Using Census Bureau Calculations of Median Earnings to Infer Age-Earnings Profiles: A Cautionary Note
Hachman, Frank, Fjeldsted, Boyd, Journal of Legal Economics
U.S. Census Bureau published estimates of median earnings by age, sex and level of educational attainment are commonly used by economic experts for the purpose of projecting future earnings losses. These published estimates, however, should be used with some caution, because of a structural bias in the Census Bureau estimation procedure. We estimate that for the March 2003 Annual Social and Economics Supplement to the Current Population Survey the bias amounts to approximately $562. That is, the 3,116 Census Bureau published estimates of the median for all the economic and demographic subgroups reported in PINC-03 overstate the "true" median, on average, by something in the neighborhood of $562.
In personal injury and wrongful death cases, U.S. Census Bureau tabulations of earnings by age, sex and level of educational attainment are commonly used for the purpose of predicting future earnings of persons without a demonstrated earning capacity or for predicting the earnings growth path for persons with a demonstrated earning capacity. Historically, perhaps the most commonly employed sources of data have been Census Bureau publications in the P-60 series entitled Money Income in the United States and in the P-20 series entitled Educational Attainment in the United States. Detailed tabulations of earnings by age, sex and educational attainment are no longer published in the regular hard copy printings in the P-60 series or in the P-20 series, but are published as files located on Census web servers. Tabulations from the 2003 March Supplement to the Current Population Survey are available in Table PINC-03. Educational Attainment - People 25 Years Old and Over, by Total Money Earnings in 2002, Work Experience in 2002, Age, Race, Hispanic Origin and Sex and also in Table PINC-04. Educational Attainment - People 18 Years Old and Over, by Total Money Earnings in 2002, Work Experience in 2002, Age, Race, Hispanic Origin and Sex. Web sites for the PINC-03 and PINC-04 Tables are listed in the references at the end of this paper. Each of these Tables actually consists of numerous subtables, which (in the case of PINC-03) the Census Bureau refers to as "Parts." The microdata underlying all of these publications is generated from the March Annual Social and Economics Supplement (ASEC) to the Current Population Survey (CPS), which was formerly known as the Annual Demographic Supplement (ADS).
Judging from economic damages reports that the authors have examined in the course of litigation proceedings, many (if not most) experts base their predictions on median earnings rather than mean earnings. This should not be surprising, since at least two of the most influential forensic economists in the history of the profession have expressed a preference for the median over the mean. In perhaps the most widely used reference work of its day, the legendary Philip Eden at 16 POF 827 (1965, 827) employed Census Bureau tabulations of median income of all males with four years of high school education from the 1960 decennial Census of Population by age group to predict the future earnings (but for the injury) of a youth with a permanent total disability. And more recently, Gerald Martin (2003, 3:24-26), in one of the most widely used modern-day reference works for forensic economic experts, sets forth a procedure for projecting age-earnings growth paths that uses median earnings data from the Census P-20 series entitled Educational Attainment in the United States or from the P-60 series entitled Money Income in the United States. (Editor's Note: Martin's work is updated annually, and not surprisingly one of the most updated sections is on Earnings. In its 17th revision, DED in 2005, has the discussion of P-20 and P-60 in Â§333.1 at pages 3-30.2 through 3-30.3 and displays a footer of Rev. 17, 7/05.)
Census Bureau Calculations of Median Earnings
But before using the published medians from the P-20 or P-60 series or from the PINC-03 or PINC-04 tabulations for the purpose of projecting future earnings, economic experts ought to be aware that the Census Bureau does not calculate medians from the CPS microdata as a straightforward middle value from an array of weighted earnings observations (i. …