This paper presents results from an unusual microdata set assembled by the authors and researchers at the Social Security Administration. The data set pools information from three sources: death certificates for residents of Washington, D.C. dying in 1967; Washington, D.C. estate tax returns; and Social Security earnings records. Under an arrangement worked out by Smith with the city of Washington and the National Center for Health Statistics, all (about 2,500) estate tax returns for 1967 decedents were matched with their death certificates. The match provided information on age, sex, race, place of birth, marital status, cause of death and assets and liabilities. Washington, D.C. has its own estate tax, which unlike the federal estate tax, starts at a very low ($1,000) filing level. A full description of this part of the data base and an estate multiplier estimate of the distribution of wealth in Washington, D.C. has been published elsewhere (Smith). This year, thanks to our colleagues, Frederick Scheuren and Wendy Alvey of the Social Security Administration, a procedure was worked out which permitted us to turn over to them our files and to obtain from them analytical results from matched records from our files and their records of covered earnings under the Social Security Act. The intended use of this data base is to estimate a lifetime savings model with earnings as a key determinant. We still may be able to do so, but the prospects look rather grim. In the spirit that science is advanced by knowing what doesn't work as well as wuiat does, we present below a few initial findings which show some promise and of a lot of statistical husbandry which bore little fruit. We shall proceed by first looking at differences in the levels of covered income reported by black and white workers, then at the wealth levels of blacks and whites, and finally at an attempt to predict the wealth of black and white workers using demographic variables and earnings records.

This paper presents estimates of the concentration of personal wealth in the United States from 1922 to 1969. These estimates lead us to conclude that the distribution of wealth (1) became significantly more equal in the 1930's and early 1940's, two periods of massive government intervention in the marketplace, and (2) has remained essentially unchanged since 1945.1 In what follows, we compare the wealth held by the richest 1.0 and 0.5 percent of the population to that of all persons. The wealth of the richest 1.0 and 0.5 percent was estimated by the estate multiplier technique.2 The wealth of all persons was derived from national balance sheets.3 The estimates presented here for the period before 1953 were developed by Robert J. Lampman using highly aggregated Internal Revenue Service (IRS) data. For 1953 and 1958 we use detailed estimates (from special IRS tabulations) by Lampman and Smith, modified slightly to take account of current knowledge. Estimates for 1962, 1965, and 1969 are new detailed estimates developed by the authors using microdata files of estate tax returns prepared by the IRS for its routine publications. Here we focus on the years since 1953. Information available from estate tax returns varies from year to year, so a number of adjustments were made to bring the estimates for individual years into conceptual alignment with one another. The alignment problem was exacerbated because the IRS has destroyed tapes of returns filed before 1963, leaving only Lampman's and Smith's printed tabulations for 1953 and 1958.4 It was impractical to reestimate the distributions for 1953 and 1958 by better methods based on current knowledge. Consequently, the estimates for 1962, 1965, and 1969 were made consistent with those for 1953 and 1958 * The Urban Institute and the Pennsylvania State University. The work reported here is part of the Urban Institute's research program on income and wealth distribution. The support of the National Science Foundation is gratefully acknowledged. 1 We wish to make clear that our concern is with temporal change and that we have sacrificed best estimates for individual years to achieve consistency over the time series. Individual figures have a downward bias of 10 to 15 percent from our best estimates of concentration. (Best estimates for 1969 may be found in Smith; similar estimates for other years will appear later.) 2 Detailed descriptions of the methodology and attendant problems can be found in Smith and Staunton Calvert, Robert J. Lampman, and Smith. I National balance sheets were constructed for a person's sector using data supplied by the Board of Governors of the Federal Reserve System. Helen Stone Tice did the basic work on these special sector balance sheets. Smith provides a detailed description of the balance sheet. I A further problem resulted from the fact that the IRS erased the age field from the 1965 tape. This was most unfortunate because of all years for which the IRS has coded estate tax returns, 1965 had the most detailed classification of information. The erased data was restored by a stochastic process which took into account the relationship between age and other characteristics observable in the files for 1962 and 1969.