Crosswalk between 1960 and 2000 PUMAs


I’m using the 1960 5% extract. I need to take the PUMAs in that sample and crosswalk them over to 2000-definition PUMAs. IPUMS offers a crosswalk here…, but I have noticed that the match is not unique. In other words, for each state & PUMA (1960), there can be multiple state & PUMA (2000) matches. I can understand that a single area might map onto multiple 2000-era PUMAs, but there is no sense of how much (a percent etc) fits into each. Lacking this, I do not understand how to use the crosswalk IPUMS provides.

Has anyone worked with this before? Help!

Thanks. Tom

Hi Tom,

When we created the 1960 PUMAs, one of our goals was to make them as consistent as possible with the 2000 PUMAs. Unfortunately, the input 1960 units (census tracts, counties, and county remainders) didn’t always match well with the 2000 PUMAs boundaries.

There were many situations where we had to actually merge 2000 PUMAs together to create a unit that matched the 1960 PUMA, and that’s what you see in the crosswalk.

I recommend actually assigning the 1960 PUMA codes (PUMA_1960) onto the 2000 data - that way you can analyze the data using a consistent footprint. If you want to analyze the 1960 data using the original 2000 PUMA boundaries, I would use the following strategy:

  1. Assign the PUMA_1960 codes to the 2000 data.

  2. For each 2000 PUMA, compute its population proportion for the PUMA_1960 code.

  3. Use the proportion to allocate data from 1960 PUMAs to the 2000 PUMAs.