Non-Expected Utility, Saving and Portfolios

Abstract

Existing findings suggest that standard, frictionless, expected-utility models have difficulty accounting for average and for median holdings of wealth and of risky assets, partly as a result of the largely unexplained limited proportion of stockholders among households. We analyze life-cycle wealth accumulation and portfolio choice under career uncertainty and quantifiable departures from expected utility maximization. Our specification nests expected utility and three types of non-expected utility: (i) Kreps-Porteus preferences that disentangle risk aversion from elasticity of substitution, (ii) Yaari's Dual Theory of Choice, and (iii) Quiggin's Rank-dependent Utility. Specifications (ii) and (iii) exhibit first-order risk aversion and kinked indifference curves. Solution of such models under multiple sources of risk presents conceptual and computational difficulties. We introduce a notion of equilibrium and a computational algorithm appropriate for such setups. Computed wealth and stockholding, based on calibrated income processes for three education categories, are compared to the 1992 Survey of Consumer Finances. Rank-dependent utility enhances the importance of precautionary effects. Contrary to priors in the literature, solutions are not typically at kinks; neither kinks nor actual solutions involve zero stockholding when income risk is recognized; and yet predictions about average wealth and risky assets tend to improve for all education categories. Mere disentangling of risk aversion from elasticity has small effects, while dual theory predictions are farther from the data and the signs of precautionary effects are reversed.