From 1940 to 2000, the long form of the Decennial Census asked respondents about their migration patterns. The Census Bureau provided the public with an array of origin-destination migration flow data products, over the decades. The county-to-county migration flows, in particular, have continued to be a staple product. For Census 2000, a migration DVD with counts of movers was produced that contained a county-to-county flow table, and tables of mover counts by characteristics.

After Census 2000, the long form data was replaced by the American Community Survey (ACS). The 2010 Census and future decennial censuses have no migration data, making the ACS a primary source of migration data. ACS started collecting data in 1996 in four test sites. The scope of the survey grew and beginning in 2005 the survey sampled housing units in all counties in the U.S. and Puerto Rico. Group quarters (e.g., college dormitories, prisons, nursing homes, military barracks) were added to the sample in 2006.

One major difference between the Census and the ACS migration questions is the time reference for the question asked about a between the previous residence and the current residence. Census 2000 asked where the person lived 5 years ago (as of April 1, 1995) while the ACS asks where the person lived 1 year ago. The time period was changed to reflect the on-going data collection of the ACS, and allows for annual estimates of geographical mobility.

The Census Bureau released county-to-county migration estimates for the first time using the ACS in March 2012. The working paper and data release provide the county-to-county flow and mover counts. This paper further analyzes the relationship of ACS migration data compared with the 2000 Census data. Because the ACS is the new source for migration data, it is important that these estimates are sound. This analysis looked for stability in the numbers, but also for reasonable changes in the data. As the estimates are determined reasonable, then mover characteristics are provided using ACS data.

We are expecting to find a larger number of movers in the 2000 Census data because it includes movers who moved between 1 and 5 years ago. We are also predicting that most of the county-to-county migration flow ratios will be relatively similar, though some outliers are expected.