Household Wealth of the Elderly under Alternative Imputation Procedures

Although many reach retirement with few resources except housing equity
and a claim to social security and Medicare, financial wealth, nonetheless,
makes an important contribution to the economic status of many of the elderly.
Most of our up-to-date information about the wealth of the elderly is based on
the Survey of Income and Program Participation (SIPP), which sometimes
adds an asset module to its core survey. As in many surveys of assets, the rate
of missing data on individual asset items is high, about 30 to 40 percent among
those with the asset. This raises the issue of the reliability of SIPP wealth measures because respondents who refuse or are unable to give a value to an asset
item may not be representative of the population. Indeed, in the Health and
Retirement Survey (HRS) it is clear that asset data are not missing at random.
Through the use of bracketing methods, which we will discuss below, the HRS
was able to reduce the rate of missing asset data substantially, and the data that
were added in this way increased mean wealth in the HRS by about 40 percent
(Smith 1995). Furthermore, because the additional data increased the mean so
much, they undoubtedly increased measures of wealth inequality.