Does Readability of Textual Disclosures in Modern Slavery Reports Pay Off? Evidence from a Regulatory Setting

Abstract

This study examines the association between textual disclosure readability in modern slavery reports and firm value. Using 212 Australian modern slavery statements for the financial year 2019–2020, a positive association is found between more readable textual disclosures in modern slavery reports and firm value. The study also finds that an optimistic tone in textual disclosures and better firm-level corporate governance accentuate this impact. Further analysis shows that the informative information component of textual disclosures in modern slavery reports is positively priced by investors. Our study’s findings contribute to the debate on why firms should consider improving the textual disclosure readability of modern slavery reports. The study also informs various regulators (e.g. Australian Border Force, Australian Securities and Investments Commission [ASIC], European Union [EU] and United Kingdom [UK] government agencies, etc.) and international organisations about firm-level efforts to promote high-quality reporting on the modern slavery risk.

Keywords:

• Modern slavery disclosure
• impression management
• obfuscation
• firm value
• Australia

JEL classification:

• M14
• M40
• M41
• M48

Modern slavery is among the most urgent human rights risks across corporate value chains, harming workers and potentially leading to reputational, legal and financial risks for companies. –– Global Reporting Initiative (GRI)¹

1. Introduction

This study’s objective is to examine the association between the textual disclosure readability of modern slavery reports and firm value. Modern slavery, characterised as an exploitative practice by powerful fund providers (Blitz & Simic, 2019), is recognised as a significant risk and an unsustainable business practice that impedes sustainable growth worldwide. In today’s Sustainable Development Goal (SDG)-driven global economy, firms in global supply chains need to be proactive to address these practices and associated issues of poor working conditions, child labour, forced labour, underpayment of wages and, above all, the promotion of social injustice.² Amidst these circumstances, several governments have enacted measures such as the California Transparency in Supply Chains Act 2012, the United Kingdom (UK) Modern Slavery Act 2015, the Australian Modern Slavery Act 2018 and the Canadian Modern Slavery Act 2023.³ These regulations mandate the regular reporting of efforts to combat modern slavery by relevant firms in these regions. Additionally, the European Union (EU) and the European Parliament have expressed significant concerns regarding modern slavery practices. Consequently, after lengthy consultations, on 24 May 2024, the EU finally approved the Corporate Sustainability Due Diligence Directive (CSDDD), aiming to exclude all products made with forced labour from the EU market.⁴

The current study is motivated by several factors. Although previous literature highlights various aspects of modern slavery disclosure (MSD), its readability and associated effects remain largely unexplored. Prior studies primarily focus on interview-based methodologies (e.g. Islam & Van Staden, 2022; Rogerson et al., 2020); industry-focused case studies (e.g. Boersma & Bedford, 2023; Dodd et al., 2023); and manual content analysis (Birkey et al., 2018; Mai et al., 2023; Pinnington et al., 2023; Schaper & Pollach, 2021). However, important qualities preferred by stakeholders, such as the readability, clarity and transparency of reported modern slavery information, are not widely examined. Therefore, it is important to investigate these aspects of modern slavery disclosures (MSD).

This study is further motivated by the need to empirically explore whether firms’ market value is impacted by their efforts to enhance the readability and transparency of modern slavery disclosures (MSDs). Specifically, the study examines how investors respond to the readability of MSDs, a topic with both theoretical and practical significance but underexplored in prior literature.⁵ Our study argues that the readability of MSDs is important for reducing information asymmetry between firms and market participants. Clear and readable disclosures provide insights into a firm’s efforts to combat human trafficking, forced labour and servitude, ensuring that firms’ decision-makers undertake necessary initiatives to provide transparency. Without these efforts, increased information volume does not enhance the reported information’s precision or trustworthiness, as information overload can increase processing costs, making it harder to extract relevant details (Bloomfield, 2002; Hirshleifer & Teoh, 2003; Miller, 2010).

Investors, limited by their rational and cognitive capabilities, may not fully process all available information, leading to suboptimal economic decisions (Hirshleifer & Teoh, 2003). Prior studies show that individual investors prefer firms with concise and clear disclosures (Lawrence, 2013), with less readable disclosures requiring more time spent by analysts to adequately follow a firm (Lehavy et al., 2011). Firms may deliberately make their MSDs less readable to conceal unethical practices or may report only positive information, in alignment with impression management theory (Cho et al., 2010; Merkl-Davies & Brennan, 2007; Subramanian et al., 1993). Given the strict enforcement of MSD regulations, it is imperative that corporate decision-makers ensure their firms’ MSDs provide both transparency and clarity. Enhancing readability not only maintains a firm’s social legitimacy but also reduces the risk of financial penalties from regulatory bodies (Searcy & Buslovich, 2014; Suchman, 1995). Poor MSD readability may also adversely affect a firm’s market value, with the information not incorporated into stock prices, leading to mispricing and inefficient resource allocation (Athanasakou et al., 2020; Hwang & Kim, 2017; Rjiba et al., 2021). Overall, when crucial information is not readable, investors and analysts find it challenging to accurately evaluate a firm’s risk exposure and ethical practices, thus influencing investment decisions and the distribution of resources in the market. Yet, the impact of MSD readability on firm value is not addressed in prior studies, creating a gap in the literature. The current study aims to bridge this gap by investigating whether MSD readability affects firms’ market value, within the Australian context.

We selected Australia as our research site and focused on the MSDs of Australian firms for several reasons. Firstly, Australia is one of the most recent countries to introduce and implement mandatory reporting regulations on modern slavery, a significant development as recent regulations may have addressed previous regulatory shortcomings. Secondly, Australian firms are obligated to provide stand-alone modern slavery statements, allowing our study to conduct readability tests focused solely on modern slavery information. Thirdly, the Australian regulations are not confined to a specific state or industry, enabling increased understanding of cross-industry differences. Finally, while only a few modern slavery-related studies are conducted in the Australian context (e.g. Christ et al., 2019; Christ & Burritt, 2023; Rao et al., 2022), the readability of MSDs and its effect on a firm’s market value remain unexplored. Given these factors, working with Australian firms in our sample provides benefits unavailable in samples from other countries with mandatory regulations.

Using a sample of 212 Australian firms for the 2019–2020 financial year, we examine the association between the readability of MSDs and firm value. We employ three proxies to measure the readability of MSDs: the Bog Index, the Gunning Fog Index and the Flesch – Kincaid Grade Level Index. Our findings indicate that more readable MSDs are positively associated with firm value, suggesting that investors positively price higher readability of these disclosures. Our results are robust using several robustness tests, including entropy balancing and propensity score matching (PSM) analyses to address endogeneity arising from observable selection bias and alternative specifications of firm value. We also examine the moderating roles of optimistic tone and better corporate governance performance in the association between more readable MSDs and firm value, with both these factors found to enhance the positive impact of more readable MSDs on firm value. We also investigate whether managers use corporate narratives to provide valuable information or to strategically distort perceptions. We find that informative disclosures in modern slavery reports are positively priced by investors. Additionally, we examine investors’ reactions to the publication of MSDs using an event study, focusing on the issuance date of modern slavery reports, and find that investors react positively to more readable disclosures in published MSDs.

This study contributes to the literature in several ways. Firstly, to the best of our knowledge, the study is the first to assess the readability of MSDs and to empirically examine its impact on firm value. Unlike prior studies that use content analysis, we focus on readability (syntactical characteristics) of MSDs, a method common in accounting and finance literature, to evaluate disclosure practices. This approach complements and extends other content analysis-based research (Christ et al., 2019; Christ & Burritt, 2023) by demonstrating the impact of MSD readability on firm value without restricting disclosure choices.⁶

Secondly, our study shifts the focus from historical, philosophical and social science perspectives on modern slavery (Cousins et al., 2020; Crane, 2013; New, 2015) to its implications in the accounting and finance fields. While previous research highlights the economic and capital market effects of corporate social responsibility (CSR) disclosures, our study examines MSDs which are mandatory with significant implications for capital markets. Modern slavery research has distinctive elements due to specific regulations and efforts by non-governmental organisations (NGOs) and human rights activists. Investors may view MSDs differently from CSR disclosures, as firms involved in modern slavery practices face severe repercussions under laws like Australia’s Criminal Code Act 1995 (Australian Government, 1995), thus increasing reputational risk and potentially impacting investments. Thirdly, our study’s findings are timely, given that firms addressing the modern slavery risk and committed to producing more readable MSDs can see tangible benefits, such as improved firm value. Moreover, our study’s findings can inform regulators, providing them with the means to encourage firms to exceed compliance obligations, thereby adding positive value for capital providers.

Finally, our findings likely to support regulators in their development of modern slavery regulations. For example, the EU approved a directive (i.e. CSDDD) on 24 May 2024 to ban all products made by forced labour from entering the EU market, with this directive to soon become law. In this context, our findings highlight the importance of better communication of information and improved readability of MSDs. This has added significance as prior research indicates evidence of tick-box practices (e.g. Mai et al., 2023; Monciardini et al., 2021); a lack of quality in disclosures (e.g. Birkey et al., 2018; Rogerson et al., 2020); and limited compliance with mandatory disclosure requirements (e.g. Monciardini et al., 2021; Schaper & Pollach, 2021). Notably, the United States (US) Securities and Exchange Commission (SEC) introduced a plain language requirement in 1998 to enhance disclosure readability, with a similar requirement likely to be of benefit for the EU and other regions implementing modern slavery regulations. This would reduce the potential for obfuscation of unethical practices through complex language. Our findings also inform broader debates about the importance of producing readable MSDs and managing modern slavery risks, providing insights for country-level regulators (e.g. Australian Border Force, European Union [EU], UK government agencies, etc.) and international organisations working to eliminate modern slavery (e.g. the International Labour Organization [ILO], Walk Free, United Nations [UN], Anti-Slavery International, European Center for Constitutional and Human Rights [ECCHR] and the Greens/European Free Alliance [EFA] Group, etc.).

The remainder of this paper is structured as follows. Sections 2 and 3 present the literature review and hypothesis development. Section 4 explains the methodology of the study, while Section 5 discusses the empirical results. Section 6 explains the additional analysis and robustness checks. Section 7 discusses the findings and concludes the paper.

2. Institutional Background and Literature Review

2.1. Institutional Background

Australia’s Modern Slavery Act 2018 marked a pivotal move in confronting the challenges of modern slavery. The Act mandated that entities operating within Australia with an annual consolidated revenue exceeding A$100 million were to annually disclose modern slavery risks in their operations and supply chains, as well as disclosing measures undertaken to mitigate these risks (Australian Government, 2018). Specifically, Section 16 of the Act outlines detailed reporting requirements, including identification of the reporting entity; a description of its structure, operations and supply chains; the risks of modern slavery within its operations and supply chains and those it controls; actions for risk assessment and mitigation; effectiveness evaluation of these actions; and a consultation process with controlled entities and others involved in the reporting (Australian Government, 2018).

Since its enactment on 1 January 2019, with a reporting deadline set six months post the reporting period’s conclusion, the Act has heightened modern slavery risk awareness among Australian businesses, leading to a significant influx of modern slavery statements. However, the effectiveness of these measures in truly mitigating modern slavery risks has been debated. A recent review of the Act’s initial three-year implementation yielded 30 proposals aimed at reinforcing its impact, including introducing non-compliance penalties; lowering the reporting revenue threshold to A$50 million; requiring reports on specific modern slavery incidents or risks; and compelling entities to establish due diligence frameworks (Australian Government, 2023). Despite these recommendations, the issue of MSD readability, which is central to the current research, remains unaddressed. Thus, the recent implementation of mandatory modern slavery reporting, coupled with the need for stand-alone statements across various industries, positions Australia as an ideal and unexplored context for our capital market research on MSD readability, distinguishing it from other countries with similar regulations.

2.2. Literature on Modern Slavery Disclosure (MSD)

As a global problem, modern slavery is receiving much attention from various stakeholders around the world. Researchers have previously examined many aspects of modern slavery from historical, philosophical and social science perspectives (Cousins et al., 2020; Crane, 2013; New, 2015). Australia, the UK and the US have introduced regulations on modern slavery risk reporting to combat these practices, their regulations garnering much attention from researchers and contributing to a growing number of studies on the advantages and disadvantages of regulating modern slavery risk reporting.

The California Transparency in Supply Chains Act 2012 (CTSCA) has been the subject of content analysis and empirical research. To illustrate, Ma et al. (2016) explore the status of the CTSCA’s implementation by examining website disclosures of 204 US-based retail and manufacturing firms that make and sell apparel, respectively, finding that more than half the firms in their sample failed to comply with the CTSCA by not posting the necessary disclosures on their websites. Similarly, Birkey et al. (2018) examine firms’ compliance with this legislation’s disclosure requirements. They document a high level of compliance, although finding firm-specific disclosures to be rather more symbolic than substantive and call for additional rules and guidance to improve disclosure quality. In addition, the authors empirically show that retail firms experienced significant negative market reaction due to the enactment of the CTSCA legislation.

The UK’s Modern Slavery Act 2015 has been examined by researchers from various perspectives. To illustrate, in their study based on interviews, Islam and Van Staden (2022) document that anti-slavery activists do not believe the minimum disclosures required are achieving the necessary transparency to address the level of public interest and to attain normativity. They also suggest that transparency and normativity could be better achieved through ongoing campaigns by anti-slavery groups which would improve the surveillance of corporate compliance with the Act. Similarly, Rogerson et al. (2020) find that universities’ disclosures on modern slavery lacked quality and were insufficiently detailed, having limited variations, thus leading to limited actions to address modern slavery risk factors.

Research on MSDs under the UK’s Modern Slavery Act reveals multifaceted insights. Flynn (2020) emphasises that compliance levels in UK firms’ statements relate significantly to factors such as firm size, prior social responsibility commitment, network involvement, industry and headquarters location. Schaper and Pollach (2021) note limited actual modern slavery case disclosure among UK firms, urging enhanced reporting on their corrective actions. Furthermore, Monciardini et al. (2021) caution that a mere focus on legal compliance could result in simply symbolic structures, thus being ineffective in addressing modern slavery risk factors. Similarly, Mai et al. (2023) and Moussa et al. (2023) both find that, while increased disclosures has occurred among Financial Times Stock Exchange (FTSE) 100 firms, these disclosures are often superficial and symbolic, lacking depth in addressing modern slavery practices.

In contrast, Flynn and Walker (2021) highlight institutional stakeholder pressure as the driver of organisational changes to combat modern slavery risks, particularly noting greater commitment among FTSE 100 Index firms than FTSE 250 Index firms. Cousins et al. (2020) explore the impact of modern slavery legislation on shareholder wealth, revealing that firms in higher labour-risk industries experience a more positive stock market reaction, suggesting potential competitive advantages for firms with a track record of addressing modern slavery risks.

Australian firms have been at the forefront of research on modern slavery risk reporting in recent years. Before the Modern Slavery Act 2018, some offered voluntary disclosures. Rao et al. (2022) document that disclosure quality is driven by firm size, whether the firm issues a stand-alone modern slavery statement and whether it is audited by a Big 4 firm, with factors such as profitability, listing status and the degree of internationalisation not related to voluntary MSD quality. Furthermore, Christ et al. (2019) note low levels of disclosure (one-third of firms) under the Australian Act, with these mainly focused on bribery, corruption and human rights. Similarly, Christ and Burritt (2023) observe diverse reporting levels. They notice that some firms exceed the compliance requirements; some report enough to achieve basic compliance, while others do not report at all. These authors then use the insights gained to develop a framework to improve current practice.

Previous studies appear to have mostly focused on content analysis, critical analysis and interviews to understand the state of MSDs following enactments of the CTSCA, the UK’s Modern Slavery Act and Australia’s Modern Slavery Act. However, a surprising lack of empirical studies is found in US, UK and Australian settings on gaining an understanding of the economic impacts of modern slavery disclosures (MSDs). To date, no studies really examine the syntactical (i.e. readability) features of MSD reports or their capital market impact. The current study attempts to fill this gap in the literature. Readability of MSDs is particularly important as businesses could use opportunistic language and provide less readable MSDs to mask their unethical practices, with this affecting stakeholders’ decision making.

2.3. Literature on Readability of Disclosure

In the literature, researchers define the concept of ‘readability’ in various ways. To illustrate, readability is defined as ‘the ease of reading’ (Smeuninx et al., 2020), while Schroeder and Gibson (1990) define readability as ‘whether a text can be read quickly and easily’. DuBay (2004) illustrates the concept as ‘the text-internal characteristic of what makes some texts easier to read than others’. Other studies focus on ‘understandability’ and ‘comprehensibility’ as two key concepts of readability (Courtis, 1998; Klare, 1963; Mc Laughlin, 1969). Some researchers argue that the ‘readability’ concept is separate from the ‘understandability’ concept (Smith & Taffler, 1992). Specifically, Smith and Taffler (1992) argue that readability is solely related to text-internal characteristics, which determine text difficulty, with understandability related to the interaction between the text and its reader and conditional on the reader’s prior knowledge. Nevertheless, readability of published information is a key concept examined by several researchers with respect to both previous financial reporting and non-financial reporting studies (Loughran & McDonald, 2014; Nazari et al., 2017; Smeuninx et al., 2020). In the current study, we define readability in light of the Bog Index, the Gunning Fog Index and the Flesch – Kincaid Grade Level Index, with these indices commonly used in the prior literature (Bonsall et al., 2017; Li, 2008).

In line with previous studies, readability in our study is regarded as a pertinent feature of any type of non-financial disclosure, including disclosure of modern slavery information. Firms that are required to publish MSDs need to provide descriptions of how they are addressing modern slavery risk factors: the description of their actions often demonstrates their level of commitment to mitigating these risk factors. Firms lacking genuine commitment to mitigating the risk of modern slavery may use various tactics to mask their current practices, such as carefully choosing their words, selecting ambiguous words or using complex sentences.

Although the readability of MSDs is yet to be empirically analysed, the literature is rich in studies seeking to understand the readability of CSR disclosures and its effects. For example, Nazari et al. (2017) document that more readable CSR reports issued by US companies listed on Standard & Poor’s (S&P) 500 Index contribute to better CSR performance by these companies. Similarly, Bai et al. (2019) examine US firms and show that increases in annual report readability play a role in reducing a firm’s future stock return synchronicity. This relationship is more pronounced for firms with low analyst coverage, those with low institutional ownership or those with high information asymmetry.

More recently, Du and Yu (2021) document that a more optimistic tone and more readable text in firms’ CSR reports contribute to better CSR performance, with the market reacting strongly to report readability and tone in released CSR reports. Similarly, Clarkson et al. (2020) find that the linguistic features of CSR reports can accurately predict firms’ CSR performance, with these features incrementally value relevant to investors. While these studies generate important insights for developing our study’s predictions, what is not known is whether more readable MSDs resulting from a mandatory regulation are associated with firm value. It is important to understand whether investors care about MSD readability. A priori, investors might treat this information differently from CSR disclosures because firms involved in modern slavery practices could face serious consequences in Australia.

3. Theoretical Underpinning and Hypothesis Development

As mentioned previously, involvement in modern slavery activities is considered an offence under Australia’s Criminal Code, specifically in Divisions 270 and 271. Consequently, a firm’s involvement in these practices not only incurs legal penalties but also poses reputational risk as investors and other market participants can be adversely affected by the firm’s reputational damage. Efforts to maintain transparency and honesty in reporting can enhance a firm’s reputation and, in turn, improve its market value. Conversely, a lack of readability in MSDs is likely to diminish a firm’s market value, as it may lead to mistrust among capital market participants. This perspective aligns with impression management theory and the substantive legitimisation aspect of legitimacy theory.

According to impression management theory, management may hide and obfuscate important information from various stakeholders (Brennan et al., 2009). Firms may also disclose information for the sake of compliance without addressing the risk factors. This behaviour, commonly known as the management obfuscation hypothesis, posits that management aims to advance its self-interest through complex, difficult-to-understand and lengthy disclosures (Bloomfield, 2008; Li, 2008). Consequently, these firms are likely to be perceived by market participants as engaging in ‘impression management’, with their symbolic, superficial or ceremonial efforts exacerbating information asymmetry in the market. With a lack of readability indicating the reported information’s lack of transparency and clarity, less readable MSDs give rise to dissatisfaction and distrust among investors and other market players (Gosselin et al., 2021).

The enhancement of the readability of MSDs by firms can also be explained through the lens of legitimacy theory, especially focusing on the core idea of pragmatic and moral legitimacy. Pragmatic legitimacy emphasises that the actions of firm decision-makers are often based on cost – benefit analyses and are primarily driven by self-interest (Khan et al., 2020; Suchman, 1995). From this viewpoint, certain business activities (e.g. readability of MSDs) are influenced by decision-makers’ cost – benefit assessments to develop a strong business case for their actions (Khan et al., 2020; Searcy & Buslovich, 2014; Suchman, 1995). Given that MSD laws are enforced through strict regulations, it is in the interest of decision-makers to report transparent and comprehensible modern slavery information. By undertaking comprehensive efforts to enhance the readability of MSDs, firms are likely to preserve their social legitimacy and avoid potential financial penalties from regulatory bodies (Khan et al., 2020). Suchman (1995) argues that under pragmatic legitimacy, decision-makers seek support from stakeholders to sustain their ongoing ‘social licence’ to operate within society. This process does not necessarily aim to satisfy all stakeholders but rather to appease and manage powerful stakeholders (e.g. regulators, the media) to ensure the firm’s survival. Applying the concept of pragmatic legitimacy, we argue that initiatives to improve the readability of MSDs are likely to lead decision-makers to enhance their perceived performance in managing modern slavery. This is not only by meeting the minimum requirements of MSD guidelines but also by striving to improve transparency and clarity in their disclosures to satisfy the expectations of influential stakeholders. This, in turn, can lead to social acceptance and support from powerful actors, positively influencing the market reputation of firms and resulting in a positive effect on their market value.

The concept of moral legitimacy offers another perspective, grounded in the belief that firms will adopt organisational practices that are morally principled, treating all stakeholders fairly regardless of their power (Deephouse et al., 2017; Suchman, 1995). It is argued that moral legitimacy, once established, reinforces the social acceptance of industry practices and guides the broader moral evaluation of decision-makers’ actions (Ashforth & Gibbs, 1990; Julian et al., 2008). We argue that firms’ substantive efforts to improve the readability of MSDs also originate from moral values and principles which help them to attract external stakeholders and maintain their legitimacy to operate. Similarly, firms with strong moral legitimacy grounds are less likely to engage in ‘window-dressing’ practices in their MSDs and are more likely to make substantive efforts to enhance their MSD readability. Over time, the readability of MSDs is expected to increase and become a standard practice within industry, reflecting the cogitative aspect of legitimacy within legitimacy theory (Suchman, 1995).

Moreover, a firm’s substantive efforts to enhance the readability of its MSDs act as a signal to the market of the firm’s actual commitment to combating slavery and ensuring transparency in its disclosures. Such firms are likely to undertake various authentic measures to mitigate slavery practices, not only within their own operations but also across their supply chains. Previous studies on readability of CSR reports (Clarkson et al., 2020; Du & Yu, 2021) confirm that firms committed to improving the readability of their CSR information enable investors to better understand the firm’s earnestness in addressing these issues, subsequently enhancing the firm’s value. Therefore, it is argued here that more readable MSDs are likely to have a positive impact on firm value.

In the Australian context, previous studies address significant issues related to modern slavery, including the extent of compliance with modern slavery regulations by Australian firms and the effectiveness of MSDs by these firms (Christ et al., 2019; Christ & Burritt, 2023). However, a gap is apparent in the literature regarding investigation of the readability of MSDs and its potential impact on firm value within the Australian context. Based on the previous discussion, it is postulated that firms presenting more readable MSDs could reduce information asymmetry among market participants, leading to a more favourable evaluation of the firm’s efforts by market players. Overall, we argue that capital market players, particularly investors and other influential stakeholders, would welcome improvements in MSD readability. Enhanced clarity and readability of MSDs by Australian firms are likely to portray a positive image of firms in the eyes of investors and other fund providers, reflecting firms’ sincere commitment and efforts in addressing modern slavery issues. Consequently, such proactive measures by a firm are expected to be endorsed by capital market participants, thereby positively influencing the firm’s value. Thus, we formulate the following hypothesis:

H1: More readable disclosure of modern slavery information is positively associated with firm value.

4. Research Design

4.1. Sample and Data

Our study’s sample includes all Australian publicly-listed firms available in the Australian Securities Exchange (ASX) 500 Index for the 2019–2020 financial year. We selected the 2019–2020 financial year as the period of study due to the first batch of modern slavery statements being released when the Australian modern slavery legislation was enacted in 2018.⁷ We manually searched for each firm’s modern slavery statement through the Modern Slavery Register maintained by the Australian Border Force and followed this up with a Google search if a statement could not be found in the register. This process yielded a total of 296 stand-alone modern slavery statements. We excluded 53 statements that did not match with other databases. We obtained financial data from the Worldscope database and stock market data from the DataStream database, while other non-financial data were sourced from LSEG ESG (previously, Refinitiv ESG) database. After removing incomplete observations from the above databases, we obtained a final sample of 212 observations for the financial year 2019–2020. Table 1, Panel A describes the sample selection procedure for our analysis.

Table 1. Sampling distribution.

Panel A: Sample selection
Modern slavery statement available from Australian Border Force website	296
Less: Firms not matched with other databases (Worldscope, Refinitiv ESG and DataStream)	(53)
Less: Firms dropped due to insufficient control variables	(31)
Final test sample	212
Panel B: Industry-wise distribution of firms in the sample
Name of Industry				Number of Firms	% of Sample
Mining and Construction				35	16.51
Food				2	0.94
Textiles, Printing and Publishing				1	0.47
Chemicals				4	1.89
Pharmaceuticals				2	0.94
Extractive				4	1.89
Manufacturing				16	7.55
Computers				17	8.02
Transportation				15	7.08
Utilities				5	2.36
Retail				27	12.74
Financial Institutions				31	14.62
Insurance and Real Estate				22	10.38
Services				27	12.74
Others				4	1.89
Total sample				212	100

Table 1, Panel B shows the industry-wise distribution of firms in our study’s sample. Our sample is dominated by firms operating in the mining and construction industry (16.51%), followed by financial institutions (14.62%), while the textiles, printing and publishing industry (0.47%) had the lowest number of observations.

4.2. Measurement of Modern Slavery Disclosure (MSD) Readability

We employed three proxies for measuring MSD readability: the Bog Index (BOG), the Gunning Fog Index (FOG) and the Flesch – Kincaid Grade Level Index (FLESCH). The Bog Index is a proprietary measure of readability developed by Editor Software’s StyleWriter plain English tool (Bonsall et al., 2017). This index takes into account multiple factors like sentence length, use of passive voice, weak verbs, overused words, complex words and jargon (Bonsall et al., 2017). The Bog Index is calculated as follows: (1) $\mathrm{Bog}\mathrm{Index}(\mathrm{BOG})=(\mathrm{Sentence}\mathrm{Bog}+\mathrm{Word}\mathrm{Bog}\text{–}\mathrm{Pep})$(1) The Fog Index is a function of two variables, that is, average words per sentence and complexity of the words (i.e. words with three or more syllables) which create a measure of readability of the text. The index proposes that, assuming everything else is equal, more syllables per word or more words per sentence make a document harder to read (Li, 2008). The Fog Index indicates the number of years of formal education that a reader of average intelligence would need to read the text once and understand that piece of writing with its word – sentence workload (Li, 2008). The Fog Index is calculated as follows: (2) $\begin{array}{r}\mathrm{Fog}\mathrm{Index}(\mathrm{FOG})=(\mathrm{Average}\mathrm{number}\mathrm{of}\mathrm{words}\mathrm{per}\mathrm{sentence}+\mathrm{percentage}\mathrm{of}\mathrm{complex}\mathrm{words})\times 0.4\end{array}$(2) where complex words are defined as words with three or more syllables.

The Flesch – Kincaid Grade Level Index is also a well-known measure of readability frequently used in prior accounting and finance studies as well as by practitioners (De Franco et al., 2015; Guay et al., 2016). The Flesch – Kincaid Grade Level Index is measured in the following way: (3) $\begin{array}{rl}\mathrm{Flesch}\text{–}\mathrm{Kincaid}\mathrm{Grade}\mathrm{Level}(\mathrm{FLESCH})& =(11.8\times \mathrm{syllables}\mathrm{per}\mathrm{word})+\\ & (0.39\times \mathrm{words}\mathrm{per}\mathrm{sentence})\text{–}15.59\end{array}$(3) Higher BOG, FOG and FLESCH values indicate poor readability. For a more intuitive interpretation of readability compared to unreadability, we multiply each of the BOG, FOG and FLESCH values by – 1 to create three new variables (RBOG, RFOG and RFLESCH) to ensure that the measure is increasing in readability. Consequently, higher scores in RBOG, RFOG and RFLESCH indicate a higher level of readability. Following De Franco et al. (2015), we construct an aggregate readability index (MR_READ) based on the average of the percentile ranks for each component (i.e. BOG, FOG and FLESCH) and then take the average across the three groups.

4.3. Empirical Models and Variable Definitions

We adopt a lead – lag approach in all our regression models to address potential endogeneity issues arising from reverse causality related to readability and firm value. Specifically, we estimate the following model to test Hypothesis 1 (H1): (4) $\begin{array}{rl}\mathrm{TOBIN}{Q}_{i,t+1}& ={\beta }_0+{\beta }_1\mathrm{READABILIT}{Y}_{i,t}+{\beta }_2\mathrm{SIZ}{E}_{i,t}+{\beta }_3\mathrm{RO}{A}_{i,t}+{\beta }_4\mathrm{LE}{V}_{i,t}+{\beta }_5\mathrm{RETVO}{L}_{i,t}\\ & +{\beta }_6\mathrm{LIQUI}{D}_{i,t}+{\beta }_7\mathrm{CAPE}{X}_{i,t}+{\beta }_8\mathrm{DI}{V}_{i,t}+{\beta }_9\mathrm{RDIN}{T}_{i,t}+{\beta }_{10}\mathrm{INTAN}{G}_{i,t}\\ & +{\beta }_{11}\mathrm{RGROWT}{H}_{i,t}+{\beta }_{12}\mathrm{INSTOW}{N}_{i,t}+{\beta }_{13}\mathrm{LENGT}{H}_{i,t}+{\beta }_{14}\mathrm{BSE}{G}_{i,t}\\ & +{\beta }_{15}\mathrm{SOC}\mathrm{\_PER}{F}_{i,t}+\sum \mathrm{INDUSTR}{Y}_{i,t}+{\epsilon }_{i,t}\end{array}$(4) where TOBINQ_t + 1 is one-year-ahead Tobin’s Q, measured as the sum of the book value of total assets plus the market value of equity minus the book value of equity divided by total assets. We use TOBINQ to measure firm value following prior studies (e.g. Cahan et al., 2016; Roll et al., 2009), as it signals, firstly, a firm’s future growth prospects and, secondly, the viability of obtaining more profits or expected performance (Luo & Bhattacharya, 2006). Moreover, as the share price is a key input in the computation of Tobin’s Q, changes in share prices potentially partly reflect investors’ reactions to the readability of modern slavery statements. READABILITY is the readability score of MSDs as discussed in Section 4.2. Appendix A provides the definitions of all variables used in this study.

Following prior studies (Bose et al., 2017; Roll et al., 2009), we control for several variables in Equation (4). We include firm size (SIZE) in Equation (4) as larger firms are more likely to enjoy economies of scale which may result in improved firm performance (Bose et al., 2017; Roll et al., 2009). We control for profitability (ROA) as more profitable firms may have more favourable investment opportunities which may contribute to higher firm value (Roll et al., 2009). Leverage may influence firm value through debtholders’ monitoring activities (Roll et al., 2009); hence, we control for leverage (LEV). Furthermore, we control for market risk (RETVOL) as firms with a higher level of market risk exert a greater impact on firm value (Bose et al., 2021). In addition, a firm’s higher share trading volume indicates greater demand for a share which may positively influence firm value (Roll et al., 2009); thus, we control for a firm’s share trading volume (LIQUID). We also control for capital intensity (CAPEX) as firms that invest more in capital expenditure are more likely to have higher investment opportunities which affect firm value (Roll et al., 2009). Firms that pay dividends may have greater free cash flow which could be used to overinvest in risky projects (Roll et al., 2009); consequently, we include dividend (DIV) in Equation (4). Investments in research and development (R&D) and intangible assets are critical components of developing intangible knowledge assets or innovations that allow businesses to operate exceptionally well (Chang & Jo, 2019). Therefore, we control for research and development (RDINT) and for intangible assets (INTANG). We control for sales revenue growth (RGROWTH) to capture its influence on firm value (Bose et al., 2017). We include institutional investors’ ownership (INSTOWN) in Equation (4) to capture institutional investors’ monitoring effects on firm value (Ferreira & Matos, 2008). We also control for the length (LENGTH) of the MSD using the natural logarithm of its total number of words. In addition, we include the number of business segments (BSEG) in Equation (4) as firms with multiple business segments may have difficulty in implementing MSDs which may affect their firm value (Li, 2008). We also include a firm’s industry-adjusted social performance (SOC_PERF) to control for its effects on firm value. Finally, we control for industry fixed effects to capture the cross-industry variation of firm value.

5. Empirical Results

5.1. Descriptive Statistics

Table 2 reports the descriptive statistics of the variables in Equation (4). The mean (median) value of Tobin’s Q (TOBINQ) is 2.148 (1.446). The average values of READABILITY measured by RBOG, RFOG and RFLESCH are −80.476, −16.505 and −15.519, respectively, suggesting that the average readability of MSDs is very poor.⁸ Furthermore, the average value of the aggregate readability index (MR_READ), based on the average of the percentile ranks for each component (i.e. BOG, FOG and FLESCH), is 0.506. The average firm size (SIZE), measured by the natural logarithm of market capitalisation, is 7.327, implying an average total market capitalisation of US$4,377.22 million (unreported) which indicates that firms in our sample are relatively large. The mean values of profitability (ROA), leverage (LEV) and intangible assets (INTANG) are about 5.70%, 22.20% and 23%, respectively, of total assets. Furthermore, the mean values of revenue growth (RGROWTH), capital intensity (CAPEX) and R&D intensity (RDINT) are 12.20%, 23.40% and 3.50%, respectively, of total sales revenue. About 89.20% of firms in our sample pay dividends (DIV). The average value of stock return volatility (RETVOL) is 2.078, while the average value of liquidity (LIQUID) is 0.817. The average value of industry adjusted social performance (SOC_PERF) is 0.001.

Table 2. Descriptive statistics.

	Observations	Mean	Std. Dev.	Median	1st Quartile	3rd Quartile
TOBINQ	212	2.148	2.362	1.446	1.035	2.117
MR_READ	212	0.506	0.206	0.511	0.370	0.643
RBOG	212	−80.476	12.351	−78.000	−88.000	−71.000
RFOG	212	−16.505	1.429	−16.600	−15.600	−17.300
RFLESCH	212	−15.519	1.359	−15.450	−14.650	−16.200
SIZE	212	7.327	1.363	7.325	6.255	8.386
ROA	212	0.057	0.098	0.049	0.021	0.092
LEV	212	0.222	0.178	0.211	0.085	0.313
RETVOL	212	2.078	0.866	2.024	1.413	2.548
LIQUID	212	0.817	0.480	0.743	0.498	1.060
CAPEX	212	0.234	0.736	0.046	0.013	0.136
DIV	212	0.892	0.312	1.000	1.000	1.000
RDINT	212	0.035	0.418	0.000	0.000	0.000
INTANG	212	0.230	0.249	0.134	0.009	0.410
RGROWTH	212	0.122	1.108	−0.007	−0.081	0.098
INSTOWN	212	0.137	0.105	0.142	0.047	0.198
LENGTH	212	7.742	0.738	7.803	7.272	8.215
BSEG	212	0.818	0.736	0.693	0.000	1.609
SOC_PERF	212	0.001	0.148	0.001	−0.065	0.071

Notes: All variables are defined in Appendix A. Std. Dev. = Standard deviation.

5.2. Correlation Analysis

Table 3 reports the Pearson’s bivariate correlation matrix of the variables in Equation (4). TOBINQ is positively correlated to MR_READ (r = 0.080); RFOG (r = 0.151); LEV (r = 0.350); RGROWTH (r = 0.197); DIV (r = 0.028); INTANG (r = 0.129); and LIQUID (r = 0.037), while TOBINQ is negatively correlated to SIZE (r = −0.141); ROA (r = −0.410); RDINT (r = −0.161); RETVOL (r = −0.065); INSTOWN (r = −0.089); LENGTH (r = −0.131); and BSEG (r = −0.112). A higher correlation is found between MR_READ, RBOG, RFOG and RFLESCH; however, we use these variables as separate proxies for readability. Furthermore, all these coefficients are significant at least at the 10% level. No coefficient exceeds a value of 0.80, with Gujarati and Porter (2009) suggesting that bivariate correlations with a value less than 0.80 do not create any multicollinearity problems. We also employ variance inflation factor (VIF) values to assess the potential for multicollinearity. For reasons of brevity, we do not report the values of VIF here; however, the unreported mean VIF value of variables used in Equation (4) is 3.55, ranging from 1.19–6.30. A VIF value higher than 10 is viewed as leading to potential multicollinearity concerns (Gujarati & Porter, 2009). Thus, our results are unlikely to suffer from multicollinearity problems.

Table 3. Correlation matrix.

		[1]	[2]	[3]	[4]	[5]	[6]	[7]	[8]	[9]	[10]
TOBINQ	[1]	1.000
MR_READ	[2]	0.080*	1.000
RBOG	[3]	0.044	0.526***	1.000
RFOG	[4]	0.151**	0.761***	0.087	1.000
RFLESCH	[5]	0.096	0.796***	0.160**	0.732***	1.000
SIZE	[6]	−0.141***	−0.181***	−0.033	−0.098	−0.067	1.000
ROA	[7]	−0.410***	0.083	−0.003	0.140**	0.127**	0.262***	1.000
LEV	[8]	0.350***	0.056	−0.012	0.057	0.064	−0.102***	−0.540***	1.000
RGROWTH	[9]	0.197***	0.070	0.076	0.060	0.015	−0.563***	−0.211***	0.036***	1.000
CAPEX	[10]	−0.011	−0.032	0.119*	−0.061	−0.085	0.287***	0.036***	−0.070***	0.139***	1.000
DIV	[11]	0.028**	−0.085	−0.137**	−0.030	−0.078	−0.131***	−0.066***	0.044***	0.168***	0.022**
RDINT	[12]	−0.161***	0.137**	0.011	0.133**	0.162**	0.632***	0.180***	−0.068***	−0.458***	0.121***
INTANG	[13]	0.129***	−0.048	−0.118*	0.108*	−0.096	−0.041***	−0.055***	−0.007	0.064***	0.017*
RETVOL	[14]	−0.065***	0.029	0.036	0.048	0.077	0.196***	0.101***	−0.053***	−0.121***	0.101***
LIQUID	[15]	0.037***	−0.100	−0.191***	0.060	−0.152**	−0.064***	−0.024**	0.001	0.067***	0.003
INSTOWN	[16]	−0.089***	−0.029	0.079	−0.022	−0.068	0.459***	0.095***	−0.042***	−0.239***	0.288***
LENGTH	[17]	−0.131**	0.011	0.192***	−0.086	−0.018	0.391***	0.069	−0.012	−0.171***	0.184***
BSEG	[18]	−0.112*	0.141**	−0.071	−0.079	−0.119*	0.093	0.058	0.089	−0.050	0.065
SOC_PERF	[19]	−0.005	−0.049	−0.078	−0.030	0.022	0.030***	−0.000	0.002	−0.005	0.004

Table 3. Continued

		[11]	[12]	[13]	[14]	[15]	[16]	[17]	[18]	[19]
TOBINQ	[1]
MR_READ	[2]
RBOG	[3]
RFOG	[4]
RFLESCH	[5]
SIZE	[6]
ROA	[7]
LEV	[8]
RGROWTH	[9]
CAPEX	[10]
DIV	[11]	1.000
RDINT	[12]	−0.126***	1.000
INTANG	[13]	−0.003	−0.100***	1.000
RETVOL	[14]	−0.107***	0.185***	0.034***	1.000
LIQUID	[15]	−0.026**	−0.099***	−0.017*	−0.015	1.000
INSTOWN	[16]	−0.067***	0.320***	−0.038***	0.140***	−0.045***	1.000
LENGTH	[17]	0.016	0.135**	−0.055	−0.007	−0.136**	0.239***	1.000
BSEG	[18]	0.002	0.033	−0.016	−0.160**	−0.013	0.180***	0.122*	1.000
SOC_PERF	[19]	−0.000	0.019**	−0.000	−0.009	−0.000	0.011	0.129**	0.015	1.000

Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A.

5.3. Regression Results

Our hypothesis (H1) predicts that more readable MSDs are positively associated with firm value. We report the regression results in Table 4. Model (1) shows regression results using only the control variables, while Models (2) – (5) show the regression results for the full model using MR_READ, RBOG, RFOG and RFLESCH as a measure of MSD readability. The coefficient of READABILITY is positive and statistically significant (β = 1.500, p-value < 0.05; β = 0.021, p-value < 0.05; β = 0.239, p-value < 0.01; and β = 0.189, p-value < 0.05) across all models from Models (2) – (5), respectively, indicating a positive relationship between MSD readability and firm value. These findings suggest that firms with more readable MSDs are positively priced by the market through increasing in firm value. Hence, our hypothesis (H1) is supported. In terms of economic significance, the estimated coefficients of Model (2) indicate that an increase of one standard deviation in MSD readability leads to an increase in firm value of 14.39% ([1.500 × 0.206]/2.148), which corresponds to an effect of US$629.88 (US$4,377.22 × 14.39%) million.⁹ Similarly, the estimated coefficients of Models (3) – (5) indicate that an increase of one standard deviation of MSD readability corresponds to an increase in firm value of 12.08% [(0.021 × 12.351)/2.148], 15.90% [(0.239 × 1.429)/2.148] and 11.96% [(0.189 × 1.359)/2.148], respectively.

Table 4. Regression results between MSD readability and firm value.

	Predicted sign	DV = TOBINQt + 1
			MR_READ	RBOG	RFOG	RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)	Model (5)
READABILITY	+	–	1.500**	0.021**	0.239***	0.189**
			(0.021)	(0.042)	(0.010)	(0.021)
SIZE	+	0.508***	0.574***	0.522***	0.561***	0.557***
		(0.003)	(0.002)	(0.003)	(0.002)	(0.001)
ROA	+	0.123***	0.123***	0.125***	0.119***	0.122***
		(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
LEV	−	−1.482*	−1.620**	−1.365*	−1.802**	−1.580*
		(0.068)	(0.047)	(0.078)	(0.028)	(0.053)
RETVOL	+/-	0.725***	0.719***	0.688***	0.738***	0.754***
		(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
LIQUID	+	−0.214	−0.166	−0.243	−0.154	−0.172
		(0.579)	(0.662)	(0.527)	(0.683)	(0.653)
CAPEX	+	−0.061	−0.064	−0.041	−0.088	−0.066
		(0.646)	(0.611)	(0.767)	(0.488)	(0.599)
DIV	−	−1.455*	−1.636**	−1.442*	−1.654**	−1.560*
		(0.070)	(0.044)	(0.061)	(0.040)	(0.052)
RDINT	+	0.251	−0.121	−0.314	−0.276	0.032
		(0.879)	(0.939)	(0.845)	(0.864)	(0.984)
INTANG	+	−0.828	−0.824	−0.839	−0.868	−0.939
		(0.194)	(0.186)	(0.186)	(0.164)	(0.132)
RGROWTH	+	0.377	0.522	0.637	0.519	0.478
		(0.557)	(0.395)	(0.313)	(0.403)	(0.436)
INSTOWN	+	−0.274	−0.480	−0.449	−0.427	−0.240
		(0.830)	(0.705)	(0.713)	(0.744)	(0.848)
LENGTH	+/-	−0.274	−0.315	−0.337	−0.254	−0.280
		(0.213)	(0.156)	(0.141)	(0.258)	(0.206)
BSEG	−	−0.276*	−0.213	−0.260	−0.234	−0.235
		(0.096)	(0.195)	(0.109)	(0.153)	(0.153)
SOC_PERF	+	−1.350	−1.335	−1.229	−1.348	−1.488
		(0.180)	(0.172)	(0.207)	(0.167)	(0.143)
Intercept	+/−	−0.603	−1.438	1.586	2.892	1.971
		(0.705)	(0.401)	(0.431)	(0.117)	(0.247)
Industry Fixed Effects		Yes	Yes	Yes	Yes	Yes
Observations		212	212	212	212	212
R-squared		0.500	0.513	0.509	0.516	0.509
Adjusted R-squared		0.423	0.436	0.431	0.439	0.431
Gujarati and Porter (2009) ΔR^2-F-statistic			5.37**	2.81*	7.25***	2.29**

Notes: This table reports the regression results for the association between MSD readability and firm value. Model (1) presents the regression results of Equation (4) excluding the variable of interest. Model (2) shows the regression results using the aggregate measure of readability. Models (3) – (5) show the regression results for Equation (4) using the reverse measures of Bog Index, Gunning Fog Index and Flesch – Kincaid Grade Level Index, respectively, as measures of MSD readability. Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

Regarding the control variables, we find that the coefficients of SIZE, ROA and RETVOL are positive and statistically significant, suggesting that larger firms, firms with greater profitability and firms with greater stock price volatility have higher firm value. On the other hand, the coefficients of LEV and DIV are negative and statistically significant, suggesting that firms with higher debt in their capital structure and firms that pay a higher amount of dividend have lower market value.

5.4. Endogeneity Analyses

Our study finds that more readable MSDs are positively associated with firm value. However, this association may be affected by endogeneity due to covariate imbalance and selection bias (Shipman et al., 2017). To mitigate these types of bias, we use the multivariate matching methods, propensity score matching (PSM) (Lennox et al., 2012) and entropy balancing analyses (Hainmueller, 2012). While PSM assigns binary weights to control observations based on matches, entropy balancing assigns continuous weights to balance the statistical properties of covariates. Although PSM is widely used, its reliability has been questioned due to potential bias (e.g. King & Nielsen, 2019). Entropy balancing, a newer technique, is seen as superior due to its fewer assumptions and reduced researcher adjustments (Hainmueller & Xu, 2013). Therefore, we employ both approaches to address endogeneity.

5.4.1. Entropy Balancing Analysis

The entropy balancing technique mitigates imbalances in firm characteristics, ensuring that our results are due to MSD readability rather than to these imbalances. We divide firms into two groups based on the medians of MR_READ, RBOG and RFLESCH.¹⁰ The treatment group (HIGH_READ/HIGH_RBOG/HIGH_RFLESCH) contains observations with MSD readability higher than the sample’s median, whereas the control group (LOW_READ/LOW_RBOG/LOW_RFLESCH) consists of observations with MSD readability lower than the sample’s median. Table 5 shows the entropy balancing results, which reweight control observations to balance mean, variance and skewness of covariates between groups. This creates a pseudo control group, reducing the risk of design bias (Hainmueller, 2012).¹¹

Table 5. Entropy balancing analysis.

Panel A: Descriptive statistics after entropy balancing using MR_READ proxy
	Treatment (HIGH_READ)		Control (LOW_READ)
	Mean	Variance	Mean	Variance
SIZE	7.104	2.030	7.103	2.029
ROA	5.777	75.890	5.778	75.930
LEV	0.243	0.039	0.243	0.039
RETVOL	2.115	0.853	2.115	0.853
LIQUID	0.794	0.260	0.794	0.259
CAPEX	0.189	0.375	0.189	0.375
DIV	0.926	0.070	0.925	0.070
RDINT	0.004	0.001	0.004	0.001
INTANG	0.238	0.063	0.238	0.063
RGROWTH	0.008	0.026	0.008	0.026
INSTOWN	0.137	0.010	0.137	0.010
LENGTH	7.716	0.627	7.715	0.627
BSEG	0.724	0.531	0.723	0.531
SOC_PERF	−0.011	0.021	−0.011	0.021
Panel B: Descriptive statistics after entropy balancing using Bog Index proxy
	Treatment (HIGH_RBOG)		Control (LOW_RBOG)
	Mean	Variance	Mean	Variance
SIZE	7.235	1.952	7.234	1.957
ROA	4.818	118.700	4.805	118.600
LEV	0.212	0.034	0.213	0.034
RETVOL	2.230	0.898	2.224	0.897
LIQUID	0.874	0.249	0.875	0.249
CAPEX	0.147	0.082	0.147	0.081
DIV	0.871	0.114	0.871	0.113
RDINT	0.006	0.001	0.006	0.001
INTANG	0.242	0.067	0.242	0.067
RGROWTH	0.032	0.052	0.032	0.052
INSTOWN	0.146	0.011	0.146	0.011
LENGTH	7.752	0.544	7.748	0.546
BSEG	0.699	0.527	0.700	0.527
SOC_PERF	−0.016	0.022	−0.016	0.022
Panel C: Descriptive statistics after entropy balancing using FLESCH proxy
	Treatment (HIGH_RFLESCH)		Control (LOW_RFLESCH)
	Mean	Variance	Mean	Variance
SIZE	7.138	1.870	7.142	1.873
ROA	6.563	84.260	6.567	84.080
LEV	0.226	0.038	0.226	0.038
RETVOL	2.050	0.732	2.051	0.733
LIQUID	0.755	0.232	0.755	0.232
CAPEX	0.172	0.500	0.172	0.499
DIV	0.938	0.059	0.938	0.058
RDINT	0.006	0.001	0.006	0.001
INTANG	0.243	0.062	0.243	0.062
RGROWTH	0.025	0.028	0.025	0.028
INSTOWN	0.127	0.010	0.127	0.010
LENGTH	7.720	0.640	7.718	0.640
BSEG	0.747	0.504	0.747	0.505
SOC_PERF	0.002	0.019	0.002	0.019
Panel D: Regression results between readability and firm value using entropy-balanced sample
	Predicted sign	DV = TOBINQt + 1
		MR_READ	RBOG	RFLESCH
		Model (1)	Model (2)	Model (3)
READABILITY	+	1.214**	0.028**	0.177**
		(0.042)	(0.012)	(0.023)
SIZE	+	0.437***	0.408**	0.442***
		(0.007)	(0.012)	(0.001)
ROA	+	0.131***	0.116**	0.132***
		(0.000)	(0.014)	(0.000)
LEV	−	−1.400**	−1.227	−0.991*
		(0.045)	(0.101)	(0.098)
RETVOL	+/-	0.563***	0.493**	0.673***
		(0.002)	(0.019)	(0.001)
LIQUID	+	0.069	0.083	−0.387
		(0.832)	(0.828)	(0.174)
CAPEX	+	−0.017	−0.297	−0.021
		(0.849)	(0.212)	(0.803)
DIV	−	−1.339**	−1.310*	−0.943*
		(0.037)	(0.061)	(0.077)
RDINT	+	25.473**	20.370*	2.189
		(0.014)	(0.075)	(0.136)
INTANG	+	−0.505	−0.898	−0.637
		(0.358)	(0.160)	(0.271)
RGROWTH	+	0.706	1.709**	0.027
		(0.176)	(0.036)	(0.948)
INSTOWN	+	−0.461	−1.224	−0.207
		(0.699)	(0.371)	(0.842)
LENGTH	+/-	−0.236	−0.248	−0.192
		(0.172)	(0.215)	(0.174)
BSEG	−	−0.222*	−0.328**	−0.143
		(0.085)	(0.043)	(0.268)
SOC_PERF	+	−1.115	−1.218	−1.039
		(0.248)	(0.213)	(0.325)
Intercept	+/−	−1.208	2.607	1.680
		(0.458)	(0.204)	(0.303)
Industry Fixed Effects		Yes	Yes	Yes
Observations		212	212	212
R-squared		0.587	0.506	0.555
Adjusted R-squared		0.521	0.427	0.484

Notes: This table shows the entropy-balancing analysis results. Panel A, B and C show the descriptive statistics after entropy balancing. Panel D shows the regression results between MSD readability and firm value using entropy balanced sample. Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

Table 5, Panels A – C present descriptive statistics for entropy-balanced samples of treatment and control groups, comparing HIGH_READ versus LOW_READ; HIGH_RBOG versus LOW_RBOG; and HIGH_RFLESCH versus LOW_RFLESCH, respectively. The results suggest that treatment and control groups are similar. Table 5, Panel D presents the second-stage regression results based on the entropy-balanced samples, showing positive and statistically significant coefficients for READABILITY across all models, corroborating our main findings as shown in Table 4.

5.4.2. Propensity Score Matching (PSM) Analysis

Although entropy balancing analysis is superior to PSM analysis, we also run PSM analysis to test the robustness of our findings. We use the same treatment and control groups from the entropy balancing analysis. To implement PSM analysis, we follow two stages. Firstly, we run logistic regression models for HIGH_READ, HIGH_RBOG, HIGH_RFOG and HIGH_RFLESCH with the same set of control variables as in Equation (4). The main principle of PSM is to use the same control variables in the first stage and second stage to provide assurance of accurate equilibrium between treatment and control observations in the matched sample (Shipman et al., 2017). Therefore, we use the same set of control variables in both models. When the treatment group (HIGH_READ, HIGH_RBOG, HIGH_RFOG and HIGH_RFLESCH) determinants are comparable between the two groups (HIGH_READ versus LOW_READ; HIGH_RBOG versus LOW_RBOG; HIGH_RFOG versus LOW_RFOG; HIGH_RFLESCH versus LOW_RFLESCH), covariate equilibrium is achieved.

The first-stage regression results are not reported here for reasons of brevity. Table 6, Panels A and B show no statistical differences in the covariates between the treatment and control groups. We match without replacement a firm’s treatment observation against a control observation. We employ the caliper matching method, matching within a caliper of 1%.¹² Using the predicted propensity scores, we match around 31–73% of observations using the four different proxies for readability. Table 6, Panel C shows the regression results using the PSM-matched sample. The coefficients of READABILITY are positive and statistically significant across Models (1) – (4), indicating that our findings hold after controlling for observable heterogeneity bias using the PSM procedure.

Table 6. Propensity score matching (PSM) analysis.

Panel A: Mean differences between treatment and control samples using MR_READ and RBOG
	READABILITY (MR_READ)			READABILITY (RBOG)
	Treatment (HIGH_READ)	Control (LOW_READ)	t-test (p-value)	Treatment (HIGH_RBOG)	Control (LOW_RBOG)	t-test (p-value)
SIZE	7.475	7.514	0.855	7.336	7.365	0.903
ROA	6.053	5.750	0.825	5.875	5.151	0.596
LEV	0.203	0.217	0.587	0.203	0.232	0.312
RETVOL	1.993	2.006	0.914	2.090	1.947	0.263
LIQUID	0.845	0.842	0.962	0.792	0.783	0.911
CAPEX	0.231	0.009	0.982	0.140	0.157	0.770
DIV	0.911	0.899	0.788	0.883	0.922	0.418
RDINT	0.004	0.007	0.439	0.007	0.005	0.645
INTANG	0.235	0.245	0.813	0.241	0.251	0.802
RGROWTH	0.040	0.066	0.605	0.035	0.018	0.648
INSTOWN	0.148	0.148	0.961	0.148	0.144	0.828
LENGTH	7.855	7.782	0.522	7.767	7.760	0.950
BSEG	0.836	0.813	0.848	0.766	0.739	0.816
SOC_PERF	0.003	0.006	0.874	−0.007	0.002	0.700
Panel B: Mean differences between treatment and control samples using RFOG and RFLESCH
	READABILITY (RFOG)			READABILITY (RFLESCH)
	Treatment (HIGH_RFOG)	Control (LOW_RFOG)	t-test (p-value)	Treatment (HIGH_RFLESCH)	Control (LOW_RFLESCH)	t-test (p-value)
SIZE	7.225	7.415	0.409	7.576	7.432	0.540
ROA	6.018	5.830	0.901	6.399	4.420	0.278
LEV	0.237	0.207	0.291	0.176	0.224	0.083
RETVOL	2.061	2.082	0.884	1.980	1.974	0.967
LIQUID	0.781	0.834	0.493	0.861	0.785	0.337
CAPEX	0.222	0.157	0.494	0.244	0.316	0.671
DIV	0.890	0.904	0.787	0.902	0.885	0.771
RDINT	0.006	0.007	0.909	0.004	0.007	0.495
INTANG	0.274	0.263	0.803	0.236	0.253	0.720
RGROWTH	0.040	0.019	0.512	0.033	0.063	0.669
INSTOWN	0.139	0.147	0.647	0.151	0.141	0.585
LENGTH	7.743	7.733	0.426	7.923	7.742	0.151
BSEG	0.823	0.825	0.992	0.771	0.925	0.269
SOC_PERF	−0.019	−0.003	0.505	0.006	−0.006	0.677

Panel C: Regression results between readability and firm value using PSM-matched sample
	Predicted sign	DV = TOBINQt + 1
		MR_READ	RBOG	RFOG	RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)
READABILITY	+	0.405**	0.020*	1.675***	0.124*
		(0.033)	(0.082)	(0.004)	(0.065)
SIZE	+	0.252	0.324**	0.305*	0.271
		(0.220)	(0.038)	(0.072)	(0.203)
ROA	+	0.101***	0.137**	0.164***	0.102***
		(0.000)	(0.039)	(0.002)	(0.000)
LEV	−	−0.133	−0.990	0.271	−0.117
		(0.847)	(0.217)	(0.729)	(0.864)
RETVOL	+/-	0.425*	0.412*	0.469**	0.462*
		(0.075)	(0.089)	(0.020)	(0.057)
LIQUID	+	0.285	−0.001	−0.373	0.280
		(0.492)	(0.999)	(0.332)	(0.502)
CAPEX	+	−0.080	−0.158	−0.061	−0.073
		(0.476)	(0.588)	(0.666)	(0.527)
DIV	−	−0.896	−1.079	−1.418	−0.908
		(0.169)	(0.214)	(0.106)	(0.175)
RDINT	+	18.202**	17.234*	18.359	17.224**
		(0.019)	(0.076)	(0.110)	(0.027)
INTANG	+	−0.482	−1.043	−1.144	−0.585
		(0.444)	(0.168)	(0.124)	(0.367)
RGROWTH	+	0.171	1.710	0.195	0.220
		(0.643)	(0.146)	(0.712)	(0.553)
INSTOWN	+	−0.192	−1.483	−0.196	−0.134
		(0.893)	(0.291)	(0.902)	(0.925)
LENGTH	+/−	−0.223	−0.270	−0.277	−0.214
		(0.217)	(0.159)	(0.290)	(0.234)
BSEG	−	−0.259*	−0.231	−0.508**	−0.256
		(0.099)	(0.180)	(0.024)	(0.104)
SOC_PERF	+	−0.806	−0.514	−1.558	−0.942
		(0.354)	(0.662)	(0.312)	(0.309)
Intercept	+/−	0.391	2.861	1.392	2.233
		(0.783)	(0.194)	(0.521)	(0.157)
Industry Fixed Effects		Yes	Yes	Yes	Yes
Observations		155	151	146	155
R-squared		0.560	0.500	0.575	0.555

Notes: This table shows the PSM analysis results. Panel A and B show the balancing of treatment and control firms using first-stage PSM results. Panel C shows the regression results between MSD readability and firm value using PSM-matched sample. Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

6. Additional Analysis and Robustness Checks

6.1. Role of Optimistic Tone

Du and Yu (2021) find that firms with a more optimistic tone in CSR reports have a higher level of CSR performance and abnormal stock returns. If an optimistic tone in MSDs increases perceived benefits for investors, we expect the positive effects of MSD readability on firm value to be higher for firms with an optimistic tone. We compute optimistic tone as the difference between total positive words minus total negative words, scaled by their sum, following Loughran and McDonald’s (2011) list of words which is extensively used by prior studies (e.g. Du & Yu, 2021; Huang et al., 2014; Jiang et al., 2019). Furthermore, for ease of interpretation, we compute MSDs’ optimistic tone (OPTIMISTIC_TONE) as an indicator variable that takes a value of 1 if the value of the firm’s optimistic tone is higher than the sample’s median, and 0 otherwise. The regression results are reported in Table 7. The coefficients of READABILITY × HIGH_OPTIMISTIC_TONE are positive and statistically significant (β = 1.753, p-value < 0.10; β = 0.475, p-value < 0.05; β = 0.364, p-value < 0.10) in Models (1), (3) and (4) using MR_READ, RFOG and RFLESCH, respectively, as measures of readability. However, no significant results are found using the Bog Index proxy for readability. These findings suggest that the positive impact of more readable MSDs on firm value is more pronounced for firms with an optimistic tone.

Table 7. Regression results between MSD readability and firm value: Role of optimistic tone.

	Predicted sign	DV = TOBINQt + 1
		MR_READ	RBOG	RFOG	RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)
READABILITY	+/−	0.640	0.026*	−0.025	0.044
		(0.485)	(0.055)	(0.865)	(0.673)
READABILITY × HIGH_OPTIMISTIC_TONE	+	1.753*	0.020	0.475**	0.364*
		(0.086)	(0.556)	(0.021)	(0.095)
HIGH_OPTIMISTIC_TONE	+/−	−0.895	−1.582	7.766**	5.581*
		(0.203)	(0.541)	(0.021)	(0.092)
SIZE	+	0.569***	0.529***	0.541***	0.570***
		(0.000)	(0.000)	(0.000)	(0.001)
ROA	+	0.124***	0.125***	0.122***	0.123***
		(0.000)	(0.000)	(0.000)	(0.001)
LEV	−	−1.515*	−1.427*	−1.831**	−1.625**
		(0.072)	(0.093)	(0.029)	(0.047)
RETVOL	+/−	0.541	0.649	0.406	0.414
		(0.264)	(0.198)	(0.396)	(0.503)
LIQUID	+	−0.061	−0.032	−0.068	−0.044
		(0.769)	(0.879)	(0.743)	(0.724)
CAPEX	+	−1.665***	−1.441***	−1.669***	−1.606**
		(0.002)	(0.007)	(0.002)	(0.046)
DIV	−	−0.173	−0.340	0.139	0.138
		(0.888)	(0.789)	(0.910)	(0.931)
RDINT	+	−0.742	−0.882	−0.691	−0.800
		(0.290)	(0.212)	(0.319)	(0.193)
INTANG	+	0.703***	0.694***	0.703***	0.764***
		(0.006)	(0.007)	(0.005)	(0.001)
RGROWTH	+	−0.097	−0.247	−0.084	−0.147
		(0.795)	(0.507)	(0.818)	(0.715)
INSTOWN	+	−0.527	−0.551	−0.785	−0.406
		(0.759)	(0.751)	(0.645)	(0.771)
LENGTH	+/-	−0.331*	−0.339*	−0.219	−0.264
		(0.098)	(0.095)	(0.266)	(0.246)
BSEG	−	−0.207	−0.265	−0.251	−0.238
		(0.272)	(0.160)	(0.173)	(0.148)
SOC_PERF	+	−1.353	−1.246	−1.370	−1.530
		(0.148)	(0.187)	(0.138)	(0.131)
Intercept	+/−	−0.886	1.999	−1.494	−0.457
		(0.636)	(0.372)	(0.610)	(0.845)
Industry Fixed Effects		Yes	Yes	Yes	Yes
Observations		212	212	212	212
R-squared		0.518	0.510	0.530	0.516

Notes: This table reports the regression results of the moderating role of an optimistic tone in the association between MSD readability and firm value. Model (1) shows the regression results using the aggregate measure of readability (MR_READ). Model (2) presents the regression results using the reverse measure of Bog Index (RBOG). Model (3) presents the regression results using the reverse measure of Gunning Fog Index (RFOG). Model (4) shows the regression results using the reverse measure of Flesch – Kincaid Grade Level Index (RFLESCH). Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

6.2. Role of Corporate Governance Performance

Studies in accounting and finance show that effective corporate governance reduces managers’ opportunistic behaviour (e.g. Gompers et al., 2003; Shleifer & Vishny, 1997); curbs earnings management (Lai & Tam, 2017); enhances accounting conservatism (Lara et al., 2009); improves disclosure transparency and accuracy; and lowers agency costs (Anderson et al., 2004). More recently, Nadeem (2022) documents that board gender diversity improves annual report readability, while Sun et al. (2022) find that effective governance reduces Chief Executive Officer (CEO) manipulation of disclosure readability. Given that effective corporate governance enhances transparency and reduces managers’ opportunistic behaviour, we expect the positive effects of more readable MSDs on firm value to be greater for firms with strong governance mechanisms. We use corporate governance performance scores from the LSEG ESG database. The scores reflect a firm’s ability to direct and control its rights and responsibilities in the best interests of its long-term shareholders, by implementing best management practices, incentives, and checks and balances. This score is the weighted average relative rating of a firm, based on the reported governance information and the resulting three governance category scores (firm-level management, shareholders and CSR strategy) that provide a comprehensive assessment of a firm’s corporate governance performance.¹³ More specifically, we create an indicator variable that takes a value of 1 if the firm-level corporate governance performance score is higher than the sample’s median and 0 otherwise.

We report the regression results in Table 8. The coefficients of READABILITY × HIGH_CGOV are positive and statistically significant (β = 0.227, p-value < 0.10; β = 0.393, p-value < 0.05; and β = 0.134, p-value < 0.05) in Models (1), (3) and (4) using MR_READ, RFOG and RFLESCH, respectively, as measures of readability. These findings suggest that the positive impact of more readable MSDs on firm value is more pronounced for firms with better corporate governance performance.

Table 8. Regression results between MSD readability and firm value: Role of corporate governance performance.

	Predicted sign	DV = TOBINQt + 1
		MR_READ	RBOG	RFOG	RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)
READABILITY	+/−	2.156**	0.023	−0.444***	0.219**
		(0.013)	(0.139)	(0.001)	(0.031)
READABILITY × HIGH_CGOV	+	0.227*	−0.010	0.393**	0.134**
		(0.089)	(0.800)	(0.035)	(0.015)
HIGH_CGOV	+/ –	−4.218	−1.236	−6.989**	−2.403***
		(0.130)	(0.670)	(0.024)	(0.004)
SIZE	+	0.579***	0.537***	0.555***	0.550***
		(0.000)	(0.000)	(0.000)	(0.000)
ROA	+	0.122***	0.122***	0.116***	0.114***
		(0.000)	(0.000)	(0.000)	(0.000)
LEV	−	−1.378*	−1.233	−1.646**	−1.574*
		(0.100)	(0.143)	(0.048)	(0.058)
RETVOL	+/-	0.529	0.699	0.504	0.605
		(0.271)	(0.166)	(0.288)	(0.204)
LIQUID	+	−0.124	−0.089	−0.182	−0.146
		(0.553)	(0.676)	(0.383)	(0.482)
CAPEX	+	−1.493***	−1.360**	−1.479***	−1.239**
		(0.005)	(0.011)	(0.005)	(0.021)
DIV	−	0.025	−0.442	−0.119	−0.279
		(0.984)	(0.727)	(0.922)	(0.818)
RDINT	+	−1.173	−1.166	−1.322*	−1.062
		(0.101)	(0.109)	(0.062)	(0.136)
INTANG	+	0.719***	0.690***	0.756***	0.710***
		(0.004)	(0.007)	(0.002)	(0.004)
RGROWTH	+	−0.269	−0.320	−0.273	−0.173
		(0.460)	(0.383)	(0.449)	(0.634)
INSTOWN	+	−0.554	−0.333	−0.558	−0.427
		(0.744)	(0.847)	(0.739)	(0.800)
SOC_PERF	+/-	−0.298	−0.323	−0.181	−0.277
		(0.133)	(0.109)	(0.358)	(0.157)
Intercept	−	−0.206	−0.259	−0.238	−0.260
		(0.266)	(0.164)	(0.192)	(0.157)
Industry Fixed Effects		Yes	Yes	Yes	Yes
Observations		212	212	212	212
R-squared		0.445	0.435	0.457	0.452

Notes: This table reports the regression results of the moderating role of better corporate governance performance in the association between MSD readability and firm value. Model (1) shows the regression results using the aggregate measure of readability (MR_READ). Model (2) presents the regression results using the reverse measure of Bog Index (RBOG). Model (3) presents the regression results using the reverse measure of Gunning Fog Index (RFOG). Model (4) shows the regression results using the reverse measure of Flesch – Kincaid Grade Level Index (RFLESCH). Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

6.3. Informative Information or Obfuscation Strategy

Prior studies indicate that management uses complex, lengthy and ambiguous disclosures to obscure poor performance and bad news (Li, 2008; Lo et al., 2017). However, other factors, such as complex business operations, multiple segment reporting, employee stock options, listing requirements, accounting standards and regulatory requirements also contribute to complex and lengthy reports (Bloomfield, 2008; Bushee et al., 2018; Clarkson et al., 2020; Loughran & McDonald, 2014). Bushee et al. (2018) argue that disclosure complexity can result from both obfuscation and efforts to provide detailed information about a firm’s fundamentals. Thus, it is unclear whether managers use narratives to reduce information asymmetry or to distort perceptions of performance. To address this, we isolate the informative component of MSD readability by regressing our readability proxies (MR_READ, RBOG, RFOG and RFLESCH) against variables explaining cross-sectional differences in MSD readability (Bushee et al., 2018; Li, 2008). Following Rjiba et al. (2021), we employ the following model to compute the informative information component of textual complexity: (5) $\begin{array}{rl}\mathrm{READABILIT}{Y}_{i,t}& ={\beta }_0+{\beta }_1\mathrm{SIZ}{E}_{i,t}+{\beta }_2\mathrm{RO}{A}_{i,t}+{\beta }_3\mathrm{LE}{V}_{i,t}+{\beta }_4\mathrm{RGROWT}{H}_{i,t}+{\beta }_5\mathrm{PPEN}{T}_{i,t}\\ & +{\beta }_6\mathrm{RDIN}{T}_{i,t}+{\beta }_7\mathrm{INTAN}{G}_{i,t}+{\beta }_8\mathrm{RETVO}{L}_{i,t}+{\beta }_9\mathrm{LIQUI}{D}_{i,t}+{\beta }_{10}\mathrm{INSTOW}{N}_{i,t}\\ & +{\beta }_{11}\mathrm{EXTFI}{N}_{i,t}+{\beta }_{12}\mathrm{BSE}{G}_{i,t}+{\beta }_{13}\mathrm{GSE}{G}_{i,t}+{\beta }_{14}\mathrm{GOODWIL}{L}_{i,t}+{\beta }_{15}\mathrm{CAPE}{X}_{i,t}\\ & +{\beta }_{16}\mathrm{ACQUISITIO}{N}_{i,t}+\sum \mathrm{INDUSTR}{Y}_{i,t}+{\epsilon }_{i,t}\end{array}$(5) All variables in Equation (5) are defined in Appendix A. We estimate Equation (5) and save residuals representing unexplained readability (RESID_MR_READ/RESID_RBOG/RESID_RFOG/RESID_RFLESCH). We then include the residuals in Equation (4) and estimate the model. As a higher value of our measure of readability (MR_READ, RBOG, RFOG and RFLESCH) indicates more readable MSDs, the residual component of Equation (5) indicates MSD informativeness.

Table 9, Panel A presents the regression results of determinants of more readable MSDs, while Panel B shows the regression results of the informative information component. The coefficients of READABILITY_RESID are positive and statistically significant using RESID_MR_READ, RESID_RBOG, RESID_RFOG and RESID_RFLESCH, respectively, suggesting that the informative information component of textual disclosures is positively priced by the market.

Table 9. Regression results between MSD readability and firm value: Role of informative information component.

Panel A: Determinants of readability
	Predicted sign	DV = MR_READ	DV = RBOG	DV = RFOG	DV = RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)
SIZE	+/−	−0.023	0.523	−0.205*	0.174
		(0.149)	(0.583)	(0.064)	(0.101)
ROA	+	0.150	−10.207	2.764**	−0.946
		(0.367)	(0.310)	(0.018)	(0.395)
LEV	+	0.123	−5.952	1.577**	−0.681
		(0.201)	(0.305)	(0.019)	(0.289)
RGROWTH	+/-	−0.166***	−14.528***	−0.843**	0.854**
		(0.002)	(0.000)	(0.025)	(0.018)
PPENT	+	−0.111	−2.073	−0.643	0.322
		(0.151)	(0.655)	(0.231)	(0.531)
RDINT	+	0.400***	31.389***	2.699***	−1.886**
		(0.004)	(0.000)	(0.005)	(0.040)
INTANG	+	−0.043	0.909	−0.050	−0.308
		(0.634)	(0.868)	(0.936)	(0.610)
RETVOL	+/−	−0.001	3.014*	−0.208	0.189
		(0.979)	(0.087)	(0.304)	(0.330)
LIQUID	+	−0.013	1.706	−0.143	0.160
		(0.748)	(0.502)	(0.624)	(0.569)
INSTOWN	+	0.116	7.803	0.425	0.491
		(0.557)	(0.513)	(0.757)	(0.710)
EXTFIN	+	0.124	5.502	0.961	−0.796
		(0.334)	(0.478)	(0.283)	(0.354)
BSEG	−	−0.039*	−0.118	−0.238	0.215
		(0.072)	(0.928)	(0.114)	(0.137)
GSEG	−	−0.026	−1.298	0.024	−0.025
		(0.362)	(0.457)	(0.904)	(0.898)
GOODWILL	+	−0.026	−2.238	−0.164	0.063
		(0.481)	(0.316)	(0.525)	(0.800)
CAPEX	+	0.011	−0.614	0.123	−0.036
		(0.662)	(0.693)	(0.492)	(0.833)
ACQUISITION	+	0.223	−7.140	−0.357	−2.320
		(0.532)	(0.741)	(0.886)	(0.332)
Intercept	+/−	0.754***	−90.701***	−14.151***	13.428***
		(0.000)	(0.000)	(0.000)	(0.000)
Industry Fixed Effects		Yes	Yes	Yes	Yes
Observations		201	201	201	201
R-squared		0.073	0.072	0.072	0.062

Panel B: Regression results between MSD readability and firm value using informative information component of readability
	Predicted sign	DV = TOBINQt + 1
		MR_READ	RBOG	RFOG	RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)
READABILITY_RESID	+	1.261**	0.020*	0.194**	0.168**
		(0.041)	(0.055)	(0.027)	(0.037)
SIZE	+	0.624***	0.623***	0.611***	0.622***
		(0.002)	(0.001)	(0.002)	(0.002)
ROA	+	0.121***	0.119***	0.121***	0.120***
		(0.001)	(0.001)	(0.001)	(0.001)
LEV	−	−1.772**	−1.822**	−1.761**	−1.763**
		(0.036)	(0.033)	(0.035)	(0.038)
RETVOL	+/−	0.820***	0.843***	0.815***	0.825***
		(0.001)	(0.001)	(0.001)	(0.001)
LIQUID	+	−0.293	−0.301	−0.303	−0.306
		(0.453)	(0.437)	(0.435)	(0.433)
CAPEX	+	−0.044	−0.035	−0.047	−0.046
		(0.736)	(0.793)	(0.713)	(0.716)
DIV	−	−1.563*	−1.404*	−1.574**	−1.511*
		(0.052)	(0.068)	(0.049)	(0.057)
RDINT	+	0.567	0.548	0.528	0.528
		(0.724)	(0.726)	(0.743)	(0.740)
INTANG	+	−0.880	−0.900	−0.880	−0.906
		(0.167)	(0.164)	(0.167)	(0.151)
RGROWTH	+	0.235	0.245	0.252	0.253
		(0.706)	(0.688)	(0.687)	(0.682)
INSTOWN	+	−0.403	−0.340	−0.411	−0.408
		(0.760)	(0.788)	(0.760)	(0.755)
LENGTH	+/−	−0.345	−0.371	−0.299	−0.317
		(0.142)	(0.123)	(0.210)	(0.179)
BSEG	−	−0.286*	−0.279	−0.287*	−0.283
		(0.100)	(0.102)	(0.097)	(0.104)
SOC_PERF	+	−1.447	−1.352	−1.441	−1.600
		(0.173)	(0.200)	(0.175)	(0.146)
Intercept	+/−	−0.846	−0.818	−1.078	−1.085
		(0.598)	(0.610)	(0.510)	(0.504)
Industry Fixed Effects		Yes	Yes	Yes	Yes
Observations		201	201	201	201
R-squared		0.519	0.518	0.520	0.517

Notes: This table reports the regression results of the informative information component strategy. Panel A presents the regression results of the determinants of MSD readability. Panel B shows the regression results of the association between the informative information component of readability with firm value. Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

6.4. Investors’ Reactions to Publication of MSDs

We also examine investors’ reactions to the publication of MSDs using an event study, around the issuance date of modern slavery reports.¹⁴ Specifically, we compute cumulative abnormal returns (CARs) over a five-day period (t =  −2 to t = +2) around the issuance date (t = 0) of modern slavery reports, using daily stock returns of the firm and overall market returns during a 200-day estimating period covering t = −210 to t =  −11. Table 10 reports the regression results. The coefficient of READABILITY is positive and statistically significant across all proxies of readability except RBOG, suggesting that our findings are qualitatively similar which supports the hypothesised effect of higher readability of MSDs on investor reactions.

Table 10. Regression results between MSD readability and cumulative abnormal returns (CARs).

	Predicted sign	DV = CAR
		MR_READ	RBOG	RFOG	RFLESCH
		Model (1)	Model (2)	Model (3)	Model (4)
READABILITY	+	0.007**	0.000	0.001**	0.001***
		(0.016)	(0.220)	(0.050)	(0.010)
SIZE	+	0.000	−0.000	0.000	0.000
		(0.744)	(0.902)	(0.915)	(0.770)
ROA	+	−0.000	−0.000	−0.000	−0.000
		(0.637)	(0.719)	(0.587)	(0.619)
LEV	−	0.010*	0.011*	0.010*	0.010*
		(0.076)	(0.061)	(0.088)	(0.070)
RETVOL	+/-	−0.001	−0.001	−0.001	0.001
		(0.492)	(0.491)	(0.509)	(0.609)
LIQUID	+	−0.002	−0.002	−0.002	−0.002
		(0.228)	(0.182)	(0.220)	(0.233)
CAPEX	+	−0.000	−0.000	−0.000	0.000
		(0.951)	(0.907)	(0.790)	(0.988)
DIV	−	−0.004	−0.003	−0.003	−0.003
		(0.296)	(0.422)	(0.318)	(0.322)
RDINT	+	−0.002	−0.001	−0.002	−0.001
		(0.677)	(0.769)	(0.672)	(0.801)
INTANG	+	−0.001	−0.001	−0.001	−0.001
		(0.694)	(0.624)	(0.621)	(0.589)
RGROWTH	+	0.000	0.000	0.000	0.000
		(0.910)	(0.970)	(0.973)	(0.991)
INSTOWN	+	0.010	0.011	0.010	0.011
		(0.184)	(0.141)	(0.153)	(0.135)
LENGTH	+/-	−0.001	−0.001	−0.001	−0.001
		(0.323)	(0.387)	(0.507)	(0.423)
BSEG	−	−0.002*	−0.002**	−0.002**	−0.002*
		(0.067)	(0.041)	(0.048)	(0.051)
SOC_PERF	+	0.006*	0.006**	0.006*	0.005*
		(0.060)	(0.047)	(0.059)	(0.096)
Intercept	+/−	0.007	0.015	0.023**	0.025***
		(0.341)	(0.127)	(0.037)	(0.008)
Industry Fixed Effects		Yes	Yes	Yes	Yes
Observations		207	207	207	207
R-squared		0.117	0.098	0.108	0.114

Notes: This table reports the regression results for the association between MSD readability and cumulative abnormal returns (CARs). Model (1) shows the regression results using the aggregate measure of readability (MR_READ). Model (2) presents the regression results using the reverse measure of Bog Index (RBOG). Models (3) and (4) show the regression results for Equation (4) using the reverse measure of Gunning Fog Index (RFOG) and Flesch – Kincaid Grade Level Index (RFLESCH), respectively, as a measure of MSD readability. Coefficient values are shown with robust standard errors, while p-values, reported in parentheses, are one-tailed when a prediction is made. Superscript ***, ** and * indicate statistical significance at 1%, 5% and 10% levels, respectively. All variables are defined in Appendix A. DV = dependent variable.

6.5. Alternative Proxies

It is important to acknowledge that Tobin’s Q may be subject to potential measurement error which can introduce bias into our analysis. To address this concern, we employ two alternative measures of Tobin’s Q (TOBINQ): – 1/TOBINQ (the inverse of TOBINQ) and the natural logarithm of TOBINQ, as previously used in studies by Ferreira and Matos (2008). The regression results are not reported here for reasons of brevity. However, the unreported results suggest our findings remain the same as shown in Table 4.

7. Discussion and Conclusion

In this study, we examine the relationship between the readability of MSDs and firm value by analysing 212 MSD reports from publicly-listed Australian firms. Our findings reveal a positive association between MSD readability and firm value, indicating that more readable MSDs provide useful information to investors that enhances firm value. This supports impression management theory in suggesting that firms face market penalties if they use less readable and ambiguous language to hide unethical practices. The positive effect of more readable MSDs on firm value implies that the capital market values transparency in these disclosures. Consistent with legitimacy theory, we argue that more readable MSDs promote transparency and clarity, upholding a firm’s social legitimacy and reducing the risk of financial penalties from regulatory authorities, leading to positive investor valuations. Furthermore, the robustness of our findings is confirmed through various tests, comprising entropy balancing analysis; PSM analysis to address endogeneity from observable selection bias; and the use of alternative measurements of Tobin’s Q to assess firm value. In addition, we show that an optimistic tone in textual disclosures and better corporate governance accentuate the impact of more readable MSDs on firm value. These findings suggest that an optimistic tone in MSDs reduces perceived risks among investors, while effective corporate governance enhances transparency and limits managers’ opportunistic behaviour in relation to producing more readable disclosures. Finally, we document that the informative information component of more readable MSDs is also priced by investors.

Our study extends the literature in several ways. Firstly, while prior studies examine various facets of modern slavery reporting based on interview-based methodologies (e.g. Islam & Van Staden, 2022; Rogerson et al., 2020); industry-focused case studies (e.g. Boersma & Bedford, 2023; Dodd et al., 2023); or manual content analysis (Birkey et al., 2018; Mai et al., 2023; Pinnington et al., 2023; Schaper & Pollach, 2021), our study examines the readability (syntactical characteristics) of MSDs and their impact on firm value. This approach, widely used in accounting and finance literature, provides empirical evidence of the MSDs’ economic impact. Secondly, our study contributes to the growing number of studies in the literature, by focusing on the moderating roles of an optimistic tone and better corporate governance in the association between MSD readability and firm value. Our findings suggest that an optimistic tone in MSDs and better corporate governance both have crucial roles in ensuring investors’ positive evaluation of MSD readability. Thirdly, we document that the informative component of more readable MSDs is valued by investors which, in turn, enhances firm value. This highlights that transparency and clarity in corporate disclosures are important for improving investor confidence and firm valuation.

Our study’s findings have important practical implications for firm management, regulators, policy makers and investors. Firstly, the findings can assist in motivating firm management to enhance the readability of MSDs, potentially improving the firm’s corporate reputation among stakeholders and the broader community. As readability fosters transparency in reporting modern slavery information, capital market players are likely to reward or penalise firms accordingly. Secondly, this study’s findings present a compelling case for managers and decision-makers to prioritise MSD readability. By making authentic efforts in this area, they can potentially increase shareholder value and make more effective use of shareholders’ funds. Thirdly, our study’s findings align with impression management theory and legitimacy theory, suggesting that firms using less readable and more ambiguous language in their MSDs to obscure unethical practices may incur market penalties. Conversely, more readable MSDs enhance transparency and clarity, supporting a firm’s social legitimacy and reducing the risk of regulatory fines, with both positively received by investors. Fourthly, the study validates the actions of decision-makers in improving MSD readability, indicating that shareholders and market participants value transparent information and support these initiatives. The findings also highlight the important role of effective corporate governance in MSDs, in terms of preventing opportunistic managerial behaviour. If firms are to be recognised in the market for improved MSD readability, they will need strong governance structures and functionalities to produce more transparent disclosures.

Our study is not without its limitations. Firstly, it focuses solely on the Australian context, with the relationship between MSD readability and firm value possibly differing in other countries. Future research could investigate this effect in an international setting. Secondly, we employ the Gunning Fog Index and the Flesch – Kincaid Index to measure MSD readability. While widely used, these measures have been criticised for their reliance on ‘complex words’ (Loughran & McDonald, 2014) and their failure to account for reader attributes, such as interest and expertise (Stone & Parker, 2013). To address these concerns, we also use the Bog Index (Bonsall et al., 2017), validating our results with each index individually and in aggregate. Thirdly, while our study supports the positive association between the readability of MSD and firm value, reverse causality is a potential issue. Financially better-off firms may produce more readable disclosures (Bushee et al., 2018). Additionally, the relationship between disclosure readability and firm performance as it relates to modern slavery compliance makes it challenging to distinguish their separate effects on firm value. Poor performance may lead to more complex and less readable disclosures (Wang et al., 2018). Future research should explore these causal limitations with methodologies that can better disentangle these effects. Lastly, given that Australia has enacted modern slavery regulations, further research could examine the potential for mandatory assurance of MSDs or could investigate their quality, especially as the trend for superficial reporting on modern slavery issues appears to be growing in other countries.

Acknowledgement

The authors would like to thank two anonymous reviewers, Francesco Mazzi (Associate Editor) and Andrei Filip (Editor-in-Chief), for their helpful comments in improving the contents of the manuscript.

Disclosure Statement

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

Data Availability statement

All data are available from sources stated in the paper.

Notes

1 https://www.globalreporting.org/media/r1ydr40k/rli-gri_advancing-modern-slavery-reporting-to-meet-stakeholder-expectations.pdf (accessed on 10 March 2024).

2 Walk Free (2023) reports that approximately 50 million people were victims of modern slavery on any given day in 2021, with around 6.4 million people living in modern slavery in Europe and Central Asia during the same period (Walk Free, 2023).

3 https://www.pwc.com/ca/en/today-s-issues/compliance-transformed/modern-slavery-reporting-requirements/legal-implications-of-canadas-modern-slavery-act.html (accessed on 10 March 2024).

4 This directive will eventually become law in the EU, imposing obligations on large companies to address human rights and environmental impacts within their operations, subsidiaries and their business chains. Additionally, the directive specifies penalties and civil liabilities for non-compliance (European Union [EU], 2024). Moreover, various organisations, including Anti-Slavery International, the European Centre for Constitutional and Human Rights (ECCHR) and the Greens/European Free Alliance (EFA) Group, are actively advocating for stronger regulations against modern slavery within the European Union (EU).

5 To date, two capital market studies have been conducted. Both employed event studies to understand the consequences of various events leading to the adoption of the UK Modern Slavery Act (Cousins et al., 2020), and the California Transparency in Supply Chains Act (Birkey et al., 2018). More recently, an event study analysing news articles on modern slavery allegations found that such allegations negatively impact financial performance (Sokat & Altay, 2023).

6 Christ and Burritt (2023) conducted inductive-thematic analysis of Australian MSDs by focusing on a specific set of paragraphs in MSDs, whereas Christ et al. (2019) used a disclosure index to focus on a specific set of disclosures related to modern slavery practices before modern slavery reporting was mandated in Australia.

7 https://modernslaveryregister.gov.au/ (accessed on 20 March 2024).

8 A value in the Gunning Fog Index greater than or equal to 18 indicates the text is unreadable; 14–18 indicates it is difficult to read; 12–14 indicates that it is ideal to read; 10–12 indicates that it is acceptable; and 8–10 suggests that it is childish (Li, 2008). Furthermore, a score in the Flesch–Kincaid Grade Level Index of 8 indicates that the document can be understandable by an average eighth grader (Li, 2008). Similarly, a higher value for the Bog Index suggests poor readability.

9 The computed magnitude is comparable to prior studies in Australia. For example, Nguyen et al. (2022) find that a one standard deviation increase in the value of CSR results in approximately a 14% increase in Tobin’s Q. Additionally, we follow Göttsche et al. (2020) to compute the magnitude of the firm value, ensuring consistency and comparability with existing research.

10 We do not use the RFOG proxy for readability for entropy balancing due to non-convergence of balancing based on statistical properties (e.g., means, variances, skewness) of covariates between treatment and control groups.

11 In our analysis, we use mean and variance for running the entropy balancing analysis because the algorithm does not converge within the specified tolerance due to their higher orders.

12 The caliper refers to the difference in predicted probabilities between the treatment and control observations (Dehejia & Wahba, 2002).

13 The management score evaluates a firm’s commitment and effectiveness towards following best practice corporate governance principles. The shareholder score assesses a firm’s effectiveness towards equal treatment of shareholders and the use of anti-takeover mechanisms. The CSR strategy score reflects how a firm practises the incorporation of economic (financial), social and environmental dimensions into its day-to-day decision-making processes.

14 We thank an anonymous reviewer for this suggestion.

Appendix A:

Definitions of variables

Table

Variables		Explanation
Panel A: Dependent variable
TOBINQt + 1	One-year-ahead Tobin’s Q	The sum of the book value of total assets plus the market value of equity minus the book value of equity divided by total assets.
Panel B: Variables of interest
READABILITY	Readability	The readability of modern slavery disclosures (MSDs).
MR_READ	Aggregate readability	An aggregate readability index based on the average of the percentile ranks for each component of readability (i.e. RBOG, RFOG and RFLESCH) and then take the average across the three groups.
RBOG	Reverse BOG Index	Readability score of MSDs using BOG Index score multiplied by – 1.
RFOG	Reverse Fog Index	Readability score of MSDs using Gunning Fog Index score multiplied by −1.
RFLESCH	Reverse Flesch – Kincaid Grade Level Index	Readability score of MSDs using Flesch – Kincaid Grade Level Index score multiplied by −1.
Panel C: Firm-level variables
SIZE	Firm size	The natural logarithm of the market value of equity.
ROA	Profitability	The ratio of net income scaled by total assets.
LEV	Leverage	The ratio of total debts divided by total assets at the end of the fiscal year.
RETVOL	Firm risk	The standard deviation of stock returns over the fiscal year.
LIQUID	Liquidity	The ratio of the number of shares traded to total shares outstanding at the end of the year.
CAPEX	Capital intensity	The ratio of capital expenditure to total sales.
DIV	Dividend	An indicator variable taking the value of 1 if the firm pays a dividend, and 0 otherwise.
RDINT	Research and development	The ratio of research and development (R&D) expenditure to total revenue.
INTANG	Intangible assets	The ratio of total intangible assets to total assets.
RGROWTH	Revenue growth	Annual growth in sales revenue.
INSTOWN	Institutional ownership	The percentage of ownership held by institutional investors.
LENGTH	Length of MSD report	The natural logarithm of the MSD’s total number of words.
BSEG	Business complexity	The number of business segments.
SOC_PERF	Social performance	Industry-adjusted social performance score where social performance data is obtained from Refinitiv ESG database.
OPTIMISTIC_TONE	Optimistic tone	An indicator variable that takes a value of 1 if the value of the firm’s optimistic tone is higher than the sample’s median, and 0 otherwise.
HIGH_CGOV	Corporate governance performance score	An indicator variable that takes a value of 1 if the firm-level corporate governance performance score is higher than the sample’s median, and 0 otherwise.
PPENT	New investment in property, plant and equipment	The ratio of net property, plant and equipment to total assets.
EXTFIN	External financing	The amount of debt or equity capital raised by the firm, scaled by total assets.
GSEG	Geographic segment	The firm’s number of geographic segments.
GOODWILL	Goodwill impairment	An indicator variable that takes a value of 1 if the firm reports a goodwill impairment charge, and 0 otherwise.
ACQUISITION	Acquisitions	The ratio of acquisitions to total assets.