Purpose: The study aims to investigate empirically the common alternative methods of measuring annual report narratives. Five alternative methods are employed, a weighted and un-weighted disclosure index and three textual coding systems, measuring the amount of space devoted to relevant disclosures.Design/methodology/approach: We investigate the forward looking voluntary disclosures of 30 UK non-financial companies. We employ descriptive analysis, correlation matrix, mean comparison T-test, rankings and multiple regression analysis of disclosure measures against determinants of corporate voluntary reporting.Findings: The results reveal that while the alternative methods of forward looking voluntary disclosure are highly correlated, yet important significant differences do emerge. In particular, it appears important to measure volume rather than simply the existence or nonexistence of each type of disclosure. Overall, we detect that the optimal method is content analysis by text-unit rather than by sentence.Originality: This paper contributes to the extant literature in forward looking disclosure by reporting important differences among alternative content analyses. However, the decision regarding whether this should be a computerised or a manual content analysis appears not to be driven by differences in the resulting measures. Rather the choice is the outcome of a trade-off between the time involved in setting up coding rules for computerised analysis versus the time saved undertaking the analysis itself.Keywords: Narrative Disclosure, Content Analysis Methods, Forward-looking Information 3
How to Measure Annual Report Narratives Disclosure? Empirical Evidence fromForward-Looking Information in the UK Prior the Financial Crisis
OverviewDisclosure is an abstract concept that cannot be measured in an unambiguous or precise manner and there is therefore a continuing debate in the literature on how to best measure disclosure and the impact of using different measures (Healy and Palepu, 2001). While a large number of alternative content analysis methods have been employed these tend to fall into one of two types, either textual content analysis or what Joseph & Taplin (2011) call disclosure abundance, which measures the amount of space devoted to relevant items and disclosure incidence or disclosure indices which instead measure the occurrence of specific items. Although there has been a debate regarding the optimal way to design disclosure indices (eg Marston & Shrives, 1991;Chow and Wong-Boren, 1987) and similar debates on how to conduct textual analysis (eg Unerman, 2000;Beattie et al., 2004;Beattie and Thomson, 2007) there has been few studies that have compared textual content analysis and disclosure indices.Cerf (1961) is considered to be the first to use a disclosure index to evaluate the extent of disclosure, and since then, disclosure indices have been used extensively in corporate disclosure studies. Un-weighted indexes treat all disclosure items as equal, irrespective of the amount of space devote...