Mandatory versus voluntary non-financial reporting: reporting practices and economic consequences

ABSTRACT

Non-financial reporting mandates are considered in many jurisdictions, with the European Union (EU) already having a mandate in place: Directive 2014/95/EU (“the Directive,” henceforth). Previous studies indicate that the Directive has achieved its goal of enhancing non-financial transparency, resulting in declines in firm value. We find that the transparency enhancements are concentrated in firms that voluntarily start non-financial reporting in anticipation of the Directive's entry-into-force date, without these firms facing net costs associated with reporting. In contrast, firms that only report when mandated are more likely to provide boilerplate disclosures, consistent with facing a negative cost–benefit trade-off of reporting, leading them to experience increased information asymmetries. Our results suggest that the flexibility embedded in and lack of enforcement of the Directive renders it ineffective in promoting non-financial transparency among firms with weak incentives to disclose meaningful information.

KEYWORDS:

• Non-financial reporting
• Directive 2014/95/EU
• boilerplate language
• firm value
• information asymmetry

1. Introduction

Non-financial reporting mandates are considered in many jurisdictions, with the European Union (EU) already having a mandate in place: Directive 2014/95/EU¹ (“the Directive,” henceforth). Previous studies indicate that the Directive has achieved its goal of enhancing non-financial transparency, leading to increases in social activities but also declines in firm value (Cuomo et al., 2022; Fiechter et al., 2022; Jackson et al., 2020). While Fiechter et al. (2022) highlight that firms, in response to the Directive, start increasing their non-financial transparency before – and in anticipation of – its entry-into-force date, these studies do not distinguish between firms that voluntarily start non-financial reporting (NFR) before the Directive's entry-into-force date (“voluntary reporters”) and firms that only report when mandated (“mandatory reporters”).

Considering the flexibility embedded in and the lack of enforcement of the Directive (see Section 2.1), we anticipate heterogeneity in reporting practices and the associated economic consequences between these two groups. Specifically, we expect that voluntary (mandatory) reporters face a positive (negative) cost–benefit trade-off of reporting and employ less (more) boilerplate language in their disclosures. Consequently, we hypothesise that voluntary (mandatory) reporters experience increases (decreases) in firm value and decreases (increases) in information asymmetries after initiating NFR.

To test our hypotheses, we focus on 497 firms from the STOXX Europe 600 subject to the disclosure requirements of the Directive. To assess these firms’ reporting practices, we collect all their reports for the financial years ending in or after 2012 and up to 2020. Subsequently, we identify firms that, in addition to non-financial disclosures mandated by national legislators, start providing (supplementary) non-financial information voluntarily in anticipation of the Directive's entry-into-force date (“voluntary reporters”) and firms that only provide such information when mandated (“mandatory reporters”).

First, we document a positive association between NFR and the use of boilerplate language for mandatory reporters and no association for voluntary reporters. Second, we employ a staggered difference-in-differences (DiD) research design and find that mandatory reporters experience a decrease in firm value (as captured by Tobin's Q) following NFR, whereas voluntary reporters do not experience changes in firm value. This contrasts with prior studies (e.g. Fiechter et al., 2022) documenting negative average firm value effects of NFR in response to the Directive. Our results show that these negative effects are concentrated among mandatory reporters. In our third set of analyses, we find that mandatory reporters exhibit increases in information asymmetries (as captured by the bid-ask spread) after initiating NFR compared to before and relative to voluntary reporters. By contrast, voluntary reporters experience reductions in information asymmetries, suggesting that their reporting is meaningful and informative. We interpret this as complementary evidence that mandatory reporters evade the spirit of the Directive by obfuscating potentially meaningful information, thereby impeding investors from efficiently processing and interpreting disclosures.

Overall, our results, all of which are robust to various sensitivity tests, suggest that the flexibility embedded in and lack of enforcement of the Directive renders it ineffective in promoting non-financial transparency among firms with weak incentives to disclose meaningful information.

The remainder of this paper is organised as follows. Section 2 presents details of the institutional context and develops the hypotheses. Section 3 discusses the study’s sample and outlines the research design. Section 4 illustrates the study’s main findings, and Section 5 presents the robustness tests. Finally, Section 6 concludes.

2. Institutional context and hypothesis development

2.1. Institutional context

On 15 April 2014, the European Parliament (EP) passed Directive 2014/95/EU to enhance the relevance, consistency, and comparability of non-financial information disclosed by undertakings across the EU (EU, 2014). The Directive recognises that member states cannot achieve this objective alone (Recital 21). Already back in April 2011, the European Commission (EC) identified in its “Single Market Act” the need to raise the transparency and comparability of non-financial information to a consistently higher level across all member states and sectors (Recital 1).

Before the Directive, the landscape of NFR policies in the EU was fragmented and heterogeneous. Specifically, 15 of the 27 EU Member States implemented national policy frameworks to promote corporate social responsibility (CSR) (EC, 2011). Most of these national policy frameworks were not legally binding but intended to serve as recommendations or guidelines, and many only required the disclosure of a few non-financial metrics (C&S, 2013). Appendix A provides an overview of the mandatory country-specific legislation in place before the Directive.² Despite significant variations in reporting requirements across European countries, we find that (i) most policies required the annual report as a disclosure venue, (ii) no policy mandated the use of internationally recognised NFR frameworks,³ and (iii) only France and the United Kingdom (UK) imposed stringent non-financial disclosure requirements.

After the implementation of the Directive, effective from reporting cycles 2018 (which refer to financial years beginning in 2017), European companies (and those headquartered in Iceland and Norway) of public interest that qualify as “large undertakings” with more than 500 employees must provide – at least – annual information on “policies, main risks, and outcomes related to environmental matters, social and employee factors, respect for human rights, anti-corruption issues, and diversity of the board of directors” (AE, 2018; EU, 2014). Surprisingly, however, the Directive does not require the use of a specific NFR framework. Article 1 states that firms’ reporting can rely on any existing national (such as standards developed by Social Value UK) or international reporting framework (such as standards set by the Global Reporting Initiative). Given the existence of a plethora of NFR frameworks (Halkos & Skouloudis, 2016), each with different focal points and varying target groups, the literature suggests that the Directive is unlikely to result in the anticipated harmonisation and comparability of non-financial disclosures (La Torre et al., 2018). Another surprising aspect of the Directive is its lack of enforcement. While member states must oversee compliance with the Directive, no official EU documentation or guidance outlines member states’ enforcement mechanisms and activities. Furthermore, the European Securities Regulator (ESMA) has not provided any guidance or reported information on enforcement activities at the country level so far (Fiechter et al., 2022).

2.2. Hypothesis development

A growing body of literature analyses the effects of Directive 2014/95/EU. Early studies focus on the impact of the Directive on firm value and stock prices. For example, Grewal et al. (2019) examine (short-window) market reactions to events leading up to the passage of the Directive. They document an average adverse market reaction concentrated in firms with weak environmental, social, and governance (ESG) performance and low ESG disclosure. Mittelbach-Hörmanseder et al. (2021) study the relationship between topic-specific CSR disclosure and firm value. They find that the relationship and magnitude of specific CSR topics in explaining firm value undergo changes after the announcement of the Directive.

More recent studies focus on whether the Directive has achieved its goal of enhancing non-financial transparency. For example, Cuomo et al. (2022) analyse firms in 17 European countries and find that the Directive increased CSR transparency. Similarly, Fiechter et al. (2022) find that firms increase their reporting transparency in response to the Directive and start doing so before – and in anticipation of – its entry-into-force date. Surprisingly, however, Fiechter et al. (2022) (and others) do not distinguish between firms that voluntarily increase their non-financial transparency before the Directive's entry-into-force date (“voluntary reporters”) and firms that only report when mandated (“mandatory reporters”). Instead, they pool voluntary and mandatory reporters and study average effects. Considering their different underlying incentives and the flexibility embedded in the Directive and its limited enforcement (see Section 2.1), we anticipate heterogeneity in reporting practices and the associated economic consequences between these two groups.

It is anticipated that reporting practices differ between voluntary and mandatory reporters, as voluntary reporting choices reflect firms’ cost–benefit trade-offs (Christensen et al., 2021). Firms with strong non-financial performance or those subject to legitimacy threats are likely to face net benefits of increased transparency through reporting and, therefore, initiate NFR even in the absence of a reporting mandate (de Villiers & Marques, 2016; Glennie & Lodhia, 2013; Mittelbach-Hörmanseder et al., 2021). Conversely, firms facing net costs of disclosure are likely to report only when mandated (Christensen et al., 2021).

Christensen et al. (2021) argue that using boilerplate language⁴ is a meaningful way to reduce the net costs associated with mandatory NFR. Hence, firms facing negative cost–benefit trade-offs of reporting meaningful information may use boilerplate language as an avoidance strategy.⁵ Prior literature indeed suggests that firms often respond to new disclosure requirements by providing boilerplate disclosures. For instance, Dyer et al. (2017) investigate the implementation of a risk factor disclosure mandate in the US. They find that firms provide boilerplate disclosures to circumvent their obligation to provide meaningful risk information. Similarly, the Business and Human Rights Resource Centre report (BHRRC, 2017) documents firms’ widespread use of boilerplate language to comply with new regulations. These findings align with Li (2008), who demonstrate that managers opportunistically use less readable language in their reports to conceal adverse information from investors.

The use of boilerplate language can be limited by requiring firms to report against established reporting frameworks that dictate what and how firms are required to report (Christensen et al., 2021). However, the Directive does not mandate a specific reporting framework; it allows companies to rely on any existing national or international reporting framework (see Section 2.1). Moreover, although companies may use the same reporting frameworks, the embedded discretion within these frameworks allows them to influence reporting outcomes based on their cost–benefit trade-offs. Consequently, Lang and Stice-Lawrence (2015) document that the effectiveness of adopting International Financial Reporting Standards (IFRS) in reducing information asymmetries depends on the level of boilerplate disclosures. Similarly, Christensen et al. (2015) conclude that the flexibility embedded in the IFRS may render it ineffective for firms with low incentives to adopt it, as evidenced by their finding that improvements in reporting quality following IFRS adoption are confined to voluntary adopters only.

Taken together, voluntary reporters are expected to face net benefits of increased transparency through reporting, incentivising them to provide transparent, non-boilerplate disclosures. In contrast, mandatory reporters are likely to face net costs of reporting. To (partially) circumvent the provision of costly, meaningful information and thus mitigate the net costs of reporting, they are expected to exploit the flexibility embedded in the Directive and provide boilerplate disclosures. Therefore, we propose our first hypothesis:

H1: There is a positive association between NFR and the use of boilerplate language for companies that only report non-financial information when mandated, whereas there is no (or a negative) association for companies that voluntarily start reporting non-financial information before the mandate.

The contrasting cost–benefit trade-off between mandatory and voluntary reporters further implies that the documented decrease in firm value, associated with the increased non-financial transparency in response to the Directive (Fiechter et al., 2022), is likely concentrated among mandatory reporters. Similarly, the improvements in non-financial transparency may be confined to voluntary reporters: studies focusing on voluntary disclosure find that firms exhibit decreases in information asymmetries following their reporting (e.g. Clarkson et al., 2013; Dhaliwal et al., 2011), while research on mandatory disclosure indicates that mandated reporting does not necessarily reduce information asymmetries but might instead increase them when the disclosures are boilerplate (e.g. Christensen et al., 2022). Hence, we analyse whether the association between NFR and firm value and information asymmetry differs between voluntary and mandatory reporters.

For the first group, voluntary reporters, the cost–benefit trade-off of reporting is likely positive. Economic and socio-political theories suggest that firms with strong non-financial performance or those subject to legitimacy threats face net benefits of increased non-financial transparency, leading them to voluntarily start NFR, even in the absence of a reporting mandate (Christensen et al., 2021; Mittelbach-Hörmanseder et al., 2021). More specifically, economic-based theories of disclosure (i.e. voluntary disclosure theory and signalling theory) suggest that better-performing firms voluntarily disclose their non-financial performance to stakeholders, thereby differentiating themselves from companies with inferior performance to gain competitive advantages (e.g. a higher firm valuation) (Healy & Palepu, 2001). These advantages may materialise as investors note the superior non-financial performance and anticipate benefits such as increased revenues (Lev et al., 2010; Sen & Bhattacharya, 2001), higher firm productivity (Roberts & Dowling, 2002; Turban & Greening, 1997), and lower future capital costs (Dhaliwal et al., 2011; Richardson & Welker, 2001).

Socio-political theories, particularly legitimacy theory, suggest that a company's existence relies on society's willingness to let it operate (O’Donovan, 2002). Therefore, even when firms’ non-financial performance is weak, companies that face legitimacy threats (e.g. firms operating in polluting industries such as mining, oil exploration, or steel manufacturing) may voluntarily disclose non-financial information to advance their image as socially responsible (Christensen et al., 2021; de Villiers & Marques, 2016; Glennie & Lodhia, 2013). The increased non-financial transparency can signal managers’ trustworthiness and engagement in social and environmental practices (Christensen, 2016; Godfrey et al., 2009; Tsang et al., 2022), mitigate investors’ estimation risks (Dhaliwal et al., 2011; Griffin & Sun, 2013), and reduce investors’ perceptions that the firm will be forced to incur future expenses related to regulation compliance, litigation, and remediation (Tsang et al., 2022). Consequently, for firms subject to legitimacy threats, the disclosure of poor non-financial performance may benefit firm value relative to non-disclosure. Accordingly, Tsang et al. (2022) document a significant positive relationship between voluntary NFR and firm value even after controlling for firms’ non-financial performance.

The theories suggesting that voluntary reporters face net benefits of increased transparency through reporting also imply that voluntary reporting is associated with reduced information asymmetries (Barry & Brown, 1985). Studies find support for this relationship in examining both financial reporting settings (e.g. Botosan & Plumlee, 2002; Leuz & Verrecchia, 2000) and non-financial reporting settings (Clarkson et al., 2013; Dhaliwal et al., 2011; Muslu et al., 2019; Plumlee et al., 2015). For example, Dhaliwal et al. (2011) document that the voluntary disclosure of CSR activities reduces a firm's cost of capital through reduced information asymmetries. Similarly, Clarkson et al. (2013) demonstrate the effectiveness of voluntary environmental reporting in mitigating information asymmetries, with these disclosures proving to be incrementally informative in predicting firm profitability and value.

For the second group, mandatory reporters, the cost–benefit trade-off of disclosure is likely negative, leading them to report only when mandated. To (partially) reduce the net costs of reporting, mandatory reporters are anticipated to evade the spirit of the Directive by providing boilerplate disclosures, thereby obfuscating potentially meaningful information (Christensen et al., 2021). Hoogervost (2013) suggests that using boilerplate language in disclosures to reduce legal exposure decreases their overall informativeness. Furthermore, previous studies show that low-quality (i.e. boilerplate) disclosures are often interpreted differently, leading to disagreements among share- and stakeholders (e.g. Christensen et al., 2022; Lang & Stice-Lawrence, 2015; Li, 2008). For example, Christensen et al. (2022) study the introduction of country-level NFR mandates. They find that providing more non-financial information in response to such mandates coincides with elevated information asymmetries between firms and rating agencies, as evidenced by the rising disagreement among ESG ratings. Consequently, we expect mandatory reporters to experience increases in information asymmetries after initiating NFR.

Nevertheless, it is unlikely that boilerplate language will completely obscure negative information or obfuscate investors such that the net costs of NFR become net benefits. Therefore, disclosures from mandatory reporters are expected to (still) convey poor signals or impair legitimacy, causing investors to anticipate political costs (Grewal et al., 2019), investments in value-decreasing projects (Brammer & Pavelin, 2006; Grewal et al., 2019; Healy & Palepu, 2001; Jensen & Meckling, 1978; Watts & Zimmerman, 1978), or decreased competitiveness due to the revelation of proprietary information (Breuer et al., 2019; Christensen et al., 2021; Grewal et al., 2019). The disclosure of negative information that a firm previously withheld based on its cost–benefit trade-off is expected to decrease firm value.

In summary, we anticipate that the association between NFR and firm value (H2a), as well as the association between NFR and information asymmetry (H2b), is heterogeneous between voluntary and mandatory reporters. Thus, we propose our second hypothesis:

H2a: There is a negative association between NFR and firm value for companies that only report non-financial information when mandated, whereas there is a positive (or no) association for companies that voluntarily start reporting non-financial information before the mandate.

H2b: There is a positive association between NFR and information asymmetry for companies that only report non-financial information when mandated, whereas there is a negative association for companies that voluntarily start reporting non-financial information before the mandate.

3. Sample and methodology

3.1. Sample

Our empirical analyses draw on the STOXX Europe 600 Index, which represents approximately 90% of European stock market capitalisation, covering firms across different industries and countries in the European region (STOXX, 2021). We start with the 600 firms from Eurostoxx and then drop companies not subject to the Directive, either because their country of residence does not need to adhere to the Directive (e.g. Switzerland) or because the company does not fall within the scope of its national-level transpositions (due to its number of employees, balance sheet total, or net turnover).⁶ The first step yields a sample of 497 firms.

In the second step, we hand-collect annual reports, integrated reports,⁷ and stand-alone non-financial reports from each of the 497 firms’ websites covering the financial years ending in or after 2012 and up to 2020. Our sample starts in 2012 because reports before this year are unavailable online for most firms. We then analyse NFR practices for each firm-year by requiring reporting “by substance rather than reference”. In other words, to qualify as a “true” non-financial reporter, firms must provide stand-alone non-financial or integrated reports prepared against internationally recognised NFR frameworks⁸ (“strict classification scheme”). This classification is based on the notion that simply referring to one (discretionarily chosen) NFR framework in an annual report does not necessarily imply informative NFR; in such cases, the reported non-financial information usually spans only one or very few pages. In contrast, non-financial information reported in stand-alone non-financial or integrated reports is comprehensive and “in substance;” the document length of our sample stand-alone non-financial reports averages 58 pages, and the non-financial segments within our sample integrated reports have an average length of 51 pages, indicating that such reports are – at least – comprehensive and not merely referencing an NFR framework. While our strict classification scheme ensures that our analyses focus on reporting practices that are comprehensive and “in substance”, it also reduces the sample size. Therefore, in robustness tests, we relax our strict classification criteria and classify firms as non-financial reporters if they refer to at least one NFR framework in their annual reports – even if they do not provide stand-alone non-financial or integrated reports (“lenient classification scheme”).

In the third step, we identify firms’ reporting practices over time and categorise them into five groups. Group A comprises firms that are never classified as non-financial reporters in any sample years; they are labelled as “non-reporters”. The number of non-reporters under the strict (lenient) classification scheme is 173 (29). Group B consists of firms classified as non-financial reporters after the entry-into-force date of the Directive but not before; these firms are labelled as “mandatory reporters”. The number of mandatory reporters under the strict (lenient) classification scheme is 51 (58). Group C consists of firms classified as non-financial reporters before the implementation of the Directive and for which a pre-reporting observation is available (which is necessary for an effective post- versus pre-analysis); these firms are labelled “voluntary reporters”. The number of voluntary reporters under the strict (lenient) classification scheme is 54 (66). Group D consists of companies always classified as non-financial reporters, making it impossible to identify their first reporting year as they have been providing non-financial reports since (and probably also before) 2012; we label these firms as “voluntary reporters, pre-reporting observation unavailable”. The number of such firms under the strict (lenient) classification scheme is 150 (249). Finally, Group E consists of all firms that cannot be categorised in any of the other groups, as they sometimes provide non-financial information and sometimes do not; we label these firms as “unidentifiable reporters”. The number of unidentifiable reporters under the strict (lenient) classification scheme is 69 (95). Figure 1 illustrates the five groups and their corresponding reporting patterns.

Figure 1. Types of reporters.

Note: The non-financial reporting patterns of firms within each group can vary. For instance, all firms for which the first non-financial report is available for a financial year that started on or after 1 January 2013 and before 1 January 2017 belong to Group C (“voluntary reporters”).

^aSee Panel A of Table 1, the STOXX Europe 600 Index encompasses 497 firms that are subject to the Directive.

^bSee Section 3.1 for the definitions of “strict classification scheme” and “lenient classification scheme”.

To test our hypotheses, we require the availability of at least one pre- and post-reporting observation to guarantee an effective pre- versus post-analysis (see Section 3.2). Consequently, we focus on firms in Groups B (mandatory reporters) and C (voluntary reporters). This approach yields a sample of 105 (124) companies under the strict (lenient) classification scheme. In the robustness tests, we also include firms belonging to Group D (“voluntary reporters, pre-reporting observation unavailable”) in our analysis.

In the last step, we obtain firm-year and firm-day level data from Refinitiv for all our dependent and independent variables. After excluding firm-year observations with missing data in Refinitiv, the final sample in our primary analyses consists of 40 voluntary reporters and 37 mandatory reporters, with a total of 603 firm-year observations. Table 1 details our sample selection procedure for the number of unique firms (Panel A) and corresponding firm-year observations (Panel B).

Table 1. Sample.

Panel A. Unique firms in sample
	Total
Number of unique firms in the STOXX Europe 600 Index	600
Less: firms headquartered a country that did not (have to) transpose the Directive	62
Less: firms to which the local laws resulting from the Directive do not apply^a	41
Number of unique firms that must comply to the new disclosure requirements	497
Less: firms not belonging to Group B (“mandatory reporters”) or Group C (“voluntary reporters”)^b	392
Less: firms with missing required financial or non-financial data in Refinitiv	28
Number of unique firms in sample	77

Panel B. Unique firm-year observations in sample
	Voluntary reporters^b	Mandatory reporters^b	Total
Unique firms in sample	40	37	77
Unique firms in sample x sample years	320	296	616
Less: missing firm-year data	11	2	13
Number of unique firm-year observations in sample	309	294	603

Panel C: Firms in sample across countries
	Voluntary reporters		Mandatory reporters		Total
Country	n	in %	n	in %	n	in %
Belgium	1	3%	4	11%	5	6%
Denmark	0	0%	1	3%	1	1%
Finland	0	0%	1	3%	1	1%
France	9	23%	9	24%	18	23%
Germany	8	20%	7	19%	15	19%
Italy	3	8%	3	8%	6	8%
Luxembourg	0	0%	2	5%	2	3%
Norway	2	5%	2	5%	4	5%
Portugal	1	3%	1	3%	2	3%
Spain	2	5%	0	0%	2	3%
Sweden	2	5%	2	5%	4	5%
United Kingdom	12	30%	5	14%	17	22%
Total	40		37		77

Panel D: Firms in sample across sectors
	Voluntary reporters		Mandatory reporters		Total
Standard Industrial Classification (SIC)	n	in %	n	in %	n	in %
10-14 Mining	2	5%	1	3%	3	4%
15-17 Construction	2	5%	2	5%	4	5%
20-39 Manufacturing	14	35%	11	30%	25	32%
40-49 Transportation, Communications, Electric, Gas and Sanitary Services	5	13%	4	11%	9	12%
50-51 Wholesale Trade	1	3%	0	0%	1	1%
52-59 Retail Trade	3	8%	2	5%	5	6%
60-67 Finance, Insurance and Real Estate	10	25%	10	27%	20	26%
70-89 Services	3	8%	7	19%	10	13%
Total	40		37		77

^a The national-level transpositions differ because the Directive allows member states to impose state-specific requirements on companies. For example, French companies with an average number of employees exceeding 500 need to comply with the transposed law, while Swedish firms with 250 employees are covered by the Directive (AE, 2018). Therefore, we have analysed whether each firm (based on the average number of employees, net turnover, and balance sheet total) falls within the mandatory scope of the local laws owing to the Directive.

^b Voluntary reporters are firms that voluntarily start non-financial reporting (NFR) before the Directive's entry-into-force date; mandatory reporters are firms that only report when mandated (see Figure 1).

Table 1 further presents the distribution of sample firms across countries (Panel C) and sectors (Panel D). Panel C shows that the number of voluntary and mandatory reporters does not vary considerably across countries, except for the UK, where we observe more than twice as many voluntary reporters as mandatory reporters in the sample.⁹ Regarding sector representation (Panel D), voluntary and mandatory reporters primarily engage in sectors “20-39 Manufacturing” and “60-67 Finance, Insurance and Real Estate”.

3.2. Research design and variables

We anticipate heterogeneity in reporting practices and the associated capital market effects between voluntary and mandatory reporters. In terms of terminology, and because we use the entry-into-force date of the Directive to separate voluntary reporters from mandatory reporters, the latter are the “treatment firms”, and the former the “control firms”. Following prior studies examining a reporting mandate, such as the mandatory adoption of the IFRS (Daske et al., 2008; Horton et al., 2013; Lang & Stice-Lawrence, 2015), we estimate a DiD model that exploits staggered non-financial reporting decisions across voluntary and mandatory reporters. Our base model for testing each of the hypotheses is as follows: (1) $\begin{array}{rl}\mathrm{Outcome}\mathrm{variabl}{e}_{\mathrm{it}}& ={\text{β}}_0+{\text{β}}_1\mathrm{POS}{T}_{\mathrm{it}}{+}_{}{\text{β}}_2\mathrm{TREA}{T}_i\mathrm{xPOS}{T}_{\mathrm{it}}\\ & +\mathrm{\Sigma }{\text{β}}_k\mathrm{Control}\mathrm{variable}{s}_{\mathrm{it}}+\mathrm{\Sigma }{\text{β}}_f\left(\mathrm{Firm},\mathrm{Year}\right)+{ϵ}_{\mathrm{it}}\end{array}$(1) Outcome variable is the consequence under study (disclosure practice or economic consequence); i denotes firm i, and t represents year t. The model enables us to draw conclusions about the incremental changes in the Outcome variable_it following the initiation of NFR for both voluntary and mandatory reporters.

For voluntary reporters, we can draw inferences by examining the coefficient β₁ on POST_it. POST_it equals one for all years t in which firm i reports non-financial information, and zero otherwise.¹⁰

For mandatory reporters, we can make inferences by examining the combined estimates of β₁ on POST_it, and β₂ on TREAT_i×POST_it. TREAT_i×POST_it is the interaction of TREAT_i (equal to one if the firm is a mandatory reporter, and zero otherwise) and POST_it. Its coefficient, β₂, represents the differential effect of NFR on the Outcome variable_it for mandatory relative to voluntary reporters, before versus after the initiation of NFR (i.e. the “difference-in-difference”). Therefore, the combined estimates of β₁, on POST_it, and β₂, on TREAT_i×POST_it, represent the incremental effect of NFR on Outcome variable_it for mandatory reporters only.

In order to ensure that the estimated coefficients do not capture the (unobservable) differences in time-invariant characteristics across mandatory and voluntary reporters, we include firm-fixed effects in Equation (1). Consequently, TREAT_i is subsumed by firm-fixed effects and not included as a separate independent variable in Equation (1). We also include year-fixed effects in Equation (1) to control for macroeconomic factors that affect all firms simultaneously.¹¹

To examine the heterogeneity in reporting practices, we compare the level of boilerplate language in the non-financial disclosures of mandatory reporters relative to voluntary reporters. As Christensen et al. (2021) suggest, we use readability scores as a proxy for boilerplate language. Given that readability scores react immediately to changes in reporting practices, in contrast to economic consequences (Barth et al., 2017), we use the incremental increase (or decrease) in readability from the last pre-reporting period (readability measure is only based on the financial report) to the first post-reporting period (readability measure is based on the financial report and the stand-alone non-financial report) as a proxy for the level of boilerplate language in non-financial disclosures. Following Caglio et al. (2020), we use the Quanteda Software¹² to compute three measures of readability: the Gunning Fog index (FOG_INDEX), Smog index (SMOG_INDEX), and Flesch-Kincaid index (FK_INDEX).¹³ In short, these indices show how difficult it is for a text in English to be understood: the higher the index, the more complex and less readable the text. As Lo et al. (2017) point out, firms may use complex language that is difficult to read to hide or obfuscate bad information, or mislead or influence readers’ understanding of reported information. The readability indices capture the extent to which firms use boilerplate language in their reporting. Consistent with Richards and van Staden (2015), we control for leverage (LEV), stock price volatility (VOLATILITY), and size (SIZE) in our readability analyses.

Since our sample includes firms from multiple countries, we also include controls for enforcement differences across these countries. Following Krueger et al. (2023) and Aresu et al. (2023), we control for differences in formal institutions by including the RULE_OF_LAW variable, and in informal institutions by adding PRESSURE, a variable derived from the Environmental Performance Index (EPI) and Human Development Index (HDI), to Equation (1). Additionally, we include both firm- and year-fixed effects.

We study the association between NFR and firm value and information asymmetry to assess the anticipated heterogeneity in economic consequences. Following Barth et al. (2017), we use Tobin's Q (TOBINS_Q) as a proxy for firm value and the bid-ask spread (BID_ASK) as a proxy for information asymmetries. Depending on the economic consequence being studied and closely following Barth et al. (2017), we include different sets of control variables, which comprise the following controls: asset growth (ASSET_GR), cash and cash equivalents (CASH), return on assets (ROA), leverage (LEV), beta (BETA), idiosyncratic risk (IRISK), size (SIZE), corporate governance score (GOV), corporate social responsibility performance score (CSR_PERF), low accounting quality (LOW_AQ), firm complexity (COMPLEX), dividend (DIVIDEND), book-to-market ratio of equity (BTM), and loss (LOSS). Consistent with our analysis of reporting practices, we additionally incorporate measures of formal (RULE_OF_LAW) and informal institutions (PRESSURE) to account for the differences in enforcement across countries. Naturally, we also include firm- and year-fixed effects.

We define all the variables in Appendix B. Table 2 provides the descriptive statistics (Panel A) and correlation matrix (Panel B) for all the variables employed in our analyses. Panel A shows that our readability scores are, on average, higher than those documented by Caglio et al. (2020). However, Caglio et al. (2020) only study financial reports, whereas we also examine non-financial reports. The higher readability scores suggest that non-financial disclosures are generally more complex.

Table 2. Summary statistics and Spearman correlations.

Panel A: Summary statistics
Variable	Mean	SD	Min	Median	Max
POST	0.48	0.50	0.00	0.00	1.00
FOG_INDEX	20.45	1.89	15.36	20.23	26.76
SMOG_INDEX	17.62	1.25	13.93	17.52	21.75
FK_INDEX	16.28	1.76	11.07	16.12	23.18
TOBINS_Q	1.83	1.32	0.76	1.42	9.70
BID_ASK	−7.16	0.70	−9.40	−7.14	−4.48
ASSET_GR	6.37	12.71	−31.35	4.77	58.75
CASH	8.03	8.47	0.08	5.51	45.63
ROA	4.80	5.99	−11.52	4.08	33.05
LEV	66.19	19.18	19.24	67.86	103.92
BETA	1.31	0.15	0.01	1.29	1.88
IRISK	0.74	0.19	0.44	0.71	1.27
SIZE	22.75	1.09	19.57	22.68	25.36
GOV	55.48	4.25	50.00	53.88	63.73
CSR_PERF	58.64	22.14	8.61	60.18	93.92
LOW_AQ	0.44	0.35	0.00	0.33	1.00
COMPLEX	5.69	2.87	1.00	6.00	10.00
DIVIDEND	0.90	0.30	0.00	1.00	1.00
BTM	0.63	0.56	0.01	0.45	3.31
LOSS	0.10	0.30	0.00	0.00	1.00
VOLATILITY	1.68	0.57	0.87	1.61	4.51
RULE_OF_LAW	78.13	6.87	63.24	79.11	89.44
PRESSURE	0.00	1.47	−6.58	0.43	2.56

Panel B: Spearman correlations
Variables	ASSET_GR	CASH	ROA	LEV	BETA	IRISK	SIZE	GOV	CSR_PERF
CASH	0.21***
ROA	0.32***	0.21***
LEV	−0.16***	−0.25***	−0.40***
BETA	−0.07*	−0.19***	−0.06	0.10**
IRISK	0.09**	0.03	−0.01	−0.09**	0.35***
SIZE	−0.21***	−0.30***	−0.11***	0.26***	0.24***	−0.24***
GOV	−0.19***	−0.15***	−0.24***	0.30***	0.14***	−0.17***	0.35***
CSR_PERF	−0.20***	−0.22***	−0.16***	0.14***	0.03	−0.14***	0.65***	0.28***
LOW_AQ	−0.17***	−0.15***	−0.21***	0.44***	0.20***	−0.10**	0.22***	0.34***	0.17***
COMPLEX	−0.04	−0.07*	−0.07*	−0.08*	0.10**	−0.05	0.16***	0.19***	0.15***
DIVIDEND	−0.13***	−0.26***	0.25***	0.00	0.23***	−0.01	0.24***	−0.01	0.10**
BTM	−0.12***	−0.24***	−0.36***	0.30***	−0.01	−0.08**	0.08**	0.34***	0.13***
LOSS	−0.12***	0.09**	−0.50***	0.05	−0.14***	0.10**	−0.12***	0.01	0.00
VOLATILITY	−0.02	0.11	−0.07	0.16*	−0.39***	−0.05	−0.16*	0.01	−0.07
RULE_OF_LAW	0.03	−0.18***	0.11***	−0.15***	−0.11***	0.03	−0.05	−0.34***	−0.16***
PRESSURE	0.03	−0.12***	0.10***	−0.08**	−0.06	0.06	−0.02	−0.30***	−0.14***

Panel B: Spearman correlations
Variables	LOW_AQ	COMPLEX	DIVIDEND	BTM	LOSS	VOLATILITY	RULE_OF_LAW
COMPLEX	−0.15***
DIVIDEND	0.16***	−0.08*
BTM	0.25***	0.30***	−0.12***
LOSS	−0.09**	0.13***	−0.41***	0.27***
VOLATILITY	−0.12	0.06	−0.20**	0.32***	0.20**
RULE_OF_LAW	−0.25***	−0.08**	−0.11***	−0.10***	0.06	−0.02
PRESSURE	−0.21***	−0.11***	−0.11***	−0.12***	0.06*	−0.04	0.65***

Note: See Appendix B for the definitions of the variables.

4. Results

First, we analyse whether heterogeneous incentives to report non-financial information (as evidenced by firms’ voluntary or mandatory NFR decisions) shape their reporting practices. We hypothesise that there is a positive association between NFR and the use of boilerplate language for mandatory reporters, whereas there is no (or a negative) association for voluntary reporters (H1).

Table 3 presents our findings for estimating Equation (1) with readability scores as the dependent variable. In line with our prediction, we find no statistical effect on using boilerplate language for voluntary reporters; the coefficient on POST is insignificant for all readability proxies. However, the coefficient on TREAT × POST is positive and significant at the 5% level for all three readability proxies. For example, providing a non-financial report as a mandatory reporter is associated with an increase in the Fog index (FOG_INDEX). As the Fog index is an inverse measure of readability, our findings suggest that mandatory reporters, in contrast to voluntary reporters, use more boilerplate language in their disclosures once they initiate NFR.

Table 3. Mandatory versus voluntary reporters, readability.

Variable	FOG_INDEX	SMOG_INDEX	FK_INDEX
POST	0.066	0.042	0.180
	(0.282)	(0.280)	(0.814)
TREAT × POST	0.777**	0.503**	0.663**
	(2.233)	(2.226)	(2.004)
LEV	0.000	0.003	0.003
	(−0.001)	(0.144)	(0.094)
VOLATILITY	0.102	0.047	0.047
	(0.261)	(0.186)	(0.126)
SIZE	0.115	0.040	−0.026
	(0.177)	(0.095)	(−0.042)
RULE_OF_LAW	0.013	−0.013	0.060
	(0.108)	(−0.168)	(0.534)
PRESSURE	−0.011	−0.006	0.014
	(−0.067)	(−0.057)	(0.095)
Fixed effects	Firm, year	Firm, year	Firm, year
No. of observations	154	154	154
R^2 adjusted	73,3%	74,6%	71,4%
R^2 within	17,2%	16,5%	19,6%

Note: See Appendix B for the definitions of the variables. t-values are in parentheses. ***, **,* denote significance at the 1%, 5%, and 10% level.

Next, we examine whether the association between NFR and firm value, as well as the association between NFR and information asymmetry, is heterogeneous between voluntary and mandatory reporters. We hypothesise that after initiating NFR, voluntary (mandatory) reporters experience increases (decreases) in firm value (H2a) and decreases (increases) in information asymmetry (H2b).

Table 4, column 1 presents our findings for estimating Equation (1) with Tobin’s Q as the dependent variable (H2a). For voluntary reporters, we do not find a significant incremental effect of NFR on firm value; the coefficient on POST is insignificant (t-value = −0.012). For mandatory reporters, however, we document a negative association between NFR and firm value. As discussed in Section 3.2, the coefficient on TREAT × POST represents the incremental effect of non-financial reporting on firm value for mandatory relative to voluntary reporters and after the initiation of NFR compared to before (i.e. the “difference-in-difference”). This incremental effect is negative (−0.258) and statistically significant at the 5% level (t-value = −2.102). Thus, mandatory reporters experience a decrease in firm value following the initiation of NFR, suggesting that, as expected, mandatory reporters face a negative cost–benefit trade-off of reporting. The coefficient estimates on our control variables are in line with prior literature. For example, larger firms with more cash holdings have a higher Tobin's Q, while firms with lower accounting quality and more asset growth have a lower Tobin's Q. In addition, a better CSR performance is – on average – negatively associated with firm value, which underlines that strong non-financial performance is costly.

Table 4. Mandatory versus voluntary reporters, firm value and bid-ask spread

Variable	TOBINS_Q	BID_ASK
POST	−0.001	−0.135**
	(−0.012)	(−2.469)
TREAT × POST	−0.258**	0.146**
	(−2.102)	(2.333)
ASSET_GR	−0.005**
	(−2.078)
CASH	0.018*
	(1.835)
ROA	−0.011
	(−1.339)
LEV	0.024***
	(5.118)
BETA	0.562*	−0.001
	(1.726)	(−0.008)
IRISK	−0.359	0.191
	(−1.159)	(1.196)
SIZE	0.480***	−0.238***
	(5.360)	(−5.036)
GOV	0.042	0.011
	(1.179)	(0.567)
CSR_PERF	−0.008**	0.001
	(−2.029)	(0.517)
LOW_AQ	−0.532***	−0.004
	(−3.529)	(−0.055)
COMPLEX	−0.017	0.000
	(−0.753)	(−0.035)
DIVIDEND	−0.115
	(−0.885)
BTM		0.064
		(0.925)
LOSS		0.056
		(0.938)
RULE_OF_LAW	0.003	−0.033***
	(0.161)	(−3.237)
PRESSURE	−0.011	0.006
	(−0.340)	(0.339)
Fixed effects	Firm, year	Firm, year
No. of observations	603	603
R^2 adjusted	83,1%	78,8%
R^2 within	15,6%	11,6%

Note. See Appendix B for the definitions of the variables. t-values are in parentheses. ***, **,* denote significance at the 1%, 5%, and 10% level.

Table 4, column 2 presents our findings for estimating Equation (1) with the bid-ask spread as the dependent variable (H2b). For voluntary reporters, we can make inferences by studying the coefficient on POST; this coefficient is negative (−0.135) and significant at the 5% level (t-value = −2.469), indicating that voluntary reporters experience reduced information asymmetries after initiating NFR. For mandatory reporters, inferences are possible by examining the combined estimates for POST and TREAT × POST (see Section 3.2). While the positive coefficient on TREAT × POST (0.146, t-value = 2.333) indicates a differential effect of NFR for mandatory compared to voluntary reporters, the net effect of NFR for mandatory reporters is also positive. Hence, our results indicate that mandatory reporters experience increased information asymmetries following the initiation of NFR.

5. Robustness tests

5.1. Sample variations

As outlined in Section 2.1, the UK and France were the only countries in our sample to impose stringent non-financial disclosure requirements before the Directive. However, even within these countries, policies that mandated NFR did not require the use of internationally recognised NFR frameworks (see Section 2.1). As a result, we are confident that our strict classification scheme (see Section 3.1; classification based on the provision of stand-alone non-financial or integrated reports prepared against internationally recognised NFR frameworks) and lenient classification scheme (see Section 3.1; classification based on the use of internationally recognised NFR frameworks) accurately identify voluntary and mandatory reporters. However, to ensure that the documented effects are not concentrated in countries at the forefront of mandatory sustainability reporting, we test the robustness of our results by excluding all firms in the UK and France separately and collectively. Panel A of Table 5 shows that our findings remain qualitatively similar when both countries are excluded.

Table 5. Robustness tests, sample variations.

Panel A: Excluding United Kingdom and French firms
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	0.104	0.208	−0.182**
	(0.354)	(1.146)	(−2.096)
TREAT × POST	0.850**	−0.561***	0.248***
	(2.264)	(−2.848)	(2.638)
Control variables	Included	Included	Included
Fixed effects	Firm, year	Firm, year	Firm, year
No. of observations	84	328	328
R^2 adjusted	85,9%	83,7%	66,9%
R^2 within	27,5%	20,6%	16,5%
Panel B: Including “voluntary reporters, pre-reporting observation unavailable”^a
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	−0.131	0.142*	−0.092*
	(−0.597)	(1.762)	(−1.650)
TREAT × POST	0.951***	−0.320***	0.132*
	(3.040)	(−3.001)	(1.773)
Control variables	Included	Included	Included
Fixed effects	Firm, year	Firm, year	Firm, year
No. of observations	295	1,692	1,692
R^2 adjusted	82,3%	86,2%	67,8%
R^2 within	20,1%	10,2%	7,8%

Note: See Appendix B for the definitions of the variables. t-values are in parentheses. ***, **,* denote significance at the 1%, 5%, and 10% level.

^a See Section 3.1 for the definition of “voluntary reporters, pre-reporting observation unavailable”.

The staggered DiD design uses voluntary reporters as control firms. As detailed in Section 3, at least one pre-reporting observation is needed to guarantee an effective pre- versus post-analysis for voluntary reporters (and mandatory reporters). As Armstrong et al. (2022) emphasise, a (staggered) DiD design exploits how potentially existing differences in the outcome variable between treatment and control groups, captured by the pre-treatment observations of the two groups, change after the treatment is administered. Otherwise, one fails to control for differences between the groups in the pre-treatment period, introducing a potential estimation bias. Due to the unavailability of pre-treatment observations for the voluntary reporters that consistently report over the entire sample period (labelled as “voluntary reporters, pre-reporting observation unavailable”, see Figure 1), identifying the effect of initiating NFR is impossible for these firms. Consequently, our primary tests compare mandatory reporters with voluntary reporters for whom at least one pre-reporting observation is available (firms belonging to Group C, see Figure 1). This implies that all firms labelled as voluntary reporters for which a pre-reporting observation is unavailable (firms belonging to Group D, see Figure 1) are excluded from our primary tests. To test the robustness of our findings to the inclusion of “voluntary reporters, pre-reporting observation unavailable” (firms belonging to Group D, see Figure 1), we rerun all the analyses using the larger group of voluntary reporters (consisting of firms belonging to Group C or Group D, see Figure 1) as the control group.¹⁴ On the downside, and as previously stated, the lack of pre-treatment observations for firms belonging to Group D means that their observations do not contribute to estimating the pre-treatment difference between treatment and control firms, which introduces a potential estimation bias. However, by including them in our analysis, the DiD effect is estimated against a larger pool of control firms. Moreover, the larger sample speaks to a broader population of firms. Panel B of Table 5 shows that all our findings remain qualitatively similar; including “voluntary reporters, pre-reporting observation unavailable” does not alter our inferences.

5.2. Lenient classification scheme

Section 3.1 describes our approach to identifying NFR practices involving a “substance over reference” criterion. To qualify as “true” non-financial reporters, firms must provide stand-alone non-financial or integrated reports prepared against internationally recognised NFR frameworks (see Section 3.1). Hence, 173 firms are classified as “non-reporters” (see Figure 1). These firms have never issued stand-alone non-financial or integrated reports, although they may have reported non-financial information in their annual reports.

The Directive allows the required non-financial information to be included in the management report (part of the annual report). In addition, no mandatory country-specific legislation in place before the Directive required the use of internationally recognised NFR frameworks (see Section 2.1). Hence, we investigate the robustness of our results by employing a less strict classification scheme based on the use of internationally recognised NFR frameworks (“lenient classification scheme”). Specifically, we identify non-financial reporting by using computer-aided text analysis to capture references to NFR frameworks that are considered appropriate for complying with the Directive by the Federation of European Accountants (FEE, 2016; La Torre et al., 2018).¹⁵ The lenient classification scheme resulted in only 29 non-reporters (down from 173; none of them provided stand-alone non-financial or integrated reports, nor did they reference any of the NFR frameworks in their annual reports), 58 mandatory reporters, and 66 voluntary reporters, as shown in Figure 1. Panel A of Table 6 presents our repeated analyses with the larger sample, showing that all findings remain consistent. These findings also hold true under the conditions outlined in Section 5.1, which involve excluding firms from the UK and France (Panel B) and including firms labelled as “voluntary reporters, pre-reporting observation unavailable” (Panel C), while using the lenient classification scheme.

Table 6. Robustness tests, lenient classification scheme.

Panel A: Lenient classification scheme^a
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	0.146	0.653	−0.095**
	(0.735)	(1.235)	(−1.978)
TREAT × POST	0.558*	−2.562***	0.145***
	(1.951)	(−4.244)	(2.640)
Control variables	Included	Included	Included
Fixed effects	Firm, year	Firm, year	Firm, year
No. of observations	186	730	730
R^2 adjusted	76,1%	59,0%	80,0%
R^2 within	13,5%	7,6%	13,3%
Panel B: Lenient classification scheme, excluding United Kingdom and French firms
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	0.193	1.115	−0.129*
	(0.788)	(1.228)	(−1.870)
TREAT × POST	0.577*	−3.719***	0.216***
	(1.794)	(−3.786)	(2.848)
Control variables	Included	Included	Included
Fixed effects	Firm, year	Firm, year	Firm, year
No. of observations	112	441	441
R^2 adjusted	86,0%	59,7%	71,6%
R^2 within	19,1%	12,4%	17,7%
Panel C: Lenient classification scheme, including “voluntary reporters, pre-reporting observation unavailable”^a
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	−0.035	0.360	−0.116**
	(−0.176)	(1.222)	(−2.371)
TREAT × POST	0.722**	−2.253***	0.136**
	(2.595)	(−5.767)	(2.109)
Control variables	Included	Included	Included
Fixed effects	Firm, year	Firm, year	Firm, year
No. of observations	358	2,074	2,074
R^2 adjusted	82,8%	66,9%	74,0%
R^2 within	14,7%	5,5%	9,1%

Note: See Appendix B for the definitions of the variables. t-values are in parentheses. ***, **,* denote significance at the 1%, 5%, and 10% level.

^a See Section 3.1 for the definitions of “lenient classification scheme” and “voluntary reporters, pre-reporting observation unavailable”.

5.3. Generalised method of moments estimation

Christensen et al. (2021) suggest that stakeholders’ reactions to firm disclosures may create a feedback loop in which firms respond to stakeholders’ responses. Therefore, unobserved heterogeneity could affect our inferences because the past values of our dependent variables (e.g. information asymmetry, that is, BID_ASK) might correlate with the informativeness of non-financial disclosures. In other words, managers who observe that shareholders’ expectations are unmet (e.g. NFR increases rather than decreases information asymmetry) may improve non-financial disclosure in the subsequent reporting cycle. To ensure the robustness of our findings to such endogeneity concerns, we evaluate the validity of our results using a generalised method of moments (GMM) estimation. The GMM approach overcomes endogeneity issues, particularly reversed causality and simultaneity, and has been utilised in similar settings (e.g. Abdullah & Tursoy, 2021; Dayanandan et al., 2016). However, it cannot control for unobserved time-invariant differences between the treatment and control groups, which may increase estimation bias and lead to a less accurate and precise estimate of the treatment effect. Therefore, we employ a staggered DiD research design as our main model and use GMM as a robustness test. The findings using the GMM approach are presented in Table 7; all our inferences are consistent with, and qualitatively similar to, our primary analyses employing the DiD design.¹⁶

Table 7. Robustness tests, generalised method of moments (GMM) estimation.

Panel A: GMM estimation – subsample of voluntary reporters^a
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	0.203	0.029	−0.034***
	(1.082)	(1.332)	(−3.047)
POSTt−1	0.096	−0.003	−0.052***
	(0.560)	(−0.125)	(−4.028)
Control variables^b	Included	Included	Included
No. of observations	309	309	309
AR(1)	0.01	0.02	0.02
AR(2)	0.36	0.37	0.11
Sargan test statistics	0.98	0.96	0.19
Panel B: GMM estimation – subsample of mandatory reporters^a
Variable	FOG_INDEX	TOBINS_Q	BID_ASK
POST	0.226**	−0.207***	0.057*
	(1.982)	(−4.008)	(1.677)
POSTt−1	0.097	−0.114**	0.101***
	(0.959)	(−2.297)	(2.806)
Control variables^b	Included	Included	Included
No. of observations	294	294	294
AR(1)	0.02	0.01	0.01
AR(2)	0.66	0.42	0.78
Sargan test statistics	0.99	0.15	0.18

Note: See Appendix B for the definitions of the variables. Z-values are in parentheses. ***, **,* denote significance at the 1%, 5%, and 10% level.

^a Voluntary reporters are firms that voluntarily start non-financial reporting (NFR) before the Directive's entry-into-force date; mandatory reporters are firms that only report when mandated (see Figure 1).

^b Consistent with prior research (Arellano & Bond, 1991; Schultz et al., 2010; Wintoki et al., 2012), we included two lags of the dependent variable in the GMM estimation to account for its persistence.

6. Conclusion

Previous studies indicate that Directive 2014/95/EU has achieved its goal of enhancing non-financial transparency, resulting in declines in firm value. Employing a staggered DiD design, we document that the transparency enhancements are concentrated in firms that voluntarily start NFR in anticipation of the Directive's entry-into-force date (“voluntary reporters”), without these firms facing net costs associated with reporting. In contrast, firms that only report when mandated (“mandatory reporters”) are more likely to provide boilerplate disclosures, consistent with facing a negative cost–benefit trade-off of reporting, leading them to experience increased information asymmetries. Our results hold when employing a GMM estimation to overcome endogeneity issues, using a lenient classification scheme instead of a strict one to identify voluntary and mandatory reporters, and excluding countries that imposed stringent non-financial disclosure requirements before the Directive.

Overall, our results suggest that the flexibility embedded in and lack of enforcement of the Directive renders it ineffective in promoting non-financial transparency among firms with weak incentives to disclose meaningful information.

Our study makes several contributions to the literature and has implications for regulators, practitioners, and stakeholders. First, we add to recent studies on the effects of sustainability reporting (e.g. Christensen et al., 2021; 2022; Cuomo et al., 2022; Fiechter et al., 2022; Jackson et al., 2020), demonstrating heterogeneity in reporting practices and the associated economic consequences between voluntary and mandatory reporters.

Second, we contribute to the literature on voluntary and mandatory financial and non-financial disclosure (e.g. Afeltra et al., 2023; Beyer et al., 2010). Prior studies link positive economic outcomes to voluntary financial and non-financial disclosures (Dhaliwal et al., 2011; Leuz & Verrecchia, 2000; Plumlee et al., 2015), and some suggest that these positive effects, which are contingent on firm incentives, may persist under mandatory reporting regimes (Daske et al., 2008; Fiechter et al., 2022). In line with Daske et al. (2013), who study a financial reporting context, a crucial insight from our study is the need for prudence in assuming that merely mandating NFR will yield positive results.

Finally, this study contributes to the literature on textual analyses in accounting, specifically regarding sustainability. Numerous studies have investigated language features of financial disclosures, such as readability (e.g. Li, 2008; Lundholm et al., 2014) and sentiment (e.g. Davis et al., 2015; Loughran & McDonald, 2011), in various contexts such as conference calls (e.g. Bushee et al., 2018; Price et al., 2012), annual reports (e.g. Guay et al., 2016; Lehavy et al., 2011), and analyst reports (e.g. De Franco et al., 2015). Building on the insights of Christensen et al. (2021), who propose that an NFR mandate will likely increase boilerplate disclosures, our research broadens the literature by scrutinising boilerplate language in non-financial reports.

Despite its contributions, this study has some limitations. First, we did not analyse firms that changed their reporting patterns during the sample period. Future research could investigate the economic consequences of inconsistent reporting behaviour. Second, although we reduced endogeneity concerns by using a staggered DiD research design and employing a GMM estimation, our study addresses associations, not causation. Specifically, companies self-select into voluntary reporting, adopting frameworks and using boilerplate language (Christensen et al., 2021; Mittelbach-Hörmanseder et al., 2021). Given the lack of exogenous shocks in the European NFR setting, it is difficult, if not impossible, to draw causal conclusions.

Notwithstanding these limitations, our study provides valuable insights into the effectiveness of the Non-Financial Reporting Directive (NFRD) in enhancing non-financial transparency. Looking ahead to the implementation of the Corporate Sustainability Reporting Directive (CSRD), it remains to be seen whether the adoption of a common set of NFR standards will level the playing field in NFR or whether – similar to the IFRS setting of 2005 – companies simply adopt a label (Daske et al., 2013) and greenwash their sustainability activities.

Acknowledgments

The authors are thankful to two anonymous referees, guest editors Diogenis Baboukardos, Silvia Gaia, Philippe Lassou, Teerooven Soobaroyen, editor Giovanna Michelon, and the 2023 European Accounting Association Annual Meeting participants, for their valuable contributions and insights.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Directive 2014/95/EU is commonly known as the Non-Financial Reporting Directive (NFRD).

2 Our overview of the mandatory country-specific legislation in place before the Directive (see Appendix A) is limited to countries where the sample companies have their headquarters.

3 Under the Swedish “Guidelines for external reporting by state-owned enterprises,” specific companies were obligated to disclose non-financial information as per the Global Reporting Initiative's G3 guidelines (see Appendix A). However, this legislation applied to a small number of state-owned enterprises, none of which are included in our sample.

4 Boilerplate language refers to text that is generic, relatively vague, and largely uninformative (Christensen et al., 2021).

5 Another option to decrease the net costs associated with non-financial reporting is not reporting. The disclosure requirements of the Directive are not strict but based on a comply or explain principle: companies explaining the reasons why they do not report would be considered non-reporters and be excluded from our analyses. However, a manual inspection of non-reporters reveals that none of our sample firms has chosen not to disclose any non-financial information but instead provide an explanation why.

6 The national-level transpositions differ because the Directive allows member states to impose state-specific requirements on companies (AE, 2018). Therefore, we have analysed whether each firm (based on the average number of employees, net turnover, and balance sheet total) falls within the scope of the local laws resulting from the Directive.

7 Integrated reporting combines non-financial and financial information into a single report (Barth et al., 2017). We classify a report as an “integrated report” if (a) the title of the report, displayed on the cover page of the report, contains the word “integrated,” and (b) the report references non-financial reporting frameworks considered appropriate for complying with the Directive by the Federation of European Accountants (see Appendix C).

8 We focus on the usage of non-financial reporting frameworks that are considered appropriate for complying with the Directive by the Federation of European Accountants, namely, those developed by the Sustainability Accounting Standards Board (SASB), International Integrated Reporting Council (IIRC), European Federation of Financial Analysts Societies (EFFAS), Global Reporting Initiative (GRI), United Nations Global Compact (UNGC), International Organization for Standardization (ISO), AccountAbility (AA), and Federation of European Accountants (FEE) (FEE, 2016; La Torre et al., 2018). All stand-alone non-financial and integrated reports in our sample refer to at least one of these frameworks.

9 Our results are robust to excluding firms in the UK; see Section 5.1.

10 Please note that POST_it does not separate the years before versus after the implementation of the Directive; instead, POST_it is a firm-level variable separating the years before versus after initiating NFR. Its coefficient, β₁, represents the incremental effect of initiating NFR on the Outcome variable_it for voluntary reporters only.

11 See Armstrong et al. (2022); a staggered DiD design – by definition – always employs firm- and time-fixed effects.

12 Quanteda is an R package for managing and analysing textual data (Benoit et al., 2018).

13 Before computing all readability measures, we manually remove – if any – (i) financial statements, (ii) notes to the financial statements, and (iii) auditor's report from the report. Using the software Quanteda, we further remove tables, graphs, titles, headings, URLs, numbers, symbols, and special characters (Caglio et al., 2020; Loughran & McDonald, 2011).

14 For “voluntary reporters, pre-reporting observation unavailable” (firms belonging to Group D, see Figure 1), the variable POST_it equals one for all years t, due to the unavailability of one or more pre-reporting observations for these firms i (where POST_it would equal zero).

15 Appendix C provides an overview of the search terms and queries utilised to detect the use of non-financial reporting frameworks that are considered appropriate for complying with the Directive by the Federation of European Accountants (FEE, 2016; La Torre et al., 2018).

16 For the robustness tests in Section 5, the untabulated Smog and Flesch-Kincaid index results are qualitatively similar to the Fog index results.

Appendices

Appendix A. Overview of the mandatory country-specific legislation in place before the Directive

Table

Country	Legislation	Year of Publication	Companies in scope	Reporting requirements	Disclosure venue	Required international non-financial reporting standards
Belgium	Article 4.1.8 of VLAREM II	1995	Companies located in the region of Flanders (i.e. not located in the regions of Wallonia and Brussels) that engage in activities that could cause disturbance to human health or the environment.	The reporting requirements vary depending on the type of activity.	Environmental report, which must be submitted to the Flemish Environmental Agency (VMM) − the company is not obligated to publicly disclose the environmental report through its disclosure venues.	None.
	The Social Balance Sheet	2003	Companies, non-profit institutions, and foundations that are required to publish annual accounts with the National Bank of Belgium. Companies, non-profit institutions, and foundations that are not required to publish annual accounts with the National Bank of Belgium with more than 20 staff members (e.g. hospitals). Foreign companies (non-profit institutions) with a branch (centre of operations) in Belgium.	Specific information about the nature and evolution of the workforce, e.g.: - Number of people employed - Personnel movements - Training.	Social balance sheet, which is part of the annual accounts that must be submitted to the National Bank of Belgium (NBB) − the company is not obligated to publicly disclose the social balance sheet through its disclosure venues.	None.
France	Grenelle Act II^a, Article 225	2010	Companies that are listed - Private companies that exceed the following 2 criteria: - average number of employees of 50 - balance sheet total or net turnover of EUR 100 million.	Disclosure of information on 42 social and environmental topics, divided into three themes: - Social (e.g. employment, labour relations, health and safety) - Environmental (e.g. pollution, waste management, energy consumption) - Commitments to sustainable development (e.g. social impacts, relations with stakeholders, human rights).	Annual report.	None.
	Grenelle Act II, Article 224	2010	Mutual funds.	Explanation on how environmental, social, and governance (ESG) objectives have been taken into account in the investment policy: which ESG criteria have been assessed and how they are embedded in the decision-making process, e.g.: - Percentage of the value of the mutual funds which take ESG criteria into account - Impact of the ESG assessment on the investment and divestment process.	Annual report.	None.
France	Grenelle Act II, Article 75	2010	Companies that exceed an average number of employees of 500.	Publication of the greenhouse gas emissions from: - Scope 1 (i.e. direct emissions) - Scope 2 (i.e. indirect emissions associated with electricity, steam, heat, or cooling for own use)	Carbon report, which must be updated at least every 3 years.	None.
	Law No 2001-397	2001	Companies that exceed an average number of employees of 50.	An analysis on the basis of relevant indicators, based in particular on quantitative elements, defined by decree and possibly supplemented by indicators that reflect the specific situation of the company, to assess, for each occupational group of the company, the situation of women and men in hiring, training, professional promotion, qualification, classification, working conditions and effective compensation. Indication of the action plan that has been followed to enforce gender equality.	Comparative situation report, which must be made available to the employees and their representatives − the company is not obligated to publicly disclose the situation report through its disclosure venues.	None.
	Corporate Governance Code of Listed Corporations	2010	Companies that are listed.	Publication of the compensation received by executive directors Information about the internal procedures to identify and monitor off-balance sheet commitments, and to evaluate the corporation's material risks.	Annual report.	None.
Germany	Bilanzrechts reformgesetz	2005	Companies that are required to publish a group management report.	Quantitative information on non-financial performance indicators that are used for internal management. Whenever the reported non-financial performance indicators are internally used under the aspect of sustainability, the connection with sustainability should be explained.	Management report (part of the annual report).	None.
Italy	Ministerial Decree	2008	Social enterprises.	Information required by the “social reporting guidelines for social enterprises (imprese sociali)”, e.g. information on employment practices, social impacts, and engagement with local communities.	Social report.	None.
	Legislative Decree no. 150/2009	2009	Public administrations.	Information on the implementation of a three-year program for transparency and integrity.	Section on the organisation's website.	None.
Norway	The Norwegian Accounting Act	1998	Companies that are legally bound to keep accounting records . Foreign companies that (i) carry out activities in Norway and (ii) are subject to Norwegian taxation.	Information on the working environment . Information on equality, e.g.: - Gender - Ethnicity - Religion. . Information about environment-related issues, and information about measures that can prevent or reduce negative environmental impact or trends.	Director's report (part of the annual report).	None.
	Act amending the the Norwegian Accounting Act	2003	Companies that qualify as large^b.	Information about what they do to integrate considerations for human rights, labour rights and social issues, the environment and anti-corruption in their business strategies, in their daily operations, and in their relations with their stakeholders. . Information about policies, principles, procedures and standards that are followed to integrate these considerations.	Annual report or sustainability report.	None.
	Public Limited Companies Act	2003	Public companies.	Information about the gender balance on the board of directors.	Corporate governance report.	None.
Sweden	Annual Accounts Act	1999	Companies that are required to publish a board of directors' report.	A brief disclosure of environmental and social information.	Board of directors' report (part of the annual report).	None.
		2005	Companies that are listed and qualify as large.	Non-financial indicators (if deemed required for understanding the company's overall position/performance), which may include information regarding employee relations and environmental issues.	Board of directors' report (part of the annual report).	None.
	Guidelines for external reporting by state-owned enterprises	2007	State-owned companies.	Non-financial information as required by the Global Reporting Initiative's G3 guidelines.	Sustainability report.	GRI^c.
	Sustainability goals for State-owned enterprises	2012	State-owned companies.	Defined sustainability goals, which should be relevant for the business activities of the company and its sustainability challenges, long-term and challenging, possible to follow up, as well as clear and easy to communicate to the public.	Sustainability report.	None.
United Kingdom	Quoted companies GHG reporting	2013	Companies listed on the London Stock Exchange main market.	Levels of greenhouse gas emissions.	Director's report (part of the annual report).	None.
	The Companies Act	2006	Companies that qualify as large or medium-sized.	Information on environmental, employee, social and community matters to the extent necessary for an understanding of the development, performance, or position of the company.	Annual report.	None.
	UK Corporate Governance Code	2012	Companies with a premium listing of equity shares in the UK.	Information on how the company applied the Code.	Annual report.	None.
			Companies that are listed.	Information on how the company applied the main principles of the Code. Confirmation that the company complied with the Code's provisions or an explanation of why it has not.	Annual report.	None.
	The Climate Change Act	2008		No (direct) reporting requirements.		None.
	The Carbon Reduction Commitment (CRC)	2010	Companies that use more than 6,000 Megawatt hours (MWh) per annum.	Emissions from all fixed-point sources (not just electricity).	Emission report which must be submitted to the UK Environmental Agency − the company is not obligated to publicly disclose the emission report through its disclosure venues.	None.

Note: This overview includes the mandatory country-specific legislation as outlined in the Carrots & Sticks 2013 report (C&S, 2013).

^a Grenelle Act II Article 225 (published in 2010) replaces two previous French sustainability reporting laws: Law No. 2001-420 (published in 2001) and The Social Review (published in 2007).

^b Companies that prepare a public report according to the Global Reporting Initiative (GRI) principles are exempt.

^c Guidelines set by the Global Reporting Initiative (GRI).

Appendix B. Variable definitions

Table

Variable	Definition	Data Source
POST	Post is a dummy variable that equals 1 for all years t in which firm i reports non-financial information, and 0 otherwise.	Firm's reports
TREAT	Treat is a dummy variable that equals 1 if firm i is a mandatory reporter, and 0 if firm i is a voluntary reporter^a.	Firm's reports
FOG_INDEX	Gunning Fog index is an inverse measure of the readability of a firm's current year's financial and non-financial disclosures, calculated as (words per sentence + percent of complex words) * 0.4.	Firm's reports
SMOG_INDEX	Smog McLaughlin readability index is an inverse measure of the readability of a firm's current year's financial and non-financial disclosures, based on Mc Laughlin (1969).	Firm's reports
FK_INDEX	Flesch-Kincaid Grade Readability index is an inverse measure of the readability of a firm's current year's financial and non-financial disclosures, calculated as (11.8*syllables per word) + (0.39 *words per sentence) − 15.59.	Firm's reports
TOBINS_Q	Tobin's Q is calculated as ((total assets – total common equity) + (common shares outstanding at financial year-end * share price at the publication date^b of the current year's report)) / total assets.	Refinitiv
BID_ASK	Bid-ask spread is the natural logarithm of the median of (daily ask price – daily bid price)/((daily ask price + daily bid price)/2) measured from the day after the publication of the current year's report to the publication date of the next year's report.	Refinitiv
ASSET_GR	Asset growth is measured as the difference between the financial year-end total assets and the lagged total assets, divided by the lagged total assets.	Refinitiv
CASH	Cash and cash equivalents, divided by average total assets.	Refinitiv
ROA	Return on assets calculated as the ratio of income before extraordinary items to total assets at the beginning of the financial year.	Refinitiv
LEV	Leverage is calculated as the ratio of total debt to total assets at the end of the financial year.	Refinitiv
BETA	Beta is the coefficient from a firm-specific regression of daily stock returns on daily market returns over the past financial year.	Refinitiv
IRISK	Idiosyncratic risk is the standard deviation of residuals from a firm-specific regression of daily stock returns on market returns over the past financial year.	Refinitiv
SIZE	Size is measured as the natural logarithm of market capitalisation at the beginning of the financial year.	Refinitiv
GOV	Corporate governance score is measured as the average of board function (AnalyticBoardFunctionsPolicyScore), board structure (AnalyticBoardStructurePolicyScore), compensation policy (AnalyticExecCompPolicyScore), and shareholder rights (AnalyticShareholderRightsPolicyScore), scores from Refinitiv.	Refinitiv
CSR_PERF	Corporate social responsibility performance score is calculated as the average of the environmental (EnvironmentPillarScore) and the social performance score (SocialPillarScore) from Refinitiv	Refinitiv
LOW_AQ	Low accounting quality is measured as the percentage over the prior four financial years of small earnings surprises. A small earnings surprise is a difference between 0 and 0.01 of net income in financial year t minus net income in financial year t − 1, scaled by total assets at the end of financial year t − 2. To measure low accounting quality, we require at least two earnings surprises.	Refinitiv
COMPLEX	Firm complexity is measured as the mean of the annual deciles of the following: (i) Earnings volatility – the natural logarithm of the standard deviation of income before extraordinary items per share from financial years t − 5 to t − 1, (ii) Number of segments – the number of product segments, or, if not available, the number of geographic segments, (iii) Return volatility – the standard deviation of daily stock returns over t − 1. If any component of firm complexity is missing, we use the average of the remaining components.	Refinitiv
DIVIDEND	Dividend is a dummy variable that is equal to 1 if the firm declares or pays a dividend in the financial year, and 0 otherwise.	Refinitiv
BTM	Book-to-market ratio of equity is calculated as book value of common shareholders' equity, divided by the number of common shares outstanding multiplied by the share price at the end of the financial year.	Refinitiv
LOSS	Loss is a dummy variable that equals 1 if income before extraordinary items is negative, and 0 otherwise.	Refinitiv
VOLATILITY	Stock price volatility is measured as the standard deviation of daily closing stock prices over the past financial year.	Refinitiv
RULE_OF_LAW	Rule of law is a country-year measure of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence.	World Bank (WGI)
PRESSURE	Pressure is a country-year measure for the degree of external social and environmental regulatory pressures, measured as the sum of the standardised Environmental Performance Index (EPI) and standardised Human Development Index (HDI)	Yale University, UNDP^c

^a Voluntary reporters are firms that voluntarily start non-financial reporting (NFR) before the Directive's entry-into-force date; mandatory reporters are firms that only report when mandated (see Figure 1).

^b The publication dates of the (non-)financial reports are not available on most companies' websites. To ensure consistency and comparability, and because some companies in the sample (of which the publication date was available) did not disclose their non-financial reports within a 3-month period after their financial year-end, we used a test-point of 6-month after the financial year-end as publication date, consistent with Landau et al. (2020) and Lang et al. (2006).

^c Yale University; Yale University, Center for Environmental Law and Policy (Wolf et al., 2022), UNDP; United Nations Development Program (UNDP, 2023).

Appendix C. Overview of used search terms and queries to detect the use of non-financial reporting frameworks

Table

Framework	Overview search terms (and queries)
SASB	[Sustainability Accounting Standards Board], [SASB]
GRI	[Global Reporting Initiative], [GRI*], [GRI-], [GRI -]
IIRC	[International Integrated Reporting Council], [IIRC], [I I R C], [], [< IR >]
EFFAS	[European Federation of Financial Analysts Societies], [EFFAS]
UNGC	[United Nations Global Compact], [UN Global Compact], [UNGC]
ISO	[ISO26000], [ISO 26000]
AA	[AA1000], [AA 1000]
FEE	[Federation of European Accountants]

Note: The search terms and queries are not case sensitive, thus (for instance) the search term [SASB] finds the word “SASB” (the same letters are capitalised) and “Sasb” (not the same letters are capitalised). SASB (Sustainability Accounting Standards Board), GRI (Global Reporting Initiative), IIRC (International Integrated Reporting Council), EFFAS (European Federation of Financial Analysts Societies), UNGC (United Nations Global Compact), ISO (International Organization for Standardization), AA (AccountAbility), FEE (Fédération des Experts-comptables Européens).

*indicates that the “whole word function” is used; thus, (for instance) the search term [GRI*] finds the word “gri”, but not the word “agriculture”. This function is turned off for the other search terms and queries; thus, (for instance) the search query [IIRC] also finds the string of words “IIRC-framework”.