A FACTOR ANALYTIC APPROACH TO AN EMPIRICAL DEFINITION OF MONEY

Abstract

IN RECENT YEARS, several attempts have been made to develop an empirical definition of money. Not only have different methodologies been employed in arriving at the empirical definitions of money, but the definitions vary widely from study to study. Friedman and Meiselman [6] define money to be that set of financial assets which best explains nominal income. Friedman and Meiselman [6, p. 181] employ two criteria in determining the set of assets which defines the money supply. The first criterion is that the sum of the assets should have the highest correlation with income. The second criterion is that the sum of the assets should have a higher correlation with income than any of the components taken separately. Using these criteria, Friedman and Meiselman conclude that a proper empirical definition of money is the sum of currency, demand deposits, and time deposits at commercial banks. Using the same dual criteria, Timberlake and Fortson [13] argue that time deposits have insignificant explanatory power in predicting income, and Kaufman [10] argues that the definition of money changes depending upon whether financial assets are related to income in preceding, concurrent, or later periods. Another approach to defining money empirically is the employment of elasticities of substitution among financial assets [3], [5]. For example, Chetty [3, p. 278] estimates elasticities of substitution and uses these elasticities to take into account Gurley's [7] suggestion that financial assets be weighted by their corresponding degree of moneyness in defining the money supply. Chetty concludes that the money supply should be a weighted sum of currency, demand desposits, time deposits, deposits in mutual savings banks, and liabilities of savings and loan associations. In this paper, a methodological and statistical approach different from the studies previously cited is employed. The analysis involves a twostage procedure. In the first stage, a set of financial assets is considered simultaneously to determine relationships among the assets within the set. The dimension of the set of assets is then reduced by the multivariate statistical technique of factor analysis. In the second stage of the analysis, the Friedman-Meiselman dual criteria are applied to the factor analytic results to determine an empirical definition of money. It will be seen that the money supply is defined to be a weighted arithmetic average of the assets considered. The factor analysis model is presented in section I, and estimated results are presented in section II of this paper. In