روشی برای تجزیه و تحلیل و ارزیابی روایات گزارش های سالانه : پروفایل توصیفی جامع و معیاری برای ویژگی های کیفیت افشا

There is a consensus that the business reporting model needs to expand to serve the changing information needs of the market and provide the information required for enhanced corporate transparency and accountability. Worldwide, regulators view narrative disclosures as the key to achieving the desired step-change in the quality of corporate reporting. In recent years, accounting researchers have increasingly focused their efforts on investigating disclosure and it is now recognised that there is an urgent need to develop disclosure metrics to facilitate research into voluntary disclosure and quality [Core, J. E. (2001). A review of the empirical disclosure literature. Journal of Accounting and Economics, 31(3), 441–456]. This paper responds to this call and contributes in two principal ways. First, the paper introduces to the academic literature a comprehensive four-dimensional framework for the holistic content analysis of accounting narratives and presents a computer-assisted methodology for implementing this framework. This procedure provides a rich descriptive profile of a company's narrative disclosures based on the coding of topic and three type attributes. Second, the paper explores the complex concept of quality, and the problematic nature of quality measurement. It makes a preliminary attempt to identify some of the attributes of quality (such as relative amount of disclosure and topic spread), suggests observable proxies for these and offers a tentative summary measure of disclosure quality.

In recent years, the nature of business has changed fundamentally. Competitive advantage increasingly involves value creation processes that rely on intangible assets not recognised in the financial statements. To serve the information needs of the market and provide the information required for corporate transparency and accountability, there is now a consensus that the business reporting model needs to expand beyond the traditional financial reporting model that emphasises backward-looking, quantified, financial information (e.g., AICPA, 1994, Elliott, 1992, FASB, 2001a, ICAEW, 2003, ICAS, 1999 and Lev, 2001; Lev & Zarowin, 1999; Wallman, 1995, Wallman, 1996 and Wallman, 1997).1 The general thrust of these articles and reports is that there is a need for more information that is forward-looking and non-financial in nature. It is recognised that much of this new information will be ‘soft’, i.e., either unquantified or unquantifiable. Of particular note is the report published by the AICPA (1994), which has become extremely influential (the Jenkins Report). This set out to improve business reporting by adopting a customer focus, i.e., by meeting the information needs of investors and creditors. The report proposed a comprehensive model of business reporting that embraced a ‘broader, integrated range of information’ (p. 131). This model comprised eight main topics (financial data, operating data, management analysis, forward-looking information, information about management and shareholders, objectives and strategy, description of business and industry structure) and many sub-topics. In response to the report, FASB set up a business reporting research project to consider the types of information that companies are voluntarily providing and the means for delivering it. Reporting on the first of these two issues, FASB stated that ‘the importance of voluntary disclosures is expected to increase in the future because of the fast pace of change in the business environment’ (2001b, p.v). Worldwide, narrative communication in annual reports is viewed as the crucial element in achieving the desired step-change in the quality of corporate reporting and regulators are focussing attention on the management discussion and analysis statement in the annual report (referred to as the MD&A in most countries and the operating and financial review (OFR) in the UK).2 In some jurisdictions, guidelines are being extended and revised; while in others, disclosures are becoming mandatory. In the US, post-Enron, MD&A regulations are being strengthened (e.g., SEC, 2003). In Canada, the Canadian Institute of Chartered Accountants (CICA) issued more detailed MD&A guidelines that set out six disclosure principles and develop a five-part integrated disclosure framework that covers strategy, key performance drivers, capabilities, results and risks (CICA, 2002). In the UK, the Accounting Standards Board issued revised OFR guidance, which draws upon the Jenkins framework (ASB, 2003). Company law reviews in both Australia and the UK are proposing mandatory OFRs for listed companies (DTI, 2002 and G100, 2003). Draft regulations were recently issued in the UK (DTI, 2004). Finally, consideration of MD&A statements was put on the IASB agenda in 2002. Meanwhile, accounting researchers have increasingly focused their efforts on investigating disclosure, in particular the determinants of disclosure and the capital market consequences (for a recent review of the empirical disclosure literature, see Healy & Palepu, 2001). Healy and Palepu observe that ‘one of the limitations of the [studies on voluntary disclosure] is the difficulty in measuring the extent of voluntary disclosure’ (2001, p. 32), while Core notes that ‘improved measures of disclosure quality also need to be developed’ (2001, p. 16). To date, two principal ways of measuring disclosure have been employed. The first approach has been to use subjective analyst disclosure quality rankings. Although this approach is not without conceptual problems, a real practical problem for US researchers is that the AIMR discontinued its rankings in 1997 (after ranking fiscal year 1995). Many other countries have never had similar rankings available. The second approach, which has a long history, has been to use researcher-constructed disclosure indices where the amount of disclosure is used as a proxy for disclosure quality (e.g., Botosan, 1997; Lang & Lundholm, 2000). Given the limitations and weaknesses of these two approaches, there is clearly a pressing need for research effort to be devoted to developing new ways of documenting disclosure practices, identifying dimensions of disclosure quality and exploring possible measurement proxies. It is suggested that developments of this nature have two main advantages. First, the development of a comprehensive disclosure profile serves as a practical tool, permitting the benchmarking of current practices. This allows inter-company, inter-industry and inter-country comparisons to be made and also allows changes over time to be monitored. Second, a richer set of objective measures relating to disclosures can permit much more powerful tests of many research questions that relate to narrative disclosures. To this end, the present paper has two main objectives. First, the paper introduces to the academic literature a methodology for generating a rich descriptive profile of a company's narrative disclosures. The basis of this profile is a comprehensive four-dimensional framework for the holistic content analysis of narratives, based on the coding of topic and three type attributes. The topic analysis is based on the Jenkins report (AICPA, 1994), which proposes a ‘comprehensive model of business reporting’. The type analysis captures the time orientation, financial/non-financial and quantitative/qualitative attributes of each text unit. A text unit is defined as a phrase containing a single piece of information. Associated with this framework is a detailed set of coding procedures and a computer-assisted methodology for implemention. Second, the paper discusses the complex concept of quality and the problematic nature of quality measurement. It makes a preliminary attempt to identify some of the attributes of quality, suggests observable proxies for these and offers a tentative summary measure of disclosure quality. The remainder of this paper is structured as follows. Section 2 describes and critiques the different extant approaches to the study of narratives in annual reports. Section 3 outlines the general principles of content analysis. Section 4 explains the new computer-assisted methodology that generates a comprehensive descriptive disclosure profile of annual report narratives. Section 5 presents an illustrative application of the procedure to the narratives in Cadbury Schweppes’ 1999 annual report. The concept of quality is explored in Section 6 and proposals are made regarding possible attributes of quality, observable proxies for some of these and aggregation into a summary measure of disclosure quality. The final section summarises and concludes.

The importance of public voluntary disclosures made by listed companies is expected to increase in the future. To date, all studies that apply content-analytic methods to narratives have been either specific or, if more general, they have been partial. That is, studies have either focused on particular topics (such as social and environmental disclosures) or have specified a broad-based set of information items ex ante and ignored any disclosures that fall outside that list. To the best of our knowledge, no study has undertaken a general, comprehensive, sequential, unit-by-unit analysis of all disclosures. This paper introduces to the academic literature a new, computer-assisted methodology for documenting the nature of a company's voluntary narrative disclosures based on a four-dimensional framework that captures, for each text unit, not only the topic but also three type attributes: time orientation; financial orientation and quantitative orientation. The topic analysis is based upon the comprehensive model of business reporting proposed in the Jenkins report (AICPA, 1994). The procedures used permit the dimensions to be analysed both individually and in combination, allowing a rich description of the nature and patterns of disclosure to emerge. The emphasis given to particular topic categories can be readily calculated. Moreover, the sequential, unit-by-unit nature of the procedures used not only provides a complete, holistic view of a company's disclosures based on frequency counts, it could also be used to analyse patterns within the flow of the text. A further advantage of this method is that, because all voluntary disclosures are included, it avoids the need to select items and hence the need to specify a user group. Specifically, a Disclosure Profile is proposed that comprises six forms of analysis: a one-way main topic analysis; nine separate one-way sub-topic analyses; one-way analyses of each of the three type dimensions; two-way cross-type analyses; three-way cross-type analysis; and full four-way analysis of topic/type interactions. This analysis can be conducted at the level of the individual company, sector or even country. With the addition of suitable benchmarks of interest (e.g., prior period, sector average), this profile represents a powerful tool for evaluating company practices both cross-sectionally and over time. In addition to this practical application, the methodology offers the potential for developing a set of metrics that could be used to facilitate academic research into voluntary disclosure (an area where there is seen to be great opportunities). A second objective of the paper was to suggest possible new measures of dimensions of disclosure quality. Recent reviews have emphasised the importance of developing improved measures of disclosure quality (Core, 2001) and the development of Qc and its components is an attempt to respond to this call. It was argued that the amount of disclosure relative to size and complexity and the spread of disclosure across main topics and sub-topics are two fundamental dimensions of quality. A composite index of disclosure quality (Qc) was created based on four measures. The suggested composite index is tentative and exploratory. Inevitably, the method is very labour-intensive. However, in the absence of alternative narrative disclosure metrics, such as subjective analyst rankings, the importance of voluntary disclosures makes it imperative that metrics be developed.22 Voluntary disclosures are important to companies, to their stakeholders/information intermediaries and to researchers as a means of advancing our understanding of major issues (such as the link between enhanced disclosure and the functioning of the capital market). Attempts to partially automate the coding have had limited success (Hussainey, 2004). A certain amount of labour-intensive manual coding is essential as the output from manual coding establishes a valid benchmark against which other procedures and proxy measures can be assessed. All research methods suffer from limitations, and in the case of the methods used in the present study, two key limitations can be identified. First, there is inherent subjectivity involved in the coding scheme used. However, the coding instrument was developed rigorously and produces high reliability measures. Second, due to the highly labour-intensive nature of the data collection and analysis procedures, generally only a small number of companies can be investigated. These two limitations must be weighed against the potential to gain new insights into disclosure practices and patterns through micro-level analysis. Researchers investigating the determinants and consequences of disclosure quality could be wasting their efforts if the primary variable of interest is not being measured with a sufficient degree of accuracy. In this regard, the use of measures that simplistically equate absolute quantity with quality is a concern. A major contribution of this paper is to suggest a new direction in the measurement of disclosure quality. It is emphasised, however, that the present study is exploratory in nature and hence the suggestions made are tentative and incomplete. Further research that builds on and extends the ideas presented in this paper is essential. The incorporation of type-based quality dimensions, and (perhaps more importantly) topic/type quality dimensions, will further refine the composite quality index. The contribution to quality made by the type attributes is rather unclear, and may depend on the topic being discussed. It would also be interesting to elicit from key user groups (or a group of acknowledged experts) their views regarding the identity and nature of quality dimensions and the appropriate weightings to be assigned to each in constructing a composite quality index. Of course, these weightings may vary depending upon the decision-context, reporting company characteristics and user characteristics.