OMEGA Int. J. of Mgmt Sci., Vol. 19, No. 4, pp. 259-274, 1991 0305-0483/91 53.00 + 0.00 Printedin GreatBritain. All rights reserved Copyright ~ 1991 Pergamon Presspie Multidimensional Scaling Applied Corporate Failure C MAR MOLINERO University of Southampton, UK to M EZZAMEL University of Manchester Institute of Science and Technology, UK (Received February 1990; in revisedform September 1990) This paper uses Multidimensional Scaling (MDS) techniques to ex#ore the rdatinnshlp between a sample of financial ratios that can be used to describe the health of a firm. It is shown that compared with conventional multivariate techniques, MDS can be used to summartse complex information in an emclent and Intuitive way. The technique allows for comparisons to be made between different sets of data and across different time pe/tods. The paper explores time invariant relationships between ratios, and the differences between failed and non-failed firms. Key words--multidimensional scaling,corporate failure, performanceratios I. INTRODUCTION THE AIM OF THIS PAPER is tO offer a new perspective on the analysis of financial ratios. The number of ratios that can be obtained from financial statements and other company data sources is potentially very large; this makes it necessary to engage in informed data reduction procedures. An attempt is made here to inform the data reduction process through a careful understanding of the association between a set of financial ratios. There is also a problem of intertemporal variation: the results of the data reduction procedures may be different if differ- ent time periods are analysed, in this paper we try to identify the basic fluctuations that may arise with the passage of time. This is done using Multidimensional Scaling (MDS) techniques. Finally, we use MDS to analyse the path that firms may take towards failure, and the way in which failed companies differ from nonfailed companies up to five years before failure. The use of MDS is justified on the grounds that it offers intuitive interpretations of statisti- cal results while, at the same time, it yields the same results that would be generated by the use of more traditional statistical approaches. MDS has the further advantage of being distribution free, thus avoiding the need to test whether the conditions under which the model is valid are satisfied; and the need to transform the original data into a related data set with better statistical properties. MDS related techniques have been used before in accounting research as an alternative to more traditional multivariate statistical methods when the data is not amenable to analysis by them [10, 23, 26]. This application is extended here to the analysis of corporate failure using financial ratios. MDS can be seen as a tool which is appropriate for exploring the characteristics of the data prior to any modelling; it produces graphical representations of the structures of the data, and thus makes it possible to acquire an intuitive understanding of its structure. This, in itself, may suggest which statistical methods are appropriate for the problem at 259