A provocative recent paper by Hubbard, Skinner, and Zeldes (1994) (henceforth, HSZ) argues that an expanded version of the Life Cycle model in which uncertainty is modelled realistically can generate patterns of wealth accumulation that are roughly consistent with average data from household surveys, and amounts of aggregate wealth that are similar to observed aggregate household wealth in the U.S. If such a model really did produce roughly correct predictions for household wealth holdings, there would be little need to study the very wealthy in detail, since they would merely be scaled-up versions of everyone else.
Behind the scenes of the HSZ model, however, all is not well. While it is true that the model can predict approximately correct average values for wealth or the wealth-to-income ratio, it achieves this average by making large but offsetting errors in predicting the underlying distribution of wealth. Specifically, the HSZ model predicts, at most ages, that the household with median wealth actually holds substantially more wealth than the median household in SCF data holds and, at the same time, the model greatly underpredicts the amount of wealth held by the households at the top of the wealth distribution.
Figure 1 presents data on the age profile of the ratio of total wealth to permanent income for the median household in a stochastic Life Cycle model very similar to that of Hubbard, Skinner, and Zeldes, The figure also presents data on the age profile of the actual median household’s wealth/permanent income ratio from the 1992 and 1995 Surveys of Consumer Finances (dashing lines) during the working lifetime. The figures make clear that the HSZ model substantially overpredicts the wealth of the median household in the SCF data.