Accountingisation and the narrative (re)turn of business model information in corporate reporting

Purpose – The debate about whether corporate reports should focus on numbers or narrative is long- standing. The recent push for business model information to be included in corporate reports has revitalised the debate. Many scholars suggest this constitutes a move towards narrative-based reporting. This study aims to investigate the debate and draws a comparison with the juxtaposition of the narrative and rational paradigms. This study also investigates how accountingisation in ﬂ uences the way business model information is presented in corporate reports. Design/methodology/approach – This study analyses data from the ﬁ nancial and non- ﬁ nancial reports from 86 globally listed companies. This study ﬁ rst uses content analysis to code the data. This study then uses a partial least squares-structural equation model to test howaccountingisation in ﬂ uences how ﬁ rms reporttheirbusinessmodel information. Findings – This study ﬁ nds that accountingisation and a rational paradigm shape how companies present information about their business model in their ﬁ nancial and non- ﬁ nancial reports. This suggests that the dominance of quantitative measures in accounting affects even the presentation of narrative-based information. Despite the much-touted shift towards qualitative reporting, this study argues that companies ﬁ ndit dif ﬁ cultto cast offtheyoke of a traditionalnumbers-based mindset. Research limitations/implications – This paper contributes to the debate on numbers- versus narrative-based corporate reporting and the workings of narrative and rational paradigms. In it, this study lays out theoretical and empirical ﬁ ndings of accountingisation. This study also makes a case for freeing corporate reportsfrom theshacklesof anaccountingisation mindset. Originality/value – This study provides new insights into how companies report information about their business models and the in ﬂ uence of narrative and rational paradigms on ﬁ nancial and non- ﬁ nancial reporting. the paradigms behind BM information in corporate ﬁ nancial and non- ﬁ nancial reports. We analyse how companies communicate their BM information in their corporate reports.

coined the expression "accountingisation" to express the phenomenon of reframing organisational self-understanding in accounting terms, thus subverting existing value systems. Similarly, Jones and Dugdale (2001, p. 38) use the term "accounting regime", meaning "a distinctive body of knowledge providing discursive representations and vocabularies that inform and are informed by, actors" perceptions of the world. We contend that accountingisation or an "accounting regime" approach expresses the rational paradigm.
In this study, we examine how accountingisation influences BM information in corporate reporting. Our analysis reveals the narrative and rational paradigms at work in different corporate reporting practices. Accounting, as a language, is an essential framework that actors draw upon to interpret economic events (Englund et al., 2011). Further, it is a language that inherently implies quantifying objectives (Robson, 1992) and an integral part of reporting practice. According to Järvinen et al. (2020), quantification implies "fake precisionism" and "limits the discussion to themes and questions preferred by company management". As a result, BM information in corporate reports can fall into an accountingisation trap, whereby numbers are privileged over narrative (Dumay and Roslender, 2013). We are interested in testing whether this consistently happens and, if so, what the implications of a focus on numbers might be.
The research design uses a two-step content analysis methodology that combines text analysis with a partial least squares-structural equation model (PLS-SEM). PLS-SEM is a standard multivariate analysis method for analysing complex models given either a priori knowledge or based on a theory S anchez, 2013). It is well-suited to smaller sample sizes. The corpus for the text analysis comprised the financial and nonfinancial reports of 86 leading global companies. The hypotheses tested posit that different paradigms explain how BM information is presented in financial and non-financial reports. These are developed in Section 2.3.
This study contributes to the debate about improving traditional corporate reporting by digging into the numbers versus narratives question. Dumay and Rooney (2016, p. 227) conclude that "balancing the use of numbers or narrative is a matter of choice based on the type of knowledge to be transferred, the actor's epistemological viewpoint or the directions of a more powerful actor". Although report preparers use different paradigms as they produce financial and non-financial reports, we demonstrate that they are inclined to quantify BM information.
Our findings reveal both the reporting approaches and the organisations' changes resulting from non-financial reporting practices. We find that, despite pressure from scholars and practitioners, the corporate practice has not ultimately shifted towards narrative-based reporting. Accountingisation has deep roots in most organisations, influencing actors who quantitatively describe reality (Martin-Sardesai et al., 2020a). As a result, we contend that accountingisation prevents non-financial reporting from developing narratives aimed at a broader category of stakeholders.
Our paper begins by setting out the context of this study, followed by hypothesis development (Section 2). Next, we describe our research methods (Section 3). Our results follow in Section 4. Section 5 concludes our paper by discussing the results and their theoretical and practical implications.
2. Literature, theory and hypothesis development 2.1 Business model information in corporate reporting BMs became popular in the mid-1990s on the cusp of the digital era as a way to explain how organisations create, deliver and capture value (Osterwalder and Pigneur, 2010). Despite initial scepticism, BMs gradually entered both the corporate reporting lexicon Page (2014) and the research agendas of scholars. Many studies to date have revealed close ties between corporate reporting, communication practices and transparency. Together, these form a blueprint for how companies do business (Nielsen, 2005). However, because corporate reporting exists to explain how businesses create value, scholars have argued that corporate reporting needs to revolve around a firm's BM (Michalak et al., 2017).
The assumed importance of BMs has led several corporate reporting bodies to promote initiatives and frameworks that guide how information on a firm's BM should be reported (European Parliament, 2014;Financial Reporting Council, 2010;IIRC-International Integrated Reporting Council, 2013). For example, the IIRC's <IR> framework positions BMs at the centre of the six capitals (i.e. financial, manufactured, intellectual, human, social and relationship and natural capital). These represent stocks of value that underlie the value creation process of firms by highlighting the connections between financial and non-financial elements (IIRC-International Integrated Reporting Council, 2013;Simnett and Huggins, 2015;Tweedie et al., 2018). Reporting on their BM allows companies to share information about their future nonfinancial value drivers and plansin contrast to backwards-looking traditional financial reports. Because including information about a firm's BM is a relatively new development in corporate reporting, we expect the content, shape and structure of corporate reports to change as time passes (Michalak et al., 2017).
The growing interest in BMs has revived the long-standing numbers-narratives reporting debate in accounting research (Roslender and Nielsen, 2017). The term 'narrative'which Mouritsen et al. (2002) define as a story of how things work, what a firm's identity is and what the reasons for its activities are -first appears in (Adelberg, 1979, p. 186). Adelberg argues that accountants should include business narratives in financial reports to communicate the "economic and social information about the events and conditions that have affected, are currently affecting, and will affect the business entity in the immediate future. Afterwards, scholars have also argued that including business narratives in reports would radically improve corporate reporting quality (Beattie et al., 2004). Narrative reporting represents a way to tackle the limits of number reporting as another and more comprehensive way to share information about how a firm creates value.
Despite these arguments favouring narrative reporting, it has not caught on and reporting quantitative information has remained the norm in accounting practice (Carnegie and West, 2005). However, scholars have continued to insist that stakeholders require a range of information about companiesinformation that may go beyond financial valuation, even though it has financial implications (Hopper, 2019). For example, financial reports that only contain quantitative information cannot answer questions about a company's response to environmental issues. Some scholars have touted BM information as a golden ticket in addressing this gap based on the idea that reporting on a company's BM complements quantitative reporting (Beattie, 2014). Further, reporting on a company's BM seems like a move towards a more integrated, narrative-based reporting practice (Beattie and Smith, 2013). Yet, as reporting BM information has increased, it has fast become an integral part of accounting and reporting practice that typically involves quantifying objectives (Power and Brennan, 2021;Unerman et al., 2018). In this study, we are interested in testing whether accountingisation influences the BM information included in financial and non-financial reporting.

Business model information
2.2 Accountingisation and the rational and narrative paradigms in corporate reporting practices As we argued earlier, the numbers versus narratives debate in corporate reporting demonstrates the differences between a rational and a narrative paradigm. The rational paradigm, which derives from Aristotle's work on logic, posits that members of society act rationally and objectively to collect and analyse information without relying on their values, interests or emotions (Baum, 1996;Fisher, 1984). Firms clarify "what is to be regarded as right and important, and who has the right to hold others (financially) accountable for their actions" (Englund et al., 2011, p. 500), thus appearing rational to their stakeholders. Additionally, as Power et al. (2003) assert, accounting information in corporate reports helps "rationalise and justify decisions ex post". Therefore, the rational paradigm, which organises information through defined methods and rules, dominates accounting practice.
Organisations strive to make "the intangible tangible" through accountingisation (Dumay, 2009). As a result, they are locked into the accounting domain and privilege the measuring approach. As Chaminade and Roberts (2003) explain, they follow the mantra of "what gets measured gets managed", assuming that measuring itself is a way to ensure follow-through. Organisations consider quantified information more accurate and comparable across organisations than qualitative information (Järvinen et al., 2020). Further, they regard that monetary information is higher quality than qualitative information (Wiseman, 1982). As such, accountingisation imbues organisations with quantitative measures and financial imperatives (Martin-Sardesai et al., 2020a).
The rational paradigm helped accountingisation become a key component of new public management and the healthcare system (Kurunmaki et al., 2003;Lapsley, 2007). That said, we observe the phenomenon in several fields. Power and Laughlin (1996, p. 447) outline that accountingisation "colonises areas of social life by creating newly internalised facts and vocabularies that potentially undermine actors' capability to question its self-evident mission". Thus, accounting "may curb our field of vision by redescribing reality into terms of calculable economic reason" (Holmgren Caicedo et al., 2019, p. 374), limiting our assessments. In recent decades, accounting values, practices and procedures pierced deep into the core of organisations, controlling their constituent parts (Kurunmaki et al., 2003;Martin-Sardesai et al., 2020a).
Financial statements represent the pinnacle of accountingisation as they provide the dominant "rational" or "numbers-based" representation of an organisation while excluding other possibilities (Power et al., 2003). Critics claim financial accounting does not merely represent and communicate reality but creates a distorted portrait of it by capturing only a limited picture of multi-dimensional organisations (Hines, 1988;Morgan, 1988). Accounting is interpretative and accountants "construct meaning and engage in all kinds of symbolically significant activities" (Morgan, 1988, p. 479). Thus, "accounting threatens to delinguistify the public realm and to absorb and transform public discourse in its own image", affecting the interaction between information and users' understanding (Power et al., 2003, p. 147). Even more importantly, the accounting process operates within the rational paradigm, which uses a hierarchical system "in which some are qualified to judge and others to follow" (Weick and Browning, 1986, p. 249). Therefore, only accounting experts can debate, interpret and assess and non-experts are considered spectators.
Accounting is the language of business (Davidson et al., 1974), but it is not the "native language" of non-experts. Corporate reports are for "the whole audience who might read it" Crowther (2012, p. 123)meaning those who are financially literate, especially in financial reporting. Financial literacy is "the ability to read and understand fundamental financial statements, including a company's balance sheet, income statement and cash flow statement as determined by the board of directors in the exercise of its business judgement" (Kouaib et al., 2018, p. 610). Therefore, accounting has implications for power and knowledge. It creates a distinction between "those who have the information, and who hold others accountable on the basis of it, and those who understand accounting language and those who do not" (Conrad, 2005, p. 6). The rational paradigm underlying accounting practices thus helps to concentrate power in organisations, limiting who has and can disclose information.
Since the late 1970s, a "narrative turn" has crept into the human sciences, proposed as a way to combat the limits of quantitative methods, to more accurately capture the complexity of human experience (Paschen and Ison, 2014) and to produce information that even the financially illiterate can understand. Narratives are not a simple description. Rather, they explain events and activities that have occurred or will occur in the future (Pentland, 1999;Wolfe and Shepherd, 2015). As such, "narrative presents something close to the identity of the firm and therefore presents some kind of raison d'ȇtre of its activities" (Mouritsen et al., 2002, p. 14). Hence, it is a rich source of insights into firms. It can explain their actions, their environmental contexts and convey meaning to the reader (Pentland, 1999;Wolfe and Shepherd, 2015).
As an alternative to the rational paradigm, Fisher (1984, p. 2) proposed the narrative paradigm, "a dialectical synthesis of two traditional strands in the history of rhetoric: the argumentative, persuasive theme and the literary, aesthetic theme". It counters the rational paradigm, which uses a direct, metric-centric approach and numerical reasoning to support arguments (Lim and Puspita, 2020). Thus, the narrative paradigm involves stories not calculations and narratives not numbers.
The narrative paradigm democratises power and knowledge as it assumes that all people have a narrative capacity, that experts are storytellers, and non-experts are active participants, not just mere spectators (Weick and Browning, 1986). According to Roberts (2004, p. 131), the narrative paradigm is "less elitist than the rational world paradigm". The capability to create and follow stories -"narrative intelligence"is considered a primary human intelligence by narrative paradigm scholars (Randall, 1999, p. 13). Everyone has a fundamental narrative intelligence, and unlike in the rational paradigm, people can become "not just literate, but literate literarily" (Bogdan, 1990). Furthermore, people can evaluate how reliable stories are by assessing both narrative probability and fidelity. Narrative probability refers to the coherence, consistency and noncontradiction of the story, perhaps compared to previous stories. Narrative fidelity refers to the degree to which the story "accords with the logic of good reasons" (Fisher, 1985, p. 349). As everyone can judge narrative fidelity and probability, everyone can assess whether a story is reliable (Fisher, 1984), determining for themselves what to believe. Unlike numbers, narratives may turn non-experts into active participants. For these reasons, the narrative paradigm is, supposedly, a better way to represent reality.
A "narrative turn" also exists in accounting and corporate reporting (Beattie, 2014). The narrative nature of non-financial reports represents a step away from a purely rational paradigm. However, the literature calls for further improving the quality of corporate reporting by focusing on the "stor [ies]" of organisations (Beattie and Smith, 2013, p. 246). Narratives explain future possibilities and explore how businesses might succeed going forward (Dumay, 2008). They help a wide variety of peopleincluding investors and financially illiterate stakeholdersgain an overall picture of a firm's performance and make judgements on their prospects. We argue that, by describing a company's BM, by showing the connections between its components and explaining the meaning of its measures, narratives provide meaningful explanations of how an organisation creates value.

Hypothesis development
This study unpacks the paradigms behind BM information in corporate financial and nonfinancial reports. We analyse how companies communicate their BM information in their corporate reports.
A financial report mainly presents a company's numbersthe counts and amounts of its key financial indicators. It is the principal kind of data managers use to demonstrate the value of a business (Roslender and Fincham, 2004). Financial reporting "should provide information to help present and potential investors and creditors and other users in assessing the amounts, timing and uncertainty of prospective cash flows" (Kothari, 2001, p. 120). Non-financial reporting, meanwhile, presents other non-financial information about a company to "help stakeholders better understand a company's overall performance, business strategy and growth perspective" (Perrini, 2006, p. 74). In this study, we consider non-financial reporting to include a wide range of voluntary and mandatory reports produced in addition to annual financial reportssuch as sustainability reports, corporate social responsibility reports, social reports and integrated reports (Stolowy and Paugam, 2018).
Financial and non-financial reports have different aims, audiences and rules. Financial reports express the rational paradigm, providing numbers-based information to financially literate users. In particular, they contain information that helps shareholders assess prospective dividends and proceeds (FASB, 1978). Their content is derived from established quantitative methods and rules, and the information is written in accounting language. Nonfinancial reports supplement and complement financial reports (Merkl-Davies and Brennan, 2017). They express the narrative paradigm by including information that is difficult to quantify for a broader audience (Beattie and Smith, 2013).
These two types of corporate reports mutually interact (Einhorn, 2005), with the content of financial reports often influencing that of non-financial reports (Bagnoli and Watts, 2007). Bagnoli and Watts (2007) find that good news in a financial report is not usually discussed at length in the counterpart non-financial report; bad news, however, will be. Others, such as Stolowy and Paugam (2018), argue that the two reports contain very different material. However, despite their connections, they are fundamentally informed by different paradigms. We expect to see some differences in the BM information included in each type of report. Therefore, the first hypothesis to be tested is: H1. Due to the different paradigms informing financial and non-financial reporting, firms will present different types of BM information in the two types of report.
Our second hypothesis goes further. In line with the narrative paradigm, non-financial reports should use narrative language that everyone can understand. However, firms are sometimes reluctant to communicate the intangible aspects of their value creation (Holland, 2004). Additionally, as BMs become an integral part of accounting and reporting practice, they may become accountingised (Dumay and Roslender, 2013), influenced by the existing, numbers-based values of the accounting field. Therefore, we contend that the BM information included in financial and non-financial reports relies primarily on numbers, even though non-financial reports supposedly reflect a narrative paradigm. Hence, our second hypothesis is: H2. Accountingisation influences how BM information is reported in both financial and non-financial reports.

Research method
We used a two-step content analysis method to test our hypotheses, first analysing the text of financial and non-financial reports, then applying a PLS-SEM to the resulting data set. PLS-SEM is a statistical data model that helps estimate and test relationships between different variables (Lee et al., 2011;S anchez, 2013). In our study, using a PLS-SEM allowed us to test whether there were significant differences in the BM information between the financial and non-financial corporate reports in our sample. Ultimately, the PLS-SEM helped us to estimate the influence of accountingisation on BM information reporting.

Sample and data collection
Our sample was selected from a list of the top 25% publicly listed companies by market capitalisation in the 2016 OSIRIS database (9,680 firms). Because these companies have global visibility, we anticipated their reports would feature more BM information, and they more likely adopt non-financial reports. Using the Neyman optimal allocation formula to get the optimal sample size, we selected 96 firms spread across a range of industrial sectors and geographic areas as a starting sample. We excluded ten firms from this initial sample because they did not have both financial and non-financial reports, leaving 86 firms in the final sample (Table 1). Of these, five enterprises produced integrated reports, four provided environmental, social and corporate governance (ESG) information in their reports, and one published an ESG report. We manually collected corporate reports from 2016 from the companies' websites and, with the aid of NVivo software, coded the contents into categories for analysis (Krippendorff, 2013). First, we analysed the reports from 2016 because it was two years after the release of the initial version of the IIRC's <IR> framework that included the reporting of BM information. Second, it is the year before the European Union Directive 95/2014 for nonfinancial reporting came into effect that requires large European companies to disclose BM information. In so doing, we avoid the potential bias caused by the mandatory effects of the European Union regulation and increase the chance to gather the effects from the voluntary adoption of the IIRC's <IR> framework. The coding process involved transcribing, recording, categorising and interpreting statements for analysis (Krippendorff, 2013). Although many studies only examine selected sections of the corporate reports (Beattie et al., 2004), we coded the full content of all 172 corporate reports following Business model information Beattie et al. (2004). They recommend that researchers begin by analysing the topics and then the attributes of each item of information.
The coding scheme was based on a merger between two popular BM frameworksthe CANVAS business model framework Osterwalder and Pigneur (2010) and the V4 business model dimensions framework (Al-Debei and Avison, 2010). Although CANVAS has become something of a de-facto standard for conceptualising BMs (Joyce and Paquin, 2016), its nine building blocks translates into too many variables to feasibly model with PLS-SEM. With only four elements, the V4 framework has fewer dimensions than CANVAS business model framework. However, its elements are derived from the information systems literature. Therefore, it can be sector-specific to analyse a broader set of industries. Thereby, we merged the nine dimensions of CANVAS into the four dimensions of V4 to create a non-sector-specific coding scheme. The specific mappings between the two frameworks are listed in Table 2.
Topic is the linear combination of the four dimensions of V4. To complete our coding scheme, two additional attributes could also be identified for each item of informationthese being the financial information presented about an organisation's BM expressed in monetary terms, such as assets or revenue, i.e. Monet, and that of a quantitative nature, such as carbon emission levels or counts towards performance indicators, i.e. Quant.
Therefore, our final scheme comprised three latent variables -Topic, Monet and Quanteach broken down into four manifest variables, being the four dimensions of the merged framework: value architecture (va), value finance (vf), value network (vn) and value proposition (vp). Latent variables (LVs) represent constructs that cannot be measured directly. They are indirectly measured by manifest variables (S anchez, 2013). Manifest variables (MVs) s are blocks of attributes, that are observed-measured and contain information of latent variables (S anchez, 2013). Thus, any given sentence could be coded to up to 12 different categories; albeit, most sentences only applied to 1 or 2. In the coding process, the coding categories were not considered mutually exclusive. A detailed description of each of the variables is given in Table 3.
Note that we took the sentence as our unit of analysis as they "are far more reliable than any other unit of analysis" Milne and Adler (1999, p. 243) and are recommended when performing manual coding (Beattie et al., 2004) [1].
We performed pilot coding using ten sets of reports and then measured the agreement between the two coders. We shared our results with all the co-authors of this paper to identify areas of coding disagreement and revised our coding rules to ensure both consistency and internal validity going forward (Beattie et al., 2004). One author coded the remaining 162 reports and discussed the results with all the authors. Finally, we considered the quantity of information coded to each dimension and attribute as the frequency variable needed to perform the PLS-SEM analysis (Appendix). value proposition q_vp . . . product/service and customers "By the end of 2016, nearly 1.8 million lamps had been sold, improving the day-to-day lives of nearly 9.5 million people" Table 3. Wold's (1974Wold's ( , 1982 PLS-SEM approach has become a foundational multivariate analysis method in several disciplines, including accounting (Hair et al., 2012;Lee et al., 2011). The PLS-SEM approach follows the spirit of a dimension reduction technique, providing a practical summary of how a set of predictors systematically explains a set of dependent variables (S anchez, 2013, p. 34). It has several advantagesamong them the ability to handle small sample sizes and the flexibility to deal with a wide range of data analysis tasks, including studies of structural systems (S anchez, 2013). The model does not depend on a priori assumptions of a particular scale or a normal data distribution (Latan and Ramli, 2014). It can also be used to generate hypotheses from qualitative methods or to test hypotheses using quantitative research (Hair et al., 2017).

Partial least squares-structural equation model
The model does depend on some pre-established relationships among LVs. LVs can be measured through formative or reflective methods depending on the cause-effect relationships between the relevant LVs and MVs. Formative correlations are those where a linear combination of MVs determine the LVs (Lee et al., 2011). Reflective correlations are the reverse, where MVs reflect the variation in the LVs (Lee et al., 2011). To establish these a priori relationships, researchers either draw from previous knowledge of the phenomenon or program the relationships according to the hypothesis they want to test. A PLS-SEM also helps to explore possible relationships that are not based on theoretical or causal justifications (Hair et al., 2017).
The three LVs of our model are topicality (Topic), monetisation (Monet) and quantification (Quant) are the three characteristics of BM information we seek to examine. From the literature, the four value variables combine linearly to confirm "topicality". Topicality is an attribute that scholars have seen as a basic condition to assess the relevance of documents (Xu and Chen, 2006). Topical relevance is "a determination of the intellectual content of a document, usually in terms of some subject classification" (Froehlich, 1994, p. 125), specifically, topicality means "how well the topic of the information retrieved matches the topic of the request" (Harter, 1992, p. 602). In our research, Topic indicates the dimensions of BM a sentence refers to and reflects the narrative concept Holland (2004) regardless of the information attributes. Monet and Quant represent the impact of accountingisation (Martin-Sardesai et al., 2020a). Unlike Topic, the four value dimensions of these variables are designed so as to test our hypotheses.
We used the PLS-SEM in the R-statistical software package "PLSPM". To build the model, we established two sub-modelsan outer model and an inner model. We used the outer model to investigate the relationship between each LV and its MVs, ultimately adopting a reflective measurement approach to test our first hypothesis.
Outer model tests "the relationships between each LV and its block of attributes" S anchez (2013, p. 23) and tries to obtain the best estimates for the LVs based on their attributes (MVs) in terms of loadings and communality. We explore loadings and communality in detail in our results section. If different attributes (MVs) explain the differences in the content and attributes of BM information between the two types of reports, the hypothesis is supported. In our case, this means that financial and non-financial reports do provide different information to meet the needs of different audiences.
Inner models test the relationships between LVs (S anchez, 2013), which we used to test our second hypothesis (Figure 1). Following the direction of the arrows, Quant explains any variance in Monet because any monetary measure must be quantified first, while Quant and Monet (both representing the accountingisation of BM information) interact and shape Topic. Through the inner model, we tested whether these relationships are positive and statistically significant. Thus, if the impact of Quant and Monet on Topic is significant for both financial and non-financial reports, then our second hypothesis is supported. Such a result would indicate that BM information is accountingised even though different paradigms inform the two types of reports.

4.1
The types of business model information in financial and non-financial reporting (hypothesis 1) To test H1, we applied the outer model to our sample financial and non-financial reports and compared the results. First, we checked whether the blocks of attributes were unidimensional. Checking this ensures that if a construct (LV) changes, then the associated indicators will also change in the same direction. We evaluated unidimensionality by calculating Dillon-Goldstein's rho, the first eigenvalue and the second eigenvalue in the correlation matrix of attributes. A Dillon-Goldstein's rho value higher than 0.7 signals unidimensionality; the first eigenvalue should be greater than 1, and the second eigenvalue should be less than 1 (S anchez, 2013). According to Chin (1998), Dillon-Goldstein's rho takes into account the extent to which the LV explains its block of indicators (MVs). Within a PLS-SEM, it is considered a better indicator than Cronbach's alpha. The unidimensionality scores of three blocks of attributes ensure that the MVs reflect the associated LV.
Using this method reveals the variables contributing to most effectively explaining the single construct are different in the two types of reporting practices (Table 4). Regarding the outer model estimation results for testing H1, the outer loadings measure the relationship between the single construct and its variables. An MV with a low loading implies that it has little shared variance with the LV (Lee et al., 2011). The cut-off for the outer loading is 0.7 (Latan and Ramli, 2014). The loadings that exceed this level are shaded. Communality represents the amount of variability explained by an LV (in percentage terms), calculated as the squared correlations of the LV to its indicators. A communality value of greater than 0.49 is considered significant (S anchez, 2013).
Our results show that only value finance (vf) has a significant and positive direct effect on Topic in financial reports. Meanwhile, value architecture (va), value finance (vf) and value proposition (vp) all have a positive and statistically significant connection with Topic and explain its variance in non-financial reports. For example, financial reports typically highlight the revenues generated by core products and services, as this extract from our data shows: Cloud and software revenue climbed to e18,424 million in 2016, an increase of 7%. Cloud and software revenue represented 84% of total revenue in 2016 (2015: 83%). Service revenue increased 2% from e3,579 million in 2015 to e3,638 million, which was 16% of total revenue, in 2016.
(SAP, Financial Report)  In non-financial reports, firms highlight the details of their products and services, their technological architectures, their organisational infrastructures and their operational configurationsfor example, as this extract from our data shows: In 2016, we will increase our sales of smartphones as we adequately respond to changes in the market environment through reinforced products and continuous streamlining of our lineup. Additionally, we will lead all market growth through cutting-edge innovation as we increase supply and continuously strengthen profitability based on a lineup that focuses on competitive product models not just in high-end markets, but also in low-and mid-priced markets.
(Samsung, Non-financial report) Therefore, while financial reporting focuses on a firm's financial performance, nonfinancial reporting provides a broader picture of performance and the BM beyond financial metrics. Our results for Monet and Quant were similar: m_vf and q_vf had the greatest hand in shaping these aspects of the financial reports. However, while we expected this result in the financial reports, we were also surprised by the significant impact these variables had on the non-financial reports. We contend this is due to the monetisation and quantification of value architecture (m_va; q_va) in the non-financial reports, alongside information provided in those reports on costs and revenues. To conclude, our results confirm H1: firms do indeed vary the BM information they present in financial and non-financial reports according to the characteristics and aims of the two forms of reports because of their different audiences.

Accountingisation's influence in financial and non-financial reporting (hypothesis 2)
To test our second hypothesiswhether accountingisation influences how BM information is reportedwe used our inner model, which measures the relationships between LVs (Quant, Monet and Topic). We conducted two separate tests for financial and non-financial reports (Figure 2).
Our analysis shows that, in both financial and non-financial reports, the LV Quant has a significant impact on Topic and Monet and Monet does not explain the variance of Topic. This means that quantitative information has a significant impact on Topic, but monetary information has not. The small impact of Monet is likely explained by the nature of BM information, which is usually about elements that may be monetised difficultly. This is the rationale we could not limit the concept of accountingisation to the monetary commiseration only because we analysed corporate factors and performance (especially in non-financial reports) that can be arduously expressed through monetary information. Thus, accountigisation manifests more prominently in the quantification instead of monetisation, so confirming that accountingisation may refer to the broad approach of using quantitative measures, which are not limited to financial measures and monetary values (Dumay, 2009;Järvinen et al., 2020;Jones and Dugdale, 2001;Martin-Sardesai et al., 2020a, 2020b.

Business model information
We expected Quant to have a significant impact on Monet as monetary information is also clearly quantitative. However, these results go further, showing the influence of numbers over non-financial information. Therefore, our second hypothesis is confirmed. Altogether, our models demonstrate that firms report on their BM information in both financial and non-financial reports with an accounting mindset, although these two types of reports are informed by different paradigms. Table 5 shows the reliability and validity results of the inner model estimations. The primary measure of reliability is the coefficient of determination (R 2 ), which assesses prediction accuracy (Hair et al., 2012;Nitzl, 2016). Specifically, R 2 indicates the amount of variance in the endogenous latent variables, which is explained by its independent latent variables. Values greater than 0.6 represent good explanatory power (Hair et al., 2014). As our results show, Quant explains most of the variance in Monet and Topic, and the overall prediction accuracy of the inner model is good.
Redundancy indicates the model's validity (Hair et al., 2014). As explained by S anchez (2013, p. 69), redundancy "reflects the ability of a set of independent latent variables [to] predict values of the indicators' endogenous construct". Values higher than 0.375 denote a large model, while the values of 0.125 and 0.25 denote fit and moderate models, respectively (Latan and Ramli, 2014). The redundancy scores for Topic in both models are very high, which indicates that Quant largely predicts its value. Monet has mid-range scores of 0.19 (financial/non-financial) and 0.29 (financial/non-financial), suggesting that Quant only partially predicts its values.
Goodness of fit (GoF) measures the overall performance of both the inner and outer models. According to Latan and Ramli (2014), the value of 0.52 for the financial reports indicates that the model is fit moderate, while the score of 0.62 for the non-financial reports denotes a large model.
Finally, we performed a bootstrap t-test to compare and test the differences between the two groups. The bootstrap t-test is a resampling method that involves "separating the data into groups and then running bootstrap samples with a replacement for each group" (S anchez, 2013, p. 120). Our results appear in Table 5. Columns 2 to 4 show the path coefficients; "diff.abs" is the absolute difference between the path coefficients for the financial and non-financial reports. The next three columns report the t-test statistics, the degrees of freedom and the associated p-value. The last column shows the significance of the difference between the path coefficients at the 5% level.
As Table 6 shows, the impact of Quant and Monet on Topic is not significantly different between the financial and non-financial reports. This fact further confirms our second hypothesis, suggesting that accountingisation influences the kind of BM information included in both financial and non-financial reports. In turn, our findings show that introducing BMs as a core feature of corporate reporting has not changed the form and paradigm through which organisations produce those reports: numbers still trump narratives. Although our model follows a more predictive than explanatory approach, we also tested the endogeneity in our inner model using the Gaussian Copula approach (Hair et al., 2019). We found an endogeneity issue with the LV Quant when considering the whole sample (172 reports) regardless of the type of reports in the test. However, this endogeneity disappears when we distinguished between financial and non-financial reports, which means that the type of report is a variable explaining both quantification and topicality. This evidence supports our findings of the difference between financial and non-financial reports and demonstrates the need to treat financial and non-financial reports in separate inner models. However, it brings us to the same conclusion about the influence of accountingisation in corporate reporting.

Conclusion
Our study is motivated by the ongoing tug-of-war between narrative and rational paradigms and the related debate on privileging numbers or narrative in corporate reporting (Dumay and Rooney, 2016;Martin-Sardesai et al., 2020b). We investigated the content and attributes of the BM information presented in corporate reports, assuming that the rational and the narrative paradigms each inform a type of report (financial or non-financial). We find that companies report on their BMs in different ways according to the report type and its associated audience. However, accountingisation heavily influences BM information in both types of reports. In other words, companies are prone to quantifying information about their BMs. This reliance on quantification reinforces Chaminade and Roberts' (2003) argument that accounting holds to a "what gets measured gets managed" rationale. By our results, firms are clearly anchored to established reporting practices grounded in the rational paradigm. Accordingly, BM information inadvertently falls into the accountingisation trap and reports continue to privilege numbers over narratives. We do not see the long-desired narrative turn in accounting or corporate reporting despite great efforts by many to induce it.

Theoretical contribution to the debate between numbers and narrative
This study extends both our empirical knowledge and our theories on the numbers-vnarratives debate, linking it to the debate between the rational versus narrative paradigm that underpins corporate BM reporting. Prior studies have explored some of these same themes, but none have been able to support their contentions with quantitative evidence. For example, Dumay and Rooney (2016) argue that using numbers or narrative is an actor's choice based on their epistemological viewpoint. Our study takes a step forward by demonstrating that it is the rational paradigm that informs corporate reporting. This paradigm leads firms to choose numbers over narrative in corporate reports. Relying on the previous theoretical contributions (Baum, 1996;Fisher, 1984;Power et al., 2003), this paradigm is used by companies to rationalise decisions and provide a semblance of reliable and objective information. That quantitative information dominates corporate reporting is This rational approach is rooted in the shared values in the organisations' culture. Accountingisation is influenced by the logic to collect, analyse and measure information and by the social and organisational practices to transform information into comparable numbers. This attitude guides and inspires preparers in both financial and non-financial reporting as accountingisation infiltrates companies at every level (Martin-Sardesai et al., 2020a). Our results show that an accounting mindset influences corporate reports and even non-financial information is quantified and reported in numbers. Thus, even though the two reports have different purposes, different content and a different audience, both remain anchored to the rational paradigm.
These findings unveil a conflicting force between the purpose of reporting and how reports are prepared. Although non-financial reporting is theoretically an expression of the narrative paradigm and targeted to a broad set of users, it too is touched by accountingisation. This tacit conflict extends over the main actors involvedthe stakeholders and preparersand their power (Conrad, 2005;Weick and Browning, 1986). The individual actor, who is responsible for collecting non-financial information that is narrative in nature, elaborates it by reducing it into quantitative information. As a result, preparers tend to exclude non-expert stakeholders allowing only (accounting) experts to debate, interpret and assess information.
The main emphasis in a narrative is the connection among fragmented pieces of information in a story (Holland, 2004). Narratives can help link the different components of a company's BM while numbers can give objective credibility to those descriptions and reduce ambiguity (Chaminade and Roberts, 2003;Holland, 2004;Järvinen et al., 2020;Robson, 1992). As Dumay and Rooney (2016) point out, neither numbers nor narratives (nor their combination) can adequately meet the needs of all an organisation's stakeholders. Yet, as narrative can enact a more inclusive reporting, we need to find a trade-off between the rational and the narrative paradigms. However, it is unlikely that inclusive reporting that incorporates narrative will become widespread without modifying current corporate reporting practices. Beattie et al. (2004) claim that we need more non-financial information about businesses while Beattie and Smith (2013) argue that BMs represent an attractive source of nonfinancial information. However, contrary to what several scholars predict (Michalak et al., 2017), reporting on a firm's BMs has not radically reshaped corporate reporting practices. Current non-financial reports contain BM information but are still produced with an oldfashioned accounting mindset. As our study demonstrates, by using an accountingisation vocabulary, report preparers follow a rational schema. Thus, the accounting mindset, which tends to infuse a narrative with numbers, remains the dominant epistemological view.
Companies have not overcome the limits of traditional reporting models. If firms cannot move from numbers to a narrativeor at least move freely between themthey will not escape the trap of accountingisation (Dumay and Roslender, 2013). They thus cannot provide a full context for understanding corporate information (Michalak et al., 2017). We need to recognise accountingisation at work in reporting on BMs and free report writersand companiesfrom the accounting mindset.

Practical implications
We began this article by mentioning "a push" to include business model information in corporate reports. One such gambit is the European Directive 95/2014 on non-financial reporting (European Parliament, 2014). However, the corresponding member state regulations have not harmonised narrative reporting practice (Strampelli, 2018). As a result, several scholars have strongly advocated for reforms to this and other similar initiatives (Abhayawansa and Guthrie, 2016). Reporting regulations, such as the European Directive 95/2014 (European Parliament, 2014), may drive firms to change their reporting practices substantively, as might other frameworks and guidelines, such as the IIRC's <IR> framework (IIRC-International Integrated Reporting Council, 2013), in their call for more significant disclosures of BM and other non-financial information (Tweedie et al., 2018). The concept of integrated thinking behind the <IR> framework may also encourage companies to understand how interdependent the financial and non-financial aspects of their business strategy are (Simnett and Huggins, 2015). But at present, reports include "what the organisation wants to disclose, rather than what is prescribed by a particular framework" (Dumay and Rooney, 2016, p. 229), suggesting it will take more to change the norm.
A change of focus within corporate reporting, from numbers to narrative, can democratise corporate reports and transform non-experts into active participants. Our findings can help researchers and policymakers unlock BMs from the accountingised corporate reporting status quo. Instead of simply introducing new reporting guidelines, we advocate the need to change the mindset behind the production of corporate reports into one that values both numbers and narratives. We know this will not happen overnight. However, unless the growing pressure from investors and financially illiterate stakeholders to make companies more accountable, responsible and transparent lifts, every company eventually will have to change its mindset.
This study has some limitations. First, we analysed a sample of corporate reports published in 2016 because we expected to find early evidence of organisations reporting BM information. Future research could compare our results with the analysis of more recent reports to determine whether much has changed over time. As the attitude of report writers to accountingisation depends on their epistemological viewpoint and social identity (Dumay and Rooney, 2016), future scholars in this field could also fruitfully explore how national, organisational and accounting cultures affect the narrative and rational paradigms. Future research can examine accountingisation in corporate reporting to explore other rationales underpinning its use, such as concealing sensitive information (Dumay et al., 2018), the institutionalisation of reporting practices (de Villiers and Alexander, 2014) and the use of calculations to support arguments (Westerdahl, 2020). Note 1. Compared to words, coding sentences allowed us to consider the context in which the idea was expressed. We acknowledge that there is some debate over using sentences versus paragraphs as a coding unit (Dumay and Cai, 2014). However, sentences are more suitable for our purposes because paragraphs can be long, they tend not to follow precise linguistic rules, and it can be arduously difficult to distinguish one paragraph from the next in a badly-formatted PDF. According to Hackston and Milne (1996, p. 86), sentences have the advantage of being 'natural units of written English which clearly exist between two punctuation marks', thus providing a more reliable measure than paragraphs. As such, using sentences for both coding and measurement provided 'complete, reliable and meaningful data for further analysis' (Milne and Adler, 1999, p. 243).