On the Cointegration of Money, Credit, Prices, and Real GDP

Abstract

While some (Friedman and Kuttner, American Economic Review, 1992) address whether money and credit separately are cointegrated with other variables, the current study finds that the natural logs of M1, domestic nonfinancial debt (DNFD), real GDP, and the implicit deflator together are cointegrated. Both maximum eigenvalue and trace test statistic results show cointegration for annual and quarterly data (seasonally adjusted and unadjusted). Further, the study finds cointegration for the lag structure of the model (with one through four lags) minimizing the Schwarz (Annals of Statistics, 1978) criterion (SC). The results generally hold whether a time trend is included in the VAR but not the cointegrating equation, or if a trend is also in the cointegrating equation. However, when using undifferenced “log levels” of all four variables [with DNFD being I(2) unlike the other I(1) variables] in seasonally adjusted form, the study finds cointegration only at confidence levels somewhat greater than 5% if a time trend is not in the cointegrating equation. Nonetheless, including a trend in the cointegrating equation produces conclusions of cointegration in both tests at the 5% level. The study uses vector error correction (VEC) models with quarterly seasonally adjusted data to test for Granger (Econometrica, 1969) causality. When specifying a VEC model with the second difference of DNFD and the first difference of the other variables, under both trend assumptions; changes in M1, real GDP, and the deflator Granger cause changes in the first difference of DNFD at the 10% level and generally at the 5% level. Using first differences of all four variables in the VEC model (with a trend in the cointegrating equation), a possible transmission mechanism appears where changes in money supply Granger cause changes in credit, in turn Granger causing changes in real GDP and prices. The results also Int Adv Econ Res (2008) 14:258–259 DOI 10.1007/s11294-008-9135-1