Money in the Utility Function: An Empirical Implementation

This paper studies household asset demands by allowing certain assets to contribute directly to utility. It estimates the parameters of an aggregate utility function which includes both consumption and liquidity services. These liquidity services depend on the level of various asset stocks. We apply these estimates to investigate the longand short-run interest elasticities of demand for money, time deposits, and Treasury bills. We also examine the impact of open market operations on interest rates, and present new estimates of the welfare cost of inflation. James M. Poterba Department of Economics W.I.T. Cambridge, MA 02139 (617) 253-6673 Julio J. Rotemberg Sloan School of Management M.I.T. Cambridge, MA 02139 (617) 253-2956 This paper studies households' demand for different assets by allowing certain assets to contribute directly to household utility. * We permit the utility function to capture the "liquidity" services of money, certain time deposits and even some government securities. Our approach yields estimates of the utility function parameters which can be used to study the effects of a variety of changes in asset returns. We investigate how asset holdings and consumption react to both temporary and permanent changes in returns, and study the effects of government financial policy. Our approach provides an integrated system of asset demands of the form which Tobin and Brainard (1968) advocate for studying the effects of government interventions in financial markets. It provides a tractable alternative to the atheoretical equations which are commonly used to study the demand for money and other assets. Those equations, which cannot be interpreted as the rational response of any economic agent to changes in the economic environment, are unlikely to remain stable when the supply of various non-monetary assets changes, Our approach to studying asset demands is somewhat controversial. Its opponents argue that assets do not yield utility directly. They explain that rate of return dominated assets such as money are held because they reduce transactions costs, which should be modelled explicitly. Unfortunately, explicit models with transactions costs are too restrictive to be useful in analyzing aggregate data. Baumol (1952) and Tobin (1956) assume that the individual receives a constant income stream and faces a constant interest rate. By assuming that the individual consumes at a constant rate, they derive the optimal timing of financial transactions. If individuals are uniformly distributed over the time of their last visit to their financial intermediary, then aggregate money holdings are a function of the representative individual's aver-