Multidimensional Scaling Measurement and Accounting Research

Abstract

Multidimensional scaling (MDS) is a measurement tool that has been utilized recently by accounting researchers (e.g., Libby [1979], Frank [1979], Belkaoui [1980], Pratt [1982], and Bailey, Bylinski, and Shields [1983]). MDS comprises a family of geometric models for multidimensional representation of data and a corresponding set of methods for fitting such models to actual data (Carroll and Arabie [1980]). As such, it has been used in accounting research primarily to help identify structures which are otherwise not obvious in the data that underlie attitudes and perceptions of accountants and users of accounting information. Indeed, the discoverer and one of the leading developers of nonmetric MDS, Roger Shepard of Stanford, argued that the single most important purpose of MDS is to discover previously unknown structures, thereby providing new scientific insight (Shepard [1974, p. 374]). In recent years, there has been an explosive growth in the number of models and methods for MDS (for a review, see Carroll and Arabie [1980]). Some models, although mathematically elegant, have little or no theoretical basis (e.g., see Carroll and Arabie [1980, p. 625]). In addition, very few of them were developed to reflect psychological theories of similarity, which is the major concept upon which these models are based. That is, MDS models generally utilize similarity data and attempt to portray psychological distance (in a geometric representation) from un-