Evaluating Micro-Survey Estimates of Wealth and Saving


This paper presents an overview of changes in household wealth accumulation and saving using wealth data from three micro-level surveys: Survey of Consumer Finances (SCF), Panel Study of Income Dynamics (PSID), and Health and Retirement Study (HRS). We provide comparisons to the macroeconomic estimates of wealth accumulation and saving, explore problems in constructing household-level valuations of wealth, and assess the value of using household-level datasets to examine wealth accumulation and saving behavior in the United States.

Our first analysis compares the macroeconomic estimates of wealth from the Flow of Funds to comparable measures from the SCF, PSID and HRS. The Flow of Funds and SCF valuations of net worth correspond closely up to 1998. Yet, after1998, the SCF reports a much more rapid acceleration of wealth, concentrated in equity-type assets. The estimates of wealth in the PSID and HRS are very similar to the SCF for the bottom 95 percent of the wealth distribution, diverging only for the top five percent of households.

Second, we evaluate the extent of bias in the wealth estimates that may have developed in the longitudinal surveys due to attrition. We conclude that both surveys remain very representative of the underlying population as judged by a comparison with the lower 95 percent of households in the SCF. We also use the longitudinal data to estimate the relationship between wealth and mortality, and adjustment factors for differential mortality that can be used to adjust the age-wealth profile obtained from cross-sectional surveys, such as the SCF. The result is greater evidence of wealth decumulation at older ages.

Finally, we use the panel nature of the PSID and HRS to construct household-level measures of wealth accumulation and partition those changes between the contribution of new saving and valuation changes. The overall changes in wealth match the macroeconomic data closely, showing a secular rise in wealth-income ratios. Although the measures of saving do demonstrate consistent differences in saving among major socio-economic groups, they do not reflect the general decline in saving rates that is apparent in the aggregate data for the past two decades.