On Comparing Data


I’ve seen a lot of people share this. Bottom line up front: this is very wrong. Judicial Watch uses two different data sets to come up with their percentages. The first data set they use is the actual registered voters as reported by different counties, I think as of September. The other number they use is “ACS 5 YR CVAP (2018).” That is “Citizen Voting Age Population (CVAP) Special Tabulation From the 2014-2018 5-Year American Community Survey.” I think it would suffice to note that this data doesn’t go beyond 2018, and we’re comparing to 2020. So, that there would be differences 2 years later is not surprising, and of course it completely invalidates any claims that numbers don’t match. You can’t compare two different data series that differ in time! It gets worse though. The CVAP spans 5 years because it’s an average of estimates over a 5 year period. There is a 120 page PDF here: https://www2.census.gov/ces/wp/2020/CES-WP-20-33.pdf that explains the statistical methodologies (plural) they use to try to come up with the numbers that they do. They talk in that PDF how they combine different data sources and different methodologies to basically come up with their best estimate of a number. They estimate (1) people (2) people over 18 (3) people over 18 that are citizens. Those are ALL estimates. Of course, that’s why the CVAP also has a margin of error (!) included in each of the rows of the report.

To sum up: Judicial Watch compared actual numbers in each county to 2 year old estimates of how many people could vote in each county, found discrepancies (well yes), and declared “353 U.S. counties had 1.8 million more registered voters than eligible voting-age citizens.”