Time Series with 2020 Geographic Integration

Will NHGIS release time series tables with geographic integration standardized to 2020 in the next month or so? If not, I assume one should use the 2010 geographic integration and apply the 2010-2020 crosswalk. Is that correct?

We do plan to release time series standardized to 2020 geographies eventually, but we haven’t set a timeline and have no immediate plans, so no, not in the next month.

Until then, my top recommendation would be to use time series standardized to 2010 if possible. If not, yes, you can use our crosswalks from 2010 to 2020 units to standardize to 2020.

If you also need 1990 and 2000 data for 2020 units, it’d be reasonable to start with our 1990 and 2000 data standardized to 2010 block groups, and then use one of our crosswalks from 2010 block groups to 2020 units to allocate the 1990 and 2000 estimates to 2020 units. The primary issues with this approach are that:

  1. Starting from block group estimates is not as accurate as starting from blocks
  2. In cases where a 2010 block group intersects multiple source 1990/2000 zones and multiple target 2020 zones, this approach would inevitably allocate some data from source zones to target zones in cases where there is in fact no overlap among the zones. (I discuss this problem in detail in this discussion of “indirect overlay” in the context of interpolation from 1990 blocks to 2010 units.) It’s better to take account of actual spatial relationships between 1990/2000 zones and 2020 zones (e.g., by using constrained direct overlay and cascading density weighting). That’s what we intend to do in constructing our time series standardized to 2020 units, and that added complexity is one reason why it may be some time before we complete that work.

Thank you Jonathan. That is very informative. So to reliably compare block group trends from 1990 to 2020, I should be using block level data where available and crosswalk from 1990 to 2020, 2000 to 2020 and 2010 to 2020. Is that correct? I understand I would need to prepare my own crosswalks except for the 2010 to 2020 which NHGIS has available.

I can’t quite provide a definitive answer because you’re getting into areas where nobody’s done the research yet! Allocating from 1990/2000 blocks to 2020 units, as you suggest, is a better way to go about things, but there are some difficulties in doing that well because we don’t have crosswalks from 1990/2000 blocks to 2020 blocks yet. Building those crosswalks is difficult because the Census Bureau never produced boundary data for 1990-2020 blocks with a consistent spatial basis for the entire period. E.g., 2020 TIGER/Line files include 2010 and 2020 block boundaries, but no 2000 or 1990 boundaries. Because the Bureau makes continual improvements to TIGER/Line feature representations, a boundary can “change” from one TIGER/Line file to another even when it hasn’t in fact changed at all “on the ground.” (There’s a section on this problem on that page detailing our 1990 block interpolation approach.)

An in-between approach would be to first use our crosswalks to allocate from 1990/2000 blocks to 2010 blocks, and then crosswalk from 2010 blocks to 2020 blocks. That still involves the second problem I noted in my previous response, but it eliminates the first problem, and the second problem is less severe with blocks than with block groups. That’d be my recommendation for now, if you don’t mind working with a few big block files!

Thank you Jonathan, the in-between approach sounds like the most reasonable method given the circumstances.