Thread by @ulleryatintell

Last December, a group called @JudicialWatch threatened 19 counties with lawsuits over alleged “extra voters.” Something felt off when I first saw the news release back then. You can read that story here, but I want to revisit thoroughly here. bit.ly/2G4N3jW

JW claimed over 370 U.S. counties had 2.5 million more more voters than people, alleging counties weren’t removing ineligible voters (inactive voters, voters who died, moved, etc.) in accordance with a federal law and inviting voter fraud.

“According to our analysis of data released by the U.S. Election Assistance Commission this year, 378 counties nationwide have more voter registrations than citizens old enough to vote,” the group stated on its website in 2019.

JW’s analysis included the “latest” census estimates of adults per county as of the letters data Dec. 11, 2019, and an @EACgov Election Administration and Voting Survey: 2018 Comprehensive report on voter activity (filled out by counties and other election jurisdictions)

None of the four PA counties, including Bucks, and one CA county had registration rates over 100%, so why were they included in the 19 legal threats with presumably hundreds more deserving counties with these extra voters? This is what bugged me.

The 5-year 2017 census estimates were the most thorough population estimates available at the time JW sent it’s letters. I compiled estimates for adults in every county in the continental U.S. and compared that with the EAC survey to recreate JW’s analysis. I started in PA

Like JW, I couldn’t find any counties in PA with >100% registration, but my estimated rates in the 4 counties were on average 4.6% lower than JW’s estimates. JW didn’t provide it’s complete report, but I think it’s important I go over my statewide estimates now.

Most counties had over 70% registration using this method, with most over 85% and none over 95%. There are also three more counties with over 90% registration that did not get a letter from JW. It’s inconsistent, and not limited to Pennsylvania.

Speaking of the method, this is generally not an ideal way to estimate how many people are registered voters. First, the survey is not uniform. Wisconsin mostly uses municipalities as election jurisdictions, sometimes spanning multiple counties.

North Dakota also entered a -88 (Generally N/A code) for all responses in the EAC survey almost across the board. Also, we're basically comparing the daily attendance sheet in a classroom to the projected enrollment next year here. It's a flawed analysis at heart

With that in mind, when I say I found some counties with over 100% registration rates, I’m in no way claiming those counties did in fact have over 100% registration, only pointing out that I found 100 fewer counties at that rate than JW did.

These are the counties I found with 100% or more registration rates by JW’s method. It’s about 280 counties across 37 states. We’ll refer these as the counties JW could have sent letters to, the ones where their “extra voter” claims would hold the most potential weight

This is the map of all U.S. counties with over 100% registration rates by my estimates that also received a letter from JW threatening legal action. San Diego, CA, had a rate of 101% by my estimates, nearly 15.6% points under JW's claim.

To review: @JudicialWatch unreleased report appears to be based on a flawed methodology. Rates claimed by JW appear to be inflated. High registration rates do not explain why they chose the 19 counties to threaten with a lawsuit. So, what do those counties have in common?

Here’s the 2016 election results by county with counties that went to Clinton in blue and Trump in red. I went with a difference in total voter versus a percentage because the percentage map didn’t really capture the population of the counties as much. Min = -10K Max=10K

And, here’s our counties that received letters from JW. All of them went for Clinton and most were places where Clinton beat Trump by more than 15,000 votes.

Here’s our map of counties that I estimated to be at or over 100% and where they fell in 2016. Trump has more There’s 56 for Clinton and 226 for Trump here.

Here they are side by side. I think it's really hard to come away from this and not suspect the counties facing a lawsuit from a "conservative, nonpartisan" watchdog might be targeted because Trump lost there. The registration rates are just a pretext.

I don't want to get into the rhetoric of voter fraud tonight, so we'll save it for tomorrow. For now, I'll just say this: they don't have to be right, they just have to be loud.

As always, data is analyzed using @PythonPr @matplotlib @geopandas in a @ProjectJupyter notebook. @threadreaderapp Unroll please.