The details on some hard current voter datasets that reinforce themselves and call into question current polling numbers that have so many key Senate and House races neck-and-neck
By Brian E. Frydenborg (Twitter @bfry1981, LinkedIn, Facebook) November 7, 2022 (with some minor grammatical/typographical/clarity fixes made November 9; would have been earlier, dear readers, but I am having my WORST case of the flu ever… get your shots!! *correction appended: this article originally misstated the year the last time midterm turnout was this high, 1912 instead of 1914)
SILVER SPRING—If polls were all we had to go by, I’d be far more worried about the current midterms culminating (more or less) tomorrow, Tuesday, Election Day. But, my weary and worried Democrats and other defenders of constitutional freedom: I come with tidings of great joy!
Let’s Talk Polls
Over the summer, polls were trending in Democrats’ favor. More recently, they have been trending in Republicans’ favor. Given the fact that by multiple measures most Republican candidates at the national- and top-statewide-levels (or almost most) are now questioning or denying the outcome of the 2020 presidential election (or quietly accepting those who do) and are thus supporting Trump’s Big Lie fascist insurrection coup effort to destroy the Constitution, free-and-fair elections, and the rule of law as the transition from political party to personality cult continues—and that most of those so-called “election deniers” are expected to win—this understandably creates anxiety among not only Democrats, but also Republicans and independents who want America to continue as a true democracy that respects process and minority rights. Collectively, the polls have gone down for Democrats in key races and have significantly lowered their chances of holding onto the House and Senate in the eyes of analysts and the predictive models they follow. With democracy itself at stake—should Republicans be able to block most of Biden’s agenda while in charge of even just the House for the next two years and then, voters blaming Biden put Trump back in White House, we may see an end to free and fair federal elections in elections after—you could say it’s time for Democrats and others willing to defend the Constitution to panic.
To those who prioritize democracy over demagoguery, though, I bear the following message: take heart, and have hope, because polling data—while the most prominently utilized data in predictive election analysis—is not the only data, and it’s possible some of that other data in certain circumstances may actually trump (sorry, couldn’t resist) the polls, and specifically in the 2022 midterms.
How? Polls are complicated: complicated to construct and complicated to interpret and judge, and even understanding what makes the best pollsters the best can be challenging. Pollsters basically base what portions of their sample—you are not going to interview everyone, literally, but a far smaller group that you hope to draw conclusions from—and/or how they will weight/adjust them towards being more appropriately male, female, rural, urban, suburban, black, white, Hispanic, younger, older, educated, less educated, etc. on a number of factors, often involving a level of guesswork and highlighting balances that pollsters think will reflect turnout considered alongside the general demographics of the country or state (especially registered or likely voters) and/or the portions of groups present in previous electorates. If they are not weighting on previous elections or the latest demographics and along census results from the American Community Survey, they may base on their own models or look at a range of models for the current election.
Furthermore, one would not include a large sample of Asian or Jewish voters in Idaho or Montana, but would include such in California or Florida, respectively. Pollsters will often try to have the proportions approach similar types of recent elections and/or other recent election cycles. For examples, midterm elections, presidential-year elections, primaries, special elections, and referenda all tend to have different demographic balances overall and there are also differences state to state, although turnout in this election is thus far breaking midterm records and thus calling into question how much previous midterms would be accurate predictors.
In addition, there is the margin of error: each poll has a +/- margin-of-error range, say, 3.5%, meaning that if, say, the numbers the polls give one candidate leading another are 51 to 48, both the 51% and the 48% could easily be 3.5% higher or lower; the margin of error says that, generally, with 95% confidence, the results will fall within that range. For polls to be “off,” the final results would have to fall outside of that +/- range. It is important to note that, given this, polls that show candidates are closer than the margin of error range should essentially be considered ties.
So what could throw polling off in an election? If, for some reason, a certain demographic group or groups was or were either significantly overrepresented or underrepresented, something that would either significantly drive up turnout or lower turnout among one group or another. Say, rural voters, or black voters, or… women.
See what I am getting at?
What I am saying for these 2022 midterms is that I am expecting there is a very good chance of a polling error missing democratic women voters’ surge inspired by the overturning of Roe v. Wade in a way that will mean victory for Democrats, who should overperform their polling predictions by at least several percentage points and therefore win most close races, that a new group of women who would otherwise not vote in a midterm will now vote (32% of eligible female voters did not vote in 2020, compared to 35% of male ones, though it should be noted that 2020 had the highest overall turnout since 1900).
Simple logic would dictate that, after the Supreme Court’s radical decision to overthrow a half-century of precedent (despite assurances hints from certain conservative justices that they would not) in the Dobbs case (and its ruling’s draft’s leak) that destroyed the constitutional protections afforded by the Roe v. Wade decision, you would see a lot more women turn out to vote than in a typical election. And this thought gave me much hope, but it was basically on a wing and a prayer along with some solid logic, and that was all I had.
Until I found more data—hard data—that suggested the polls here are off and off because they are undercounting female votes.
…Consecrated into Form
Enough with the abstract, then; let’s get into the hard data that has since given concrete form to my abstract hopes and hunches!
When I was thinking about all this, I asked myself: when was one of the last times pollsters underestimated turnout among a particular demographic group that turned out in significantly higher portions and that this caused an upset-win for the side not favored in the polls? In 2016, to name one example, with rural white voters turning out in very high numbers for Trump and their participation at that level was not anticipated by most pollsters, giving him his wins in three key swing states that were heavily favored for Clinton.
In related votes after Dobbs this year, there are multiple serious data points in actual electoral contests backing up my main thesis. First, with the Kansas referendum on allowing a lift on current protections in the state constitution for abortion rights, there had been just one poll beforehand, predicting the vote to allow tampering with abortion rights would win by four points; it was voted down by 18 points, a 22-point swing against expectations and a triumph for abortion rights.
And there have been multiple special congressional elections since, with Democrats overperforming their expectations by an average of nine percentage-points across four special elections from June-August (in the one that resulted in a Democratic victory, in New York’s 19th District, Democratic victor Pat Ryan focused on protecting abortion right as a top issue; and this leaves out a fifth special election, Alaska’s ranked-choice election, which a Democrat won and will be be discussed later).
Ok, but an abortion rights referenda and five congressional special elections are not the same as the midterms. What could indicate more specifically that female turnout would be significantly higher in this midterm election than others and that pollsters would miss this, overrepresenting Republican voters in poll tallies and underrepresenting Democratic votes, particularly women?
As I noted, polling is generally based on tinkering around with normal turnouts or models for the current year. In this case, looking at women in recent elections according to exit polls, they already generally turn out in slightly-higher numbers than men (across all major racial categories) and thus form more of the electorate, with pollsters already taking this into account.
Let’s look at a competitive swing state for this year’s midterms, Arizona. In the 2018 midterms, women in Arizona were 53% of the electorate for the U.S. Senate race to 47% for men, and 52% to 48% for men in the 2020 presidential election. One recent poll I saw in Arizona for this year’s midterms has the same portions as the last midterm there. Other midterm polls in Arizona have proportional female-male sample-population breakdowns closer to the more recent presidential election, but all the Arizona polls I saw in which I could see the breakdowns reflected some sort of preexisting gender imbalance in favor of more women voting and close to the gender breakdowns of recent elections.
To pick another state, for midterms in Ohio in 2018, it was 51% women as voters in the U.S. Senate race there to 49% men and, for president there in 2020, it was 53% women as a share of voters to 47% men (here’s two polls I checked that are close to matching the latter and one that’s in between). Try the same for more Arizona or Ohio polls, or especially other 2022 battleground states (I have), and you will mostly (perhaps always?) see the same.
Nationally, for both the presidency and the U.S. House in 2020, it was 52% women to 48% men as a share of the vote, the same for midterm House results nationally in 2018, the same for the presidential election in 2016 (current generic ballot national polls also show the same or a close gender gap in favor of women).
In general, even if there seems like there might be a slightly larger-than-average gender gap, I have seen these presented as unweighted (sometimes you just get more of respondents in a certain category randomly) and it isn’t clear that this gap was not mitigated by weighting.
Thus, most polls in the U.S. are now reflect something of that 53-51-percent-female to 47-49-percent-male breakdown in their samples and/or are adjusted by weighting to reflect this. In other words, there is already a built-in “women vote in most elections more than men” factor with most polls and has been for some time.
This means that any new surge in women voting in this midterm—particularly women registering who are far more prone to be Democrats and/or young, which would far more predispose them to prefer laws/policies that allow women to decide their own bodily and reproductive autonomy without (or just minimal) government regulation—would be missed by the current crop of polls.
The next question I am pretty sure you have on your mind is—“Well, whatever Mr. Smarty-Pants Blah-Blah, do you have any actual data that there are far more women—specifically women who would lean pro-choice—registering to vote than men for this midterm now?”
The answer is “YES! Yes I DO!”
The Current Hard Data on Voters in the Current 2022 Midterms
For the following I must thank—of all outlets—Teen Vogue, specifically an article by Fortesa Latifi (if you doubt her awesomeness, just know that her Twitter background image is of Tony Soprano in his pool with his beloved ducks), who introduced me to the unique work of political data professional Tom Bonier (CEO of TargetSmart, a political data operation) and his New York Times op-ed. Tom has been providing some invaluable takes on the current midterms and they bear much weight in supporting my thesis of a big polling miss.
Notably, he has been detecting huge rises in the portion of female voters registering to vote in the period since the Dobbs decision has been an issue.
Tom also compares this to 2020 election data, hardly a year where women were weak in turning out (they outvoted men by four points, 52% to 48%) and finds that in 2022, differences over the same period of time in 2020 were drastic, with far higher portions of women registering than in 2020 and with a significant portion of states in 2020 even having men outregister women.
Specifically, Bonier notes that for our current year they looked at 45 states and that in 41 of those 45 states, women increased their share of voter registration after Dobbs (and the four states that did not had automatic registration). With its unique ballot measure on abortion, Kansas led the way, but also among the highest states were Alaska (which had just elected its first Democrat in a half-century—and first Native Alaskan—Mary Peltola, to the U.S. House and even over Sarah Palin, but just to finish the recently-deceased Republican representative’s term for a few months; she is up for reelection on Tuesday and looks to win again), and the three key swing states of Pennsylvania, Arizona, and Ohio.
You can see the breakdown from this recent September here:
If many of those look like huge gaps, it is because they are.
And for a relative sense of how big these are, here Tom provides the data at the same time from 2020:
Not only is the gender gap far larger in favor of females compared to 2020, but about one-third of states had a gap that favored men at the time. Those gaps in favor of men have disappeared in 2020 except for literally three states, two of which (Georgia and Oregon) have automatic registration (and in Georgia, women were requesting more of a share of mail-in ballots than they did in 2020).
Bonier helpfully takes updated numbers (with even larger gaps) from October 2022 imposed on October 2020 numbers, and the differences are all the more striking:
And these gaps are not fading over time:
And if there is any doubt that the women forming the gap are overwhelming registering to preserve their rights to choose, to bodily autonomy, and to reproductive freedom, the same period shows not only major increases in Democratic share of registrations and major drops in share of registrations for Republicans but also a big bump in the portion of under-age-25 voters registering compared to the same period in 2020: in other words, young women are flocking to register to vote and to vote as Democrats relative to other elections.
In fact, in 31 out of the 45 states analyzed by Bonier in September, under-25s were increasing their vote share after Dobbs. And these are not polls that are estimates. These are sets of hard voter registration data.
To quote Tom Bonier, “Women Are So Fired Up to Vote, I’ve Never Seen Anything Like It.”
Reinforcing Current Voter Data
There are other key sets of related statistics that only reinforce my thesis.
Early and absentee voter turnout overall and in many states are up significantly since the 2018 midterms: as of the tally for the end of Sunday (and this will update today), up over 8.2% and over three million votes in absolute terms (Republicans’ share of early voting is down from 2018), with nearly 3.1 million more Democrats having voted early in this midterm than in 2018 by the tally the Sunday before that midterm (you can find roughly similar differences in many other states, including key swing states like: Arizona, Georgia, Ohio, Pennsylvania, and Wisconsin—but North Carolina, for example, bucks that trend—just to name a few).
In the past few election cycles, the early game has heavily favored Democrats, most famously in 2020, so it is interesting to note that Democrats are improving on their best area. But, you might ask, could that signify a weakening of their weakest spot: in-person voting on election day, an offset that might negate or surpass whatever proportional gains they are making in early/absentee voting?
Here’s where things get interesting: Pew, one of the most consistently reliable sources of polling research, notes that, compared to 2020, 14% more—34% vs 20% back in October 2020—Democrats polled in October indicated they would vote in person on Election Day, almost one-and-a-half times greater, compared to only a four-percent intended-increase on Election Day participation by Republicans (they were already high, at 50% in 2020). It would be one thing if Democrats’ margins over early/mail voting Republicans were offset significantly by some sort of matching inverse behavior from Republicans, but this is not happening.
Instead, while Democrats are increasing nationally and in many key Senate-race states their already larger portion of early/mail-in/absentee voting despite a major decrease in their overall portion of their votes cast this way, they are also increasing majorly their presence where they were weakest: at polling stations on Election Day; Republicans, meanwhile are also decreasing their vote share of their overall votes with early/absentee/mail-in votes, though less so, but are also far less so increasing their Election Day turnout, only by four percent to the Democrats’ 14%.
Provided the Pew data is accurate (and it usually is), this means Democrats are pretty much set to gain ground on Republicans’ in both early/mail-in and in-person Election Day voting.
Additionally, consider what was just discussed in terms of overall turnout: early voter turnout is setting recent records in many states and for the U.S. as a whole in 2022 for a midterm election and the overall vote is expected to surpass 2018, which was the biggest proportional midterm turnout of voters since 1914. If the voting margins in favor of Democrats over Republicans for early voting are higher now than they were in recent elections, and, on top of that, a much higher portion of Democrats are set to vote in-person on election day in 2022, a higher turnout seems capable of reinforcing both Democrats’ advantage with the first and its mitigation of the gap regarding the second.
This may seem a bit confusing so I will try phrasing this another way: in what are currently polling as very close elections (within those margins of error!) and knowing there is already an absolute increase in early/absentee votes for Democrats by 3.1 million votes compared to the last midterm (compared to an decease of about 0.95 million Republicans) as of tallies from closing on Sunday-two-days-before-the-midterms, and with Democrats set to increase their overall portion of their vote on Election Day by 14% (compared, again, to just 4% for Republicans), combined, this more than “suggests” a greater anticipated turnout for Democrats than polls do.
And this effect goes for those polls asking people to rank issues: abortion is also being underrepresented there because the women who prioritize it are also being undersampled and/or underweighted.
Does All This Really Mean Democrats Should be Favored?
Let’s unpack all of this: early voting and mail-in voting overwhelmingly favors Democrats; if they are outperforming their portions relative to Republicans in early voting compared to 2020 and 2018, that’s a good sign for Democrats. But with COVID not as much of a problem, Democrats voting early and mail-in are down as a proportion of total Democratic votes cast. You might be thinking: “Wait, even if the margin of Democrats voting early or absentee is higher relative to Republicans, if a lower portion of Democrats are voting early, would that not hurt Democrats, since Republicans are much stronger on Election Day?” No, again, because Democrats are making up for it by voting in-person a lot more on Election Day this year. So, again, this means that Democrats are outperforming both their margins in terms of the share of overall early voting and their share of the overall in-person Election Day voting, improving their margin where they are weakest and weakening the GOP’s advantage where it is strongest. And with more women, Democrats, and young people overall registering, and with Pew’s October survey only having a four-percent-higher share of Republicans’ total votes coming from Election Day in-person voting compared to 2020, will Republicans have enough to overcome these other powerful trends in favor of Democrats that I have highlighted?
I think not.
Taken together, all this suggests Democrats will represent a higher share of the overall votes this election cycle than in previous ones, and, alongside the far higher post-Dobbs portions of women vs. men registering to vote, the also higher portion of Democrats registering, and the higher-than-usual under-25 crowd registering, well, this adds up to some serious math in favor of the Democrats.
Polling is a lot of guesswork, but early voting data and voter registration numbers are hard numbers that are not projections based on samples: in other words, that data is based on actual behavior and factual in a way polling is not; a poll could put together a weighted sample that does not actually reflect the election turnout, as discussed earlier, but the voter registration and early voting data are simply what they are.
So this means that there are multiple data points of compelling, hard evidence based on real-world numbers and not estimates, that the current set of polls—in particular failing to account for a mass mobilization of women that should have women forming a significantly higher portion of the overall electorate than elections in the past—are significantly underrepresenting the female vote as a portion of the overall turnout and, thus, are favoring Republicans by at least several percentage points across the board.
By significantly, I mean enough to swing most key races in most key swing states, as those races are neck-and-neck and are basically polling ties within the margins of error. This means you can expect the portion of votes not only to be, in this sense, significantly higher for Democrats than the polls are indicating, but that, in those close elections, most of those races should break in the Democrats’ favor, with the gender gap making a serious—and the—difference for the Democrats in most of these marquee races, for, even though the level of the gender gap varies, in almost every state, it still favors women (and pro-choice-type women) and far more than it did in 2020, when Democrats won the presidency, Senate, and House.
All this, in the end, is heartening to me. In many past elections, people were not fired up or fired up enough. They didn’t vote because they didn’t feel enough “enthusiasm” or were sore losers that preferred another candidate did not win their party’s nomination: they were asleep at the wheel of their own democracy. Well, after 2016, by the 2018 midterms, they woke up, and by 2020, they drove the car out of the ditch they had crashed it into back in 2016, so it would be a damn shame for them to go right back into that ditch by rewarding the people who sought to overthrow the government in a coup, resulting in the first non-peaceful presidential transfer of power in U.S. history, going back all the way to 1797 overall and 1801 between parties, to hand the people supporting and excusing that ongoing insurrection attempt the very keys to the halls of power after they literally smashed those halls’ windows and smeared feces on their walls as they sought out our elected leaders with deadly intent. The initial results of this midterm election are evidence that we will not reward the traitors.
What Would It Take for Me to Be Wrong?
In an ideal world, people not of a particular group could lead proportionately in supporting a different group. In the real world, members of any group better be prepared to stick up for their own rights more intensely than anyone else. In an ideal world, we could count on men to dismantle patriarchy as much as women, even more so since they have a larger part in its construction and implementation. In the real world, more women than men are going to have to try and try harder than men in order for patriarchy to be dismantled.
To be clear: I have faith in women. I think they have been awakened in the way Japan’s Admiral Yamamoto apparently feared America would be awakened after Pearl Harbor.
I generally find Bill Burr to be funny but also sometimes crass and offensive, and you can determine for yourselves what you think of him in this clip but his point within about the WNBA’s issues with selling tickets—that not as many women and feminists attend WNBA games as men and macho-types attend NBA games—stands.
I hope I didn’t lose any of you with the tough love from Bill Burr, but the point I am making here, ladies, is that we still have a free and fair election system (they may make it harder on purpose for some of you specifically to vote, but they still can’t stop you or your vote if properly cast from being counted) and that, if I am wrong and this data somehow doesn’t mean a big surge in women voting to protect the rights of women to bodily autonomy and reproductive freedom, it’s not going be because of patriarchy, it’s going to be because, when faced with a threat from patriarchy, far, far too many women simply shrugged, didn’t mobilize to vote, or just prioritized other “issues” they foolishly perceived Republicans to be “better” on; as I noted, these elections are close, and it won’t take an insane number of you to make that difference between victory and defeat for Democrats: the vast majority of those 32% of women who did not vote in 2020 don’t even have to vote, just enough, and your rights are preserved. If women fail to do so when it is so easily in their power, too many of them will have surrendered their rights without a real fight.
To be clear, some women are the enemies of women’s rights: about a third to 40% of women think abortion should be illegal in all or most cases; (Pew and Gallup, respectively). There are more men that think this (though not dramatically), but those are hardly insignificant minorities of women. And enough men support you that only a modest increase in pro-choice women voting in this election could have a real impact.
So, again, I have faith in women and that they are going to embarrass the pollsters, but if I am wrong, well, that’s basically the only explanation, sadly, all things being equal (other than the polls somehow being skewed significantly and wrongly in favor of Democrats across the board): that not enough additional women voted, that too many thought it was another “normal” election and did not take their own destinies into their own hands when they could, that only a minor surge resulted that did not have enough impact. I wish with all my heart that all men supported a women’s right to choose, but please do not rely on us to protect your rights for you, ladies, you vote for your rights!
That may sound harsh, but if my analysis is accurate and Dobbs overturning Roe does not mobilize a significant number of new female midterm voters determined to protect abortion rights, and if Democrats come out on the short end along with abortion rights for all American women, then that would be a crushing disappointment (and I can say the same for everyone equally of all genders when it comes to protecting democracy from fascism).
Having said that, any men on the fence or who didn’t vote but can, please join those of us already doing our part…
Conclusion: Democrats Should Outperform Their Polls and Hold Congress! BUT VOTE ANYWAY!!
Having expressed my reservations and covered my ass, I really am confident the Democrats will hold onto Congress after all the votes are counted and at least increase their position in the Senate, and that a surge in women voters unanticipated by the polls will be the main reason why.
Because folks, the normal 53-51 female vs 47-49 male breakdown just isn’t going capture what is happening this year. That’s what most polls, even in the states where there have been huge increases in the portion of women registering over men, are sticking to (in fact, every poll I have looked at where they display this information clearly is within or very close to these pre-Dobbs margins). So you can safely take many of the polls you are looking at and add at least a few points to Democrats, take a few percent away from Republicans—that’s if these polls are generally accurate apart from this glaring issue—and you will have your actual outcome. And polls are estimates, but the voter registration data is actual registration data.
Nate Silver’s FiveThirtyEight recently noted that if there was a significant across-the-board polling error, it could mean either a blowout by Republicans or actual gains by Democrats, depending in which direction. Well, given what we know from what I’ve told you here, we can safely assume the latter is more likely, and that is what my premise has been: an across-the-board if varying polling error that is inflating what GOP performance will be and deflating what Democratic turnout will be.
Which sounds great if you’re not ignorant or a fascist. As I noted a while ago, we’re past normal right-left “issues,” for the survival of our democracy is stake (and I’m sorry, but poo-poo to anyone saying it was stupid for Democrats to campaign in part on saving freaking’ democracy!): to quote Gen. Ulysses. S. Grant: “There are but two parties now, traitors and patriots and I want hereafter to be ranked with the latter, and I trust, the stronger party.”
I think preserving women’s rights to bodily autonomy and reproductive freedom and standing up to fascism and authoritarianism are deeply allied fights. And, again, I think the hard voter data I outline here favors Democrats in these midterms. If we do win, THANK YOU LADIES!
In the end, though, just make sure you vote!
*correction appended: this article originally misstated the year the last time midterm turnout was this high, 1912 instead of 1914
© 2022 Brian E. Frydenborg all rights reserved, permission required for republication, attributed quotations welcome
Also see Brian’s eBook, A Song of Gas and Politics: How Ukraine Is at the Center of Trump-Russia, or, Ukrainegate: A “New” Phase in the Trump-Russia Saga Made from Recycled Materials, available for Amazon Kindle and Barnes & Noble Nook (preview here).
If you appreciate Brian’s unique content, you can support him and his work by donating here; because of YOU, Real Context News surpassed one million content views on January 1, 2023.
Feel free to share and repost this article on LinkedIn, Facebook, and Twitter. If you think your site or another would be a good place for this or would like to have Brian generate content for you, your site, or your organization, please do not hesitate to reach out to him!