1990-2000 crosswalk (census tracts)

Hi there, I’m trying to harmonize 1990 data to 2000 tract census boundaries. I eventually want to get to census tract level data but understand I should start with the smallest geographic units for which data is available. However, I only see crosswalks between various years and 2010. Do you know of any crosswalks between 1990 and 2000 I could use? If not, would it be reliable/recommended to harmonize the 1990 data to 2010, then use a reverse crosswalk to harmonize 2010 to 2000?

You’re correct that NHGIS has no crosswalks between 1990 and 2000. The Census Bureau’s relationship files between 1990 and 2000 tracts conveniently include an estimate of the proportion of 1990 tract population in each intersection with a 2000 tract. This proportion is a reasonable interpolation weight for allocating 1990 tract population counts to 2000 tracts, but as you note, it would be more accurate generally to produce 1990 estimates for 2000 tracts by starting from smaller geographic units, like blocks or block group parts. Still, if you require 1990 data in 2000 tract boundaries, this would be the simplest option and may be accurate enough; it’s hard to say for every case!

Regarding your second question, NHGIS doesn’t have a reverse crosswalk from 2010 to 2000, so it wouldn’t be possible to use NHGIS crosswalks in this way. You could potentially get one from another source, but I think it’d be better to use the Census Bureau’s 1990-2000 relationship file.

That said, I’d also suggest you consider using 2010 census tracts as your unit of analysis. Then you could use NHGIS crosswalks from 1990 to 2010 and from 2000 to 2010 to produce estimates from the smallest possible units, which I’d guess to be more accurate than any simple option for estimating 1990 data for 2000 tracts. (For many short-form census statistics, you could also get 1990 and 2000 data for 2010 tracts from NHGIS geographically standardized time series tables without using a crosswalk.)

Also, based on your other recent forum post, I see you’d like to associate 2000 data with CBSAs. If you conducted your analysis with 2010 census tracts, then you could get associations between 2010 tracts and 2009 CBSAs (the CBSAs used in 2010 census data) by downloading a 2010 census tract data file from NHGIS. Tract data files from decennial census summary files contain geographic codes for all higher levels that tracts nest within, including CBSAs.