From posts to pricing: circular intellectual capital disclosure via social media and the cost of debt

The last decades have been characterised by a progressive shift from an industrial to a knowledge-based economy, in which intangible resources have gained primary relevance as pivotal drivers of economic success and wealth creation (Lev et al., 2005; Dumay and Guthrie, 2019; Nicolò, 2020). Within this scenario, the concept of Intellectual Capital (IC) has emerged in both academic and professional domains as the aggregation of firms' intangible resources–such as knowledge, experience, skills, information systems, brands, patents, organisational routines, and relationships–which, when effectively managed, enable organisations to achieve competitive advantage and generate long-term value (Sonnier, 2008; Nicolò, 2020; Abdallah et al., 2024). IC has progressively become a central construct within accounting and management research, evolving through distinct stages of scholarly development (Dumay and Garanina, 2013; Dumay and Guthrie, 2019; Dumay et al., 2020). The dominant leitmotif of this research stream has been the attempt to overcome the limitations of traditional accounting systems by identifying more appropriate ways to measure, manage, and disclose IC resources to improve internal decision-making processes and organisational performance while simultaneously satisfying investors' and other stakeholders' information needs (Striukova et al., 2008; Duff, 2018; Nicolò et al., 2021).

However, increasing concerns related to global challenges–such as environmental degradation, natural resource depletion, climate change, and biodiversity loss–have prompted a further evolution of IC research. As a result, scholars have embraced a broader perspective to explore how IC can be mobilised beyond corporate economic value creation, to support wider environmental objectives (Massaro et al., 2018; Dumay et al., 2017, 2020). Accordingly, the notion of Green Intellectual Capital (GIC) emerged to incorporate environmental knowledge resources into the traditional IC framework (Chen, 2008; Chang and Chen, 2012; Rehman et al., 2021). GIC refers to the aggregate of intangible assets related to environmental protection and green innovation that organisations leverage to enhance their environmental performance and achieve sustainable competitive advantage (Chen, 2008; Rehman et al., 2021; Shehzad et al., 2023).

More recently, driven by the urgent need to accelerate the ecological transition and decarbonisation processes–as emphasised by supranational initiatives such as the United Nations (UN) 2030 Agenda, the Paris Climate Agreement, and the European Green Deal–the circular economy has gained significant momentum. It represents a transformative paradigm aimed at decoupling economic growth from resource depletion, climate change, and environmental degradation (Kirchherr et al., 2017; Esposito et al., 2023; Massari and Giannoccaro, 2023). The circular economy seeks to replace the prevailing unsustainable linear economic model based on the “take–make–dispose” approach with a closed-loop system designed to extend product lifecycles and minimise the consumption of resources, materials, and energy (Urbinati et al., 2017; Roberts et al., 2023; Farrukh et al., 2023). Grounded in three core principles–reduce, reuse, and recycle–the circular economy fosters a virtuous transition towards more sustainable patterns of production and consumption (Kirchherr et al., 2017; Urbinati et al., 2017; Esposito et al., 2025).

However, the effective implementation of circular practices along the value chain crucially depends on the proper management and strategic exploitation of a wide range of intangible resources, including employees' skills, capabilities, and specialised knowledge, robust internal information systems, consolidated organisational routines, and stable relationships with external partners and institutions (Raimo et al., 2025). For this reason, scholars have recently introduced the concept of Circular Intellectual Capital (CIC), which embeds circular economy principles within the traditional IC taxonomy, emphasising the role of intangible resources in enabling circular strategies and value creation (Raimo et al., 2025).

Articulated into three main dimensions–circular human capital, circular organisational capital, and circular relational capital – the CIC refers to the bundle of intangible resources that enable firms to advance circular economy practices and contribute to sustainable development (Raimo et al., 2025). In this sense, CIC builds upon and extends the GIC framework by highlighting the strategic operationalisation of IC to foster a systemic and holistic transformation of value creation, centred on the redesign of production and business models, the optimisation of resource reuse, and the strengthening of interdependencies across human, organisational, and relational capital (Farrukh et al., 2023; Raimo et al., 2025). It follows that conveying information on CIC may represent a key strategic instrument through which firms respond to investors' and other stakeholders' emerging demand about the interconnections between intangible resources, value creation processes, and circular economy trajectories (Farrukh et al., 2023; Raimo et al., 2025). So, by means of CIC disclosure, companies can demonstrate how they strategically deploy their IC to generate value that extends beyond organisational boundaries, thereby producing positive externalities for the broader ecosystem.

However, while the academic debate on IC disclosure has largely reached maturity, the relationship between IC and the circular economy remains underexplored from an accounting and reporting perspective, leaving a critical research gap. Accordingly, this study seeks to bridge this gap by responding to the call by Raimo et al. (2025) for further research into CIC disclosure practices and their potential benefits, including improved access to finance.

Accounting literature has extensively discussed how lenders increasingly rely on non-financial information to assess borrowers' default and reputational risks and, consequently, their ability to repay debt (Eliwa et al., 2021; Raimo et al., 2021; L'Abate et al., 2024a). A growing body of studies shows that higher levels of non-financial disclosure reduce information asymmetries between lenders and firms, thereby lowering the costs incurred by loan providers when evaluating firms' sustainability-related performance (e.g. Eliwa et al., 2021; Raimo et al., 2021; Maaloul et al., 2023). As a result, greater transparency is associated with lower perceived risk and, in turn, with reduced borrowing costs (e.g. Hamrouni et al., 2019; Maaloul et al., 2023; Alves and Meneses, 2024). Nevertheless, whether and how CIC disclosure affects firms' cost of debt has not yet been empirically investigated.

Against this background, the present study aims to fill this relevant research gap by advancing the understanding of CIC disclosure practices conveyed through social media and examining their implications for firms' cost of debt. Drawing on signalling theory, the study analyses a sample of 252 listed firms included in the MSCI Europe Index, as retrieved from the LSEG database (formerly Refinitiv/Thomson Reuters/Asset4). All posts published in 2024 on X (formerly Twitter) by the sampled firms were collected using Apify Tweet Scraper V2. A Large Language Model (LLM)-assisted content analysis was then employed to identify and classify posts containing CIC-related information, in line with the framework developed by Raimo et al. (2025). Specifically, a Python algorithm embedding the OpenAI API was developed to operationalise the conceptual definitions of the three CIC dimensions and systematically classify posts according to predefined semantic rules. Subsequently, two regression models were estimated to test the impact of CIC disclosure – proxied by two independent variables–on firms' cost of debt, measured by the debt component of the weighted average cost of capital (WACC) as computed by the LSEG database.

The decision to focus on social media is motivated by the pervasive diffusion of Web 2.0 digital technologies, which has triggered a profound transformation in accounting and reporting communication practices (Ramassa and Di Fabio, 2016; Arnaboldi et al., 2017; Hogarth et al., 2024). Within this evolving landscape, platforms such as Facebook, Twitter (rebranded as “X” in July 2023), and LinkedIn have enabled organisations to engage with stakeholders beyond the monologic and one-directional logic of traditional corporate reports, fostering more interactive, timely, and continuous communication patterns (Arnaboldi et al., 2017; Amin et al., 2021; Pizzi et al., 2024). Consistent with mainstream IC research, social media therefore represent a relevant communication channel through which firms can disseminate more accessible, cost-effective, and comprehensive information on IC resources whose value is often insufficiently captured by traditional annual reporting (Nicolò et al., 2021; Bryl et al., 2022; Schiuma et al., 2024).

The study's results show that European listed companies display heterogeneous levels of engagement in CIC-related communication via X. While some firms actively exploit this channel to disclose how IC supports circular value creation processes, others show limited or no CIC activity. More importantly, the empirical evidence reveals a negative and statistically significant association between CIC disclosure and the cost of debt. Specifically, firms that engage more intensively in CIC communication via X tend to benefit from lower borrowing costs, consistent with the predictions of signalling theory.

This study provides manifold contributions to several distinct streams of literature. First, it advances mainstream IC literature by providing hitherto undocumented evidence regarding how firms exploit social media platforms to disseminate information concerning the nexus between intangible assets and the circular economy. In doing so, it extends the foundational work of Raimo et al. (2025) by delivering empirical insights into the financial consequences of conveying CIC disclosure via digital media. Second, it extends prior research documenting that higher levels of non-financial disclosure–encompassing ESG (e.g. Hamrouni et al., 2019; Raimo et al., 2021; Eliwa et al., 2021; Alves and Meneses, 2024; Malik and Kashiramka, 2024; Nicolò et al., 2025a), CSR (e.g. Bhuiyan and Nguyen, 2020), circular economy (L'Abate et al., 2024a), and traditional IC disclosure (Orens et al., 2009) - are associated with a lower cost of debt, translating these dynamics to the specific context of CIC disclosure. Thirdly, this study expands the application scope of signalling theory. In this regard, it conceptualises the dissemination of CIC disclosure via X as a strategic signal deployed to mitigate informational frictions, ultimately facilitating more favourable financing terms in debt markets.

The remainder of this study is structured as follows. The next section presents a literature review on IC and GIC complemented by a brief discussion of prior research on the relationship between non-financial disclosure and the cost of debt. Section 3 is devoted to the theoretical background and development of hypotheses. Section 4 outlines the research methodology. Section 5 presents and discusses the findings, and the final section offers concluding remarks.

IC research has evolved through several distinct stages (Dumay and Garanina, 2013; Dumay et al., 2017, 2020; Dumay and Guthrie, 2019).

The first stage is rooted in practitioners' contributions during the 1980 and 1990s, which sought to raise awareness and develop a shared understanding of the role of IC in driving value creation and competitive advantage within organisations (Petty and Guthrie, 2000; Massaro et al., 2018). This phase was characterised by the development of taxonomies and guidelines aimed at “making the invisible visible”, thereby fostering the measurement and reporting of IC resources that were not recognised in traditional financial statements (Petty and Guthrie, 2000; Nicolò, 2020).

The second stage consolidated first-stage efforts within the academic domain (Dumay and Garanina, 2013; Secundo et al., 2017; Nicolò, 2020). During this phase, more than 50 IC classifications and measurement methods were developed and empirically tested, particularly in studies examining the extent of IC disclosure in annual reports (e.g. Bozzolan et al., 2003; Oliveira et al., 2006; Guthrie et al., 2006; Sonnier, 2008) and the relationship between IC, financial performance, and value creation (e.g. Firer and Williams, 2003; Chen, 2008).

While the first two stages were instrumental in establishing a shared IC terminology and providing a strong impetus to IC research–mainly from a measurement and reporting perspective–the third stage marked a shift towards a more critical examination of IC practices in action (Guthrie et al., 2012; Dumay and Garanina, 2013; Massaro et al., 2018). Accordingly, in this stage, a growing need emerged to unveil how managers and employees concretely leverage IC to support organisational processes (e.g. Cuganesan et al., 2007; Dumay and Chiucchi, 2013; Chiucchi and Dumay, 2015).

The fourth stage further extends the concept of IC beyond firm boundaries to encompass broader ecosystems, including countries, cities, and communities (Dumay and Garanina, 2013; Massaro et al., 2018; Secundo et al., 2017). As noted by Dameri and Ricciardi (2015, p. 681), this stage recognises that IC is pivotal not only for creating economic value at the organisational level, “but also, and even more importantly, for addressing the paramount ecological, social, and demographic challenges faced by contemporary societies”. In this perspective, scholars have increasingly emphasised the interconnections between IC and sustainability (e.g. Lopez-Gamero et al., 2011; Massaro et al., 2018), as well as the role of IC in enabling smart city governance and performance (e.g. Lombardi et al., 2012; Dameri and Ricciardi, 2015; Matos et al., 2017) and fostering national competitiveness and economic growth (e.g. Käpylä et al., 2012; Seleim and Bontis, 2013). Closely related to the fourth stage of IC research is the emergence of Green Intellectual Capital (GIC), defined by Chen (2008, p. 277) as “the total stocks of all kinds of intangible assets, knowledge, capabilities, and relationships related to environmental protection or green innovation at both the individual and organisational levels within a company”. Building on this conceptualisation, a distinct stream of research has developed, focussing on the antecedents and outcomes of GIC.

Overall, scholars emphasise that GIC constitutes a vital strategic asset that organisations leverage to mitigate environmental and operational costs, attract and retain talent, comply with increasingly stringent environmental regulations, and respond to emerging stakeholder expectations regarding ecological issues (Chang and Chen, 2012; Yusliza et al., 2020; Srouji et al., 2025). In doing so, GIC enables firms to unlock new market opportunities and advance sustainability-oriented competitive advantages (Chang and Chen, 2012; Yusliza et al., 2020; Srouji et al., 2025). Empirically, some scholars identified Corporate Social Responsibility (CSR) and environmental consciousness as drivers of GIC (e.g. Chang and Chen, 2012; Chaudhry et al., 2016; Sudibyo and Sutanto, 2020). A substantial body of empirical evidence has documented that GIC and its core components–green human capital, green structural capital, and green relational capital–positively contribute to firms' competitive advantage (e.g. Chen, 2008; Chaudry et al., 2016; Anik and Sulistyo, 2021; Naseem et al., 2024). Prior studies have also highlighted the significant role of GIC in enhancing multiple dimensions of corporate green performance (Wang and Juo, 2021; Shehzad et al., 2023; Ibrahim et al., 2026), as well as green innovation (Jirakraisiri et al., 2021; Wang and Juo, 2021; Ibrahim et al., 2026) and green human resource management practices (Yong et al., 2019). Also, both Rehman et al. (2021) and Shehzad et al. (2023) revealed that the impact of GIC on green performance is mediated by green innovation, while Asiaei et al. (2022) demonstrated that environmental management accounting acts as an intervening mechanism between GIC and environmental performance. In addition, Yusliza et al. (2020) provided evidence that GIC positively influences economic, ecological and social performance in Malaysian manufacturing firms, whereas Yusoff et al. (2019) found that GIC exerts a significant positive effect on the business sustainability of Malaysian SMEs. Attuned, Srouji et al. (2025) found that CSR amplifies the positive impact of GIC on the environmental sustainability performance of companies listed on the Amman Stock Exchange.

Only a limited strand of literature has examined GIC from a reporting perspective. In particular, Fawad et al. (2025) underscore that the effective deployment of GIC is crucial for Chinese listed firms to enhance transparency on Environmental, Social and Governance (ESG) issues, thereby signalling a concrete commitment to sustainability to their stakeholders. By contrast, Ferri et al. (2025) specifically analysed the level of GIC disclosure provided by Italian listed firms in their sustainability reports, showing that such disclosure is negatively associated with financial performance.

An emerging stream of research has built upon the concept of GIC to illuminate an evolving construct that captures the interconnections between intangible resources and circular economy practices (Farrukh et al., 2023; Raimo et al., 2025). The circular economy is commonly defined as “an economic model aimed at the efficient use of resources through waste minimisation, long-term value retention, reduction of primary resource inputs, and the establishment of closed loops for products, components, and materials, within the boundaries of environmental protection and socio-economic benefits” (Morseletto, 2020, p. 1). Its operationalisation entails substantial efforts at the business level, as firms are required to adopt regenerative and restorative solutions that enable the design of closed-loop systems for the use and management of materials, resources, and products, thereby extending their presence in the market for as long as possible (Urbinati et al., 2017; Barnabè and Nazir, 2020; Roberts et al., 2023). Thus, grounded in the three core principles of reuse, reduce, and recycle, the circular economy entails a profound redefinition of corporate vision, strategies, and operational processes aimed at minimising environmental impacts–particularly in terms of emissions and waste generated through the production of physical goods–thereby offering a concrete contribution to sustainable development (Morseletto, 2020; Sartal et al., 2020). Accordingly, scholars (Khan et al., 2021; Farrukh et al., 2023; Raimo et al., 2025) have highlighted that IC and its green derivation may represent a strategic lever for supporting the effective implementation of circular economy–based systems within organisations. Specialised knowledge, green-related expertise, robust environmental management systems and data infrastructures, innovative green technologies, and strategic relationships and partnerships with external stakeholders constitute essential intangible resources which, if adequately managed, can enable companies to fully integrate circular economy principles into their business models (Raimo et al., 2025). For this reason, Raimo et al. (2025) introduced the concept of Circular Intellectual Capital (CIC) to identify and systematise the pool of intangible resources that facilitate sustainable and circular practices. The CIC framework is articulated around three interrelated components: circular human capital, circular organisational capital, and circular relational capital. Circular human capital refers to the collective reservoir of employees' knowledge, skills, competencies, experience, mindsets, creativity, and commitment that are specifically oriented towards the adoption and implementation of circular economy principles and practices. Circular organisational capital encompasses the set of internal structures, routines, processes, governance mechanisms, and information systems that enable the organisation to embed circular economy principles into its operations and strategic decision-making. Finally, circular relational capital captures the network of relationships, partnerships, and collaborative arrangements with external stakeholders that support the firm's circular transition (Raimo et al., 2025).

Building on this conceptualisation, Raimo et al. (2025) examined the extent of CIC disclosure provided by a sample of Italian listed companies through their non-financial reports, as well as its main determinants. Their findings reveal that, although Italian firms are still lagging behind in systematically integrating CIC-related information into non-financial reporting, larger and more profitable companies are significantly more inclined to disclose such information. However, this stream of research is still in its embryonic stage. Accordingly, this paper seeks to contribute to the emerging literature by offering novel insights into CIC disclosure practices from an innovative and comparative perspective. More specifically, the study advances academic understanding of whether and how companies included in the MSCI Europe Index disclose CIC-related information through their official X accounts, and examines the impact of such disclosure on firms' cost of debt.

Signalling theory represents one of the most influential theoretical frameworks in accounting research for explaining firms' incentives to disclose financial and non-financial information beyond mandatory regulatory requirements (Spence, 1973; Ross, 1977; Uyar et al., 2020). The theory is grounded in the concept of information asymmetry, which arises when one party possesses superior information, in terms of either quantity or quality, relative to another (Spence, 1973; Mahoney et al., 2013; Uyar et al., 2020). In credit markets, information asymmetry gives rise to adverse selection problems, whereby lenders are unable to perfectly distinguish between high-quality (low-risk) and low-quality (high-risk) borrowers. This friction impairs their capacity to accurately price the cost of debt (Leland and Pyle, 1977; Dhaliwal et al., 2011; Guidara et al., 2014), given that sound lending decisions rely on a rigorous evaluation of borrowers' creditworthiness, including the factors dictating their repayment capacity and default likelihood (Sengupta, 1998; Raimo et al., 2021; Malik and Kashiramka, 2024). Consequently, heightened information opacity and a lack of corporate transparency hamper this evaluation process, increasing the probability that a borrower is misclassified as high-risk (Guidara et al., 2014; Malik and Kashiramka, 2024). This pervasive uncertainty prompts lenders to safeguard their capital by demanding higher risk premia, imposing stricter contractual restrictions, and introducing tighter covenants, thereby driving up the overall cost of debt (Hamrouni et al., 2019; Raimo et al., 2021; Nicolò et al., 2025a).

According to signalling theory, voluntary disclosure serves as a strategic signal that companies (borrowers) rely upon to enhance their credibility and highlight their performance, thereby reducing information asymmetries vis-à-vis lending institutions (Ross, 1977; Guidara et al., 2014). Particularly, non-financial information has attracted growing attention in debt markets, as factors such as climate change–related risks and opportunities, circular economy practices, biodiversity preservation, and waste management increasingly constitute integral components of banks' assessments of borrowers' creditworthiness, given their impact on firm's future cash flow (Agnese and Giacomini, 2023; Alves and Meneses, 2024). Consequently, aligned with signalling theory, firms face pronounced incentives to deliver higher-quality and more extensive disclosures that transparently demonstrate their commitment to sustainability. Such transparency is aimed at conveying a positive corporate image and securing debt capital at lower costs (An et al., 2011; Uyar et al., 2020). Increasing the volume and breath of non-financial disclosure mitigates the information-processing costs incurred by financial institutions when evaluating borrowers' sustainability performance, alongside their broader implications for cash flows and default risk. Ultimately, this reduction in informational frictions manifests in more favourable interest rates (Orens et al., 2009; Hamrouni et al., 2019; Nicolò et al., 2025a). Several studies provide empirical support for these arguments, documenting that higher levels of non-financial disclosure–including ESG (e.g. Hamrouni et al., 2019; Raimo et al., 2021; Eliwa et al., 2021; Alves and Meneses, 2024; Malik and Kashiramka, 2024; Nicolò et al., 2025a), CSR (e.g. Bhuiyan and Nguyen, 2020), circular economy (L'Abate et al., 2024a), and IC disclosure (Orens et al., 2009) - are associated with a lower cost of debt.

As stated by Sengupta (1998, p. 459), “a policy of timely and detailed disclosure reduces lenders' and underwriters' perception of default risk for the disclosing firm, thereby reducing its cost of debt”. In this regard, social media represent effective communication tools that enable firms to disseminate non-financial information in a more timely, accessible, and cost-efficient manner (Bryl et al., 2022; L'Abate et al., 2024b; Nicolò et al., 2025b). The emergence of social media has disrupted the traditional corporate reporting landscape, creating opportunities for a more interactive and dynamic dialogue between firms and a wider range of stakeholders at minimal marginal cost (Lodhia et al., 2020; Nicolò et al., 2025b). Accordingly, based on Web 2.0 technologies, these platforms have facilitated a paradigm shift from one-way transmissional models of communication to two-way transactional patterns that enable immediate stakeholder engagement across a wide range of topics, including IC, circular economy practices, and other non-financial issues (Lodhia et al., 2020; L'Abate et al., 2024b; Nicolò et al., 2025b).

The potential of social media is particularly relevant for IC resources, whose economic value and internal composition are largely overlooked in traditional annual reports due to the restrictive recognition criteria imposed by international accounting standards such as IAS/IFRS (Sonnier, 2008; Nicolò et al., 2021). For this reason, social networking platforms such as X may help overcome these limitations by allowing firms to signal how they exploit IC resources to create value while preserving the ecosystem through the implementation of circular economy practices, thereby differentiating themselves from lower-quality competitors (L'Abate et al., 2024a; Schiuma et al., 2024). In particular, by providing more timely and accessible information via X on how IC intersects with circular economy principles within value creation processes, firms may reassure banks and debt-holders about their financial stability and their capacity to meet both short- and long-term financial commitments while contributing to sustainable development (Orens et al., 2009; L'Abate et al., 2024a; Nicolò et al., 2025a). This, in turn, may enable them to secure more favourable borrowing conditions, resulting in a lower cost of debt.

Therefore, based on the above theoretical arguments and broader empirical evidence, the following hypothesis is posited:

This study relies on listed companies included in the Morgan Stanley Capital International (MSCI) Europe Index, which have an active X profile for the year 2024. This Index encompasses approximately 85% of the free float-adjusted market capitalisation of the European developed equity market and is extensively employed in academic research for cross-country and cross-industry investigations (e.g. Hampl and Vágnerová Linnertová, 2025; Nicolò et al., 2025a). As such, it offers a broad and heterogeneous landscape of well-established firms operating within the European setting, which are comparatively more inclined to implement sustainability disclosure practices via different communication tools – including social media–and benefit from enhanced access to capital and debt markets (Hampl and Vágnerová Linnertová, 2025; Nicolò et al., 2025a).

The dataset was retrieved from the LSEG database (formerly Refinitiv/Thomson Reuters/ASSET4), one of the most widely used and reliable databases in accounting research and one of the largest providers of financial and non-financial corporate and market data (Refinitiv, 2022), serving more than 40,000 customers and 400,000 end users across 190 countries (Pucheta-Martínez and Gallego-Álvarez, 2019; Alves and Meneses, 2024; Nicolò and Andrades-Peña, 2024).

Accordingly, the sampling process unfolded in several stages (see Table 1). First, the initial sample comprised 472 companies listed in the MSCI Europe Index available in LSEG. Second, companies without an official account (113), with a non-public or protected profile (15), with duplicated profiles (7), or with no posts on their X profile during 2024 (9) were excluded. Third, 76 companies with missing LSEG data for the dependent and control variables were removed. After this screening process, the final sample consists of 252 companies from 16 countries, as shown in Table 2.

The dependent variable of this study is the marginal Cost of Debt (CoD). CoD reflects the interest rate required by lenders and thus the cost a company incurs when raising external debt financing (Petruzzella et al., 2026). Since debt may originate from either public or private channels, and “in either case, the cost of debt is the applicable interest rate” (Sharfman and Fernando, 2008, p. 572).

Based on previous studies (Alves and Meneses, 2024), CoD is retrieved from the LSEG database, which provides standardised and internationally comparable financial metrics. Using the marginal cost of debt is methodologically advantageous, as it captures the immediate valuation effects of shifts in firm disclosure levels more rapidly than historical average costs (Alves and Meneses, 2024). Specifically, the CoD reflects the incremental cost a firm incurs when issuing new debt. LSEG calculates this metric by combining the weighted cost of short-term debt and the weighted cost of long-term debt, mapped across the 1-year and 10-year points of an appropriate credit curve. Structurally, it corresponds to the pre-tax debt component of the WACC, as computed by the LSEG platform (Alves and Meneses, 2024).

An automated content analysis assisted by a Large Language Model (LLM) (Siano, 2025) was adopted as a research method to assess the level of CIC disclosure provided by sampled companies through X. Compared with other social media platforms, X shows one of the highest corporate adoption rates for information disclosure, with approximately 321 million monthly active users (Zhang, 2015; Amin et al., 2021; Bryl et al., 2022). This makes it a leading platform for professional communication, including non-financial disclosure practices (Bryl et al., 2022; L'Abate et al., 2024b; Nicolò and Cervilla-Bellido, 2025). In addition, most content on X is public by default, facilitating open conversations and the exchange of opinions in the public domain (Nicolò et al., 2025b). Moreover, its relatively strict character limit encourages concise and clear communication, thereby potentially reducing ambiguity and limiting opportunities for greenwashing (Bryl et al., 2022; Petruzzella et al., 2024; Nicolò et al., 2025b). These features make X an ideal social media platform alternative to traditional annual and sustainability reports to investigate corporate CIC disclosure practices.

Prior literature has largely employed a manual approach to content analysis for searching, collecting and classifying IC information via annual reports, websites and social media, and for building related disclosure indices (e.g. Striukova et al., 2008; Duff, 2018; Nicolò et al., 2021; Schiuma et al., 2024). However, in recent years, semi-automated and automated content-analysis procedures have emerged to mitigate the limitations of the manual approach, mainly related to subjectivity, coding complexity and its time-consuming nature (Chakraborty and Bhattacharjee, 2020; Frankel et al., 2022). These include dictionary-based natural language processing (NLP) tools and machine learning (ML) approaches, as well as more recent generative methods (Chakraborty and Bhattacharjee, 2020; Shimamura et al., 2025). NLP and ML approaches–such as bag-of-words models, topic modelling, and supervised classifiers–allow large volumes of text to be processed at scale while limiting human intervention (Lewis and Young, 2019; Damiano and Picciotto, 2025). However, these methods also present notable limitations when applied to non-financial disclosure, given its semantic complexity, contextual nuance, and heterogeneous terminology (Huang et al., 2023; Siano, 2025). In essence, they typically rely on word frequencies and treat terms largely in isolation from their position within the document, which weakens their ability to recover contextual relationships (Loughran and McDonald, 2016; Siano, 2025). Moreover, they often reduce textual content to single-dimensional attributes (Loughran and McDonald, 2016), overlooking rhetorical intent, latent meaning, and interdependencies among concepts, including those related to non-financial issues (Shimamura et al., 2025; Siano, 2025).

In contrast, as deep neural networks trained on extensive text corpora, LLMs can perform a wide range of natural language tasks, learning semantic and syntactic relationships among words while capturing both distributional and context-dependent meaning across large volumes of text (Shimamura et al., 2025; Siano, 2025). LLMs use distributional representations (embeddings) together with self-attention mechanisms to model semantic similarity and long-range contextual dependencies, thereby mitigating many limitations of frequency-based or dictionary-driven representations (Frankel et al., 2022). When properly prompted, configured, and validated, they can enable more accurate and granular characterisation of disclosure texts, especially in dynamic communication environments such as social media, which are characterised by high informational complexity (Huang et al., 2023).

With these premises, this study adopted an LLM-assisted automated content analysis approach (Siano, 2025) to build two independent variables representing the level of CIC disclosure provided by sampled companies via the X social media platform. The process unfolded through several steps (Figure 1).

First, Apify Tweet Scraper was used as a web-scraping tool to collect all posts published in 2024 by the sampled companies. In line with prior studies (L'Abate et al., 2024b; Schiuma et al., 2024; Nicolò et al., 2025b), only original corporate posts were retained, while reposts were excluded because they cannot be treated as official corporate disclosure. This process yielded an initial dataset of 95,473 posts. The unit of analysis was the individual textual post (Krippendorff, 2004; Bryl et al., 2022). Accordingly, posts containing non-textual content (e.g. images, videos, or external links) (36,489), as well as posts containing only mentions and hashtags (16,849), were excluded (Bryl et al., 2022; Nicolò et al., 2025b). As a result, the final dataset comprised 42,135 posts (see Table 3).

A Python-based computational algorithm embedding the OpenAI API (GPT-4o architecture) was developed to conduct an automated semantic content analysis of the sampled firms' posts and systematically evaluate CIC disclosure. The classification protocol was anchored in the theoretical CIC framework proposed by Raimo et al. (2025), which was operationalised as a rigorous coding instrument comprising 36 items distributed across three core dimensions: circular human capital, circular organisational capital, and circular relational capital, each anchored to its respective conceptual definition (see  Appendix). To guarantee methodological consistency and transparency, a predefined set of semantic and prompt-engineering rules was integrated into the algorithm. These rules provided explicit operational criteria for identifying key CIC concepts, classifying textual data, and generating justification rationales for each coding decision in strict compliance with the established framework (see Table 4). Crucially, these operational guidelines demarcated the semantic boundaries of each CIC dimension, enabling the model to accurately differentiate between adjacent or conceptually overlapping categories.

To validate the model's reliability, several pilot tests were conducted on random subsamples of posts, during which the authors manually verified the coherence of the LLM outputs against the coding framework and the predefined semantic rules. This validation process followed an iterative approach, enabling the continuous refinement of prompt constraints and the recalibration of the algorithm whenever ambiguous or borderline classifications emerged. This iterative auditing established a systematic mechanism to evaluate the convergence and agreement between human expert coding and the AI-driven classification protocol, thereby safeguarding the overall robustness and replicability of the textual analysis.

Recognising that a single social media post could encapsulate more than one CIC disclosure item, consistent with Bryl et al. (2022), a multi-label classification approach was adopted. Accordingly, a specific “CIC-mention” logic was operationalised, whereby each distinct thematic reference to a CIC item within a post was codified as an individual mention. This granular approach allowed the algorithm to systematically identify and capture multiple references to different IC items within a single textual unit. Consequently, the aggregate number of CIC mentions could exceed the total volume of CIC-related posts, thereby accurately reflecting the multi-faceted and dense nature of non-financial communication on digital platforms.

At the end of the process, the algorithm returned for each company, the set of posts classified as CIC-related and the corresponding CIC mentions by dimension. Overall, 16,278 CIC-related posts and 35,483 CIC-related mentions were identified and classified according to the items included in the coding framework developed by Raimo et al. (2025). As a result, two independent variables were derived: (1) CIC disclosure (CICD), measured as the natural logarithm of the number of posts including information on one or more CIC items; and (2) CIC mentions (CICM), proxied by the natural logarithm of mentions of CIC items contained in posts.

Consistent with prior research, a set of control variables was incorporated in this study to strengthen the reliability and goodness of the econometric analyses.

In particular, the following control variables were taken into account. SIZE is a measure of firm dimension and is computed using the natural logarithm of total assets (Hamrouni et al., 2019; Eliwa et al., 2021). LEV captures the firm's degree of indebtedness and is operationalised as the ratio of total debt to total assets (Hamrouni et al., 2019; Eliwa et al., 2021). ROE is a measure of firm profitability and is calculated as the ratio of net income to shareholders' equity (Michelon and Parbonetti, 2012; Malik and Kashiramka, 2024). B_SIZE captures board size and is operationalised as the total number of directors serving on the board in a given fiscal year (Michelon and Parbonetti, 2012; L'Abate et al., 2024b). B_IND reflects the degree of board independence and is calculated as the proportion of independent directors relative to the total number of board members (Bhuiyan and Nguyen, 2020; Michelon and Parbonetti, 2012). CEO_D is a dummy variable that takes the value of 1 if the CEO simultaneously serves as Chairperson of the board, and 0 otherwise (Michelon and Parbonetti, 2012; Nicolò and Andrades-Peña, 2024). ESG_S is a composite measure capturing a firm's performance across environmental, social, and governance dimensions (Nicolò and Andrades-Peña, 2024). Finally, ENV_S is a dummy variable equal to 1 if the firm operates in an environmentally sensitive industry, and 0 otherwise (Branco and Rodrigues, 2008; Nicolò et al., 2025b).

Table 5 resumes the name, definition, description and source of dependent, independent and control variables.

This study uses two Ordinary Least Squares (OLS) regression models to test the research hypothesis. As reflected in Equation (1), the first model includes the independent variable CICD, while the second model, as reflected in Equation (2), includes the independent variable CICM.

Model 1)

Model 2)

Table 6 shows the descriptive statistics for dependent, independent, and control variables.

Regarding the dependent variable, CoD has a mean of 0.029 (2.9%), with values ranging between 0 and 0.05 (approximately 5%), indicating some dispersion in borrowing costs across firms.

With respect to the independent variables, CICD has an average of 64.6, with values ranging from 0 to 192. Similarly, CICM shows a mean of 140.81, ranging from 0 to 473. For the sake of clarity, both CICD and CICM are reported in descriptive statistics in absolute terms, representing the raw number of CIC-related posts and mentions. The wide dispersion observed for both proxies indicates that firms differ substantially in the intensity of CIC-related communication through X, with some companies showing no CIC activity and others engaging extensively.

Concerning industry-related characteristics, ENV_S indicates that approximately 49% of the sampled firms operate within environmentally sensitive sectors, while the remaining 51% belong to non-environmentally sensitive sectors.

Looking at firm-level characteristics, SIZE has a mean of approximately 24.3, with values ranging from about 20.5 to 28.7. Also, ROE has an average of 0.134 (13.4%), while LEV exhibits a mean value of 0.25 (25%).

Turning to governance-related variables, B_SIZE average values suggest that boards are, on average, composed of around twelve directors and include approximately 67% independent directors. Furthermore, in 19% of the sampled companies, the CEO also serves as chair of the board. Finally, ESG_S shows a mean value of about 72.9, indicating that, on average, companies achieve a relatively high sustainability score, although a high dispersion exists within the sample.

Table 7 presents the correlation matrix for the dependent and independent variables.

All correlation coefficients among the independent variables remain below the conventional threshold of 0.8, with the exception of CICD and CICM (0.985), which represent the two main independent variables included in different regression models. So, correlation analysis outcomes exclude potential multicollinearity drawbacks (Mertens et al., 2017; Gujarati et al., 2020).

Table 8 presents the results of the two OLS regression models estimated to test Hypothesis 1.

In both models, the F-statistics are statistically significant (p < 0.01), and the adjusted R² values equal 16.4 and 16.3, respectively, suggesting an adequate explanatory power. Also, all VIF values are well below the conventional threshold of 10, with the highest VIF equal to 1.683 (B_SIZE), thus confirming the absence of multicollinearity concerns. Furthermore, White's heteroscedasticity test pinpoints p-values that are not significant, thus eliminating heteroscedasticity problems.

The results of the OLS regression models indicate a negative and statistically significant association between both independent variables representing CIC disclosure and CoD. Specifically, in Model (1), the coefficient of CICD is negative and significant (p < 0.01), and in Model (2), CICM also exhibits a negative and highly significant coefficient (p < 0.01). These findings support Hypothesis 1, indicating that higher levels of CIC disclosure provided via social media led to a reduction in companies' cost of debt. They are also consistent with prior literature documenting that higher levels of non-financial disclosure are associated with a lower cost of debt (e.g. Hamrouni et al., 2019; Raimo et al., 2021; Eliwa et al., 2021; Alves and Meneses, 2024; Malik and Kashiramka, 2024; Nicolò et al., 2025a).

This finding aligns with signalling theory, which posits that voluntary disclosure can mitigate information asymmetries in credit markets, thereby yielding financial benefits through a reduced cost of debt. In contexts where lenders cannot perfectly observe borrowers' underlying risk profiles or their capacity to generate sustainable future cash flows, transparency extending beyond traditional financial metrics becomes particularly critical. In the absence of a comprehensive informational package, lenders may struggle to assess a firm's sustainability-related performance and its direct link to future earnings and cash flows. This heightened uncertainty increases the perceived risk of lending capital, leading to adverse effects such as stricter contractual restrictions and covenants, which ultimately boost the cost of debt. Consequently, reliable and timely non-financial disclosures are vital for lenders to construct a holistic view of creditworthiness, capturing the underlying factors that influence loan repayment capabilities and default likelihood.⁹

In this context, CIC disclosure can operate as a credibility-enhancing signal that reduces informational opacity and alleviates the adverse selection problems inherent in credit markets. This transmission mechanism occurs because debt-holders increasingly incorporate sustainability initiatives and circular business models into their formal risk appraisals. In the absence of such signals, lenders–unable to accurately differentiate between high-risk and low-risk borrowers–typically implement protective measures, such as demanding higher default risk premiums. Consequently, by voluntarily communicating how IC is deployed to support and enhance circular economy practices, firms provide critical insights into their capacity to meet both short- and long-term financial commitments while simultaneously contributing to sustainable development. In this sense, unpacking how a firm leverages its IC to promote circularity affords lenders a comprehensive understanding of its organisational adaptability, operational resilience, and long-term viability. Accordingly, debt-holders can conduct a more precise evaluation of their borrowers' creditworthiness and forward-looking risk exposure, systematically accounting for all sustainability-driven strategic levers and their respective financial implications.

In this scenario, the observed negative association between CICD/CICM and CoD, underscores the pivotal role of social media platforms in fostering continuous dialogue between corporations and their stakeholders, including lenders. By leveraging widely accessible public platforms such as X, firms can dynamically signal their commitment to circular practices aimed at retaining the maximum value of products, components, and materials, thereby mitigating waste and emissions. Crucially, this communication channel operates free of geographical constraints and entails negligible transactional costs. The resulting heightened transparency enhances corporate reputation while simultaneously reducing uncertainty and risk perception among lenders. Consequently, financial institutions are incentivised to extend more favourable financing conditions. Therefore, in line with theoretical expectations, CIC disclosure via X functions as a strategic signal through which companies enhance their credibility and mitigate informational asymmetries, thereby reaping tangible financial benefits in the form of a reduced cost of debt.

Regarding control variables, LEV is positively and significantly associated with CoD (p < 0.05), indicating that more leveraged firms face higher borrowing costs, consistent with the view that higher debt exposure increases perceived default risk and amplifies lenders' concerns regarding firms' ability to meet their financial obligations. Also, SIZE exhibits a positive and slightly significant coefficient (p < 0.1), suggesting that, within the sample, larger firms do not necessarily benefit from lower borrowing costs after controlling for other characteristics. One possible interpretation of this result is that larger firms may be exposed to more complex risk structures, which could offset potential size-related advantages in debt markets.

Governance-related variables further contribute to explaining variations in CoD. B_SIZE is negatively and significantly associated with CoD (p < 0.01), indicating that firms with larger boards tend to face lower borrowing costs. This finding may evidence that larger boards ensure an enhanced monitoring capacity and stronger oversight mechanisms that allow for mitigating information asymmetries, thereby increasing lenders' confidence in managerial decisions. Similarly, B_IND shows a negative and marginally significant association with CoD (p < 0.1), pinpointing that a higher proportion of independent directors improves governance quality and transparency, mitigating perceived risk. Additionally, CEO_D is negatively and significantly related to CoD (p < 0.1 and p < 0.05). This finding may underline that, rather than signalling governance weakness, leadership unity facilitates more coherent strategic decision-making and communication, reducing uncertainty from the lenders' perspective.

Finally, results indicate that ROE, ESG_S, and ENV_S do not exert any statistically significant influence on CoD, suggesting that short-term financial performance does not materially affect lenders' pricing decisions once leverage, governance characteristics, and disclosure-related variables are taken into account.

This study provides an innovative contribution to the IC and circular economy literature, as well as to the stream of research examining the financial implications of non-financial disclosure. Accordingly, it offers fresh insights into how MSCI Europe companies leverage the potential of social media to communicate how intangible assets support circular economy practices, ultimately shedding light on the implications of such disclosure for the cost of debt.

The findings indicate that European listed companies devote particular attention to disseminating CIC-related information via social media. However, heterogeneous patterns emerge: while some firms are highly active on X, others still lag behind in leveraging its communicative potential for CIC disclosure purposes.

The study's empirical findings indicate that higher levels of CIC disclosure via social media are significantly associated with a lower cost of debt. This evidence suggests that CIC disclosure operates as an informational signal that enhances lenders' capacity to conduct a holistic assessment of firms' risk profiles, thereby directly influencing borrowing conditions. In this manner, disclosures regarding how a firm deploys its intangible assets to enable circular practices, mitigate waste, and curb emissions complement traditional financial accounts, providing debt-holders with a more comprehensive overview of corporate performance. This heightened transparency mitigates informational asymmetries and reduces the likelihood of borrowers being perceived as high-risk, ultimately creating the economic latitude to secure more favourable interest rates. Within this framework, social media platforms such as X emerge as pivotal communication tools that allow firms to disseminate their CIC-related performance in an accessible, timely, and cost-effective manner, thereby facilitating lenders' ongoing evaluation of corporate creditworthiness.

From a theoretical perspective, the study advances IC research by providing empirical evidence that integrating circular economy principles into IC frameworks is not only conceptually meaningful but also economically consequential. It positions CIC as a financially material construct in debt markets, extending prior research that has provided empirical evidence documenting the financial effects of ESG, CSR, and IC disclosure. Moreover, by contextualising signalling theory within circular economy communication, the study deepens understanding of how disclosures concerning the nexus between IC and circular economy can shape capital allocation decisions.

The findings also entail significant practical implications for corporate managers, policymakers, standard-setters, and academic institutions. First, they suggest that providing extensive disclosures on how IC supports circular economy strategies is not merely a reputational exercise, but a vehicle for generating tangible financial returns. Companies that align their circular economy initiatives with coherent communication practices can strengthen their credibility in debt markets, thereby securing more favourable borrowing conditions. Consequently, these dynamics underscore the operational need to appoint or train qualified personnel, such as social media managers, who possess the specific expertise required to design targeted communication strategies. These strategies must satisfy stakeholders' informational demands while avoiding the creation of redundant or purely promotional content. To achieve this, social media managers should work in close coordination with environmental or sustainability managers, sustainability committee members, production executives, and corporate accountants to collectively calibrate the appropriate scope and quality of CIC disclosure disseminated via digital platforms.

Furthermore, the study's findings offer valuable insights for policymakers and regulatory bodies, emphasising the need to encourage firms to leverage social media channels for sustainability disclosure. Promoting the dissemination of information regarding how intangible assets intersect with circular economy practices can significantly optimise corporate dialogue with investors and wider stakeholder groups, ultimately enhancing the informational efficiency and overall functioning of capital markets. In this vein, it is crucial to underscore that social media platforms should be conceptualised as complementary to, rather than substitutes for, traditional financial and sustainability reports. Consequently, policymakers and regulators should actively encourage firms to leverage social media as channels through which to provide value-relevant and material information. This includes insights that are either not mandatorily required in conventional corporate reports, or that can be communicated in a less complex, more accessible, and familiar manner via publicly available tools–as is precisely the case for CIC disclosures.

Lastly, the study offers significant insights for lenders and financial institutions. In an economic landscape increasingly shaped by transition risks and mounting social and regulatory pressures to adopt sustainability-oriented practices, CIC disclosure can serve as a vital complementary information source. Integrating these disclosures into traditional credit risk appraisals enables financial institutions to conduct a more holistic assessment of borrowers' long-term operational resilience and their overall exposure to environmental and strategic risks.

Notwithstanding these contributions, the study has limitations that open avenues for future research. First, the analysis focuses on European listed companies included in the MSCI Europe Index and on a single year of observation. Future studies could extend the time horizon and conduct longitudinal or cross-country comparative analyses to assess the persistence and generalisability of the relationship. Second, although the LLM-assisted approach enhances scalability and semantic sensitivity, classification outcomes remain contingent on the operationalisation of the coding framework and embedded rules. Future research could compare alternative AI architectures and/or adopt more hybrid human–AI validation procedures to further strengthen measurement reliability. Third, the study relies on a cost-of-debt measure provided by the LSEG database. Given that multiple proxies for the cost of debt exist, future studies could employ alternative measures from other data providers to triangulate results.

The authors whose names are listed above certify that they have NO affiliations with or involvement in any organisation or entity with any financial interest (such as honoraria; educational grants; participation in speakers' bureaus; membership, employment, consultancies, stock ownership, or other equity interest; and expert testimony or patent-licencing arrangements), or non-financial interest (such as personal or professional relationships, affiliations, knowledge or beliefs) in the subject matter or materials discussed in this manuscript.