The analysis utilizes 1990 and 2000 decennial census data and the American Community Survey (ACS). Collected by the U.S. Census Bureau, the ACS is the largest household survey in the United States, with a sample of more than 3 million addresses. It covers the topics previously covered in the long form of the decennial census. The ACS is designed to provide estimates of the size and characteristics of the nation’s resident population, which includes persons living in households and group quarters. The Census Bureau began collecting the ACS in 2001, but the sample was not expanded to include group quarters until 2006.

The microdata files utilized were provided by Integrated Public Use Microdata Series (IPUMS) from the University of Minnesota. IPUMS assigns uniform codes, to the extent possible, to data collected in the ACS. The 1990 and 2000 decennial census results are based on the 5% samples.

The analysis begins in 1990 because that was the first census to specifically identify unmarried partners separately from roommates. 2019 is the latest available ACS data. The Census Bureau recently announced that it will not release a standard 2020 ACS micro data sample due to collection difficulties during the pandemic.

The census and ACS do not identify all adults in a cohabiting relationship. Only cohabiting relationships that include the household head are identified. Cohabiting adults in relationships that do not involve the head of the household are classified as unpartnered in this analysis. The 2019 Annual Social and Economic Supplement collected by the Census Bureau illuminates how extensive the misclassification is. It includes cohabiting relationships not involving the household head as well as those that do. Among prime-working-age cohabiting adults (ages 25 to 54), about 6% are in cohabiting relationships that do not include the household head. Cohabiters who are not the unmarried partner of the head or the household head account for less than 2% of the unpartnered population.