Seeking crosswalk strategy, if feasible

NHGIS
1

INTRO: The non-profit I work with has been engaged in collaborative community-level social and environmental work in northern California for over 30 years. For that work we devised a socioeconomic scale (SES) that consists of five variables drawn from the 1990 decennial census. Fortunately, these variable are essentially unchanged in ACS data, so we continue to use the SES with communities. Most recently (2024), each of these communities was engaged to define its “cultural” boundary as an aggregation of several contiguous block groups (BGs). For this work, SES scores were straightforward to calculate. Now, however, we would like to look back in time (1990 to present) to determine if there are temporal trends in SES scores leading up to the present day.

SEEKING: I’m hoping to receive some expert comments on the above. In particular, do you believe it is feasible to use multiple geographic crosswalks to compare SES scores in 1990, 2000, 2010, and 2020?

P.S. I am aware of the statistical downside of drawing from both decennial and ACS datasets, but welcome comments anyway!

2

It is definitely feasible to do this.

To begin, you’d need to determine the vintage of area units for which you’d like to have data. It sounds like your units of interests are aggregations of block groups. Block group definitions change with every decennial census (representing distinct “vintages” of block groups). Because your project defined community boundaries in 2024, I assume that these boundaries are based on 2020 block groups specifically.

Our NHGIS geographic crosswalks support allocations of count data from 2010 block groups to 2020 block groups, but we have no crosswalks going directly from 1990 or 2000 to 2020 block groups. We intend to add those in the future, but we don’t have a timeline for that yet.

For now, you could instead “string together” crosswalks, first using our crosswalks to allocate 1990 and 2000 counts to 2010 block groups, and then using our 2010->2020 crosswalk to allocate those counts forward to 2020 block groups. The steps would be:

  1. Use the crosswalk from 1990 block group parts* to 2010 block groups to generate 1990 estimates for 2010 block groups
  2. Use the crosswalk from 2000 block group parts* to 2010 block groups to generate 2000 estimates for 2010 block groups
  3. Combine these 1990 and 2000 estimates with “2010 data”** for 2010 block groups from the American Community Survey (ACS)
  4. Join this combined dataset with the NHGIS crosswalk from 2010 block groups to 2020 block groups and use the crosswalk to allocate the “2010 data” as well as the 1990 and 2000 estimates to 2020 block groups

*We don’t have crosswalks from 1990 or 2000 block groups because it’s always best to crosswalk from the smallest possible units, and, to our knowledge, all of the 1990 and 2000 data that the Census Bureau published for block groups is also available for “block group parts,” which are generally smaller than block groups. I expect you should be able to acquire data for your SES measures for both 1990 and 2000 block group parts through NHGIS. See this section of our crosswalks documentation for help with that.

**The Census Bureau provides no SES data at the block group level for the year 2010 alone. There’s 2010 decennial census data at the block group level, but that covers only general demographic and housing characteristics (e.g., age, sex, race, ethnicity, housing tenure, household size, …). You can instead get ACS data for SES characteristics at the block group level, but that data covers different 5-year periods, not single years. To produce a time series with a decadal interval including the year 2010, I’d recommend using data for the 2008-2012 period (which is centered on 2010) to “stand in” for 2010 1-year data. Likewise, you might consider using 2018-2022 or 2019-2023 ACS 5-year data for your latest data (rather than 2020 census data or 2016-2020 5-year data).

Note: Our crosswalks are designed for count data. If any of your SES measures are for medians or quotients, see this forum post for ways you could use our crosswalks to handle those.

Feel free to follow up if you have more questions.

3

Thanks beyond measure, Jonathan! You’re outline is precisely what was needed. I expect to have a couple questions as I delve into things, but not many…looks like you’ve anticipated our needs. And we are grateful that IPUMS NHGIS exists and has great stewards! - Jeff