Two issues are addressed. First, in regard to the proper measure of wealth,current marketable wealth, expected lifetime resources, financial wealth, family wealth, per capita wealth, and equivalence-adjusted wealth can all be appropriate for different purposes. Time trends are found to differ considerably among alternative wealth indices. Second, in regard to the construction of a reliable time-series from different sources, estimates of the size distribution of household wealth are found to be quite sensitive to the sampling frame (because of the skewness of the distribution), to the accounting framework, to patterns of response error (because of portfolio variation with wealth class), and to observational unit. © 1990.