Crosswalk between 2010 and 2020 Census tracts

NHGIS
1

Hello. I’m interested in mapping changes in ACS data at census tract level in San Francisco between 2010 and 2020. On a map it seems like a fair number of tracts map on to each other one on one (like 0607501700), some seem split (like 06075020600 into 06075020601 and 06075020602) However, when I used the Census relationship files to get the crosswalk - there seemed to be some weird inconsistencies that I was not sure how to handle. Like for example the tract 06075012602 in 2010 is tied to 06075012602, 06075010202 and 06075012700 in 2020. But looking at the map, it seems like really minor changes and for the most part 06075012602 in 2020 is 99% of 06075012602 in 2010. So I believe my questions are

  1. Do these kinds of minor tract changes happen or is this incorrect?
  2. If there is a mismatch, Is there a crosswalk that ignores these minor mismatches?

Thank you.

2

These types of small boundary changes are legitimate and very common. The Census Bureau makes many updates to tract boundaries from one census “vintage” to another; some are substantial while many others are very small. Unfortunately, there isn’t an easy way to be certain that a small boundary change really involved no population or housing, so it would be risky to use a crosswalk that completely ignores all small boundary changes.

Have you investigated the NHGIS geographic crosswalks? As explained on the NHGIS crosswalks page, the NHGIS tract crosswalks include weights based on block-level characteristics within tracts. For the example you cite, 2010 Tract 06075012602, the NHGIS crosswalk from 2010 tracts to 2020 tracts includes relationships to the same three 2020 tracts as in the Census relationship files, but the NHGIS weights are 0 for both of the small intersections, and the weight is 1 for the main association between 2010 Tract 06075012602 and 2020 Tract 06075012602. The 0 weights indicate that there was no population or housing in the blocks where the small tract boundary changes occurred, so it’s appropriate to allocate all of this 2010 tract’s characteristics to a single 2020 tract.

Even better, instead of using a tract-to-tract crosswalk, I’d recommend you use an NHGIS crosswalk from 2010 blocks to 2020 tracts with decennial census data, or if your data of interest aren’t available for blocks, then use a crosswalk from 2010 block groups to 2020 tracts. As explained here, using tract-to-tract crosswalks results in unnecessary errors that can be avoided if you start with base data from smaller units.