No, The Census is Not "Proving" Election Fraud
The latest allegation from the election fraud theorists is that Census data is “confirming” that Biden stole the 2020 election. This is boosted by Robert Barnes, an attorney who has eagerly taken on the mantle of finally uncovering the great election fraud of 2020.
Barnes argues that recently released Census data provides proof of "irregularities" (There goes that slippery term again, does it mean literal fraud or just unfairness? Kept ambiguous). The claim is this:
According to the Census, the recorded number of people voting in 2020 was tallied at 154,628,000. On the other hand, official results place the number of actual ballots cast slightly north of 158 million. That’s a discrepancy of nearly four million votes.
First thing I noticed when I came across this is that the 2020 census was conducted throughout the year, so I had no idea how they collected post-election data for the entire country. Turns out they didn't. Instead, the Census has relied on something called the "Current Population Survey" which is a monthly survey done by the Census using roughly 60,000 households as a sample. The CPS has methodology here, specifically starting pg 8.
The Census asked 54,000 "households" (not individuals, important distinction) questions about their voting patterns, and then extrapolated national statistics from that sample size. They got really damn close (coming in within 3% error rate), and had a much better track record in years prior. It's clear they have robust statistical methodology given the circumstances. But I really don't understand what level of precision people expect from that type of extrapolation. This is especially so given that 2020 voting patterns were dramatically different (voter participation was record-high), and the Census was still using 2010 weights when analyzing its sample (in the middle of what many consider a political realignment).
There are plenty of other issues with this methodology which would not have been present and accounted for in years prior. Two obvious ones stem from covid relocation (some people went and lived with their parents during covid lockdowns), and increased political division potentially mitigating transparency among household members. You can just imagine a typical household, where the "head" (I'm guessing this defaults to the father) is tasked with answering questions about the voting habits of everyone they live with, and it seems clear the reliability of that vector would be diminished at least to some extent given how weird 2020 was. Not to mention that a common argument during the 2020 election was that polls showing a Biden lead were unreliable because of the "shy Trump voter" factor. Are we supposed to expect that factor doesn't also exist among families?
The headline would be "Census statistical extrapolation from 54k household sample size doesn't precisely match official voter tallies". Does anyone think that's noteworthy news?
But even assuming this was a legitimate observation, what exactly is the theory here? What exactly is this observation about census data supposed to imply? It would be helpful if the election fraud theorists actually committed and pre-registered to a single cohesive theory, rather than just pirouette between whatever the latest "hmmmmm" speculation could support. One can dream.
Between an official tally registered by bipartisan and localized election officials, and a statistical extrapolation with clear methodological deficiencies, I'm not clear why the latter should be more of a reliable indicator than the former. That is, of course, unless you have already concluded that Trump won the 2020 election and you are just trying to swim upstream towards that fabled conclusion and fueling your journey with motivated reasoning.