A multivariate time series analysis of commodity, money, and credit markets

Abstract

<p>This dissertation centered around an empirical specification of relationships among five macroeconomic variables--three quantities and two relative prices representing commodity, money, and credit markets;Granger's definition of temporal causation was the framework for investigating the interrelations of the economic time series of interest, i.e., real GNP, the GNP price deflator, the Ml money stock, total nonfinancial debt, and the 4-6 month commercial paper rate;In order to obtain the dynamic forecasts of the variables, multivariate time series models were constructed. The methodology utilized was the multivariate autoregressive moving average (ARMA) approach for building multivariate time series models. This methodology is proposed by Tiao and Box (1981) and is an extension of Box and Jenkins (1970) methodology for constructing univariate time series models;To make inferences about causal relationships, a statistical procedure proposed by Ashley et al. (1980) was employed. This approach, which is explicitly designed to test causal hypotheses in a time series context, is more faithful to the definition of Granger causality since it is based on the out-of-sample forecasting performance of the models;The causal relationships discovered in this study are as follows: The rate of money and total nonfinancial debt are Granger causally prior to the growth rate of prices or the inflation rate. The inflation rate Granger causes the nominal interest rates. And, there is a feedback relationship between the nominal interest rates and the growth rate of real income.</p>