How are census tracts standardized across years?

I am looking to compare tract-level data between 1950 and 1980. I downloaded the data that uses the 2000 Tiger/Line + definition. My question is how the data are constructed so that I can appropriately difference two values, e.g. 1980 population minus 1950 population.

The 2000 Tiger/Line + definition leaves me to assume that NHGIS has standardized 1950 and 1980 census tracts to 2000 census tracts using some crosswalk. Is this right? If so, where can I find the documentation for this crosswalk? And in this case there should be a 1:1 complete matching of tracts across both years.

Otherwise, is it the case that the data are given in the census tract definitions of their respective year? In this case, I will need to create my own crosswalk from 1950 to 1980 census tracts.


In short, yes, it is the case that “the data are given in the census tract definitions of their respective year.”

At this time, NHGIS has no geographically standardized data or crosswalks for tracts before 1990. Our geographically standardized time series tables include 1990-2020 population data for 2010 tracts and other 2010 geographic units. Our geographic crosswalks cover the same period.

TIGER/Line files are a Census Bureau product that provides spatial definitions of various census geographies as well as roads, water features, etc. NHGIS files that have a 2000 TIGER/Line basis correspond to geometries in the 2000 version of TIGER/Line files. E.g., the coastlines follow the 2000 TIGER/Line version of coastlines, and where tract boundaries correspond to roads, those boundaries will match the 2000 TIGER/Line version of the road geometry. They will generally not match the newer, improved geometry provided in later TIGER/Line versions. An NHGIS tract file with a 2000 TIGER/Line basis does not mean the tract boundaries will match 2000 tract boundaries. They are historical tract boundaries based on 2000 TIGER/Line geometries wherever possible.

Thanks for the response! While there may be other work I am missing, I am linking a paper that has one approach to harmonizing tracts across time.

Jonathan Schroeder has published on geographical standardization, and I think his paper on target density weighting may be of interest. His bibliography, along with the papers listed in this 2017 paper, lists a number of papers (largely from the geography literature) that have tackled this issue. This work has highlighted the limitations of areal weighting, which is what Lee and Lin use for the pre-1970 data. I think (and Jonathan may chime in) that the target density weighting method may also be useful to extend over long time periods.

To standardize tract data across more than 2 census years (e.g., 1950 to 1980), I’ve used what I call “cascading density weighting.” There’s a high-level description of that method in our documentation for our 1990 data standardized to 2010 census geographies. As noted there, I provided a complete specification of the method (along with assessment and discussion) in Chapter 3 of my dissertation.