Introduction

In this study, we investigate the extent to which corporate social responsibility (CSR) disclosure by French companies is changing over time. France is one of the few countries to have enacted legislation requiring the disclosure of social and environmental information. Passed in 2001, the Nouvelles Régulations Économiques (NRE) #2001-420, in theory, requires French firms to provide information regarding social and environmental performance in their financial reports. However, as noted by Delbard (2008), initial compliance with the NRE requirements was quite low, a phenomenon consistent with Larrinaga et al.’s (2002) findings for Spanish electric utilities’ disclosure following passage of that country’s Plan General de Contabilidad (PGC) which mandated certain environmental disclosures. Bebbington et al. (2012) argue that the low level of Spanish company compliance with PGC was due to a lack of normativity; that is, the degree to which actors see rules as binding. They further note (p. 90) that “formal legislation alone may not be sufficient to create a norm,” and then identify how the influence of outside actors, in particular the Association of Chartered Certified Accountants (ACCA) through its Environmental Accounting Awards, appeared to foster a sense of higher normativity for environmental reporting in the U.K. even without the existence of formal laws. Bebbington et al. (2012) document that levels of environmental disclosure by U.K. utility companies were significantly higher that those of their Spanish counterparts.

While Bebbington et al.’s (2012) primary intent is to explore the differing ways in which normativity for environmental reporting developed across Spain and the U.K., their empirical analysis is limited to an examination of differences in levels of disclosure across companies from the two countries over the period surrounding the adoption of the PGC in 1998. In contrast, we examine the extent to which CSR disclosure has changed over time. This is relevant to Bebbington et al.’s (2012) exploration in that they also argue (p. 79) that normativity is subject to change over time whereby diffusion leads to a point “where norms are internalised and acquire a taken-for-granted quality.” We assume that the lack of early compliance with NRE requirements in France suggests a lack of normativity at that time, but that observations “of an increased rate of compliance every year” (Delbard 2008, p. 402) suggest movement toward diffusion, or even a taken-for-granted quality over time. Importantly, we attempt to assess whether normativity in French CSR disclosure, or at least the presumed movement toward it, reflects improved CSR disclosure. In order to explore this question we analyze the CSR disclosures in both financial reports and, where issued, stand-alone CSR reports for a sample of 81 publicly traded French firms specifically focusing on changes in the space devoted to, and the quality of, these disclosures.

Comparing disclosure from the 2004 and 2010 reporting periods we find, first, substantial evidence that the amount of space allocated to CSR disclosure by our sample of French firms increased significantly. These increases exist across both financial report and stand-alone report issuers. We also find that the breadth of CSR disclosure—that is, the provision of CSR information across different areas of social and environmental disclosure also increased, although this is limited to financial report-only disclosers. In a unique contribution to the CSR literature, we next assess the changes in informational quality—attributes of disclosure recommended by both standard-setting boards relative to financial disclosure and the Global Reporting Initiative relative to CSR information—and find that: while overall informational quality is slightly improved, quality remains quite low and seems to be a function of increasing disclosure across CSR areas in financial reports as opposed to substantial increases in informational quality overall. We further find that the disclosure of negative CSR performance information, an argued indicator of balanced reporting, has decreased for our sample firms. Finally, in an attempt to determine whether CSR reporting has moved toward increased transparency, a goal of the NRE, we test whether legitimacy concerns appear to influence differences in disclosure and whether that has changed over time. Our results show that disclosure in 2004 was highly associated with legitimization variables and that those relations, for the most part, remain unchanged in 2010.

Overall, our analysis suggests that CSR disclosure in France is increasing with respect to the space being allocated to it, a finding we argue indicates that the normativity of CSR reporting is, if not yet taken-for-granted, at least in a stage of diffusion. However, the reporting that appears to become the norm does not appear to be substantially improved with respect to quality and still seems to be more about legitimization than transparency. As such, we argue that the original goals of the NRE remain unmet. The next section of our paper presents the justification for our study and lays out our expectations. Section 3 reports the methods and analysis while our results are presented in Section 4. Finally, in Section 5 we offer discussion of our findings and present concluding comments.

Justification for the Study

CSR Reporting in France and NRE

Prior literature in CSR reporting in the French context primarily consists of several conceptual or theoretical studies investigating the issues related to voluntary corporate social disclosure (Capron and Quairel 2004; Quairel 2004; Rivière-Giordano 2007; Giordano-Spring and Rivière-Giordano 2008), empirical studies examining the determinants (Igalens 2007; Damak-Ayadi 2010) or users (Déjean and Martinez 2009) of such disclosure, and other longitudinal analyses of CSR disclosure strategies of a particular case firm (Cho 2009; Déjean and Oxibar 2010).

With passage of the NRE in 2001, France became one of the few countries in the world to mandate CSR disclosure.Footnote 1 As noted by Delbard (2008, p. 400), “the law compels companies to inform their stakeholders on their social and environmental impact” by providing “transparent and comprehensive information… on the social and environmental consequences of their activities,” and it was, in theory, required to be disclosed in companies’ annual reports. However, as also discussed by Delbard (2008) initial compliance with the mandate was extremely low. According to a report commissioned by the French government, only about 35 % of companies met the requirements of the new law in 2003 and wide discrepancies in its application—ranging from a few companies noted as having “best practices” to many firms minimally meeting the demands—were noted (Delbard 2008, p. 401).

Much of the problem regarding poor compliance with NRE can be attributed to the lack of specificity of the law. Delbard (2008, p. 400) notes that the legislation did not include “any specific constraints in terms of standards, norms, pollution thresholds or any new form of regulatory requirement,” and, typical of what he calls French “soft laws,” it did not clearly identify sanctions for non-compliance.

Normativity

The failure of the NRE to immediately improve CSR disclosure in France is consistent with Larrinaga et al.’s (2002) findings regarding Spain’s attempts to improve environmental disclosure by utilities through its PGC in 1998. Bebbington et al. (2012) explore this latter case in their exposition on environmental reporting and the role of normativity. They argue (p. 79) that normativity, the degree to which rules and practices become accepted and standardized, is not an inherent quality of legal mandates, but instead must meet the “test of legitimacy” whereby “norms are congruent with previous practice; a hierarchy of secondary rules define how the rule is to be made and applied; and the rule is well designed for its intended purpose.” Bebbington et al. argue that the Spanish attempt at mandating environmental disclosure failed to meet these criteria, and thus lacked normativity. In contrast, they show that environmental reporting by utilities in the U.K. over the period when the PGC took effect was far superior to their Spanish counterparts and argue that this is due to a degree of normativity for reporting fostered not by legal mandate, but largely through the guidance of the ACCA over time. Bebbington et al. (2012) further argue that normativity arises through a three-stage process. As they state (p. 79):

[It] starts with emergence of norms, characterized by the innovation of norm entrepreneurs, followed by diffusion leading to a “tipping point” after which the norm cascades to reach a point at the end of the life cycle where norms are internalised and acquire a taken-for-granted quality.

We argue that the low levels of French company compliance with the NRE suggest that the law, like the PGC in Spain, lacked normativity at its inception. However, Delbard (2008, p. 402) notes that, rather than giving up on the law, the French government continued to support its stand claiming it had to be seen as “a continuous improvement process” that it hoped would lead to improving disclosure practice over time. Delbard further claims (p. 402) that this stance appears justified as compliance with the law is improving every year. This would appear to be consistent with Bebbington et al.’s (2012) conceptualization of evolution in normativity over time, and suggests that movement toward normativity is, if not at the taken-for-granted level, perhaps in a stage of diffusion. Accordingly, the intent of our investigation is to empirically investigate if, indeed, CSR disclosure in France has changed since the years immediately following implementation of the NRE guidelines—both with respect to the amount of space being allocated to it and, importantly given the NRE’s goals of increased transparency, with respect to the quality of the information being provided.

Measuring Changes in CSR Disclosure

Changes in the Amount of Disclosure (Space)

There is considerable debate in the social and environmental accounting literature over methods used to measure the disclosure of CSR information. Both Al-Tuwaijri et al. (2004) and Smith and Taffler (2000) distinguish these measurement techniques across two basic groups—space measures and content scores. Space measures in essence attempt to quantify disclosure extensiveness in terms of a “unit of analysis” (Milne and Adler 1999). Disclosures have been measured by counting the number of words (see Deegan and Rankin 1996; Neu et al. 1998), sentences (see Buhr 1998; Hackston and Milne 1996; Tsang 1998), or pages (see Guthrie and Parker 1989; Patten 1992, 1995). Space-based disclosure measures have also been calculated as the percentage of pages (Gray et al. 1995a; O’Dwyer and Gray 1998) or the percentage of total disclosuresFootnote 2 (Trotman and Bradley 1981). Smith and Taffler (2000, p. 627) refer to this approach as “form oriented” (objective) analysis. This count method focuses solely on the extent or, more specifically, the amount of space allocated to CSR disclosures. While, as we discuss below, space is not the same thing as quality, greater space allocation does imply increased information provision. As such, if CSR disclosure in France was in a state of diffusion over the mid-to-late 2000s, we would expect to find that the amount of space allocated to the disclosure would be increasing over time.

Changes in Quality—Breadth of Disclosure

In contrast to the use of space measures, other studies use a disclosure scoring index based on a content or “meaning oriented” (subjective) analysis (Smith and Taffler 2000, p. 627). With this technique, the examination focuses primarily on the underlying themes or topics that are textually present in the disclosures of interest.Footnote 3 Researchers identify the areas of CSR information they are interested in measuring and attempt to determine whether these topics are addressed or discussed by corporate managers in the reporting media of choice. Accordingly, the researchers utilize a scoring index categorizing those themes and assess the presence or the absence of each identified item in the disclosures using a “yes/no” (or 1, 0) coding methodology. After their quantification, an aggregate score, generally labeled as the disclosure score variable, is determined for each firm in the sample (Barth et al. 1997; Cho and Patten 2007; Cho et al. 2006; Ingram and Frazier 1980; Patten and Trompeter 2003). Such measures have been particularly prevalent in the environmental disclosure literature and the metrics have ranged from rather limited scales (Blacconiere and Patten 1994 use a five-item scheme) to more detailed and extensive indices (Clarkson et al.’s 2008 scale includes a possible 95 points). Since the emphasis of NRE regulations was on improving the disclosure of both environmental and social aspects of CSR, we would also expect the breadth of disclosure across categories of disclosure to widen during a period of diffusion. Based on prior research (Cormier and Magnan 2003; Cormier et al. 2005; Ernst & Ernst 1978; Gray et al. 1995b), we identified five categories of CSR disclosure for our analysis. These are (1) environment and energy,Footnote 4 (2) business practices, (3) human resources, (4) community, and (5) products. Figure 1 provides more detail on what aspects of disclosure each category represents.

Fig. 1
figure 1

Categories of CSR information

Full size image

Changes in Quality—Informational Quality Assessment

However, while both the amount (“how much”) and the themes (“what”) of CSR disclosure are potentially important for firm managers and report users, neither is necessarily a measure of the quality of the information that is being provided.Footnote 5 Accordingly, other studies, again primarily within the environmental disclosure arena, have modified the traditional content analysis scoring method in an attempt to better capture what they argue are higher quality informational items. For example, some researchers assign different levels or weights to the scoring according to whether the disclosure contains monetary, quantitative, or qualitative terms (Al-Tuwaijri et al. 2004; Choi 1999; Wiseman 1982), or whether the disclosures are descriptive, vague, or immaterial (Hughes et al. 2001). While these weighting schemes do capture some additional aspect of information quality, their use has been largely limited to assessments of only environmental information (although see Guidry and Patten 2010) and they focus on only one aspect of (assumed) improved informational quality.

Given the limitations with the prior attempts to better measure the informational quality of CSR disclosures, we explore in this study an alternative method for capturing the quality of the CSR reporting package. It focuses more broadly on what Solomon (2000, p. 33) refers to as the “qualitative characteristics for decision useful information,” and is rooted in the conceptual frameworks of financial reporting standard-setting bodies including the IASB and the FASB (of the U.S.).Footnote 6 Solomon (2000) argues that because both financial reporting and CSR reportingFootnote 7 are aimed at providing information to facilitate decision making by interested parties, the beneficial informational qualities identified by the financial reporting boards ought to be applicable to a large extent to CSR disclosure. It also appears that the GRI, the organization probably most responsible for the advancement of corporate sustainability reporting over the past decade (Ballou et al. 2006; Gray 2006), agrees in that its discussion of desirable qualitative dimensions of reported information closely aligns to those of the FASB and IASB.

Based on a review of GRI guidelines and the conceptual frameworks of both the IASB and the FASB, we identified five major quality characteristics for CSR information: relevance, comparability, verifiability, clarity, and neutrality. We discuss each of these in more detail below.

Relevance

Financial reporting conceptual frameworks typically define relevance as “information that is capable of making a difference in the decisions made by users” (FASB 2010, p. 17). The GRI, in its G3 reporting guidelines, similarly suggests that the extent to which a user will consider a piece of information significant in the decision-making process determines the relevance of CSR reporting. However, as noted by Jeffrey and Perkins (2011), the GRI, in discussing CSR information relevance, stresses the importance of stakeholder inclusiveness. By consulting its stakeholders an organization can better identify their information needs and desires. Further, the GRI notes that, because an ideal report should include CSR information that is useful and relevant for both the organization and its users, disclosure of the organization’s activities and related risks (and risk factors) can also contribute to the relevance of information.

Comparability

The FASB (2010, p. 19) defines comparability as “the qualitative characteristic that enables users to identify and understand similarities in, and differences among items.” The IASB (www.iasplus.com/standard/framework.htm) further argues that information “is more useful if it can be compared with similar information about other entities and with similar information about the same entity for another period or another date.” The GRI notes that providing comparative data allows both external and internal stakeholders to compare performance and assess any progress made in various areas and activities. In short, information comparability makes it possible, within a pre-defined report perimeter, to assess the evolution of data over time, or to evaluate them consistently and objectively with those of other companies.

Verifiability

Both the IASB and the FASB note that the quality of verifiability means that “different knowledgeable and independent observers could reach consensus, although not necessarily complete agreement, that a particular depiction” faithfully represents the underlying phenomena it is attempting to capture (FASB 2010, p. 20). In other words, verifiable information is data and/or claims that an independent agent could likely tie back to the underlying phenomena.

Clarity

The GRI notes that clarity—referred to in the financial reporting conceptual frameworks as understandability, is an information characteristic that allows various groups of users to properly understand and exploit such information, and ultimately utilize it for analysis or decision making. While the standardized nature of accounting information inherently confers a large degree of clarity, CSR-type information can be made more clear and intelligible to a wide group of users by providing a clear definition of presented data and indicators, as well as an explanation about the methods of elaboration, calculation, and/or reporting mechanisms.

Neutrality

Our final qualitative aspect is neutrality. The FASB (1980, p. 41) states neutrality means that “the primary concern should be the relevance and reliability of the information” not the effect that it may have on a particular interest. The GRI refers to this concept as “balance” and notes that both the positive and the negative aspects of organizations’ activities should be reported to allow for a reasoned assessment of overall performance. However, as noted by Li et al. (1997) in discussing environmental liability information, firms are reluctant to disclose bad news owing to potentially increased proprietary costs. As such, Patten (2000) argues that when companies do disclose negative items of CSR data, they have an incentive to offset that disclosure via the provision of other more positive pieces of information.Footnote 8 Thus, to the extent that quality matters in providing an offsetting image, disclosure of negative CSR information might also be associated with higher overall quality in the reporting package. Given this potential impact on the reporting, and due to our empirical tests related to disclosure quality (discussed below), we treat neutrality as a separate component of CSR informational quality.

As noted by Delbard (2008), a major intent of the NRE was to increase the provision of transparent and comprehensive information on firms’ social and environmental activities. We argue that in order to increase transparency, the informational quality of disclosure should be enhanced. As such, we extend prior research in the CSR domain by examining for changes in the informational quality of disclosure over time.

Quality and Legitimacy

Regardless of whether prior studies used space-based or item-specific content measures, many were interested in using the metrics to examine what drives differences in the disclosure. Gray, Owen, and Adams (1996, p. 45) note that legitimacy theory is a “systems-oriented” theory that underlies the explorations of “the role of information and disclosure in the relationship(s) between organizations, the State, individuals, and groups.” Deegan (2002, p. 292) further states that systems-oriented theories recognize disclosure as an “important means by which management can influence external perceptions about their organization.” Where firms face exposures to the social and political processes through which social legitimacy is monitored and bestowed (Patten 1991; 1992), they have an incentive to use disclosure to enhance their legitimacy. As such, differences in social and political exposures are expected to be associated with differences in CSR disclosure. Deegan (2002, p. 297) notes that a large body of research studies provide evidence that “corporate social and environmental disclosure strategies have been linked to legitimizing intentions” (also see Deegan 2007). More specifically, this prior body of work documents that differences in the extent of CSR disclosure (some studies use space-based measures while others rely on item-specific content schemes) tend to be associated with factors assumed to capture differences in exposures to social and political pressures. As such, disclosure is shown to be more about reducing exposures than about providing meaningful and transparent information on social and environmental aspects of the firms. As we discuss in more detail below, the primary legitimacy factors include firm size and industry membership.

Given the substantial body of evidence documenting the relation between social and political exposures and the extent of CSR disclosure, we attempt in our investigation to determine if the use of disclosure as a legitimating device changes over our period of investigation. We argue that if disclosure is moving toward transparency, the influence of legitimacy-based factors should decrease. We examine this both relative to changes in the space allocated to CSR disclosure and the informational quality of the disclosures.

Research Methods and Analysis

Sample Selection

In order to determine whether CSR disclosure by French firms is changing over the period subsequent to the adoption of the NRE, we focus on a sample of companies listed on the French SBF 120 index in 2004 and compare those disclosures to information provided in the same companies’ 2010 reports. To be included in our sample, firms had to meet the following criteria:

  1. 1.

    They had to be listed on the French SBF 120 index for fiscal year 2004.

  2. 2.

    They had to have disclosed social and/or environmental information in either a stand-alone sustainability-type report or as part of their annual report in 2004.

  3. 3.

    They had to have required data for both 2004 and 2010.

Of the 120 SBF index companies in 2004, three did not publish their reports on their websites and did not respond positively to our request for these documents. Among the 117 companies for which reports were collected, 26 did not publish any social or environmental information in their reports, and ten did not have required data for both sample years.Footnote 9 Therefore, our final sample consists of 81 companies. The number of sample firms with stand-alone reports increased from 24 in 2004 to 31 in 2010. However, only twelve of the companies issued a stand-alone report for both years. In contrast, 43 sample firms included disclosures in only their financial reports for both years. Although our focus is on the change in disclosure across the entire reporting package (both financial reports and, where issued, stand-alone reports), we also provide comparative data across subsets of companies based on the consistency of reporting medium.

Data Collection

To collect data for analysis, the sample company reports for both 2004 and 2010 were reviewed and coded by two members of the research team relative to the provision of CSR information. Where discrepancies across reviewers existed, the differences were discussed and reconciled. We started by identifying the space devoted to disclosure across the environmental and social categories noted above and measured this disclosure in pages, where the total was rounded up to the next highest page. We next calculated the breadth of disclosure by identifying how many of the five respective areas (environmental, business practices, human resources, community, and products) were discussed in the sample reports. This breadth measure could, therefore, range from zero to five.

Finally, we measured disclosure quality based on the informational characteristics discussed above. More specifically, we examined the sample company disclosures and awarded one point for each of the following attributes:

Relevance

  • Identification of stakeholders and their needs.

  • Dialog with stakeholders to define their needs.

  • Analysis of risk factors specific to the company.

Comparability

  • Temporal features and comparisons.

  • Comparisons with other companies or to external norms.

  • Clear definition of the report perimeter and boundaries.

Verifiability

  • Verifiability of information presented.

Clarity

  • Clear definition of presented data and indicators.

  • Explanation about the methods of elaboration, calculation, and/or reporting mechanisms.

Since we are trying to identify the disclosure quality for the entire CSR reporting package, we examined for the presence of each of the nine characteristics relative to each of the five areas of CSR disclosure included in our analysis. Thus, the maximum disclosure quality score is 45 points (nine attributes over five areas). Since not all companies provide information across all five of the disclosure areas, we also calculated an average disclosure quality per area score, measured as the total quality score divided by the breadth of disclosure metric. Finally, we separately measured neutrality in the reporting by examining for the disclosure of negative performance information across each of the CSR areas.

Legitimacy Analysis

In this stage of the investigation, we attempt to identify if CSR disclosure is influenced by legitimacy concerns and whether those relations are reduced in 2010 relative to the 2004 reporting. We start by testing the 2004 reporting for associations between disclosure, using both space and quality measures, and several factors used in prior research as legitimacy measures. The specific legitimacy variables included are firm size, membership in a socially exposed industry, and the disclosure of negative performance information.Footnote 10 A number of prior studies (Cowen et al. 1987; Hackston and Milne 1996; Patten 1991, 2002b) document that the extent of social and/or environmental disclosure is positively associated with firm size and membership in industries whose activities have greater social or ecological impacts. Larger firms, presumably due to greater visibility are assumed to face greater social and political pressures, and, as such, may use disclosure as a tool to reduce those exposures (Patten 1991, 2002b). We measure firm size as the log-transformed amount of total revenues for each of our sample companies and for each of the appropriate years. Firms from socially exposed industries are also assumed to face greater social and political pressures (Brammer and Millington 2005; Patten 1991) and thus also have an incentive to use disclosure to reduce these exposures. For our analysis, we follow Brammer and Millington (2005) and designate companies from the chemical, extractive, paper, pharmaceutical, alcoholic beverage, and defense industries as being socially exposed.

Our final legitimacy variable centers on the concurrent disclosure of negative social or environmental information. Patten (2000) documents that mandatory increases in the disclosure of hazardous waste-related disclosures (negative information) by U.S. companies in the early 1990s were accompanied by associated increases in the provision of other, more positive, environmental information. Given the reluctance of firms to disclose negative information (Li et al. 1997), we anticipate that companies choosing to do so have an incentive to offset that negative information both with more extensive disclosure and with a higher quality disclosure package. Sample firms disclosing any piece of negative social or environmental performance information are coded one, while the negative disclosure variable (NegDisc) for non-disclosing companies is coded zero. Overall, we estimate the following multiple regression modelFootnote 11 for each measure of disclosure (total pages, breadth, informational quality scores, and quality per area of disclosure scores) for 2004:

$${\text{Discl}}_i = a_ 1+ B_ 1{\text{Size}}_i + B_ 2{\text{SEI}}_i + B_ 3{\text{NegDisc}}_i$$

Based on prior studies, we expect each of the legitimacy variables to be positively related to difference in the disclosure measures.

In order to identify whether the relation between the legitimacy factors and disclosure is changing, we pool the 2004 and 2010 data and add interaction terms to the model. The expanded model is stated as:

$$\begin{aligned} {\text{Disc}}_i =\, & a_ 1+ B_ 1{\text{Size}}_i + B_ 2{\text{SEI}}_i + B_ 3{\text{NegDisc}}_{\text{i}} + B_ 4{\text{Yr}} \\ & + B_ 5{\text{Yr}}*{\text{Size}}_i + B_ 6{\text{Yr}}*{\text{SEI}}_i + B_ 7{\text{Yr}}*{\text{NegDisc}}_i \\ \end{aligned}$$

If disclosure is moving away from legitimization we would expect the parameters on the interaction variables—B 5, B 6, and B 7—to be significantly negative.Footnote 12

Results

Descriptive Statistics

Table 1 summarizes descriptive statistics for all model variables. As indicated in the table, our sample companies are relatively large with a mean sales revenue of 13,515 million Euros in 2004 (15,632 million Euros in 2010) and profitable with a mean ROA of 3 % in 2004 (4 % in 2010). The mean leverage of the firms stands at 18 % of total assets in 2004 (20 % in 2010), whereas the ratio of asset newness is 44 % in 2004 (43 % in 2010) and the capital intensity is 5 % for both years. There were eight firms (in both years) from socially exposed industries. Finally, the mean CSR disclosure scores are 17.47, 2.80, 7.42, and 2.40 in terms of total pages, breadth, information quality score, and quality per area in 2004, respectively (36.71, 3.68, 10.69, and 2.58, respectively, in 2010).

Table 1 Descriptive statistics
Full size table

Pages of Disclosure

Table 2 summarizes the mean number of pages of CSR disclosure for our sample firms across the 2004 and 2010 periods. As noted in Panel A of the table, the mean number of pages of CSR disclosure increased from 17.47 in 2004 to 36.71 in 2010.Footnote 13 This difference is highly significant. However, as noted above, the disclosure media for the sample firms overall varied across the two periods. In particular, more firms issued stand-alone reports in 2010 than had been released in 2004,Footnote 14 and, as such, the difference in space devoted to CSR disclosure could be a function of this factor. To more carefully assess the extent of changes in space across consistent reporting media, we separately tested for changes in the pages of disclosure for a sub-group of firms that issued no stand-alone report in either 2004 or 2010 (designated as sub-group 1), and for firms that issued stand-alone reports in both years (designated as sub-group 2). Panels B and C, respectively, report the results of these additional analyses. As shown in Panel B of Table 2, the total pages of CSR disclosure for companies relying on only financial report disclosure increased from just under 12 pages, on average, in 2004, to 18.47 pages in the 2010 financial reports. This difference is highly significant. Similarly, the pages of CSR disclosure for sub-group 2, those companies issuing stand-alone reports in both 2004 and 2010, also increased significantly. Mean page counts increased from 37.82 to 80.45 over the two periods.Footnote 15 Overall, the assessments of changes in the space devoted to CSR disclosure consistently indicate that French firms increased the disclosure of this information significantly, and thus would appear to support arguments that CSR disclosure is attaining a higher level of normativity over time.

Table 2 Mean pages of CSR disclosure 2004 versus 2010
Full size table

Quality of Disclosure

Breadth of Disclosure

Data related to the extent of CSR disclosure across categories are presented in Table 3. As noted in Panel A of the table, the number of companies including disclosures of environmental and energy information (60 in 2004 and 67 in 2010) and human resources (63 in 2004 and 69 in 2010) showed only minor improvements (neither change is statistically significant). In contrast, the number of firms presenting business practice, community, and product CSR information increased significantly (based on Chi square tests of the proportion of disclosers) across the 2 years of analysis. Not surprisingly given the increases across all five areas of disclosure, the mean breadth scores, reported in Panel B of Table 3, rose from 2.80 in 2004 to 3.68 in 2010. This difference is highly significant. However, as also indicated in Panel B of the table, the extent of change appears to differ across reporting media. Sub-group 1 companies (those relying only on financial report disclosure in both years) show significantly increased breadth of disclosure scores (2.33–3.05 from 2004 to 2009), while the sub-group 2 companies that issued stand-alone reports in both 2004 and 2009 do not exhibit statistically significant increases. However, this may largely be due to the already high breadth of disclosure (an average of 4.58 out of a possible five) in 2004. While perhaps slightly less robust than the results for the pages of disclosure, the findings with respect to breadth of disclosure further support the claim that CSR disclosure normativity in France is increasing.

Table 3 Breadth of CSR disclosure 2004 versus 2010
Full size table

Informational Quality

We next turn to our analysis of the informational quality of the CSR disclosures. Table 2 summarizes the attributes and areas of CSR disclosure quality for our sample firms (Panel A depicts the data for 2004 disclosures whereas 2010 data are summarized in Panel B). Individual cells identify the number of sample firms that are awarded points for each characteristic within each area of disclosure. The final column of the table indicates the sample’s average score (out of a possible five points) for each of the quality characteristics while the final row shows the average quality score (out of a possible nine points) by area of disclosure. The bottom right cell of the table shows the mean overall CSR informational quality score.

Certainly, one of the most striking features reflected in Table 4 is the low informational quality of the disclosures. The mean informational quality score is only 7.42 (out of a possible 45) for the reporting packages in 2004, and still only 10.69 based on the 2010 analysis. For 2004, the informational quality item most reflected in the French disclosures is verifiability with a sample mean of 2.21 (out of a possible five), and only two other attributes, the disclosure of temporal comparisons and clarity—definitions, exhibit an average score higher than one. With respect to the areas of CSR disclosure, human resource disclosures (an average score of 2.45 out of a possible nine) and environmental disclosures (with a mean score of 2.42) exhibited the highest quality. Informational quality scores across all five reporting areas showed increases in 2010 relative to the 2004 scores; although consistent with the earlier period, human resource (an average score of 3.38) and environmental (mean of 2.80) disclosures continued to show the highest quality. The comparisons across quality attributes are more mixed. Whereas four factors (relevance—identification, relevance—dialogue, comparability—temporal, and clarity—definition) showed increased scores; on average, the number of firms including relevance—analysis, comparability—others, comparability—perimeters, verifiability, and clarity—methods was lower in 2010 than it had been for the 2004 reporting packages. More detailed analyses of changes in informational quality are reported below.

Table 4 Quality of disclosure by attribute and area
Full size table

Table 5 presents comparisons of mean informational quality scores across 2004 and 2010 for the total sample (Panel A), sub-group 1 (Panel B), and sub-group 2 (Panel C). As indicated in Panel A, the difference in overall informational quality scores between 2004 (mean of 7.42) and 2010 (average score of 10.69) is statistically significant at the .001 level, two-tailed. Similarly, mean scores across 2004 and 2010 for sub-group 1 firms disclosing only in financial reports (means of 5.19 and 8.30, respectively) are significantly (at p = .008, two-tailed) different. However, informational quality scores for the twelve companies issuing stand-alone reports in both 2004 and 2010, while higher, on average, than those in sub-group 1, show little improvement over time. The mean informational quality score for these firms was 13.83 in 2004 and rose to only 14.75 by 2010. This difference is not statistically significant at conventional levels. Similarly, the average informational disclosure score per area of disclosure (for the total sample), highlighted in Panel D of Table 5, while slightly higher in 2010 relative to 2004 (2.58 vs. 2.40), does not reflect a change that is statistically significant.

Table 5 Mean CSR disclosure information quality scores 2004 versus 2010
Full size table

Overall, analysis of informational quality scores for the CSR disclosures of our sample of French firms for 2004 and 2010 provides results that suggest that if normativity in reporting is being approached, it appears to be doing so without a substantial increase in informational quality. While scores for the overall reporting package show improvement, differences appear to be driven by financial report-only disclosers. Further, the finding that informational quality scores per area of disclosure have not changed significantly suggests that the improved informational quality overall is likely attributable to expansions in the areas of CSR disclosure (again limited to financial report-only disclosers) rather than to changes in the quality attributes of the information itself.

Disclosure of Negative Information

Finally, Table 6 identifies the disclosure of negative CSR performance information by our sample of French companies. As noted in the table, 24 firms disclosed at least one piece of negative social or environmental performance data in their 2004 reports whereas only 17 companies did so in 2010. The decline is almost exclusively due to the decrease in the number of companies that disclosed negative environmental information. While 15 sample firms included negative environmental performance information in 2004, only five did so in their 2010 reports. One possible explanation for this finding is that environmental performance across the sample firms is improving in almost all areas. Unfortunately, we are aware of no publicly available, wide-scale environmental performance metrics for French companies, and as such we cannot assess the validity of this explanation.

Table 6 Companies disclosing negative performance information
Full size table

In sum, our analyses indicate that the space allocated to CSR information by French firms significantly increased between 2004 and 2010 but the information quality largely did not improve. These results suggest that such disclosure may be playing an increasingly important part of firm corporate communications in favor of its image and reputation. The “managerial capture” of this reporting goes against the NRE goal of democratic transparency in that stakeholders can be deceived by large amounts of—but not very informative—disclosure (Owen et al. 2000; O’Dwyer 2003).Footnote 16

Legitimacy Analysis

Our last stage of analysis focuses on the degree to which differences in disclosure for our sample of French firms appears, if at all, to be associated with attempts at legitimization and whether that has changed over time. Table 7 presents the results of our regression analyses on the four disclosure measures (total pages, breadth, informational quality, and quality per area) for 2004. As shown in the table, all three legitimacy variables—firm size, socially exposed industry membership, and the disclosure of negative social or environmental performance information—are positively and significantly (at p < .10, one-tailed or better) related to differences in both the pages of CSR disclosure and breadth of the reporting. Firm size and negative disclosure are also significantly related to differences in informational quality and the quality per area of disclosure scores; whereas the SEI variable, although still positive in both cases, is not significant at conventional levels. Overall, these results are consistent with other legitimacy-based explorations of CSR disclosure (Hackston and Milne 1996; Patten 1991, 2000), and suggest that the French companies were using disclosure as a tool of legitimization at that time.

Table 7 Regression results for the relation between CSR disclosure and legitimacy variables—2004
Full size table

Table 8 presents the results of the tests for changes in the relation between the legitimacy variables and the disclosure metrics for 2010 relative to 2004. The analysis of the total pages of disclosure is the only model yielding any statistically significant interaction effect. The Yr*Size interaction variable is significantly, but positively associated with the space measure. This indicates that firm size plays an even greater role in explaining differences in disclosure space in 2010 than it had in 2004. In contrast, the Yr*SEI variable is negatively and significantly (at p < .10, two-tailed) related to pages of disclosure. Thus, socially exposed industry membership is less of a factor for disclosure space in 2010 than it had been in 2004. For the other three disclosure models, however, none of the interaction terms is statistically significant (and there is no consistent pattern of direction). Overall, therefore, our evidence suggests that legitimacy-based influences on CSR disclosure are not being reduced over time.

Table 8 Regression results for tests of the change in relation between CSR disclosure and legitimacy variables
Full size table

Conclusion

In this analysis, we examined the extent to which CSR disclosure provided by publicly traded French firms has changed over time. Using the notion of normativity as presented by Bebbington et al. (2012), we empirically tested whether the argument of normativity leading to a diffusion stage and further a taken-for-granted level of practice held across our sample of French firms. We find that space allocated to CSR disclosure in both financial and stand-alone reports by our sample firms significantly increased when comparing 2004 disclosure to the 2010 reporting year. We also document that, for financial report issuers only, the breadth of CSR disclosure has increased. Each of these increases appears to suggest that disclosure is at least moving toward normativity. Our contribution to, and extension of, prior CSR disclosure research also lays in our assessment of the changes in the informational quality of corporate CSR reporting based on attributes derived from recommendations of the International Accounting Standards Board, the Financial Accounting Standards Board, and the Global Reporting Initiative. Our results show that despite a slight improvement in the informational quality of disclosure, quality remains low and appears to be associated with increasing breadth of topical coverage in financial reports, and that the disclosure quality per area of disclosure has not changed significantly. Further, we find that negative CSR performance disclosure has decreased over time. Finally, our analysis shows that differences in disclosure, measured using space, breadth, or informational quality measures, appeared to be largely driven by legitimacy concerns in 2004, and that those relations remain largely unchanged in 2010.

Overall, our results suggest that while France’s regulatory reporting requirements appear to be moving toward normativity given the increases in space and breadth of disclosure, they have not led to a proliferation of high-quality CSR reporting packages by French corporations. As such, it appears that the NRE goal of increased transparency in CSR disclosure remains unmet.

Of course, given its exploratory nature, our study has several limitations that must be noted. We examine only two separate years of reporting rather than disclosure over a continuous time period and our investigation is limited to 81 firms. As such, we cannot observe the detailed patterns of changes in space and informational quality in CSR disclosure over time or for the entire population of potentially affected companies. Related to this point, some sample firms chose not to disclose CSR information in either 2004 or 2010, suggesting that uncontrolled but related factors (self-selection bias) may be influencing disclosure choice, and these could potentially also influence differences in quality. Next, while we can observe a change in stages of normativity, we are not able to make a direct causal relation and explanation between this change of stage and the disclosure changes in pattern. We also focus on only French firms. Corporations in other countries may do a better job of incorporating quality into their CSR reporting (see, e.g., Spain and the UK in Bebbington et al. 2012). Finally, we acknowledge that our informational quality disclosure metric is based on a subjective evaluation of how the attributes discussed by the IASB, the FASB, and the GRI can be captured and measured. Refinements and extensions of the metric could yield more insight into the strengths and weaknesses of corporate CSR disclosure. We would encourage extensions of our work along each of these limiting dimensions.