Polls miss what history tells us plainly
When I was an Army intelligence officer, I expressed notable disdain for futuristic systems that required power and infrastructure to operate properly. This is not to say that technological advancement hasn’t sustained the U.S. as the premier fighting force in the world, but certain disciplines are timeless and work in all conditions.
Then, just as today, I preferred primitive methods of analyzing intelligence. I wanted to see it on a physical map and plot it on a pattern analysis wheel. I didn’t need the power to work for my products to come to life. Early on in our tour in Afghanistan, we got rocketed. No one was injured, but we took note that it was the 15th of the month, exactly. The next month, another rocket, and on the 15th. This pattern continued, and the points of origin were all similar – from a village to our south. After a few months went by, I mentioned to the commander that we had developed an ironclad case that the enemy was likely being paid to launch one rocket at us every month. History was my guide.
We were ready the next time around, and one of our Apache teams blew the little bastard straight to hell. That was the end of the rocket attacks on Shindand Air Base, and First Lieutenant Keshel played a role in snuffing them out:
Now I deal with an even more dangerous and nefarious enemy – the modern incarnation of the Democrat Party. They wage war not with Chinese rockets, but with lies, deceit, ballot fraud, and practically every institution in the world in their back pockets. It is incumbent upon me, the intelligence officer, to use all methodologies available to me to help those in the fight (like the Apaches that took the shot to end our rocket attack problem) know when to hold ’em and when to fold ’em.
The world obsesses over polling, and I’ve written so many virtual words about that topic that we could piece together my second book based solely on these rants. Still, the greatest indicator of future performance remains past performance. You can throw all of these stupid polls out, and what you’re going to be left with to help understand the baseline of the 2026 midterms are two things:
- Historical midterm results since 1934
- The 2018 midterm elections, specifically
The first reminds the observer that the president’s party has lost seats in the U.S. House in 20 of 23 midterms since 1934, for an average loss of 27 seats. Those results also show Senate losses in 16 of 23, but the fate of the Senate rests on which seats are in play. Simply put, knowing about the challenges facing a sitting president in midterms and expecting a wave election favoring his party is like not packing winter clothes when invading Russia. It’s stupid.
The second, the focus on 2018, is where true specificity kicks in. The 2022 midterm, held under a Democrat president, produced different headwinds than we can expect this year based on 92 years of history (like I tell my kids, history has already happened and doesn’t change, so learn it). It is not relevant to margins in House races in 2024, nor will it be in 2026. 2014 was Obama’s final accounting and more of a bloodbath than he faced in 2010, his first midterm. 2018 is our itch to scratch, and it tells us a lot about the shifting of coalitions in America taking place at the onset of the Trump era. This brings us good and bad news, as you will read.
Let’s look at the bad news first:
Republicans gained three seats – MN-1, MN-8, and PA-14. Minnesota continued with its 2016 coalition shift, giving the GOP two Republican seats vacated by the Democrats that held them, and Pennsylvania faced a brand new map ordered by the courts. Other than that, it was a dismal election (not counting the Senate, which featured a net GOP gain). Not one incumbent Democrat was defeated by a Republican.
Democrat incumbents are shielded this year as well, and I think only four have a risk of losing:
- Don Davis in NC-1 (Republican leaner)
- Kathy Castor in FL-14 (Republican leaner)
- Marcy Kaptur in OH-9 (decisive seat)
- Marie Gluesenkamp Perez in WA-3 (decisive seat)
Democrats won 13 open seats that were won by Republicans concurrent with Trump in 2016, and knocked off 30 incumbents. There were some losses in rural-heavy seats that had no business being losses, like IA-1, NM-2, and SC-1, but by and large, one particular group killed the GOP’s House majority:
The people who had no issue voting against Barack Obama twice, but once a tough-talking New Yorker became the face of the Republican Party – the duty to repel Democrats became optional. We were forced to obsess over the new shock term, suburban women, and listen to virtue signaling and pearl clutching from high income, college educated voters. Basically the same people who have been whining about Bo French for four days.
Someone needs to explain it to me, because I make good money, have an MBA, wrote a bestselling book, and have an extensive library yet my hatred of the Democrat agenda and its enablers has only intensified as I’ve gotten older. Either way, suburban voters absolutely crushed Paul Ryan’s majority…but they did me the future favor of giving me all the warning signs so I can properly prepare for this year.
Here is how they warned us:
They told us they had their feelings hurt by Trump back in 2016.
Let’s use CO-6 as an example. Mike Coffman held the seat for a decade and ran for reelection in 2018. Colorado had transformed rapidly in that timeframe, starting in the suburban counties that had voted for George W. Bush. Trump lost the district by nine points in 2016, as Coffman carried it by 8.3% with a bit of help from a libertarian, which gave him a higher margin of victory despite having a modest majority of the vote.
In 2018, Democrats nominated a veteran named Jason Crow to challenge Coffman, and in Colorado’s all-mail election environment, he won by double digits (11.3%).
The swing was 19.6% from 2016 to 2018 in the House. But the most interesting and valuable piece of this puzzle to me as an analyst is that we wound up within 2.3% of the presidential result from 2016. Voters split tickets in 2016 and kept their GOP House rep, and then in keeping with the tides of standard midterm environments, punished the president’s party with a nearly mirrored outcome in 2018. All the warning signs were there.
CO-6 remains under Crow’s watch to this day, and it is a safe Democrat seat. Democrats took the liberty to remove Weld County from the modern boundaries and add more communist precincts around metro Denver, but the equivalent of this seat is no longer a GOP liability to defend. Two other seats people are watching this year are CO-3 (Hurd, R-inc.) and CO-4 (Boebert, R-inc). CO-3 represents the vast western portion of the state, and CO-4 the eastern equivalent in the plains. In 2018, Democrats swung CO-3 6.2% in their favor, and CO-4 10.6%. The same swings in 2026 would tip CO-3 by just over a point, but narrowly miss knocking off Boebert. Hurd is now an incumbent, but there’s a reason I have his seat called a leaner. In a wave, it could go.
The point of the above paragraph is to remind you that specificity matters. Let me give you another classic example of a Trump 2016-18 suburban backstabber, and another that is gone for good:
Democrats now have all of Northern Virginia (NoVA) under lock and key, including VA-10 (now more competitive thanks to ambitious redraws). Barbara Comstock took this seat in the 2014 midterm red wave, held it by almost six points in 2016 concurrent with Trump’s run – running nearly 16 points apart from Trump. 2018’s suburban swings led to a nearly mirrored 2016 POTUS-2018 House result as I showed in CO-6.
This is another seat that is no longer on the GOP’s liability list.
Application for 2026
Paid subscribers will see many more graphics like those shown above. The purpose of today’s article is to show how specificity matters in 2026 targeting. Here are some 2024 facts:
- Trump won 13 U.S. House districts that have a Democrat representative. These seats are likely to have less swing to them than others, and I have 8 of these seats listed as leaners or decisive seats.
- There are only three seats that Trump lost that have a Republican representative. In 2018, there were 25 seats on the table that were lost by Trump and won by the GOP House nominee. 22 of them were lost in 2018.
Anyone looking to replicate 2018 swings and flips without accounting for the re-sorting of coalitions and elimination of nearly impossible to defend GOP House seats of a bygone era is missing a key analytical component. My coming analyses will integrate these findings in an effort to be as accurate and truthful as possible about the state of political affairs in America today.
Seth Keshel, MBA is a Senior Fellow at the Gold Institute for International Strategy and a retired Army Captain. He publishes Captain K’s Corner on Substack.
