovereign risk. Finally, we control for economic development as an indicator of the current business climate, using the growth rate of gross domestic product (GDP), labeled GDP growth. All independent variables were measured at least 1 month before the dependent variable, in order to allow time for the information to be incorporated.

| Empirical analysis
Our empirical approach includes fixed effects at the industry and quarter levels, to control for industry specific conditions and cross-sectional shocks. We confirm that this fixed effects model is appropriate by means of a Hausman test (Cameron & Trivedi, 2005). Based on tests for heteroscedasticity of residuals, we estimate robust standard errors clustered by firm.

| RESULTS
Our sample is an unbalanced panel of quarterly observations from 614 firms between 2011 and 2016, resulting in 11,201 firm-quarter observations. Table 3 shows the distributions of all continuous variables. Table 4 shows the correlation matrix for all continuous variables in the regression. Noteworthy is the correlation between OEKOM and ASSET4 ratings of 0.52, which indicates that the CSR ratings differ substantially between these two rating agencies. The initial sample is determined by the intersection of firms rated by OEKOM and the availability of dependent and control variables. The sample for ASSET4 is a subsample that covers 97% of the original sample with a total of 10,970 observations.  Model 1, shown in Table 5, demonstrates a baseline model with only the control variables. Most of the control variables are highly significant and point in the expected direction: leverage increases default risk, while firm size and ROA reduce default risk. The adjusted R-squared is at 53% of variance explained, indicating that the control vector captures a substantial part of the variance. Model Abbreviation: GDP, gross domestic product; ROA, return on assets. *p < .05; **p < .01; ***p < .001.

T A B L E 3 Distributions
2 includes the OEKOM rating as an additional explanatory variable. The coefficient for OEKOM is negative (−0.15) and highly significant (p > .01), indicating that firms with a high OEKOM score have lower default risk. Model 3 shows the same relationship for ASSET4 scores, where the coefficient is also negative (−0.49) and highly significant (p < .001). Thus, Hypothesis 1 is supported: A higher CSR rating score is associated with lower default risk. Model 4 tests the moderating influence of regulatory distance with reference to Germany. It includes the main effect of regulatory distance and the interaction effect between regulatory distance and OEKOM. The coefficient of the interaction term is positive (0.01) and significant (p < .05). This supports Hypothesis 2 for the German context: the greater the regulatory distance between the respective home countries of a firm and a CSR rating agency, the weaker the risk-reducing effect of a good CSR rating.
Model 5 tests the moderating influence of cultural distance with reference to Germany. It presents the main effect of cultural distance, along with the interaction term between OEKOM and cultural distance, which is positive (0.02) and significant (p < .05). This result indicates that increasing cultural distance diminishes the effect of OEKOM scores on default risk and supports Hypothesis 3: the greater the cultural distance between the respective home countries of a firm and a CSR rating agency, the weaker the risk-reducing effect of a good CSR rating.
Models 6 and 7 show the interaction of ASSET4 scores with regulatory and cultural distance with reference to Switzerland. The interaction terms are not significant in either model. While this outcome does not empirically prove that institutional distance has no effect for a rating agency that operates cross-nationally, it is in line with our expectations, given that ASSET4 is a rating agency that is not embedded in its original Swiss context.
To illustrate the moderation effect for the German context, Figure 1 shows the moderating effect of cultural distance on the relationship between the OEKOM rating and default risk. The graph is F I G U R E 1 Moderation effect of cultural distance on the relationship between OEKOM and default risk generated by predicting the dependent variable based on the cultural interaction model for several levels of cultural distance. The independent variables, besides OEKOM rating and cultural distance, are kept constant at the sample mean. Figure 1 clearly shows how the negative relationship between the OEKOM rating and default risk fades with increasing cultural distance. For a cultural distance of zero, that is, for German firms, the relationship between the OEKOM rating and default risk is negative. For firms in Canada, which has a cultural distance of 10.17, the relationship is still negative, but the slope is only half as steep. For firms with a cultural distance above a value of 20, such as firms in Japan, the relationship is even slightly positive.

| Robustness checks and further analyses
We performed two robustness checks. First, the number of firms rated by OEKOM increased from 2014 onwards because before 2014 OEKOM only rated firms that cleared a certain CSR performance threshold in a prescreening. Starting in 2014, OEKOM rated all companies, regardless of their CSR performance. The prescreening creates a potential concern that the sample is skewed toward firms with high CSR ratings. However, rerunning the analysis with only the data from 2014 to 2016 produces virtually the same results, even with slightly higher significance levels. Second, even though we controlled for fixed effects on the firm and quarter levels, there is still potential for unobserved firm characteristics to bias the results. However, rerunning the analysis with firm-level fixed effects produces very similar results for the interaction terms, making us confident that our findings are not subject to unobserved heterogeneity.
In addition to these robustness checks, we investigated whether there were important differences between different rating components of OEKOM, that is, its social and environmental aspects. The moderation through regulatory and cultural distance is more pronounced for the social aspect. This suggests that environmental CSR practices vary less across contexts as compared with social CSR practices.

| DISCUSSION
This article investigates how CSR rating agencies-functioning as legitimating agencies-shape the relationship between CSR, legitimacy, and default risk. Including CSR rating agencies themselves in the theoretical reasoning generates new insights regarding the boundary conditions of the CSR insurance effect and the role of CSR rating agencies as institutional intermediaries. Furthermore, it provides important implications for investors, managers, and researchers.
The first contribution of this article is to demonstrate that the insurance effect of CSR is partially, but not entirely, moderated by context-specific conceptualizations of CSR. The empirical literature so far suggests that performing against CSR ratings has a risk-mitigating effect (Godfrey et al., 2009;Koh et al., 2013;Shiu & Yang, 2017). In contrast, the international business literature holds that legitimacy and its associated benefits arise from conforming to context-specific CSR practices (Campbell et al., 2012;Crilly et al., 2016;Ioannou & Serafeim, 2012). In a cross-national setting, this raises the question of whether firms can reduce default risk by conforming to local norms or to global CSR performance standards, or both.
This study provides a differentiated answer to this question. On the one hand, it shows that CSR ratings are generally associated with lower default risk. Despite fundamentally different rating approaches, CSR ratings from both OEKOM and ASSET4 are associated with lower default risk. This suggests that the insurance effect of CSR ratings is not entirely dependent on the way in which CSR is interpreted and measured. Instead, there seems to be a set of widely accepted CSR practices that decrease default risk in any institutional context. On the other hand, the study shows that OEKOM CSR ratings reduce default risk more for German firms, and less for firms that originate in countries with a different institutional context. The effect decreases with increasing regulatory and cultural distance from Germany. This suggests that the insurance effect is moderated by differences in the institutional context because different norms and expectations result in different interpretations of CSR. Thus, conforming to CSR performance standards reduces default risk to the extent that these standards reflect widely accepted CSR practices, or CSR practices that are accepted in the firm's local context.
We interpret these results with the concept of signal fit: CSR ratings reduce default risk to the extent that the CSR rating signals a firm's underlying legitimacy. We show that there are variations in signal fit, depending on which CSR rating agency is chosen, to what extent it is embedded in a particular institutional context, and in which institutional context the company operates. When studies investigate CSR ratings in a domestic setting, where both rating agency and rated firm are within the same institutional context, signal fit is likely to be high. Most of the studies on the insurance effect of CSR use samples of U.S. firms and rely on ratings from U.S.-based rating agency Kinder, Lydenberg, and Domini (KLD). However, in cross-national studies, such as the present article, the concept of signal fit highlights that the same CSR rating might signal different levels of legitimacy across different institutional contexts.
Inserting the concept of signal fit into the relationship between CSR ratings and default risk has the potential to inform further theorizing on boundary conditions that emerge in cross-national settings. The concept of signal fit contributes a way to integrate empirical findings that are based on CSR ratings-and thus implicitly assume that there is a CSR performance standard-with a perspective that views CSR as inherently context specific.
The second contribution of this article is to highlight that CSR rating agencies do not simply provide objective measures of CSR. In addition to providing a measure, rating agencies provide an interpretation of CSR that is implicit in their ratings. Our results suggest that a rating agency's country of origin and its embeddedness in that country influence its interpretation of CSR along with the meaning of the ratings that it provides.
We show this by comparing two different CSR rating agencies: OEKOM, which is strongly embedded in the institutional context of Germany, and ASSET4, which has a very international presence and no strong roots in any particular country. Our analysis shows that the effect of CSR ratings on default risk differs substantially in a cross-national setting. While both CSR ratings have an effect on default risk, we find a sensitivity to regulatory and cultural distance only for OEKOM ratings.
This finding extends prior research on CSR ratings as institutional intermediaries (Chatterji et al., 2016;Doh et al., 2010;Slager et al., 2012). One of the important features of institutional intermediaries is that they standardize constructs, thereby allowing measurement and orientation, ultimately facilitating transparency and competition (Durand & McGuire, 2005;Rao, 1994;Ruef & Scott, 1998). However, these processes have typically been studied within a homogenous institutional context, where uniform institutional pressures favor convergence toward one common interpretation.
In our setting, CSR rating agencies are faced with the challenge of attempting to standardize a construct that is interpreted differently in different contexts. The first option for dealing with this challenge is the one that OEKOM pursues, which appears to rate firms against an interpretation of CSR that is congruent with the German context. This means that OEKOM ratings are not ideal for gauging the legitimacy of firms in Spain or Japan, for instance. However, it does mean that OEKOM ratings reflect firm performance against one consistent CSR performance standard. The second option is that used by ASSET4, which appears to rate firms relative to an interpretation of CSR that represents the firm's own context. This kind of rating, which reflects the norms and expectations that the rated firm is actually exposed in its institutional context, can signal a firm's legitimacy more reliably than an OEKOM rating. However, an ASSET4 rating relies on a method that shifts the standards of CSR performance between different countries. Chatterji et al. (2016) demonstrate that CSR ratings by different rating agencies continue to diverge because these agencies have different views of what CSR means. We offer an explanation of why different rating agencies have different views and how this, in turn, affects the relevance of CSR ratings for determining legitimacy and risk. First, this study suggests that a rating agency's embeddedness in its original institutional context has an important influence on how it operationalizes CSR in its methodology. Thus, different CSR rating agencies may provide different scores to the same company, because they come from different institutional contexts and have different perspectives on the same company. Second, this study suggests that in a cross-national setting, rating agencies face a trade-off between measuring CSR against one universal standard or against multiple context-dependent standards. Taken together, the origin of CSR rating agencies and the corresponding differences in rating methodologies effectively explain the discrepancies between CSR ratings.

| Limitations
There are several limitations that should be kept in mind when interpreting the results of this study. We find support for our moderating hypotheses based on data from only one CSR rating agency, OEKOM, which is strongly embedded in its original institutional context. We compare this to ASSET4, a CSR rating agency that operates independently of its original institutional context, and do not find the same effects. While these results are consistent with our hypothesis that a CSR rating agency's institutional embeddedness has implications for the legitimating effect of its ratings, they leave open the possibility that endogenous factors specific to our research case are driving the results. Unfortunately, even if we included all existing CSR rating agencies in our analysis, the sample would be too small to fully eradicate this concern. Nevertheless, the study provides theoretical reasons why institutional distance interferes with the legitimating effect of CSR ratings and presents evidence that this may indeed be the case.
Moreover, while measures of institutional distance are widely accepted in the literature, there are challenges to their validity . Most of the theoretical and methodological shortcomings apply to situations where interactions between specific actors are investigated, for example, headquarter-subsidiary relations. In our research setting, we do not focus on such interactions. Thus, the illusion of symmetry, causality, and discordance  can be ignored. However, we must acknowledge that large corporations often have a strong international presence, which means that there is some ambiguity regarding the national context against which institutional distance should be measured. Nevertheless, the context that tends to be most important for firm survival, and thus for our measure of default risk, is likely to remain the context of the country of incorporation, which we use in this study. Finally, our measure of cultural distance is linear and reflects a single point in time. Thus, we are not able to explore the possibility that institutional distances may change over time, as well as the possibility that there may be cultural thresholds or tipping points that change the dynamics of this relationship.

| Implications
The findings have profound implications for empirical research in the CSR domain. Given that there are differences between different CSR ratings, which are related to the rating agencies' interpretations of CSR, the choice of one CSR rating agency is likely to influence results. Researchers should be aware that choosing a CSR rating agency means buying into its specific interpretation of CSR. Rather than avoiding the discussion around the meaning of CSR (Gond & Crane, 2010), researchers should assess whether the rating agency's approach suits their research objectives. The problem arises in particular in cross-national settings. ASSET4 ratings, which have become popular for usage in recent research (Cheng et al., 2014;Ioannou & Serafeim, 2012), seem useful for international samples because they adapt the meaning of CSR to the regional context. However, the fact that ASSET4 has different standards in different regions also raises questions about how exactly these standards differ. The bottom line is that researchers should be aware that CSR rating agencies provide not only a measure but also an interpretation of CSR.
A promising avenue for future research lies in reapplying the concept of signal fit to other relationships in cross-national settings. For example, certifications by the ISO may provide different signals in different contexts-even though the certification process is highly standardized. A second avenue for future research lies in investigating the history of CSR rating agencies. While our study hints at the institutional pressures that shape CSR rating agencies as organizations and their rating methodologies, a case study approach could provide a more detailed understanding of the mechanisms that shape and differentiate CSR rating agencies and their ratings.
The results also have important implications for investors and managers. Investors, who are interested in CSR from an instrumental perspective and hold globally diversified portfolios, might be better served with a rating agency such as ASSET4. Regardless of context, it reflects whether a firm is in line with its stakeholder's expectations. But investors with an ethical mission, such as churches, may prefer a rating such as OEKOM, which applies a single standard universally. From an ethical point of view, it may not be acceptable that certain aspects of CSR, for example, gender equality, are considered only to the extent that they are important in the firm's local context.
For managers the results have two implications. First, there seems to be a set of CSR practices that is consistently part of different CSR rating methodologies. These widely accepted CSR practices seem very promising in terms of increasing a firm's legitimacy and lowering its default risk. In addition, there seem to be CSR practices that are more context specific. Managers could further increase organizational legitimacy by adding these CSR practices in the contexts where they are appropriate. In order to learn which CSR practices are appropriate, managers may be able to learn most from locally embedded rating agencies. For instance, regarding CSR practices in Germany, OEKOM is likely to have superior information. A general CSR strategy focused on widely accepted practices, combined with local CSR strategies that are informed by local legitimating agencies, is likely to lead to an approach to CSR that is optimal in terms of reducing default risk.

| CONCLUSION
There is a fundamental tension between the view that CSR rating agencies formalize and standardize CSR (Chatterji et al., 2016;Doh et al., 2010;Slager et al., 2012), and the view that CSR is interpreted differently in different contexts (Campbell, 2007;Crilly et al., 2016;Ioannou & Serafeim, 2012). This study synthesizes these views by framing CSR rating agencies as institutional intermediaries that create and formalize interpretations of CSR, which may be more or less embedded in a particular institutional context. Our results suggest that, on the one hand, there are widely accepted CSR practices that are always part of different interpretations of CSR by various rating agencies. These practices reduce default risk consistently across different institutional contexts. On the other hand, our results suggest that when a CSR rating agency is strongly embedded in its original institutional context, its CSR ratings have a stronger negative effect on default risk for firms in the rating agency's home country than for firms from countries that have very different regulatory approaches or cultural backgrounds. These results suggest that even though CSR rating agencies standardize CSR in their rating methodologies, these standards can be context specific and thus may reinforce the variety of CSR practices in cross-national settings.