A Closer Look at Long-Run U.S. Money Demand : Linear or Nonlinear Error-Correction With M0, M1, or M2?

We study annual U.S. data from 1869 or 1900 to 1999. We find evidence for a well-specified and stable model of money demand with data from 1946 to 1999. We carry out diagnostic and stability tests, including linearity tests. A linear error-correction model with the monetary base performs better than a model with M1. A specification with M2 is not supported. We use real gross national product as the scale variable and a short-term interest rate as the opportunity cost measure. We estimate an income elasticity of 0.86 and an interest rate elasticity of -0.44 for the monetary base