Looking at the Polls and Wishing I Weren’t

I considered just dumping the blog below this one because it is from last week and I didn’t get around to posting it on time but it shows my work and how I decide what to review when I’m looking at polls. Aggregators like Real Clear Politics are useful because they’ve already hunted down all the wild polls, but averaging is just dumb. I will ignore the average. There are so few good enough polls available right now that I doubt I’ll take either the median or mode but if I do, I won’t be showing the work. You can find that in the blog attached below if you’re interested.

Just to get it out of the way, here’s the RCP aggregate for 10/28/24:

And the culled polls. None of this matters as this is all popular vote. We want the swing states but here ya go:

Arizona has a lot of old polls and only three current ones. CNN’s sample size is much too small. Atlas Intel is one point out of the margin of error which is insignificant. Wash. Not a tie, and not conclusive. Sure, I’m as tempted as the next snarky, cranky old lady to just pick the best looking one and run with that but that is just silly. These polls just don’t tell us anything.

Okay, from here on out I’m dropping the title and just showing the polls. Y’all can follow the link if you somehow get confused about where I got the information.

Nevada! And never mind, nothing to see here. Nothing out of the margin of error even if we ignore CNN’s insanely small sample.

Wisconsin – ugh, seriously? Are they all this bad?

Michigan – why, yes, I believe they are.

Pennsylvania – y’all are lucky I like you people. I’m already fed up with this whole project. Spot the blatant nonsense for yourselves. For those too lazy or disinterested to look at the graphics that hard, none of this stuff is significantly out of the margin of error. I’m only looking at the most recent polls. Evidently, the pollsters are as fed up as I am because they don’t seem to be doing a lot of polling.

North Carolina – are you kidding me?

Georgia – et tu, Georgia?

Well, that was a little slice of something. What ticks me off is that we’re going to spend the next four years hearing how right or wrong these polls were. In all likelihood, everyone will complain that the polls were wrong. They aren’t wrong; they are inconclusive.

Now, the reason the polls are inconclusive is more interesting. Skipping the cheating theory because only a danged fool would deliberately misconstrue their polling this close to an election we’re left with two main possibilities: The race really is extremely close or the polling is just completely inadequate.

I’m in Camp Inadequacy, myself.

Forget all the margin of error and sample size stuff – it’s important but let’s just boil this down to the simplest of terms – tiny polls only work when the pollsters have a really good idea how the various groups of voters will behave. Tiny polls are like broken scales. You know they are broken but by how much you’re having to guess.

Those guesses are called weighting. Basically, the pollsters are sticking a thumb on the scale not to mess it up but to get it to measure correctly. When they know pretty much how various groups will behave, they know how much extra weight to put in.

All of that goes out the window when the voters get all volatile. People are the hardest things in the world to measure because they get to change their minds any time they please. When they are restless or upset, they do a lot of that changing. Messes up the pollsters’ understanding of how they will behave and makes the polling next to impossible to get right.

But they still try. They need a guidepost to judge how much they should weight their polls. That guidepost is usually the preceding election.

The 2020 election which was won by extremely narrow margins in every swing state. That is being reflected in the current polling. It’s probably causing the polls to be much closer than they should be.

Probably. It’ll be interesting to see if any of the polls shown above represent the actual election poll numbers. If they are, then the race was probably really just tight.

Probably.

We’ll never know for absolute sure. I’ll be happy if we are about 70% sure. I don’t expect to be happy.

There’s another possibility I should mention. I’m assuming that the tiny number of current polls has more to do with pollsters not wanting to be the ones to tell the Democrats Trump is winning. They also don’t want to discourage voters, especially Democrats. Polls that show a significant Trump lead might just be dumped instead of reported.

A week before the election that seems a bit strange.

Maybe the pollsters simply cannot get enough people to respond. We call that ‘mortality’ meaning the folks that won’t participate. Pollsters working with such small samples need to get as representative a grouping as they can. If only Republicans will answer the questions the poll is worthless. Mortality is a huge polling problem normally, but with Democrats becoming demoralized and Republicans distrustful of anything to do with the media, it’s just possible that pollsters aren’t reporting polls because they can’t get enough people in a given group to answer them.

FYI, if so, that would NOT improve the quality of the polls that they do manage to get done.

One of the several reasons major pollsters like Gallup, Zogby and Pew don’t play in this end of the pool. The water is just too shallow for anyone to swim in.

Okay, I’m done with the polling for this cycle. Unless one of these things suddenly has significant results and no major issues, I’m waiting for the one poll we should be able to trust.

The 2024 Election.

See y’all next Tuesday night! I’ll be the cranky one in the BBQ apron!

PS: The original blog is included below as promised. I won’t be recording it so you’ll need to go to Quill Sword dot net to read it for yourself.

Here’s the second blog as promised. Yeah, it’s computer audio. I figured I’d do that much.

Polls be Dumb – But Trump’s Winning Anyway

Okay, so I’m going to post all three of the RCP graphics I created purely to show my work.

From the first you can see RCP had a bunch of polls on the list.

Nine of them are ancient. We only care about the election, not what people thought last week. So, most of these get dumped. I kept Fox because of the 10/14 end date. Frankly, it’s not a good poll and normally I’d drop it for that tiny sample size but with so few available that are young enough I kept it. Take with large grain of salt.

If I were only using one of these it would be Rasmussen purely for that larger sample size. But looking them over collectively, Trump appears to be leading. Let’s see what the median and mode tell us.

Well, the first thing is that I should have dropped Fox. Oh well. Median is the number in the middle and no, I didn’t bother with the decimal on either as it is meaningless. Not sure what I mean? Good, it really will only bother the math geeks. Trump has the higher median, 49, but we’re still in the margin of error. That makes the median a wash.

I find the higher mode for Trump interesting, but not significant. Still a wash.

If I cared about the popular vote nationwide, I’d drop everything but Rasmussen. Trump’s ahead. Yay.

But I don’t care about the nationwide popular vote and neither should you. We want to predict a winner here, right?

Before we leave the popular vote altogether, I do think it can tell us one thing – pollsters are scared of the results they are getting. Notice how few reported results in the last week. They’re calling and using the internet, how hard is it to update weekly? Well, it’s not hard at all. They aren’t doing it in order to cover their collective backsides.

They don’t want to indirectly help Trump – or have the Democrats accuse them of helping Trump – so they don’t want to release polls favoring him. There are only so many tricks you can pull to make a poll say what you want and none of them can be safely pulled two weeks ahead of the general election. Whoever TIPP is, the remaining pollsters are happy to let them stick their professional necks out. The rest are reporting older results that look better.

I really need a new monitor. I owe RCP an apology – they did not have Texas as a toss-up last week.

So, let’s look at the toss ups. I’m only looking at the newest poll on each states list. I’ll note it if I drop the newest one for some other reason but if it’s half decent, I’m just using the newest.

Arizona: CBS News – Trump +3 (51)

Georgia: AmGreatness/TIPP – Trump +1 (49)

Michigan: (deferring to Rasmussen based on sample size) Rasmussen – Tie

Minnesota: Omitted, no polls for October

Nevada: Rasmussen – Trump +2 (49)

North Carolina: Carolina Journal/Cygnal – Tie

Pennsylvania: Atlas Intel – Trump +3 (50)

Wisconsin: Rasmussen Reports – Trump +2 (49)

Nebraska CD2: Omitted, no polls for October (only one electoral vote)

Okay, so here’s where the science becomes art. Obviously, Trump is in the lead in almost all of these. While I didn’t bother to note it, most are in or right beside the margin of error. So, inconclusive if we just go by the numbers.

But there’s this little problem. None of these things have ever really gotten Trump’s races correct. Historically, they tend to underestimate Trump’s performance.

No, Trump does not ‘overperform’. That’s a lie pollsters tell so they don’t have to admit they simply got it wrong.

History isn’t determinative. Just because the pollsters got it wrong every other time doesn’t mean they necessarily got it wrong this time. BUT it doesn’t exactly give us any confidence that they actually do have the race correctly pegged, now does it?

We also have Trump’s still impressive rally sizes. So, should we assume that the polls are underestimating Trump? In my opinion, it’s a safe assumption. But let’s not kid ourselves, it is an assumption. It can be wrong.

My final reason for thinking that the polls are underestimating Trump is demonstrated in that second graphic – only four pollsters reported for this past week. There’s no technical difficulty anywhere but North Carolina that could account for that. No, they are almost certainly withholding unflattering results. The only results they are likely to do that with are those showing Trump in a decisive lead.

Trump has a rally in Madison Square Garden on October 27th. That should be very interesting. There are reports of ticket buyers buying purely to not attend in order to make the rally look smaller. Unfortunately for them, it’s first come, first served and in New York. If there are people who want to attend, they can just show up.

Given Trump’s usual draws, I expect a packed crowd.

If Trump can draw a massive crowd in a blue state like New York a second time, the race is his.

So, all that polling stuff boils down to rally attendance telling us more than the polls do.

Told you. Polls be dumb!

Spread the word!

Author: Archena

Cranky old lady with two degrees in Political Science and she ain't afraid to use 'em!