Disclosure and determinants studies: An extension using the Divisive Clustering Method (DIV)

Abstract Past accounting research contains an extensive range of disclosure and determinants studies. But these studies have one major methodological drawback: the disclosure analysis is often restricted to determination of the disclosure index, that is, the sum of disclosed items, weighted or unweighted. The disclosure profile (which reflects the structure of published information) is generally not part of the research design. The objective of this paper is to introduce a divisive (descendant) clustering method, which splits the sample into homogeneous sub-groups corresponding to disclosure patterns (or profiles), for clearer determination of the financial characteristics of each group. This methodology is illustrated by a study of disclosure on provisions by large French firms. The results show that the disclosure pattern is related to provision intensity, size, leverage and market expectation, but not to profit, return and industry. This new research method is a valuable complementary tool for expanding on disclosure and determinants studies, moving from disclosure levels to disclosure patterns.