The Part 3 journey-to-work “flow” tables DO INCLUDE the Census Bureau’s (primary) allocation of workers to county-of-work, and place-of-work (and by extension, POWPUMA-of-work, since all POWPUMAs are single counties or groups of counties). So, the user can readily match the CTPP commuter flows with “standard” American Community Survey 2012-16 tables (retrieved via data.census.gov or the r-package “tidycensus”). 

On the other hand, the Part 3 CTPP tables DO NOT INCLUDE the “extended” allocation of workers, that is, allocated (imputed) to the tract, TAZ or TAD of workplace. I really can’t point to the documentation where this is made any clearer.

So, I examined the county-to-county total commuters (CTPP Table A302301) for 9 states, plus the Northern California Mega-Region (24 counties). I then compared the county-to-county total commuters to the tract-to-tract summary level, summed to county-to-county level. 

The basic finding is that the tract-to-tract total commuter file is missing about 20 percent of the workers. They were *not* allocated (imputed) to the tract-of-work level. (I can think of dozens of strategies to handle this predicament.)

Here’s a table summarizing these results.

I’ve also uploaded my R-scripts for three states: Idaho, Alaska, and Delaware, to my GIST/Github:https://gist.github.com/chuckpurvis

Can somebody double-check my work? Check it using datafiles downloaded from the Beyond 2020 CTPP and/or datafiles pulled using the r-package CTPPr. Sometimes CTPPr doesn’t work for VERY large data pulls, or even smaller areas (like Alaska state for whatever reason.)

Hopefully this is worth discussing.

cheers and Happy Pi Day!

Chuck