Scottish Journal of Political Economy, Vol. zyxwvutsr 43, zyxwvut No. 2. May zyxwvut 19% Q Scottish Economic Society 19%. Published by Blackwell Publishers Ltd. 108 Cowley Road. Oxford OX4 zyx IF. UK and 238 Main Street. Cambridge. MA 02142, USA FISCAL POLICY AND AGGREGATE CONSUMPTION: NEW EVIDENCE FROM THE UNITED STATES zyx Julia Darby and Jim Malley* ABSTRACT In this paper we estimate the marginal rate of substitution between aggregate per-capita consumption and per-capita government expenditure on goods and services using zyxwvut US quarterly data over the period 1953 to 1993. This estimate is an important input to any attempt to assess the overall effectiveness of fiscal policy since it directly affects the size of the fiscal policy multiplier. Other recent consumption studies which incorporate the effects of government expenditure have failed to establish a stable estimate of the marginal rate of substitution. We argue that this failure results from imposing the unrealistic assumption that this parameter is constant. In contrast, we allow the marginal rate of substitution to depend on both the level and composition of government spending. I INTRODUCTION A central concern of macroeconomics is the extent to which changes in government expenditure translate into changes in output. Empirical analysis of fiscal policy multipliers is generally conducted using macroeconomic models (some recent US and UK examples include, Fair, 1994; Taylor, 1993; Church et al., 1993; Wren-Lewis et al., 1996). These models typically incorporate indirect crowding-out effects induced by interest rate changes, real balance effects and exchange rate movements. In addition, there are numerous single-equation studies which estimate the extent of indirect crowding-out effects (for a review of this literature, see Barro, 1990). In contrast, the extent to which government expenditure substitutes for, and hence directly crowds-out , consumption has been relatively neglected in empirical research. ' Exceptions to this include a small number of US single-equation studies (see, Feldstein, 1982; Kormendi, 1983; Aschauer, 1985; Bean, 1986; Graham, 1993). ' Whilst relatively neglected in empirical work, theoretical studies have frequently incorporated this mechanism (see, for example, Hall, 1980; Barro, 1981; and most recently Aiyagari zyxwvuts er al, 1992; Baxter and King, 1993). *University of Glasgow 129