Estimating Money Demand in Log-First-Difference Form

An estimate of the demand for money is provided that achieves a reduction in the instability that characterizes recent estimates. Through examination of the time-series properties of the variables in the function, and adjustments to produce stationarity in them, an improved form of the demand-for-money function is developed. A comprehensive comparison of short- and long-term forecasts for demand for money in log-level and log-first-difference forms is conducted. Finally, several issues are investigated using the more stable form. The data used cover the period 1959-1983. It is concluded that estimation in log-first-difference form: 1. is preferable to the log-level form because of its stable time series characteristics, 2. is superior to log-level form for long-term forecasts, 3. implies a faster and more reasonable speed of adjustment than does the log-level form, and 4. implies that the nominal adjustment mechanism and real gross national product as a scale variable are appropriate.