There was an interesting comment about the 5 fake counties that I thought I should post here in full here as a tentative correction. He makes a solid case that the anomalous counties could have been the result of a transposition error in the FIPS database that assigns county IDs based first by state and then by a numerical code. (Here's a list of county ID Numbers) (transition.fcc.gov/oet/…). While I find this explanation plausible, it is unfortunate that this did not come from Decision Desk HQ.

Anyway, this is the comment by Nate J Gardner:

——————

This appears to be an error processing FIPS codes. The counties you mentioned:

Burke County, "FL"

Burke County, GA FIPS is 13033.

Escambia County, FL FIPS is 12033.

Allen County, IN is 18003

Alexander County, NC is 37003

Oklahoma County, OK: 40109

Menominee County, MI: 26109

FIPS codes are broken down into states (first two digits) and counties (last 3 digits). It appears someone made a mistake in processing these three examples. You could check the population data for the corresponding counties.

I can't explain the PA case this way though:

Stafford County, VA: 51179

Stafford County, KS: 20185

Neither correlate to a PA county FIPS code.

Also plausible that Oklahoma comes from a mistake for Oakland County?

Not sure on Stafford.

Burke County, GA (13033): 24,596 people

Escambia County, FL (12033): 321,905 people

Allen County, IN (18003): 385,410 people

Alexander County, NC (37003): 36,444 people

Oklahoma County, OK (40109): 796,292 people

Menominee County, MI (26109): 23,502 people

Stafford County, VA (51179): 156,927 people

Stafford County, KS (20185): 4,067 people

Oakland County, MI (26125): 1,274,395 people

In any case, it looks like an incorrect join where the wrong state was used processing a FIPS code that had already been parsed into two parts, plus something else going on with Stafford.

ETA's proposed mechanism for tabulator vote flipping is much more plausible than fake counties being used as a buffer. Counties don't get registered in tabulators. They simply count and validate votes. They don't process data related to multiple counties, either. It's the Election Management System that aggregates results into county-wide reporting. These also stop at the county level typically. The state then aggregates the results of the counties. There isn’t really room in this threat model where automated fake counties would make sense to be used in an attack. There isn’t a way to store extra votes at the tabulator level in a different county, and if fake counties reported their data to the state, the state would throw the data out immediately. The proposed mechanism here just doesn’t fit, I'm afraid.