Analyzing Data Over Time, Across Census Tract Changes


Since census tracts can split, merge, or otherwise change at the decennial census (so most recently, with the 2020 census), I'm wondering if anyone has been successful analyzing census tract data from year to year as census tracts change (i.e., in a line chart), and if so, how you went about doing that?



  • IPUMS NHGIS has multiple resources to help with this. Our geographically standardized time series tables provide 1990-2020 decennial census data for 2010 census tracts, block groups, and other units. (We plan to extend these sometime in the future to provide data for 2020 tracts, too.) For ACS data, we provide geographic crosswalks between 2010 and 2020 geographies, from block groups to tracts and other units, including high-quality interpolation weights that you can use to allocate data from one year's units to another. The crosswalks page includes extensive guidelines and documentation.

    All NHGIS resources are available for free to registered IPUMS users, requiring only that you agree to cite us and to request permission before redistributing the data.

  • Regarding your hope to analyze tract data "from year to year," you should know that no Census Bureau data products directly support the measurement of annual changes in census tracts. That would require a massive sample of the population every year. The ACS is big, but not that big, which is why ACS data for small areas are available only in 5-year pooled samples. As such, you can use ACS data to compare tract characteristics in two non-overlapping 5-year periods, but you can't directly derive measures of annual change. This Census Bureau blog post provides more background and guidance about this.

  • Social Explorer has data up  back to 1970 for tracts on the 2010 boundaries, which had slight changes as of 2012.  We plan to move these data to 2020 boundaries soon.  You should realize that tract boundaries are delineated by a cooperative program between the Census bureau and Local Authorities.  We used the fractions that John Logan of Brown created for his Longitudinal Tract file (with his permission).  For more recent years  2000, there is block correspondence file, which can be used to allocate based upon land area, or based upon population.  For the earlier years, one can use areal allocation.  The original allocation of this sort was done by Michael White and the Urban institute.  These data are used extensively to tract change.  Obviously, they are estimates.  Social Explorer is available at more than 400 institutions of higher education on a license for students, faculty and staff.  We also support short trials.  Go to for more information.  Or write to us at

    Andy Beveridge.  

  • Thanks for this info, .  I'm sorry, I should have clarified that the main data I'm trying to analyze over time by census tract is our own data and not the ACS data.  So analyzing our own data for each year isn't a problem for us, but the changing census tract boundaries make this difficult.  I can see the crosswalk files being useful for this, so I will definitely check those out!