Raw Signal & the Generic Ballot

Lately, there’s been a lot more chatter about the 2026 midterm elections. August is always a dry spell as candidates finish up their spring primary election crush and sweat out the summer parade season. August always feels like the month where everyone who works in politics takes a big, long breath. Then we hit September and everything starts to crank up to 100 with what feels like no end in sight.

Since you’re likely to start seeing a lot more news about polls and the horse race to finish out the last ~30 days of the election, we wanted to talk about what a poll actually measures, what the generic congressional ballot tells us, and finish off with why a single poll with +/-3% margin of error is effectively a coin flip.

What Does a Poll Actually Measure?

And why should you care about it? Well, a poll measures one moment in time. It takes a small sample of folks in a targeted demographic and then attempts to measure their vibes before extrapolating it out and asserting broader assumptions to the broader public. More simply, polling attempts to say “Likely voters feel bad about the way Governor Cox and President Trump are governing.”

Ok, but what does that actually tell us? We’d argue - not much.

Here’s the breakdown, this was based on a real poll conducted by Noble Predictive Insights “conducted from September 8-12, 2026, surveyed 602 registered voters in Utah, yielding a margin of error of ± 3.99%.” This poll showed that Governor Cox saw a precipitous drop in approval from “56% a year ago to 42% today” and that President Trump saw an almost equal drop from 54% approval last year to 43% approval today.

If a poll is measuring how people feel when they were interviewed then what you’re missing from this poll is the price of diesel today. While the poll was in the field, diesel was sitting at $5.90/gallon, whereas the price of diesel as of this writing is $6.51. A $.61 cent diesel price increase in about two weeks is going to come with some major consequences at the grocery store, during your midnight online shopping binge, and for your favorite restaurant. Our economy runs on diesel and price shocks like this have a lag effect before they become reality.

Let’s take another example, Peter Crosby (D), who is running against Blake Moore (R) for congressional house district 2, ran a poll “in early September…using commercial SMS polling software” and went on to get a headline that says “Recent partisan polling suggests that 2nd Congressional District race may be a dead heat”. Ok - we have problems with so much of this already. First off, if you aren’t releasing the crosstabs - we cannot trust that you aren’t cherry picking the data to make you look good. Second, SMS polling software is the least effective at getting high-quality results. Finally, the poll write-up shows that Moore has 32.41% compared to Crosby’s 30.7%.

While it’s undoubtedly hard to know how voters will vote in this new district, it is true that in Moore’s last few elections, the Democrat’s average performance hovered around 34%. So, if you’re comparing to previous election cycles, Crosby is currently on track to underperform the historical baseline! Not a great look in a year that’s potentially a blue wave. Plus you factor in the massive discrepancy in fundraising with Moore sitting on a staggering $1,886,961.70 compared to Crosby’s measly $23,283.12 and this poll shows its true colors. It was a cheap text-message blast meant to grab a quick headline and make the race look closer than it actually is. It worked well enough to get a headline in the Cache Valley Daily but didn’t merit a mention from the more trusted SLTrib or Deseret News. It may have caught our attention, but as a reliable measure of voter intent it’s nothing but political theater.

The main takeaway here is that polls are a snapshot in time, you need to read the crosstabs so that you aren’t being told what to think by external parties, and candidate polls conducted via sms rank the lowest on our fundamentals measurements (and everyone else’s too) so you shouldn’t take them too seriously.

What’s the Generic Congressional Ballot?

The Generic Congressional Ballot is a standardized polling question that aims to measure a general feeling towards congress by asking “If the election were held today, would you vote for a Republican or a Democrat” give or take some phrasing on a poll-by-poll case.

This is then used as a general barometer to measure sentiment for or against the two major political parties. Currently, Tiny Model’s Generic Congressional Ballot shows the Democrats leading by 6 points.

Over at Silver Bulletin, Democrats have an 8 point lead. That’s wild considering in 2024 when Trump was elected to the presidency and Republicans gained control of the House and the Senate, Republicans had a 2% lead. So, an 8 point lead today means Democrats are +10 points putting places into play like Iowa, Ohio, Texas (?!), Kansas, and Alaska - all places where President Trump won.

So, What’s the Deal with +/-3% Being Effectively a Coin Flip?

Statistics, baby! Here’s the secret you rarely hear about polls, a margin of error attaches to each candidate individually, not the gap between them. If Candidate A has a 3% margin of error, and Candidate B has a 3% margin of error, the uncertainty around the leader actually stretches out into 6%. That means that when a candidate trumpets that they have a 1 point lead in a tight race, the math is actually telling you it’s a literal coin toss.

A few other factors matter here too, like the margin of error inside a sampling error, which means that if you actually answer your phone and talk to a pollster for 30 minutes you’re one of the very few who will do so. Factor in your grumpiness about $7 cauliflower and the screaming toddler in your backseat and you may find that you change your mind come election day. There’s also the fact that candidate success ranges often overlap so if you are candidate A and you are polled at 51% success, your range is actually more like 48%-54% (3% on either side!) and the same applies to your opponent, candidate B at 49% with a range of 46%-52%. There’s A LOT of overlap there. Factor in the margin of error, random chance producing a lead, an unexpected October surprise, and the fickleness of voters and you are left with the feeling that nobody actually knows anything anymore.

Back At The Kitchen Counter

Bringing it all back to the Kitchen Counter, we encourage you to think of a single poll like a pinch of salt added to a pot of soup or a dash of bitters added to a drink. Sure, those additional bits of flavor can add up over time to complete a dish and bring a candidate across the finish line. But more frequently, they are a small part of a bigger whole. We don’t advise you to judge an entire soup on its base stock until it’s finished simmering and all the ingredients have been added. Sure, you can try, but in the end salt is just salt and my margarita would taste great without that salt rim.