This paper shows that standard life-cycle models have difficulties accounting for two features characterizing household wealth holdings in U.S. data. First, life-cycle models predict that households with similar lifetime incomes should hold similar amounts of wealth at the outset of retirement. In the data, by contrast, there is tremendous wealth inequality within lifetime income deciles. Second, life-cycle models imply that consumption and wealth are less intergenerationally persistent than

more »

... nings. This is the opposite of the persistence pattern observed in the data. Extending the model to incorporate accidental or intended bequests, entrepreneurship, and preference heterogeneity does not help account for observed patterns of wealth inequality. These findings suggest that standard lifecycle theory fails to account for an important source of wealth inequality, which is intergenerationally persistent. * Financial support from the College of Business at Arizona State University is gratefully acknowledged. For helpful comments I am grateful to Jose-Victor Rios-Rull and to seminar participants at Arizona State University, the University of Pennsylvania, and the 2002 Midwest Macro Meetings. Some studies suggest that preference heterogeneity be an important source of wealth inequality (e.g., Samwick 1998; Guvenen 2002) . If preferences are transmitted from parents to children, this could also improve the life-cycle model's ability to replicate observed degrees of intergenerational consumption and wealth persistence. However, extending the baseline life-cycle model to accommodate heterogeneity in time or risk preferences does not improve the model's performance. While wealth inequality within lifetime income deciles increases, the model achieves this success in a way that is strongly at variance with the data. Instead of generating wealth rich households in all lifetime income deciles, the model implies that large fractions of households possess zero or negative wealth. Moreover, preference heterogeneity makes no contribution towards accounting for the intergenerational persistence of consumption and wealth, even if preferences are strongly transmitted from parents to children. Finally, I examine the role of entrepreneurship. A develop a simple extension of the basic life-cycle model which offers households random opportunities for investment in high return self-employment projects. This model overpredicts wealth inequality within lifetime income classes, even among households who are never self-employed. However, the model achieves this success in ways that differ sharply from data. The model predicts a tight association between earnings growth and retirement wealth holdings. Households who experienced positive earnings shocks prior to retirement hold large amounts of wealth, whereas the majority of households holds very little wealth. By contrast, in the data, the correlation between earnings growth and retirement wealth is essentially zero. In addition, the model implies that more than one-third of households hold no wealth (compared with 11% in the data). Selfemployment also makes almost no contribution to the intergenerational persistence of consumption or wealth. These findings lead me to conclude that life-cycle models lack an important source of wealth inequality. The fact that the model economies imply too little intergenerational persistence suggests that the missing source of wealth inequality is transmitted from parents to children. Identifying the origins of wealth inequality among similar households is an important task for future research. The last part of the paper studies whether bequests contribute to wealth inequality. Several recent studies suggest that bequests increase wealth inequality as rich parents transfer large amounts of wealth to their children. Based on a stochastic life-cycle model, Gokhale et al. (2001) find that accidental bequests contribute to retirement wealth inequality, at least in the presence of a social security system. De Nardi (2000) and Laitner (2002) find that intended bequests increase wealth inequality. Consistent with these earlier findings, bequests contribute to overall wealth inequality in the model economies studied in this paper. However, the higher wealth inequality does not result from the accumulation of large estates that are passed on from generation to generation. Instead, bequests temporarily reduce the wealth holdings of young households who expect to inherit in the future. As a result, bequests have little effect on wealth inequality among older households who have already received their inheritance. These findings suggest that even confiscatory estate taxation may not be an effective instrument for promoting wealth equality. Related literature. In addition to the research cited earlier, this paper touches on issues addressed in several strands of the existing literature. Numerous papers study the implications of life-cycle models for wealth inequality. Instead of summarizing this literature, I refer the reader to Castaneda et al. (2000). Rangazas (2000) shows that life-cycle models have difficulties accounting for wealth inequality within age classes, even if households are altruistic. Since his model abstracts from precautionary savings and accidental bequests, it is of interest to reexamine his finding in a richer model. I show in section 4 that Rangazas' findings continue to hold in the model economies studied here. Little is known about the intergenerational persistence properties of life-cycle models, except in simple, analytically tractable special cases. In a two-period overlapping generations model, Charles and Hurst (2001) show that wealth is as intergenerationally persistent as income. However, life-cycle models