Weighting on Recall Vote
After my “Oh NO, Not Again” post Tuesday, the NYT/Sienna poll of Florida was published. It projected a -13 percent margin for Harris-Walz. And that translated into a potential 1.5 million vote defeat in the Sunshine State
That, in turn, drove the net margin in the twenty Mega-Democratic states from 11.9 million votes in 2020 to a projected 5.0 million votes in 2024. And that gave Trump-Vance ‘the win’ in the popular vote
My heart sank.
But this morning, Nate Cohen of the NYT tried to explain away their abnormal survey result. The NYT/Sienna polls wasn’t just an outlier. It was DOUBLE the worst results since Labor Day. Cohen admitted that none of the 11 other polls of Florida in since then put Trump-Vance ahead by more than six points
So Cohen argued that those “11 recent polls aren’t traditional ‘gold standard’ surveys... They usually employ a methodological choice called ‘weighting on recall vote’ and we do not.”
He then explained: “‘weighting on recall vote’ is a technique in which pollsters ask respondents how they voted in the last election, and then weight the number of Biden ’20 or Trump ’20 voters to match the outcome of that election.”
So why poll at all?
The 2020 election stats provide us with 157 million data points nationwide, including total turnout in Florida, Biden-Harris and Trump-Pence vote tallies in Florida, and their net margin in Florida. and that data has a zero margin of error
But Cohen claimed those non-gold standard pollsters “believe the technique might help them avoid underestimating Mr. Trump yet again.”
OMG! Now, the NYT/Sienna poll is going to save us from overestimating Trump’s strength… by DOUBLING his margin?
WTF!!
That’s double-speak. It distorts and reverse the meaning of what polling is supposed to do: give us a snapshot of where the contest is right now.
Not four years ago, not two years ago, not two months ago. Not 30 days from now. But NOW!
And this week we learned, inadvertently, that:
(a) 2020 election stats are being used to weight the polling results in Trump’s favor, right now; and, consequently,
(b) the 2024 contest has reverted to 2020 stats without factoring in any of the intervening events since those votes were cast.
So, in effect, that new “weighting on recall vote” ignores Trump’s responsibility for a million American deaths during the Covid pandemic, his Federal Reserve Chair’s five trillion dollar increase in the money supply and the resultant spike in inflation, the lingering impact on 30 million real unemployed and underemployed working and middle class Americans, his role in appointing the Supreme Court Just who overturned Roe v. Wade, and his failed coup d’ etat on January 6th.
It also minimizes the impact — actually, it erases the impact from our memory — of the massive recovery efforts of the Biden-Harris administration, with nary a MAGA vote, to save lives, put folks back to work, and create the strongest, fairest economy in three generations.
Chew on this for a moment. It’s no wonder the polls are NOT moving. They’re mired, mathematically, in the not-so-distant past.
So let’s give Nate Cohen the chance to dig himself out of this quagmire and quandary. He ends his column today with: “In the past, almost every poll would have had a chance to capture this kind of shift [in Florida voter registration]. In fact, that’s the whole point of polling. Now, many polls are designed to ensure they don’t show it at all.”
Really? Is that all you got?
If so, one has to wonder how far off those non- and pure- gold standard polls would be, IF they weren’t weighted at all. Or, alternatively, the polls were “weighted by calls refused?”
Each poll conducted makes over forty thousand dials to achieve their “random” sample. A big data voter match of all those forty thousand voters would tell us far more about an electorate than manipulating 600 to 1,200 completed interviews ever could.
I’m being sarcastic.
But the weighting of polls by geography, demographics, economics, education, and now voting history means nothing is sacred nor secret. And every question answered adds to Big Data’s trove of information about each of us, individually.
But I digress. A bit.
Back in 1948, voters WITHOUT telephones led the Chicago (Daily) Tribune to run with its famous and erroneous banner headline: Dewey Defeats Truman.
Perhaps, those “calls refused” are the equivalent of phone-less households?
Maybe, just maybe, those “calls refused” are skewing the accuracy of today’s polls in ways that neither algorithms nor the actual manipulating of the results can correct for.
And we should wait, not weight, on votes cast and counted.
Vote! Vote!! Vote!!!