1991 Volume 42 Issue 3 Pages 254-274
Using a battery of specification error tests, I find several regime shifts in the Japanese money demand since early 1960's to the present in both M1 and M2CD. I conclude that the real partial adjustment model a la Goldfeld with appropriate interest rate variable, to be solid in terms of sign conditions, significance of estimated coefficients and tests of misspecifications. Estimated M2CD function displays a remarkably solid forecasting ability, and it tracks the rapid growth of Marshallian K2 after 1986.