Election Update 10/18/24

 Okay, so, weekly update for my presidential predictions.

How election day is likely to work this year

So, normally on election day, I do my prediction over two days. I construct the chart the day before, give my semi final prediction, and then i check the last polls the day of and update my final predictions before the polls close.

However, this year, since I switched my formal to excel/sheets, I'm going to be inclined to do things slightly differently this year. I'm going to have a semi final prediction I put out the night before, which will act as my final prediction in case something happens on election day which stops me from getting a prediction out, and then I'll release an updated prediction the next day with whatever additional polls come in overnight. All predictions will be locked in before polls close on election day, and then we'll see how things go. 

I just wanted to get that straight. With that said, let's do predictions.

Presidential election

So, yikes. This isn't good, we're back to the same old 226-312 we had when Biden was running, although Harris's odds are much better than Biden's (remember, Biden peaked at like 33%). Still, this isn't looking good for harris, which begs the question, what does this mean?

Well, it seems like it's no mistake that harris has lost a few steps since taking over the role as candidate. She was really strong at first, but then reality set in as we had the convention (which is where the new car smell wore off for me), and with recent interviews, yeah it really does feel like we're stuck with Biden again, just less senile. This is because Harris won't disavow herself of the Biden administration and promise a clean break from it. She's running like your normal new democratic candidate from 1992, running to the center, promising to be no different than the last guy (who no one wants again), and she's running around touting her dick cheney endorsement and promising to put republicans in her administration. Gee, thanks. You know Harris, no one actually WANTS that. i know the brain dead DNC consultants keep telling you this, but no. We want something else.

Heck, I was watching the news earlier talking about how trump's "no tax on tips" policy is landing and how harris basically stole it for herself too, and i think it captures a populism that harris could have done her own way.

It seems like what people really want is a better living standard. You know what would give that to them? A UBI. Or a NIT. Or your LIFT act. You know better, Harris, come on. You know these policies work, but you won't push them. Even I figured this election we should back off given inflation woes and how giving people money would be framed as causing inflation, but at the same time...you know what also would cause inflation? Tax cuts. Fiscally, it's all the same. it doesnt matter if the negative on the government's balance sheet comes from more spending or lowering taxes. And you know what? Harris could promise an income grant to all americans and in this era where people wanna recover from inflation, maybe it would go like hot cakes, as long as we frame it properly.

But we do need to have that discussion. It's like the whole negative income tax vs UBI thing. They do the same thing. There's different philosophies behind the ideas, but functionally, what's the difference between a UBI that promises $15000 a year with a 20% tax rate, or an NIT that has a $15000 minimum income and a clawback rate of 20 cents for every dollar earned, with a 20% tax on those making above $75000? They're the same policies, just done differently. Conservatives balk at the idea of giving people money from the public treasury and there's even that quote about how it will bring about the doom of a nation, but a tax cut is functionally the same. It's just framed as giving people their own money back. But functionally, they have the same net effects on the government budget, and peoples' pocketbooks. And people do like free money if it's framed the right way. So maybe Harris should be more populist on that front. Just saying. But I digress. 

The point is, Harris is dropping the ball. She's not resonating as well as she could and to some extent trump outflanks her, at least in optics, and that's not good. This is why we're losing.

Another thing I did want to discuss relates to the poll bombing thing. As I keep saying, I use RCP. I've used RCP since 2008. Back in the day, it was the only game in town, and looking at its history, it was one seen as the "dow jones of political polling", but, it's come under fire in recent years as being biased toward trump, and the averages it gives seem to be more right leaning than other sites. if anything, RCP is an outlier. And it's because they dont add a lot of polls to their averages.

I go with RCP because of their track record, being relatively "conservative" in updating my methods (as in, i dont like change, if it aint broke dont fix it), and the fact that the other sites try to inject their "expert analysis" into things (see: 538), and I dont want other peoples' analysis. I want the data, and I want to analyze it and do my own projections. That's why i do this.

As such, while I dont think I'll change my aggregator this time, as we're in october and i dont want to radically change my methods right before an election (remember how it screwed me up in 2020?), but I'll say this: if RCP ends up grossly overestimating trump, i'm gonna try something else in the future. I notice RCP doesnt account for a lot of polls that are out there, although debatably many of them arent great, but it also throws in crap from right wing pollsters, which is where the accusation of bias comes from. Again, last time I feel i was too reactive to this and decided to "unskew" the averages causing me to be WAY off, so I won't shift my methodology at this time, but I'll be reconsidering my methodology for future elections if this whole thing goes south and Harris ends up grossly overperforming.

At the same time, to be fair, the other aggregators are reporting similar movement as RCP, they're jsut maybe 1 point to the left of it, with the race being closer to 50-50, whereas I'm at 41-59 Trump. So maybe it's not a big deal, but yeah, RCP is the only ones actually giving us 226-312 right now. Most are giving some mix of the states going different ways, but yeah. We'll see who ends up correct. To be fair this election is so close that even if harris wins, methodology might not mean much, polling error will simply be larger than the difference that different aggregators gives us. But I will be considering shifting where i get my polling data from if I'm not happy with my final prediction's results.

But yeah. Not good for Harris, Trump has the energy right now, and the race has been trending in his direction for 3 weeks now and it's scary. At this rate, I kind of do expect Trump to win, although with these margins and probabilities, anything can still happen, and it's still anyone's game. 

Here's the trend model:

Senate

So the blue wall continues to erode in the senate. I've been watching polling average tick downward for a while now, and now Ohio is being rated as a tossup. In all fairness, RCP just dumped a bunch of old polls which may have played a role in that outcome, hence another reason why I'm getting frustrated with them and why I can't help but feel like they may be manipulating polling averages to have a more conservative slant on the data, but again, don't want to really deal with that until after the election. We'll see how they do and if they end up dropping the ball, we'll see what other aggregates are out there to use by 2028. 

But yeah back to the data, 50-50 is still the "most likely" result. Now, you might be wondering why, fi that's the case, the republicans have a greater chance than a tie. It's because of this: by most likely, I mean the median outcome. The outcome where if the polling data is 100% accurate, this is what will happen. Right now, it's a 50-50 tie. Dems still get Ohio, other states beyond that go R. Of course, in terms of the overall spectrum of possibilities, 44% of them have republican outcomes and only 43% have ties. What matters is where the middle of that spectrum is. In a presidential race where a tie will only come up if 269-269 comes up in my linear model, ties just dont manifest very often. But because of the nature of the senate and how things seem to trend toward 50-50, well, that outcome does come up quite often. 

Speaking of which, let's look at the simulators. 

Simulations

So, as I've been saying, I've been upscaling my simple simulator to spit out 1000 outcomes at once. I'll look at whether the efficiency of such a monstrosity is worth it in another article (as an expansion of my previous article discussion how automation sometimes makes more work), but suffice to say...it does give me a lot of data really quickly.

So for the presidential:

Here we have a random sample of 1000 outcomes with us getting 524 Trump ones, 452 Harris ones, and 24 ties. This would be akin to a 52 Trump, 45 Harris, and 2 Tie situation in a normal sample of 100 outcomes. So, the simulation finally spits out outcomes that lean in Trump's favor, despite Trump leaning more red for much longer. I was considering swapping to just a big 7 version of my simulator to focus on swing states, but I won't, and here's why. While Texas and Florida going blue every so often do throw off my model a little, let's go back to 2016. I primarily looked at tossup states that year like florida, north carolina, nevada, ME2, and new hampshire that cycle. If I saw the simulator giving me trump outcomes where trump was spitting out wisconsin, michigan, and pennsylvania in a minority of simulations going red, i wouldve believed that couldn't have happened in reality much. But it did. My simulator liked to especially show michigan, colorado, and pennsylvania defecting in the minority of results where trump won. My simulator showed a weakness in hillary's campaign strategy that most overlooked, and then it happened. So...if this is showing a potential weakness in trump's strategy where we could see a southern red wall flip blue in the case of a harris overperformance...I say, let it ride. Not like it's throwing my similator off big time. As you can see it's only happening about 7.5% of the time. 

But yeah. This simulator is, nevertheless, flawed. If anything, it's gonna be too conservative. My official forecast assumes a trend model where if one state overperforms others will too, but this one treats all states as unrelated trials. It kind of just brute forces different combinations the elections CAN go. Many of the outcomes are probably utter nonsense if the trend model is accurate. But it does show what CAN happen when randomness is allowed to do its thing. And we currently get a nice bell curve of outcomes centered around 270. The median and the mean actually give me 268-270 outcomes in trump's favor. Heck, it's so conservative, results that show all 7 swing states go in one direction or another are a statistically insigificant minority for the most part. The 98th percentile in Harris's favor is a 320-218 outcome which is probably 319-219+ME2. For Trump, the 98th percentile outcome is the 312-226 outcome. So yeah. It really does expect a lot of states to flip both ways and mostly generates closer outcomes, rather than more extreme ones. And because of texas and florida, it DOES favor dems slightly because of that.

Still, this outcome in net doesn't give me much different results than 538's model. They currently have 519 Trump outcomes, 478 Harris outcomes, and 3 ties. So once again, let it ride. We can judge it after the election. Heck, after the election, I'm literally going to see how often the actual correct result comes up. But yeah. 

Anyway, I did a senate one too. And MAN the Ohio race going tossup screwed up the odds for the dems. 

Here we get 777 republican outcomes, 31 democratic outcomes, and 192 ties. If we had a sample of 100, it would look 78R, 3D, 19T. Not all that different than we've been seeing. Maybe a bit more bullish on republicans lately though. 

And as we can see, Ohio is responsible for a large proportion of the flips. 44% to be exact. As we can see, the 98th percentile result for the dems is basically a 51-49 win, the median and mean results are 49-51. And the 98th percentile win for republicans is a 46-54 win in their favoe. 

So yeah, senate looking bleak for the dems here. 

Will I do a house forecast?

I'm not sure but I'm leaning toward skipping house forecasts. It's just too hard to do a model when half of the "toss up" districts lack data. I tried developing a methodology of guessing it from the general congressional vote, in a similar way to how I estimated how well different dem candidates would do based on their popular vote and how that would relate to the electoral college vote. I got convincing results, 215-220 where the republicans win. But then I tried backporting it to 2022 and 2020 to verify its accuracy and it was WAAAY off. To be fair, this could have, once again, been RCP's fault, but it also could've been a completely bad methodology. 

I'll say this. It's gonna be close. The 538 has the house at a 53% chance of republicans taking it. The model i tried to derive I think had a 58% chance of republicans taking it, so maybe it wasn't bad but yeah. It's gonna be a toss up. I know that much. Right now, RCP has 196 relatively safe dem districts, 207 rep ones, and you have 32 tossups. If you have half the tossups go either way, with 16 going to each, you get 212D-223R. Trying to weigh it toward the dems using the GCV gave me 215-220. I don't know. it's a toss up. It's gonna be close. 

Looking at the polls, 18 of the 32 dont seem to have polls. 10 seem to lean dem, 3 lean R, and 1 is a tie. Extrapolating based on that, I'd give the dems 219, and the republicans 218. 

Again, different methodologies do different things, and there quite frankly isn't enough data for me to even have a workable model that doesn't just spit out 50-50 due to lack of polling data. 

So just treat my house forecast as 50-50 because I know it's close, and I don't know the actual results. 

Conclusion

So yeah, lots of discussion this week, but the big takeaway is that the dems seem to be losing ground. Even if RCP is potentially biased, all polling trackers are showing clear movement in the republican direction, it's just that those other models are 0.5-2 points more optimistic for harris in the first place. Right now, harris isn't in a good place, she is losing ground, her odds are dwindling, and it seems to be impacting dems down ballot in the senate as well. While the race is still anyone's game, and it's functionally a tossup, the fact that the dems have lost this much is concerning to me. Remember, 2016 was a tossup too but one of these guys still needs to win. And if I went by who has the energy right now, it's trump. And yeah, that's my forecast. Still anyone's game, but I give the edge to trump and the republicans.