LONG RUN CONSUMPTION AND INVESTMENT POLICIES: ADAPTIVE STRATEGIES AND UNIFORM RULES

The purpose of this thesis is to examine the problem faced by an individual who must choose a portfolio risk policy and a consumption policy in each of several successive time periods, taking into account the effects of current strategies on future opportunities. The thesis discusses the results of applying the methods of numerical dynamic programming to a formulation of this problem for which no analytical solutions have been found.;The results of applying the numerical model to several sets of preference and portfolio parameters are discussed. It was always assumed that the investor's preferences for single-period outcomes displayed proportional risk aversion that increased with wealth. The optimal solutions had the following common properties: (1) portfolio risk declined with wealth, approaching an asymptote at high levels of wealth which corresponded to the solution of a simple one-period portfolio problem; (2) portfolio risk declined as the number of periods remaining decreased; (3) the fraction of current wealth consumed declined with wealth, approaching an asymptote that depended on the number of periods remaining; (4) consumption and portfolio risk were insensitive to time remaining when many periods remained; (5) intertemporal risk aversion strongly influenced consumption and portfolio risk only when wealth was low and when many periods remained. This last result suggests that analytic solutions that neglect intertemporal risk aversion may yield acceptable strategies over a wide range of values for wealth and time remaining.;The thesis also considers the appropriateness of substituting certain uniform portfolio strategies, invariant with respect to wealth and time, for the optimal strategies. A uniform portfolio strategy is found which yields acceptable results except when wealth is very low or when many periods remain.;Finally, the thesis discusses the prospects for applying these results to actual investment management problems. It is concluded that the full multiperiod model could be applied directly to the problem of an actual investor, though at some expense and difficulty. The uniform rule proposed herein could be applied much more easily and though inferior to the optimal strategy could yield a significant improvement over naive decision rules.;Compared to formulations that have been solved analytically, the model presented here makes less restrictive assumptions about the consumption preference structure of the individual. First, the investor's preferences may embody an aversion to "all or nothing" outcomes for consumption patterns. The investor may be unwilling to risk many years of low consumption for a chance to realize many years of high consumption. This property has been called "intertemporal risk aversion." Second, the investor's preferences for outcomes within any single period may depend on his wealth in that period. None of the existing analytic solutions to multiperiod consumption and investment problems have embodied both of these properties at once.