Introduction

The stripped–down version of the life–cycle model implies that older consumers should start decumulating assets right after retirement at a rate sufficient to achieve zero wealth by the time of death. Of course, this extreme pattern of behavior only obtains under highly restrictive assumptions, such as lack of uncertainty about the time of death, the absence of health uncertainty and the absence of a bequest motive. Relaxing these assumptions might still lead to decumulation of assets as households get older, though at a slower rate.

Several studies, for various countries, provide evidence on this issue. By and large, there does appear to be asset decumulation by the elderly though at a much lower rate than would be implied by a life–cycle model without uncertainty or bequests. Thus, as is pointed out by Ando and Kennickell (1987), the traditional characterization of the no–uncertainty–no–bequests life–cycle model is clearly rejected by the data. While there is a fair consensus on this statement, there is much less agreement on the reason for the observed slowness of decumulation. As already mentioned, two main explanations have been proposed: the precautionary saving induced by (uninsurable) uncertainty about the time of death or by the possibility of major catastrophes in old age that require large outlays; and the desire to pass part of the accumulated assets on to one's heirs.