Errors in Age and Sex Data in the Census Bureau's Public Use Microdata Samples

• Activity Number: 309
• Type: Topic Contributed
• Date/Time: Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics
• Author(s): Betsey Stevenson*+ and Trent Alexander and Michael Davern
• Companies: University of Pennsylvania and Minnesota Population Center and University of Minnesota
• Keywords: Census Control Total ; Weighting ; Aging ; American Community Survey ; Census 2000 ; PUMS

Abstract

This paper describes the origins and implications of a potentially serious error in the Census Bureau's Public Use Microdata Samples from the 2000 census and the American Community Survey of 2000--2006. For persons aged 65 and up, PUMS population estimates for single years of age are significantly different from figures in the Census Bureau's complete-count population data. For all ages 0--64, PUMS estimates are never more than 1 percent higher or lower than published counts. But for those aged 65 and up, PUMS estimates for each year of age differ from published population counts by up to 17 percent. These errors affect all ages from 65 up and vary by gender in a way that causes substantial distortion in the gender ratios by age. Our paper describes the problem, and discusses the impact that this erroneous data could have on research and policy.