A causal analysis of several economic time series

Econometric modelling has become an increasing popular method of economic analysis. This process has traditionally involved the theoretical specification of a model, followed by the empirical estimation of the parameters of the model. In order to obtain unbiased and consistent parameter estimates a number of assumptions must be satisfied. Often an assumption about the exogeneity of a variable in the model must be made. If this assumption about exogeneity is incorrect, then a number of potential problems can occur in the parameter estimation process. Several econometric testing procedures have been developed in response to the problem of theoretically assuming exogeneity. These tests are referred to as exogeneity tests or causality tests. Sims' regression procedure, Sargent's regression procedure, and the Haugh cross-correlation method are the basic tests available. All of these tests claim as their basis the Granger definition of causality;The objectives of this study are: (1) to statistically test and analyze the causal relationships among six economic time series, (2) to compare the three basic causality tests to see if they produce consistent and conclusive results;The data set used in the analysis consisted of 523 observations from September 18, 1968 to September 20, 1978. The observations are nonseasonally adjusted weekly averages of daily observations. The data were obtained from the Board of Governors of the Federal Reserve System;The Sims' and Sargent's procedures were applied to the time series in this data sample. The cross-correlation procedure was not used because an earlier study applied this procedure to a similar but observationally smaller data set. The application of the three procedures to the same time series allowed for a comparison of the three different causality tests;The regression tests produced results which contained some inconsistencies. The causal results were generally consistent with the following three causal hypotheses: (1) Unidirectional causality from the federal funds rate to the 3-month Treasury bill rate. (2) Unidirectional causality from the interest rates to time deposits. (3) Unidirectional causality from the money supply to reserves. The application of the cross-correlation procedure in an earlier study to the same time series generated considerably different results. This difference and the inconsistencies among the results of the regression tests suggest that bivariate causality testing should be done with careful consideration.