Institutionalization of the Contents of Sustainability Assurance Services: A Comparison Between Italy and United States

A descriptive-exploratory analysis of assurance practices is presented in this paper, by analysing the patterns of sustainability assurance reporting in two national contexts with different levels of assurance activity (Italy and the U.S.) over a period of 11 years (2003/2013). The study is based on theoretical insights drawn from institutional sociology and normativity production. It is framed both in the Italian situation, where assurance statements consistently include a narrow set of formal and procedural communications, and in an unsettled situation in the U.S., where assurance activity is more incipient, but where experimentation in substantial assurance disclosure practices has more room to develop. Its main implication is that the diffusion of sustainability assurance and the creation of sustainability assurance disclosure norms are not without its cost: information quality does not increase with patterned practice. The results also point towards the noteworthy role of specific professionals in the earlier and later stages of assurance practice norms. They reveal a significant influence of non-Big4 firms (mainly certification bodies and consulting and engineering firms) in the diffusion of sustainability assurance disclosure norms. In contrast, the Big4 firms appear to be positively associated with the narrowing down of the assurance focus to a selected subset of this activity in later stages of the development of the assurance norm. In this regard, this study provides insight into the circumstantial—but relevant—carrier role of the Big4 firms in determining what “assurance” means.


Introduction
Over the past decade, sustainability reporting (hereafter SR) has begun to be equated with standard business practice among large global companies. KPMG (2013) reported that nearly 100% of the largest 250 global companies, as well as the top 100 companies in several countries, were committed to Sustainability Reporting (hereafter SR), affirming that "the debate on whether companies should report on sustainability… or not is dead and buried" (KPMG 2013, p. 10). However, both researchers (Boiral 2013;Boiral et al. 2017;Dando and Swift 2003;Gray 2010;Levy et al. 2010;Milne and Gray 2013) and practitioners (KPMG 2013) have expressed concern over the reliability, the materiality, and the quality of SR, at the expense of accountability and transparency to stakeholder (O'Dwyer and Owen 2007).
Along the same lines, it has been suggested both by researchers (Hodge et al. 2009;Kolk and Perego 2010) and by practitioners (ACCA 2004) that sustainability assurance, through independent and qualified external revision of the reports, can confer credibility and validity on the information in SR. In this regard, sustainability assurance has been construed as an indication of SR credibility and it is, therefore, important for discharging the social responsibility of firms.
The literature suggests that it is possible to increase the credibility and the quality of SR, to improve stakeholder confidence in the accuracy and the validity of the information that is disclosed. The risk of perceiving SR as addressing only the needs of the most powerful stakeholders is also reduced through the adoption of independent external assurance (Adams and Evans 2004;Deegan et al. 2006;Hodge et al. 2009;Kolk and Perego 2010;O'Dwyer and Owen 2007;Perego and Kolk 2012;Simnett et al. 2009;Zadek et al. 2004). The literature refers to several further benefits, as it is seen to contribute to decreasing agency costs (Fuhrmann et al. 2017); providing stability to markets and their operations (Kend 2015); reducing information asymmetries (Fuhrmann et al. 2017); and improving overall performance in relation to existing policies, commitments, and risks (Zadek et al. 2004).
Consistent with these assumed benefits, international evidence indicates an increase in the number of corporations providing third-party assurance of their sustainability reports (e.g., Casey and Grenier 2015;Peters and Romi 2015). For example, KPMG (2015) reported that 63% of the top 250 global companies producing sustainability reports provided an assurance statement as part of the process.
In the previous research, the factors that influence the decisions of firms to assure their reports have been studied. The findings of various studies showed that corporate characteristics, such as social and environmental sensitivities (Casey and Grenier 2015;Peters and Romi 2015;Simnett et al. 2009); size (Branco et al. 2014;Cho et al. 2014;Kolk and Perego 2010;Simnett et al. 2009); listed status (Zorio et al. 2013); and profitability (Branco et al. 2014), appear to significantly determine assurance adoption. However, an investigation of corporate attributes alone cannot explain the different assurance activity when different countries are compared: this practice appears to have attained higher levels in Denmark, Italy, and Spain, where over 60% of reports are assured (KPMG 2011), while, in contrast, it is less prevalent in countries such as the U.S. or Canada.
An exploration of the divergent assurance practices in different countries can enlighten us to the functions that assurance services perform (such as conferring credibility on the SR) and the obstacles that this practice might encounter. In that respect, the enhancement of transparency (and hence social responsibility) is assumed to be a function that SR assurance performs. Hence, the aim of this investigation is to cast light on the actual functions of sustainability assurance, by conducting a descriptive-exploratory study of assurance disclosure practices analysing the patterns of sustainability reporting assurance content in two national contexts with different levels of assurance activity (Italy and the U.S.) over a period of 11 years (2003/2013). The interest in these two national contexts is connected with their differences in terms of country-level institutional characteristics. Italy is traditionally considered a common law country with a stakeholder orientation and a particular "attentiveness to social relationship as a whole" (Albareda et al. 2006, p. 395). Its approach to CSR policies has been defined as an "Agora model", which pays special attention to the involvement and dialogue of stakeholders (Habisch et al. 2011). On the contrary, the U.S. context is characterized by high litigation costs (e.g. Simnett et al. 2009;Kolk and Perego 2010) where corporations operate within an economic and legal environment dominated by a "managerial focus on shareholder's interests" (Peters and Romi 2015, p. 142).
In view of the above, the study sets out to explore, from an institutional theory perspective, whether and how specific assurance disclosure practices became norms in specific constituencies. In doing so, this investigation is making three sets of contributions.
First, this paper focuses on the norm itself. Previous literature has approached the question of the international diversity of sustainability assurance through cross-sectional studies that inquire into the association between sustainability assurance and a set of country-specific institutional factors. Some studies found a positive relation with stakeholderoriented countries (Kolk and Perego 2010;Simnett et al. 2009;Zhou et al. 2016) and higher institutional pressure for corporate sustainability (Kolk and Perego 2010). Evidence is also mixed with respect to the legal systems and the strengths of each country (Kolk and Perego 2010;Simnett et al. 2009;Zhou et al. 2016). However, sustainability assurance is a recent, fragile, and contested practice, where the boundaries between attestation and advice and between technical procedures and sustainability are fluid, with assurors seeking to define the framework of sustainability assurance (O'Dwyer 2011). An often-disregarded question in previous research is how sustainability assurance developed so quickly to become a business norm in some countries rather than in others. In this regard, one contribution of this paper to the literature is its focus on the norm itself, by performing a longitudinal study (Casey and Grenier 2015) that pays more attention to the institutional process of normativity production (Bebbington et al. 2012;Finnemore and Sikkink 1998).
Second, this research illustrates how the institutionalization of sustainability assurance is coupled with patterned sustainability reporting practices, but not necessarily with disclosure quality. To that end, an analytical distinction is drawn in the paper between the quality and the convergence of sustainability assurance statements. Previous research has illustrated the diversity of sustainability assurance practices in terms of extension and quality of the assurance statements (Dando and Swift 2003;Hummel et al. 2017;Manetti and Becatti 2009;O'Dwyer and Owen 2005;Perego 2009;Perego and Kolk 2012;Zorio et al. 2013). Likewise, Perego and Kolk (2012) using an international panel of Fortune Global 250 firms over the years 1999, 2002, 2005, and 2008, analysed the quality of assurance statements by means of content analysis based on the framework provided by O'Dwyer and Owen (2005). Their results showed an improvement in the quality of assurance statements over time, associated with the type of assurance provider. Overall, these studies analyse the quality of assurance statements. Unlike those studies, the interest of this research is to explore the convergence of sustainability assurance practice into norms and whether the emergence of those sustainability assurance norms is coupled with specific disclosure practices. More specifically, this study shows how the crystallization of a sustainability assurance norm in Italy is coupled with a narrower focus on assurance disclosure, while the more unsettled and embryonic assurance practice in the U.S. is coupled with more ambitious assurance disclosure practices (see O'Dwyer 2011).
Third, as a further contribution, this investigation explores the role of assurors (the Big4 firms and non-Big4 firms) in the production of assurance disclosure norms. The credibility that assurance provides to sustainability reports depends not only on the company that seeks assurance, but also on other actors such as the assurors (Hummel et al. 2017). The assuror interacting with firms can build, through professional practice, norms that determine the content of SR assurance statements. In that respect, assurance practices could be subjected to "professional capture", which is a more nuanced understanding of the general capture argument (Smith et al. 2011). Previous literature has documented the association of different types of assurors with assurance quality. However, the focus of this paper is on the association of different types of assurors with the convergence of sustainability assurance practice into norms.
The rest of the paper is structured as follows. The following section presents a theoretical framework for the study of sustainability assurance and develops a set of research questions on the evolution of assurance norms based on a constructivist understanding of regulation and institutional sociology. The research method (based on content analysis of all the sustainability assurance reports published in Italy and the U.S. by large companies between 2003 and 2013) is described in the third section. The fourth section presents the empirical results. Finally, the fifth section discusses the results and offers some concluding comments and implications derived from this research.

The Institutionalization of SR Assurance
The patterns of sustainability assurance reporting in two countries are analysed from an institutional theory perspective in this paper, to explore whether and how specific assurance disclosure practices became norms in specific constituencies. This aim leads to research questions that are amenable to an institutional analysis of the dynamics of regulation (see Bebbington et al. 2012 andChelli et al. 2016 for a review of this literature).

Institutional Theory and Normativity Production
Theoretical perspectives based on institutional sociology and normativity production have been used to explain how recent CSR reporting practices have emerged and developed to become norms (Bebbington et al. 2012;Levy et al. 2010).
The sociological roots of institutional theories emphasize the diffusion of conformity pressures leading to the emergence of well-established social order in organizational behaviour (Aerts et al. 2006). Considering that the focus of the present investigation is the study of standardized patterns of behaviour, institutional theory provides an appropriate conceptual platform to explain how assurance has become a norm for business in a specific constituency. In this regard, it is important to stress that the focus of this paper is on the assurance norm itself, understood as the degree of convergence of SR assurance disclosure practices, rather than disclosure quality. Disclosure quality or, at least, an approximation of disclosure quality, is associated with the amount of disclosure or the spread of disclosure across a list of items (see e.g. Beattie et al. 2004). Individual assurance statements can be judged by their quality; average quality can also be computed for a set of assurance statements. In contrast, the focus of this study is on the convergence (standardization) of assurance statements, i.e. the degree to which they are alike in a particular jurisdiction, regardless of their greater or lesser quality.
Two basic tenets of the constructivist understanding of normativity production are important in this exploration of assurance norms: dynamics matter and the sources of normativity, as well as the actors that participate in it, are multifarious (Bebbington et al. 2012).
On the one hand, the evolution of assurance practices must be considered, i.e. a complete explanation of assurance norms cannot be obtained through a cross-sectional interrogation of assurance practices. This constructivist understanding differs from the pervasive legitimacy arguments in SR research based on the social consonance of actors with relevant rules. While legitimacy theory is interested in organizational legitimacy (see, for example, Cho and Patten 2007), the constructivist understanding of norms focuses on normative legitimacy. In this regard, the focus of O'Dwyer and colleagues (O'Dwyer 2011;O'Dwyer et al. 2011) is on assurance norms themselves, when they explore the endeavours of the sustainability assurance practitioners to legitimize this emerging area of practice in the eyes of key audiences, despite the inherent fragility of assurance stemming from the difficult task of rendering SR auditable (Gendron et al. 2007).
The notion of the life-cycle of norms has been proposed in this theoretical perspective to inquire into the dynamics of norms. This theoretical perspective suggests that the critical aspect for a practice to become a norm is its diffusion until a tipping point is reached where it is adopted by a critical mass of pertinent actors, after which the norm cascades towards full internalization, acquiring a taken-for-granted status. The crucial change is that the practice depends on enforcement and/or beliefs over its legitimacy before that tipping point, while beyond that point it becomes "natural" for all the actors (Bebbington et al. 2012;Finnemore and Sikkink 1998).
On the other hand, the sources of normativity are multifarious, with norms emanating from a plurality of regulating actors (not necessarily a hierarchical state). Along the same lines, a distinction is made between state and non-state norms, allowing for a more fluid understanding of norms. For example, previous research has problematized regulation showing that environmental disclosure (state) rules are not matched by environmental reporting disclosure practices (norms) (e.g. Peters and Romi 2013). Bebbington et al. (2012) defined normativity production such as the process through which different actors understand specific rules (norms) as binding. Translated into SR assurance, this concept allows us to explore the emergence of disclosure norms and the role played by the participant actors (particularly, assurors).

Research Questions
Referring to sustainability assurance, KPMG (2013) audaciously affirmed that "the tipping point has been crossed, with over half the world's largest companies" (p. 12) disclosing such information. The notion that assurance is gaining broader acceptance means that our analysis may be framed in terms of the dynamics of assurance disclosure practice, and more particularly in terms of the different pace at which assurance content has developed in Italy and in the U.S. Previous evidence has shown that, since the early 2000s, Italian companies have developed significant assurance activity, assuring over 60% of all sustainability reports (Kolk and Perego 2010;Rossi and Tarquinio 2017;Simnett et al. 2009) with an assurance market dominated by the largest (Big4) accounting firms. In contrast, only about 6% of the sustainability reports produced by U.S. firms were assured in the same period. Accordingly, any potential sustainability assurance disclosure norm is likely to be ahead of the life-cycle of norms in Italy when compared with the U.S.
If SR assurance is at a later stage in the life-cycle of norms in Italy, then the legitimacy of the norm and the pressure for convergence would be higher in that country than in the U.S. But beyond its mere existence, the understanding of assurance norms and the specific practices that they include requires us to examine the content of assurance statements (Bebbington et al. 2012 developed a similar study for environmental reporting). The early stages of the norm life-cycle through an institutional lens suggest that they are characterized by innovation and by the assembly of shared meanings and understandings of the emerging practice (Djelic and Quack 2008). Previous literature suggests that assurance practices in the U.S. seem to be at an initial stage with innovation carried out by a limited number of organizations (Casey and Grenier 2015;Cho et al. 2014;Peters and Romi 2015;Simnett et al. 2009). Innovation will at an initial stage be characterized by "advancing new ideas, solutions, and practices" (Scott 2008, p. 126) and, therefore, by more diversity in disclosure practices. The unsettled and innovative character of assurance hypothesized for initial stages in the life-cycle of assurance disclosure norms is likely to result in divergent content in the assurance statements. The life-cycle concludes with the crystallization of the norm, i.e. with convergence towards defined structures and patterned practices that acquire a permanent character (Djelic and Quack 2008). Patterned assurance practices are likewise approached through the exploration of the content of assurance statements, i.e. how different assurance statements converge to convey the same kind of information. Our first exploratory research question, to ascertain patterns of convergence in the assurance disclosure practice of both Italy and the U.S., is as follows: RQ1 Do particular SR assurance disclosure practices, in terms of the content of assurance statements, converge towards defined structures over time?
As mentioned above, one of the basic tenets of this institutional perspective is the multiplicity of actors that participate in the production of normativity (Bebbington et al. 2012). As in the notion of the arena that Smith et al. (2011) advanced, formal and informal rules can emerge from the interaction of actors with different characteristics and interests. Actors might include rule enforcers and issue-amplifiers in arena studies, which resonate with the institutional notions of norm entrepreneurs and carriers; notions that are more consistent with the dynamics of the life-cycle of norms. Concerning norm entrepreneurs, they have, for example, attracted some interest to explain SR (Etzion and Ferraro 2010;Levy et al. 2010). Norm entrepreneurs innovate and propose specific practices in the earlier stages of the life-cycle of norms. This role has arguably been played in sustainability assurance by AccountAbility and the International Auditing and Assurance Standards Board (IAASB) with the issuance of their standards AA1000 and ISAE3000 that have dictated the general disclosure rules (Dando and Swift 2003;Deegan et al. 2006;Manetti and Becatti 2009;O'Dwyer and Owen 2005). Their role is better captured by this notion than by the notion of rule enforcers, since their power is not derived from delegation by political institutions (Smith et al. 2011), but from rule-setting in a context of more diffuse power structures (Bebbington et al. 2012).
Carriers participate in processes of innovation diffusion and the proposal of practices in the later stages of norm lifecycle, contributing to the generalization of particular forms and interpretations of the practices (Sahlin-Andersson and Engwall 2002; Scott 2008). The institutionalization of sustainability assurance would also require carriers to transport practices and values between fields and organizations. The carriers identified include consultants, auditors, and professional associations (CorporateRegister 2008;Perego 2009).
Institutional theory suggests that carriers are not innocuous: carriers actively interpret and edit the ideas and the practices that they transport. The accounting literature has illustrated this for accounting firms (Greenwood and Suddaby 2006). In the case of sustainability assurance, O'Dwyer (2011) has shown how the Big4 firms have sought legitimacy in the eyes of key audiences (see also O'Dwyer et al. 2011), by narrowing the approach towards the construction of sustainability assurance practice.
Sustainability assurance has been provided by two different professional groups: Big4 firms and non-Big4 firms, such as individual accountants, specialists such as engineering firms, certification bodies, and others (e.g. stakeholders' panels and NGOs) (CorporateRegister 2008;Perego 2009;Perego and Kolk 2012). Previous research has inquired into the kind-of-assuror choice (e.g. country characteristics; Kolk and Perego 2010;Perego 2009;Zhou et al. 2016), showing an association between Big4 assurance and stakeholder-oriented countries. Previous studies have also analysed how the kind-of-assuror choice affects assurance quality (e.g. Hummel et al. 2017;Mock et al. 2013;O'Dwyer and Owen 2005;Perego 2009;Perego and Kolk 2012;Zorio et al. 2013).
An understanding of the Big4 firms as assurance carriers suggests that assurance norms are initially designed by institutions such as AccountAbility and IAASB. Triggered by other contextual factors, these norms are translated and edited by the Big4 firms. Considering the previous literature, institutional theory would suggest that the Big4 firms do not play a role in the initial design and the launch of the sustainability assurance norm, but are more likely to play a relevant role in the diffusion of the practices, at a subsequent stage in the life-cycle of norms, once other contextual factors are triggered. As Boiral and Gendron (2011) and Smith et al. (2011) warned, assurors might have incentives to take control (capture) assurance practices, in line with their professional interests mediating their translations and editions of assurance practices. In this regard, following O'Dwyer et al. (2011), assurors would narrow the assurance approach over time, thereby becoming the engine of sustainability assurance uniformity.
In view of the relevant role that 'Big4 firms' appear to play in the development of sustainability assurance practice, this investigation analyses them in contrast with 'non-Big4 firms'. Based on this discussion, the second research question addresses the association of assurors with the different stages in the production of SR assurance norms: R2 Are different types of assurance providers associated with the earlier and later stages of the life-cycle of SR assurance norms?

Sample
The nature of the research questions requires a qualitative and longitudinal analysis of sustainability assurance which, as such, means that the inquiry is narrowed down to certain countries. The choice of Italy and the U.S. is justified by their contrasting situation with regard to assurance practices (Kolk and Perego 2010;Simnett et al. 2009). The sample for this study includes all the SR assurance statements published by large Italian and U.S. organizations over 11 years, from 2003 until 2013. These organizations belonged to a wide range of industry sectors (see Appendix 1). Listed organizations represented 50 percent and 70 percent of the total amount of Italian and U.S. organizations publishing SR assurance statements, respectively. The choice of 2003 is dictated by the emergence of sustainability assurance. In March 2003, one of the dominant international assurance standards, the AA1000 Assurance Standard 'AA1000AS' was issued by AccountAbility (2003). The second dominant standard, the IAASB's International Standards on Assurance Engagement 'ISAE3000' (IAASB 2003) was applied as from January 2005 (CorporateRegister 2008(CorporateRegister , 2013. Italian and U.S. sustainability reports were identified by reference to the GRI database. We are aware of the criticisms of GRI (Gray 2006), but this choice is justified on two grounds: (i) through its SR guidelines the GRI has become a dominant global player in the area (Levy et al. 2010;Milne and Gray 2013); and, (ii) the GRI database is the reference point for the identification of published sustainability reports and is regularly used in SR research to identify reporters (e.g. Boiral 2013). Sustainability reports published by large companies 1 were selected for two main reasons: (i) a positive association between corporate size and the presence of SR assurance has been identified in the previous literature (e.g. Branco et al. 2014;Cho et al. 2014;Mock et al. 2013;Peters and Romi 2015;Simnett et al. 2009); and, (ii) institutional theory predicts that larger companies are more likely to create norms in the activities they developed (Lieberman and Asaba 2006). 375 Italian and 1057 U.S. sustainability reports produced by large (both listed and unlisted) companies were identified in the GRI databases (Appendix 1). Subsequently, we searched for those reports that included an assurance statement. According to the GRI databases, 180 Italian and 125 U.S. companies declared having assured their SR. However, the GRI database proved to be inaccurate in some instances and we found no assurance statements in 22 of those cases for Italy and 29 for the U.S. The missing assurance statements were requested by email from the companies, to ensure that the substance of SR assurance disclosure practice was included in the analysis, obtaining one further assurance statement from a U.S. company 2 (for more detail, see footnotes in Table 1). The final sample for this study included 158 Italian and 97 U.S. SR assurance statements, corresponding, respectively, to 42% and 9% of the sustainability reports (see Table 1). Italian and U.S. companies assuring the SRs did not differ substantially in terms of industry (see Appendix 1) or size (average total assets in the year of their first sustainability report were €145 million for Italy and $102 million for the U.S.).

Content Analysis and Research Instrument
An in-depth analysis of SR assurance disclosure practices requires an examination of the content and characteristics of this practice. To that end, a thematic content analysis of the SR assurance statements was performed (Gray et al. 1995;Jones and Shoemaker 1994). A similar approach has been followed in the previous literature (Hummel et al. 2017;O'Dwyer and Owen 2005;Perego and Kolk 2012;Rossi and Tarquinio 2017;Zorio et al. 2013).
We followed the research instrument developed by Rossi and Tarquinio (2017), which adapted the framework proposed by O'Dwyer and Owen (2005) and Perego and Kolk (2012), considering the most recent versions of the assurance standards AA1000AS (2008) and ISAE3000 (2013). This instrument (see Appendix 2) consists of an index containing 29 assurance disclosure items that belong to seven types of assurance disclosure thematics: (i) assurance provider characteristics; (ii) representation by the responsible party; (iii) nature and extent of the planning process; (iv) formal requirements; (v) particular conclusion on the specific accounting system; (vi) assurance form and criteria; and (vii) comments and advice. Content analysis requires an inspection of the assurance reports for the presence or absence of disclosure across the set of information items (Guidry and Patten 2010). Electronic copies of all the 255 SR assurance reports were used for that purpose. The coding procedure consisted in assigning 1 for the presence (and 0 for its absence) of each specific sub-item.
Content analysis validity and reliability (Potter and Levine Donnerstein 1999) depended on the above-mentioned coding procedures, designed for thematic content analysis, and the fact that internationally recognized frameworks in sustainability assurance are the source of the disclosure items. We applied Cronbach's alpha to test the internal validity of the research instrument developed by Rossi and Tarquinio (2017). In other words, we checked whether all the 29 assurance disclosure items measured a single construct (internal consistency of the research instrument). Cronbach's alpha value was 0.80 yielding a value higher than 0.70, a level suggested by Peterson (1994) as sufficient to confirm the validity of the research instrument. In addition, two coders participated in the analysis (Berg and Lune 2012) for greater reliability, in such a way that all the reports were coded by one of the authors from among which a second coder independently analysed 20%. Coding reproducibility was tested with Krippendorff's alpha (0.86; above the recommended level of 0.80), showing an acceptable level of agreement between the two coders (Hayes and Krippendorff 2007). Each discrepancy was discussed and agreed among all the authors, the result of which was followed in the rest of the coding.

Research Design
RQ1 was explored with the assistance of the Herfindahl-Hirschman Index (hereafter HHI), a concentration ratio used in such research fields as computer sciences (Lu et al. 2017), management (Alcacer and Chung 2014), and accounting (Christensen and Kent 2016). In the present research, the HHI allows us to explore the degree of convergence of SR assurance practices in Italy and the U.S, in terms of the content of assurance statements. The HHI range is the whole (0,1) interval. High HHI values indicate a high level of concentration of assurance practice in particular disclosure items and, conversely, low values indicate a high level of variation in assurance practice.
HHI was calculated for each country as follows: where HHI represents the Herfindahl-Hirschman Index, n is the number of companies that disclose a SR assurance statement, and p i is the proportion of companies that disclose the same ith combination of the 29 items according to Rossi and Tarquinio (2017). This analysis is complemented with a graphical representation of the distribution of assurance elements and characteristics in both countries, plotting the items that are either rarely or typically disclosed in each country.
To explore RQ2, the HHI was recalculated, splitting each country sample into assurance reports attested by Big4 firms and by non-Big4 firms. In this way, we were able to explore whether and how different assuror types participated in the earlier and later stages of the life-cycle of SR assurance disclosure norms.
This method of analysis (calculating the HHI for sample subsets) was also followed to test the robustness of the results in RQ1. To that end, we tested whether the findings displayed in RQ1 were contingent on variables such as environmental and social sensitiveness and size, which the literature has found to influence sustainability assurance. 3 Previous studies have found mixed evidence concerning the relative propensity of firms in "environmentally sensitive" industries to have their sustainability reports assured (Casey and Grenier 2015;Cho et al. 2014;Peters and Romi 2015;Kolk and Perego 2010;Simnett et al. 2009). Accordingly, we split our sample into environmentally and socially sensitive and non-sensitive industries in each country and ran an HHI estimation for the different subsamples. In our analysis, sensitive industries include energy (GICS code 1010), materials (1510), and utilities (5510)-considered environmentally sensitive industries by Aerts and Cormier (2009) and Cho and Patten (2007), as well as chemical (151,010), capital goods (2010), retailing (2550), and food, beverages, and tobacco (3020)-deemed as "socially sensitive" by Bouten et al. (2012).
As a positive association between corporate size and assurance has been found in the previous literature (Branco et al. 2014;Cho et al. 2014;Mock et al. 2013;Peters and Romi 2015;Simnett et al. 2009), the robustness of the results of RQ1 was also tested through the estimation of the HHI in both Italy and the U.S. for two subsamples containing, respectively, larger and smaller companies. In each country, the larger companies were those that in the year of their first sustainability report had greater total assets than the average total assets for all the companies in the 2003-2013 3 Information on industry classification and total assets was obtained from the AIDA database (see http://www.bvdin fo.com/en-us/ourprodu cts/compa nyinf ormat ion/natio nal-produ cts/aida) for the Italian companies and from the Compustat database for the U.S. companies. period (average total assets Italy = €145 million; U.S. = $102 million).

Results
A total of 82 large Italian and 382 large U.S. companies produced at least one sustainability report between 2003 and 2013. Table 1 presents the sustainability and the assurance reports identified per year (Italy in panel A and the U.S. in panel B). Table 1 shows a steady increase of sustainability assurance activity in Italy for the period under analysis, and in the U.S. as well, although at a slower pace. Table 1 also informs us that sustainability assurance, with less than two statements per year per country, was a rather marginal practice over the period 2003/2006. However, this activity accelerated in 2007, with an assurance statement included in over 40% of Italian sustainability reports (panel A). The same figures from the U.S. were less impressive, although 12% of the American sustainability reporters claimed that they had included an assurance statement in 2007. In 2011, this proportion reached 58% in Italy and stabilized in both countries in later years at around 50% for Italy, and 13% for the U.S. These data confirm previous evidence noting more established sustainability assurance practice in Italy, in terms of the number of adopters, than in the U.S.

Convergence of Disclosure Practices (RQ1)
RQ1 explores whether SR assurance practices converge, in terms of the content of assurance statements, over time in Italy and the U.S. Figure 1 plots the HHI values, providing an indication of the sustainability assurance disclosure practice and its concentration around similar disclosure items in those countries.
Considering the paucity of SR assurance adopters until the year 2007, HHI is only computed from 2007 to 2013. The results show that the concentration of assurance disclosure around a set of specific items systematically increased for Italy from 2008 up until 2013, reaching HHI > 0.300 in 2013. HHI for the U.S. before 2010 (about 0.5) was higher than for Italy (about 0.10), something that can be explained by the reduced number of observations for the U.S. in the first years. However, from 2011 (when the percentage of U.S. sustainability reports assured increased to about 10%; see Table 1), HHI values showed a sharp decrease in convergence levels, with HHI values attaining values close to zero. Low HHI values are indicative of the fact that a mere three U.S. assurance reports included the same disclosure items over the whole period. In contrast, from 2011 (when the percentage of Italian sustainability reports assured was higher than 50%; see Table 1), HHI values indicated an increasing convergence towards patterned disclosure practice in Italy. In 2013, Italian HHI exceeded 0.30 points. Overall, these results showed clear identifiable trends as assurance practices increased in both countries.
Since RQ1 is dynamic in nature, Fig. 2 analyses the evolution of SR assurance disclosure by plotting the concentration over 2007/2011 on the coordinate axis and the concentration over 2011/2013 on the ordinate axis. Figure 2 visually confirms the previous conclusions. A process of convergence of the assurance content was not discernible in the U.S. However, in Italy, while eight items are plotted in the more uneven disclosure area (20% > frequency > 80%) in 2007/2011, further SR assurance disclosure practice in the 2011/2013 period "pushed" those items to either nondisclosure (< 20%) or norm status (> 80%). For example, information on the competence to provide assurance services (item 1.3) was disclosed in around 40% of the statements in the first period but further practice in the second period proved that it was an item worth disclosing in an ideal assurance statement (> 80%). In contrast, the consideration of completeness issues in sustainability reports (item 6.4) was a common disclosure in assurance statements in the first period (around 70%) that was subsequently abandoned (< 20%).
The distribution of disclosure items in both countries, displayed in Fig. 2, is consistent with the results of Fig. 1. One indication of concentration could be the proportion of items that are either rarely or typically disclosed in each country relative to the total. In this regard, a high proportion of items included in more than 80% or less than 20% of assurance statements would be indicative of patterned assurance practices. In contrast, a high proportion of items disclosed between 20% and 80% of the assurance statements would denote more uneven disclosure practices. Over the period 2011/2013, the first proportion, indicative of more homogeneous reporting, amounted to 26 out of 29 disclosure  This exploration of RQ1 allows us to conclude that, according to institutional arguments, assurance disclosure practice in Italy has converged around a subset of the practice investigated through the content analysis, while the rest of those practices appear to be marginalized. Patterned assurance disclosure practices include some of the minimum requirements of sustainability assurance (independence, competence and scope, among others). This convergence appears to be associated with a later stage in the life-cycle of norms.
At the same time, assurance disclosure practice in the U.S. appeared unsettled, with companies experimenting and innovating with different assurance sub-practices. This situation is characteristic of an initial stage in the life-cycle of norms. For example, while specialists and certification bodies assurors in the U.S. tend to produce elaborate assurance statements, individual accountants apply a more limited approach. In this regard, it is worth highlighting that assurance practice in the U.S. appears keener to introduce demanding disclosure items, such as materiality and stakeholder participation in assurance, that, in contrast, appear to be repressed by the Italian assurance norm.
Finally, from an accountability perspective and considering corporate discharge of social responsibility, the exploration found that a set of more challenging disclosure items were generally ignored in assurance disclosure practice in both countries, in such a way that the norm was not to disclose them. Those items included the evaluation of the information systems used for SR, the consideration of completeness and performance issues, and the comments and advice on SR progress. A finding that is consistent with previous studies (Deegan et al. 2006;Manetti and Becatti 2009;O'Dwyer and Owen 2005).

Assurance Providers and the Life-Cycle of SR Assurance Norms (RQ2)
RQ2 aims to explore the type of assurors associated with different stages of the SR assurance norm life-cycle. Specifically, RQ2 inquires into the role of different types of assurors (Big4 firms and non-Big4 firms) in the earlier and later stages of the life-cycle of SR assurance disclosure norms. Table 2 distinguishes three categories of assurance providers (CorporateRegister 2008;Perego and Kolk 2012), namely Big4 firms, non-Big4 firms (including individual accountants, certification bodies, specialists-both consultants and engineering firms, and reinsurance firms), and non-specified.
Italian SR assurance statements were mostly produced by Big4 firms, representing more than 80% of the market over the last years of the study (Table 2). In contrast, the non-big4 firms produced most SR assurance statements in the U.S., representing more than 60% of the market in the last years (considering that the type of assuror is not specified in 23% of the cases); notably, the activity of specialists (consultants and engineering firms) and certification bodies such as Bureau Veritas, DNV, and SGS.
Focusing on the role of different types of assurors in the life-cycle of SR assurance disclosure norms, Fig. 3 plots the HHI of concentration of the items reported in assurance statements, by type of assuror. This analysis allows us to explore whether different genres of assurors participate in the emergence and diffusion stages of SR assurance disclosure norms, providing greater insight into how SR assurance normativity production takes place.
HHI values plotted in the lower part of Fig. 3 increase from 0.25 in 2007 to 0.50 in 2013, showing a trend of convergence in the contents of Italian assurance statements. In contrast, HHI values indicate that Italian assurance statements prepared by Big4 firms appeared not to follow a clear pattern before 2011. However, since 2011, the Italian Big4 assurance statements have started to follow the same trend as non-Big4 firms, converging around a subset of the assurance disclosure practices (HHI values not only reaching but surpassing the value of non-Big4 firms in 2012).
Non-Big4 firms produced the pioneering U.S. SR assurance statements (2007/2010). Unlike in the case of Italy, non-Big4 firms in the U.S. appear not to be associated with the convergence of the contents of the assurance statements around a subset of practices, but rather with an unsettled situation (the HHI concentration index value dropped from 0.50 in 2007 to 0.05 in 2013). Big4 firms were not active in the U.S. assurance market until 2012 and, then, they seem to have followed a similar trend to the non-Big4 ones, with HHI decreasing from around 0.35 in 2011 to 0.10 in 2013.
The Italian case is consistent with the suggested significant role played by the assurors in the diffusion of assurance disclosure practices in a specific direction. Despite not representing a significant part of the market in the last years, non-Big4 firms appear to have been associated with the convergence of the contents of the assurance statement in Italy since the earlier period. In contrast, the Big4 firms were not associated with similar levels of convergence until 2011. These convergence trends in Italy suggest that the Big4 would have followed the practices initiated by the non-Big4 firms, probably editing and diffusing them, contributing to the emergence of a specific SR assurance norm.
The lack of assurance statement content convergence in the U.S. is associated with both Big4 and non-Big4 types of assurors. One plausible explanation for the non-role of assurors as carriers of any assurance norms is that convergence is difficult, because of the heterogeneity of assurors that characterize the U.S. market (Perego and Kolk 2012).
In relation to Big4 firms, the results appear counterintuitive, because these firms (large accounting firms) would be expected to play a more homogenizing role internationally, restraining international differences between Italy and the U.S. In this regard, it is important to note that Deloitte, E&Y, KPMG, and PWC are assuring sustainability reports in both Italy and the U.S. Nevertheless, the institutional account of the life-cycle of these norms provides an explanation to this apparent paradox: the Big4 association with the development of the Italian assurance norm can be explained by their role as carriers rather than by their nature as dominant firms. The results in both Italy and the U.S. are consistent with the carrier role of the Big4 firms. In Italy, where non-Big4 firms (among other actors) launched a process of convergence of the content of assurance statements, the Big4 firms appeared to follow suit and to transmit the values and the practices that produce an assurance norm around the subset described earlier. Conversely, the Big4 firms in the U.S. played no carrier role at all, apparently because there were no emergent values or practices to be carried. We might speculate that the Big4 firms could have the incentives to carry ideas and practices from other (e.g. European Commission 2003) constituencies. However, carriers interpret and edit the norms they carry, in line with their interests, among other factors. In this regard, the non-role of Big4 firms could be associated with the avoidance of national contexts characterized by high litigation costs such as the U.S. This explanation would be consistent with Perego and Kolk (2012) who noted that such an approach might hamper the development of assurance services in some national contexts.

Additional Analysis
To test the robustness of the results in RQ1, we examined the contingency of the findings on size and environmental and social sensitiveness by plotting HHI curves derived from different subsamples, as per the procedure described in the method section. The results in Figs. 4 and 5 show that, with some yearly variability, the HHI curves derived from different subsamples undergo similar trends (mostly replicating the curves in Fig. 1), ruling out the possibility that size or environmental and social sensitiveness could explain the assurance convergence.

Discussion and Concluding Comments
This paper has presented an exploratory and longitudinal analysis of SR assurance disclosure practices in Italy and the U.S., which seeks to inquire into whether and how disclosure practice and norms become institutionalized. In contrast to previous literature on sustainability assurance, the focus of this study is not on assurance quality. Disclosure quality consists of the amount of disclosure and the spread of disclosure, evaluated against a set of informative items. This study, however, focuses on a putative assurance norm, understood as the degree of convergence of SR assurance disclosure practices, notwithstanding their quality. The choice of Italy and the U.S. is driven by their contrasting situations regarding their respective levels of SR assurance services; Italy is among the countries with the highest and the U.S. among those with the lowest assurance activity.
Two questions approached through an institutional lens have been investigated: (a) which is the relative pace at which assurance norms are developing in Italy and the U.S., in terms of convergence of contents of assurance statements (disclosure practices); and (b) the role of different types of assurors, the Big4 firms and non-Big4 firms (individual accountants, specialists, certification bodies and others), in the production of SR assurance norms.
The results of the study are in line with the theoretical framework proposed for this research. Consistent with previous evidence, this research has confirmed very different levels of assurance activity in both countries, providing an appropriate contrasting empirical site for the study of assurance disclosure practices. In particular, the theoretical framework draws attention to the importance of investigating the nature of this practice, to analyse the institutionalization of SR assurance reporting, in terms of its elements and characteristics, and to explore the institutionalization of SR assurance. In this regard, although the volume of disclosures made in SR assurance statements were comparable in both countries, an in-depth analysis revealed that SR assurance disclosure practice converged in Italy around a set of specific disclosure items, in such a way that by the end of the period under analysis, almost all Italian assurance statements were disclosing the same information. This concentration of the disclosure practice was perceptible in the content analysis. For example, it revealed that most assurors in Italy used templates for the wording and structure of substantial parts of the assurance statements produced for companies in different sectors/years. Italian assurance was framed around a set of disclosure items, focusing on the characteristics of the assuror, the nature of the work performed, and formal requirements. These findings are consistent with previous literature suggesting that these disclosure practices, imported from financial auditing, attempt to create comfortable conditions for this still fragile practice (Boiral and Gendron 2011;Boiral et al. 2017;Gendron et al. 2007;O'Dwyer 2011;Smith et al. 2011).
In contrast, the U.S. shows how an unsettled situation, typified by a lower percentage of sustainability reports that are assured, lower participation among the Big4 firms, and a higher variation in SR assurance disclosure practices, is associated with companies willing to experiment with different assurance sub-practices. This description, which is consistent with the observations of practitioners (Scheneider 2013), is characteristic of an initial stage in the life-cycle of norms. However, despite the purportedly lower level of assurance in the U.S., the results show a higher level of disclosure on substantive issues, such as assumptions, stakeholder engagement, evidence, and materiality. These results mean that patterned practice is not increasing transparency (Perego 2009), suggesting that the restrictive version of SR assurance is motivated by managerial interests, rather than by broader accountability and transparency to stakeholders (O'Dwyer and Owen 2007).
With respect to the role played by the Big4 firms in the institutionalization of assurance disclosure practices, the results of this study have indicated that their activity is positively associated with the subset of specific assurance disclosure practices that constitute the Italian assurance standard. While at an initial stage (2009)(2010)(2011), the Big4 firms followed the practices of non-Big4 firms in Italy, the Big4 firms later became the main carriers, editing the assurance disclosure norms to narrow down their focus to a selected subset of this activity. It should be noted that SR assurance was initially a business for domestic specialist firms that considered assurance akin to yet another certification process. Later on, assurance started to resemble more of a financial audit and was perceived as a market opportunity by the Big4 firms (Boiral and Gendron 2011;Manetti and Becatti 2009). Even though the non-Big4 firms initiated the promotion of SR assurance, the Big4 firms were the ones that subsequently edited the assurance norms to narrow down its focus to a selected subset of this activity. Hence, it is interesting to note that the Big4 firms not only promoted ISAE3000, the auditing-oriented standard, but edited it in such a way that a set of disclosures proposed in this standard were introduced in the Italian assurance norm (e.g. 1.1 Independence), while other items were ignored (e.g. 6.3 Materiality and inclusivity).
In contrast, the role of the Big4 firms is not significant in the U.S., where non-Big4 firms (especially specialist consultants, engineering firms, and certification bodies) dominate the assurance market and tend to experiment more and, for example, provide services associated with advice and evaluation of the information systems.
The findings have suggested that international differences are not only observable for the assurance activity in which non-Big4 firms participate, but also for the activities of the Big4 firms. Non-Big4 firms such as engineering and specialist consultants are often locally based. However, considering the international integration of the Big4 firms, it would be reasonable to expect that those firms could play a more active role in transnational SR assurance uniformity. Our findings appear to contradict this expectation and are in line with Suddaby et al. (2007), who found that despite their level of international integration, the Big4 firms are playing different roles in different countries. The carrier role of the Big4 firms is circumstantial rather than based on their attributes. Our analysis shows that carriers play an important role in later stages of the life-cycle of norms (in Italy), but play an insubstantial role at earlier stages (in the U.S.).

Societal and Managerial Implications
The main implication of this study is that the diffusion of SR assurance and the creation of SR assurance disclosure norms are not without cost. The results find no reason to suggest that transparency is increased by patterned practice; a finding that has implications for the understanding of the production of norms in the context of Corporate Social Responsibility reporting. In that respect, the results have shown that SR assurance standardization and the participation of the Big4 firms as carriers are associated with a narrow SR assurance disclosure practice.
We might conjecture that if the SR assurance market developed in the U.S. and the Big4 firms were to develop an appetite for this new market, then a process of narrowing down and standardizing SR assurance would be likely in U.S. assurance practice. The results are also consistent with assurance disclosure practice that is largely imported from financial auditing, shifting from a criterion of sustainability to one that gives weight to the norms shared by financial auditors (Boiral et al. 2017;Smith et al. 2011).
Our findings in this research therefore suggest the existence of some form of professional capture (O'Dwyer 2003;Smith et al. 2011) in the translation and edition of Italian norms of assurance. This capture is enabled by the fragility of sustainability assurance, derived in turn from the inherent difficulty of evaluating the materiality, completeness, and reliability of SR, which includes information from sources as diverse as the supply chain or biodiversity. The Big4 firms use their "position to play a role as powerful social actors" (Smith et al. 2011, p. 428) in the production of assurance norms that can make assurance legitimate, but will construct what "assurance" means, enabling or preventing assurance to perform its function of enhancing transparency and corporate social responsibility.

Limitations and Potential Future Research
A limitation of this study derives from its focus on sustainability reports published by large companies: the patterns of institutionalization of sustainability assurance reports found in this study cannot be generalized to small-and mediumsized companies. A further limitation is that the focus of the study on SR assurance disclosure practices provides neither explanations nor reason for both the emergence of disclosure norms and the capture process. A promising area of future research might be the extension of this longitudinal study with qualitative data. Semi-structured interviews with key actors in the Italian and U.S. assurance fields could provide rich information on the dynamics involved in the interaction between different actors producing norms and/or capturing this activity (Smith et al. 2011). (1) AA1000AS required the "Scope of the engagement", in this sense the assurance provider before accepting an engagement shall be satisfied that all the requirements of the standard can be met and that the reporting organization is acting in good faith. In particular, the assurance provider shall be satisfied that the engagement subject matter is appropriate, and the practitioner will have access to sufficient evidence to support findings and conclusions (2) There are two types of AA1000AS (2008) sustainability assurance engagement Type 1 Accountability principles: the assurance provider shall evaluate the nature and the extent of the organization's adherence to all three AA1000 Accountability Principles; Type 2 Accountability Principles and Performance Information: the assurance provider shall evaluate the nature and the extent of the organization's adherence to the AA1000 Accountability Principles, as for the Type 1, and shall also evaluate the reliability of specific sustainability performance information (3) Consistent with extant ISAE 3000, two levels of assurance are possible for engagements: reasonable assurance and limited assurance. The standard indicates the use of positive form for the reasonable level and the use of a positive form for the limited level (4) The AA1000AS indicates that an assurance engagement may be carried out to provide a high level of assurance or a moderate level of assurance. Since a different subject matter may be addressed in one assurance engagement, a high level of assurance may be provided for some subject matter while a moderate level of assurance may be provided for other subject matter in the same assurance statement (5) The presence of any limitation in the scope of the disclosure on sustainability, the assurance engagement, or the evidence gathering shall be addressed in the assurance statement and reflected in the report to management, if prepared