Time-Dependent Portfolio Adjustment: Yet Another Look at the Dynamics

The interest semi-elasticity of money demand has been a long standing puzzle in the monetary economics literature. Researchers consistently have estimated low short-run semi-elasticities, usually around 1, and high long-run semi-elasticities of 10. Given the crucial role of interest semi-elasticity in determining the welfare costs of inflation and the effectiviness of tax cuts, we must understand why these short- and long-run estimates are so different. To explore this issue, I formulate and estimate a model of the demand for money that simultaneously accounts for low short- and high long-run semi-elasticities. In my formulation, re-balancing money holdings between money for purchases and money for financial investment is costly. I model this re-balancing cost by assuming that households re-optimize their money holdings subject to an exogenous probability. In the log-linearized version of my model, velocity depends on both its own past value and households' present and future expectations of the interest rate. I use this equilibrium condition to estimate my model's parameters by employing generalized method of moments. My estimates for the short-run and long-run interest semi-elasticities are 0.96 and 12.62, respectively. When I apply my model of money demand to explain the increase in the volatility of real balances after 1980, my model indicates that the late-1970s financial innovations, which facilitated portfolio re-balancing, lie behind this rise.