So...yeah. Polling has been under fire A LOT this year. Some people have said because polls were off a few points in previous election cycles, that that means polling is useless, to which I respond: "learn statistics, dingus!". We've had accusations of right wing pollsters putting their finger on the scales, which have proved to be...mostly BS. And even if it wasn't, after what happened in 2020, I'm not about to change how I calculate things because of those accusations. And then there's the "herding" accusation. Which Kyle addresses in his segment today. Basically, the argument is that pollsters are so afraid of being wrong, that we're getting what's called "herding." The pollsters end up all having suspiciously uniform polls that defy our statistical understanding of polling error, where they're all getting the same results, and the results are all roughly 50-50. Basically, no one ones to be that one outlier who got it wrong, so they're all manipulating their results to show the race being tied.
Eh...yeah, I do think that could happen. Now, am I gonna do anything about it? No. My actual argument is "throw it on the pile" and let the polling averages figure it out. I'm not doing weird things to data myself to compensate for what other people do to the data. I tried that last time, it blew up in my face. No. not doing it. But I do respect the argument.
Honestly, this is actually why those "right wing pollsters" arent manipulating the averages. because their results are literally conforming to the averages. Whatever the polling average is, they report the same average, and then other pollsters copy that, and yeah, it's like a whole class cheating on a test by looking at what the kid next to them is doing. Except then that kid is looking at what the kid next to them is doing, so the whole class outside of like 2-3 honest ones get the same answer.
It could also be why the race has trended toward 50-50. Because everyone is so afraid of being wrong so they're just going 50-50 so they can turn around and say no matter what they did....the polling was technically correct. They dont want to be embarrassed either by a 2020 underestimation of the republicans, or an overestimation in 2022. So they're just kinda hedging their bets either way so they break even.
And yeah...it can happen. Now, what do I think the reality is? I don't know. I just follow the polling. I dont DO polling, Im not a polling agency, I'm a jack### with a spreadsheet that takes data other people collect and analyze it. If they did something to it where it's wrong, i don't know. I rely on their data to get an idea of where things are, and if the data is wrong, then I'm gonna be wrong too.
Do I have a better method of knowing the truth? No. I don't trust allen lichtman's 13 keys, that's for darn sure, and I also don't trust vibes. Because vibes are subjective. And my own biases show through. if I were to go by vibes, i dont even know anyway. Because this seems like an election where the electorate is fired up, which means I expect it to be like both 2020 and 2022. BUT...with that said, in 2020, the polling way overestimated biden, and in 2022, it overestimated the republicans. We got almost the same result both times, but at the same time, that's the thing. What changed between those two election cycles was the polling. The right wing pollsters bombing the averages ended up being right as most pollsters systemically underestimated the republican vote. But then people overcompensated for it in 2022 and we ended up with dems overperforming several points.
In 2024, we could go either way, and narratives speak to it going either way. And there are so many arguments that can be thrown in either direction. I could point to the national vote being roughly tied as evidence the electoral college is down a few points, but given both the electoral college and the tipping point states have the same general polling average right now, putting trump ahead by less than a point, that could ALL be herding. Maybe the reason the national averages are matching the swing states is because of herding. In that case, we could argue HARRIS is being underestimated.
But that's the thing, we don't know. And honestly, we can point to early voting showing strong showing for republicans, but at the same time, we can also point to the fact that the early voting is less polarized than it was in 2020 due to covid not being a factor to argue against that. The fact is WE DONT KNOW. WE ARE FLYING BLIND! I'm just taking the data, making my best guesses based on the data, and then giving predictions. And it is literally a tossup.
Like here's the thing. If the race is 48-48 nationally, that can mean harris has anywhere between 52 and 44 points, and so does trump. So...that means the actual outcome can be anything between 52 trump 44 harris, and 52 harris 44 trump. It's a literal 16 point spread, based on a 4 point margin of error. And that can look radically different.
You know what, I was gonna save this until election day, but I was going to devise a version of my chart that I can add over/underperformance variables to and show you what the map looks like with it going in any direction. This will be similar to my early 2016 predictions where instead of doing probabilities, I showed what an over/underperformance would look like.
Current prediction
So this is where we stand right now. Wisconsin and Michigan are tilting blue, Pennsylvania, Nevada, and North Carolina are tilting red, and we're looking at a 43% chance for Harris, a 57% chance for Trump, with a 251-287 electoral count. This is IF THE POLLS ARE DEAD ON. Which they very well might be.
Remember, there's a reason I say that if the race is within a point on the tipping point, it's a tossup. This is what the race looks like if Harris overperforms by a single point:
Technically, this is 1.01 points, but you get the idea. I didn't wanna leave NC a tossup. So yeah, let's assume Harris overperforms a point and is able to win any state within a point. Basically, you get 292-246. NC, NV, and PA flip blue. This is why this race is a tossup. Because this is VERY EASILY happen. Keep in mind, polling is the median prediction. It's the 48-48 that can be 52-44 or 44-52. And we don't know where on that spectrum it will be. All things considered, polling actually seems to do a good job, although I admit the industry could be doing weird stuff this year that leads to an over or underperformance.
Say we get a 2016 or 2020 like scenario were Trump overperforms by 3-4 on average. What would the map look like?
Welcome back, Joe Biden. If the polls are wrong on election day, and Trump overperforms, this is what the map's gonna look like. 226-312. Heck, even with a Trump +1 map, you'd STILL get 226-312. It's a very likely combination. Because it assumed trump wins all 7 swing states, but stops short of taking any blue wall states. Look at minnesota, this is why my simulator keeps randomly spitting out red minnesota outcomes.
But yeah, if trump overperforms on election day, this is what the map will look like.
Now, at the same time, things could flip back to Harris. Say this is more like a 2022 year where the red wave is a red mirage and harris actually kicks butt? I could see it happening. Polling is WAY more red than it was in 2016 and 2020 and maybe it isnt so much shifts in the electorate, but rather pollsters overestimating Trump to avoid a 2016-2020 repeat. What would the map look like in THAT scenario?
I doubt the pink kansas thing would happen as that's likely polling error, but yeah. This is the "2022 dems overperform, everything is great, let's all be arrogant resistlibs on the internet laughing at the bedwetters" scenario. And it could happen. This is why some say the most likely scenario is actually 226-312 OR 319-219. Either can happen. As can anything in between.
Now, let's take things to the next level. Here's what I consider the 98th percentile trump outcome, basically at what I consider the margin of error. This is that "52-44 in favor of trump" scenario I'm talking about.
This is what i consider the ABSOLUTE BEST CASE SCENARIO for trump.
And here's the same for Harris:
ANY of these scenarios can happen. And we can get ANYTHING in between those two last maps. I'd say maps between the R+4 and D+4 scenarios are what's LIKELY to happen, but anything beyond that CAN still happen. So yeah. When I give like a 43-57 Trump forecast with like 251-287 outcome. That's just the middle scenario. I have no idea what the actual outcome will be.
And yeah, if pollsters are just doing polls wrong, well, buckle in, we can get some crazy crap happening that isn't expected. We could get ANY of the above scenarios. We just don't know which one until election day.
No comments:
Post a Comment