A factor analytic study of financial data

Abstract

This study employs the multivariate statistical technique of maximum likelihood factor analysis to investigate the ability of financial variables to group themselves empirically into interpretable dimensions. The financial variables used in this study were selected based on their joint frequency of appearance in previous factor analytic research and in currently popular business finance texts. Five segments of the U.S. economy were factor analyzed for each of two expansionary time periods, 1973 and 1979, and two recessionary time periods, 1975 and 1980. The five segments sampled are: mining; manufacturing; transportation, communications and utilities; wholesale and retail trade; and services. The two expansionary and two recessionary time periods were chosen to allow investigation of the stability of the factor analytic results both over time and over changes in economic conditions. As an example of a possible practical application, the results from the factor analytic portion of the study were then employed in a multiple discriminant analysis of bond ratings in the manufacturing and transportation, communications and utilities segments for the years 1975 and 1979. The results of this study indicate that financial variables do group themselves empirically into easily interpretable dimensions. Moreover, these dimensions are stable over time and over changes in the economic cycle. There are differences, however, in the types of financial dimensions found in the various segments. This may be an indication of the importance of segment affiliation. Using the Lachenbruch or jackknife validation technique, an average overall classification accuracy of 61.39 percent was obtained for the multivariate discriminant analyses. The multiple discriminant results indicate that the important discriminating variables change both across segments and across changes in economic conditions.