The Readability of Sustainability Reporting and its Interaction with Company Performance

The relationship between corporate performance and the language used in corporate reporting is long-contested. The ‘obfuscation hypothesis’, for instance, which posits that that poorer-performing companies will attempt to conceal bad news through less-readable annual reports, has seen contested evidence amongst scholars. At the same time, however, the strict legal scrutiny that financial annual reporting faces has also been reported to inhibit any such trends. In view of this, it is interesting to examine the increasingly prominent, yet still primarily voluntary and less regulated genre of corporate sustainability reporting. Despite its increasing prominence, few studies have quantitatively examined readability, linguistic variation and manipulation within this genre. Based on a manually assembled 2.75 million word corpus of sustainability reports in five varieties of English and four industries, we aim to scrutinise the correlation between both financial and non-financial (environmental and social) company performance data drawn from Thomson Reuters Datastream, and various linguistic features of corporate reports, readability prime among them, in order to discover the extent to which this genre might face linguistic manipulation as well (or even more). We rely on Natural Language Processing techniques to parametrise the reports used in this study, drawing on tried-and-tested automatic readability formulae but also automatically parsing and tagging our corpus for parts of speech using the Stanford CoreNLP package. This allows for an in-depth analysis of syntactic depth, extent of passivisation, and use of causal structures, among many other linguistic aspects, using HENDI, an automatic readability suite. Intermediate conclusions show that industry does not have a significant impact on readability, whereas language variety, through analysis based on both readability formulae and deeper-level syntactic data, does. To further assess the impact of financial and non-financial performance, we will use General Linear Models and regression.

Jaar van publicatie:2016

Toegankelijkheid:Closed