Hopium? This blog is suggesting an incredible margin of victory for Harris.
VDH is the website. They are outright calling respected meta-polls FiveThirtyEight and RCP completely wrong. Their overall argument is that the Senate-race is incredibly favored in the Democrat’s favor.
I don’t know if I necessarily believe that argument. But its still interesting to think about. Discuss?
If you didn’t take the time to read the entire article you should. Even if you don’t have statistical training, you don’t have to agree with all parts or any of it, but it’s at least an effort to put together a backed up thesis on what’s happening. It’s a good article and we’ll written, but it has some major blindspots that need addressing.
First:
There are more than 3 possible explanations, and anytime you see some one trying to box in what you are allowed or not allowed to consider as possible, your hackles should raise.
Second:
The article makes little mention or discussion of the candidates, their policies, their approach to campaigning. Which is fine, the article isn’t necessarily trying to say why things are in such a way, but so much as saying, things maybe aren’t in the way they appear. And that’s a totally fair approach to take, but it assumes naivety or moreso, a kind of uniformity around the candidates and the campaigns themselves as if these people running are random effects. This is also baked into much of the statistical approaches they use in the article, which are all effectively coming out of the parameteric/ traditional stochastic world view of how to do these things. And that’s fine, but it’s an important assumption to key in on.
Third:
The point about flooding the zone has been thoroughly debunked. Sure there are more polls being released but even Times Sienna, Quinipiac, etc are showing largely the same things. Harris isn’t winning in the polls. That just is what it is. So when they say “we should listen to the data”, then they don’t, well, that’s at least a yellow card.
Something that needs to always be addressed is the possibility that Trump “is a different kind of candidate”, which touches on points 1,2. I wish I had all the data worked up and on mobile it’s just not worth my time to do so, but we need to talk about something specific about Trump, Trumpism, and polling. Look at Trump’s performance on the ground in elections relative to polling in 16, and 20, and compare that to candidates he endorsed and campaigned for in 18 and 22. Trump has a unique property where they singularly defy their polling, and this property is not transferable. Trump massively outperformed his polling in 2016, and also in 2020. 2020 should not have been close, yet it was. Likewise, Trump seems to only be able to influence down ballot races with their presence. People are not showing up for Trump endorsements the way they show up for Trump. It’s not about the party it’s about Trump. He is able to drive out demographics that don’t /aren’t represented in extant sampling approaches to polling. His campaign strategy is to find blocks of voters that arent voting and to get them to show up (which was also the Bernie strategy). He’ll always defy polling using this approach because his strategy is to literally grow the electorate.
The same kinds of issues need to be addressed on the Kamala side of the house too. These candidates don’t sit in vacuum tubes insulated from the world. Biden was dead in the water literally a year before the primary got started: the insistence on running a clearly failing President significantly damaged the Democrats chances this cycle. Democrats didn’t get to have Primary debates, which are literally billions in free marketing and an opportunity for a party to present their vision of how to proceed. It allows you to control and steer the narrative before getting going on the campaign proper, as well as to test messaging and get a better read on the electorate. Not primarying Biden will go down as one of the dumbest political malpractices of all time. Meanwhile, Biden has shown to be uniquely feckless in the face of its vassal state effectively defying orders, and committing to a policy of genocide and ethnic cleansing. The US is absolutely on the wrong side of history right now, and the Harris campaign owns that because they are the party in power. Not only that, but the Harris campaign has effectively doubled and tripled, down on thus strategy by aligning so closely with the neoconservative movement (not even popular with Republicans) that also advocates for a first strike, settler colonialism approach to global politics. The reason I bring this up, is because there are explanations for why polling looks the way that it is that don’t require massive leaps in logic or a fundamental understanding of politics, but maybe do require you to adopt a more process based or non-parametric approach to analyzing election data.
It’s a good analysis and maybe it’s right. But what I can tell you right now, is that analyses that only rely on prior conditions have a hard enough time modeling the past, let alone the future. There is a reason why our climate models are missing the massive heating that seems to defy explanations in our current approach. There is a reason why insurance businesses globally have been caught completely off guard by the scale, severity, and frequency of natural catastrophes . It’s because the past isn’t a reliable indicator of the future, and if that’s all you are relying upon, you’ll only ever be able to make predictions in typical elections for the data set you are basing them on.
Watching early vote exit polls is kind of a tough game to play prognosticator on, but it begins to give us a sense of what the polls mean, because the info is a lot more concrete than polls. Basically, polls have a segment of responses that are undecided, meanwhile exit polls don’t. The idea as I understand it is that you can contrast exit polls with polls in order to discern what that undecided vote really seems to be breaking for.
In 2016, that undecided segment broke hard for Trump. It hasn’t in any election since.
Here’s what exit polls so far say about Trump (vs 2020) and Harris (vs Biden & Obama):
Obviously, again, exit polls are subject to swings and changes over time and so it’s all contingent on this continuing, but right now the early votes exit polls are at severe contrast with the aggregators. Like, embarrassingly severe.
One remaining thing from the exit polls worth mentioning - the last minute surge of support for Trump in 2020 was largely because the Republican leadership was stalwart in telling everyone to vote only on election day. That isn’t happening this year, though, which means that Republicans aren’t going to be able to expect the same kind of last-minute surge this year. Meanwhile, the opposite seems true for Harris: a lot of early votes for Harris are first time voters or infrequent voters, and not from the pool of 2020 early voters.
So, at this point the early vote is around 40m, or 25% of total votes in 2020. In order to get back to the “surprising Trump upswell” that we’re all worried about, this trend would have to not only stop, but AGGRESSIVELY reverse course. Either that or all the exit polls are horribly wrong.
And that’s the #1 problem with polling.
Look at 538 for PA:
https://projects.fivethirtyeight.com/polls/president-general/2024/pennsylvania/
Here’s what people miss… Go down the LEFT hand column.
860LV
1,586LV
840LV
812LV
812LV
866RV
866RV
794RV
583LV
794RV
1,084LV
1,256LV
2,048LV
2,048LV
600LV
600LV
“LV” - “Likely Voters”.
Of these 16 polls currently up, 12 of them are trying to determine who is “likely” to vote, and no 2 polls use the same definition.
Generally “Did you vote in the last 2 elections?”
Yeah, every time I wind up looking deeply at polls I find more questions than answers. I recognize they’re a snapshot of a segment, not representative of the whole segment but sort of a sampling of it.
For example, the 3 polls there from Franklin, and the 4 from Morning Consult: the same methodology and around the same sample size, conducted at the same time frame. Each poll with different outcomes from their sample set.
I also recognize that as long as X% are “undecided”, the poll can’t really show anything other than trend motions. And these polls are actually kind of static. Like if you plot them all out, they don’t seem to have an upper or downward trend trajectory.
It’s frustratingly ambiguous stuff.
You mention Trump as up in 18-39 and 50-64 but down in “a bunch of other age groups.” How many other age groups are there? 40-49 and 65+ seem like only way to back them up with any significance.
Oh shoot, sorry, I meant 18-29. The groups are:
It’s worth mentioning that these groups are not equal! 18-29 is usually a very low representation, where 40-49 is pretty big, and 50-64 / 65+ are huge.
Boomers potentially saving us from Trump was not on my 2024 bingo card (I know that is from a higher base of existing Trump support but still).