Election prediction 8/23/24

 Not sure if I should do these Saturday since I'm getting weekend polls btfoing me, but then I also see data coming in on Sundays now. I guess Friday is as good of a day as any. Let's get to it.

Presidential

In the macro, not a lot changed. We've seen a slight regression from 50-50 to 48% Harris and 52% Trump, with PA still being the tipping point. The overall map is very competitive, it's about as competitive as it gets with 5 of the 7 big swing states being effective tossups and 2 of them being not far from tossup range. And I finally got some more data on stuff like NE2, New Mexico etc. Data looks solid there. Even Florida and Texas are reasonably in play but I couldn't expect them to flip any time soon. Maine CD2 is back in play, with it being D+5. Given how it was off in 2020, idk how much I trust that though. The poll also had Maine proper at D+17 so it's removed from this chart. The landscape really has shifted in the dems' favor over the past month. Although momentum has stalled over the past week, with Harris's odds regressing slightly. Only time will tell if that's a temporary setback, or her reaching her peak. 

As for the electoral map overall, currently the best guess for it 252-286 Trump.

Presidential Multiparty

So, I'm not even sure how reliable this is given RFK just dropped out of the race and endorsed Trump today. The map is a bit more bullish for Harris on the third party front due to RFK's presence in the race mostly taking potential votes from Trump (most democratic defectors went back because of Harris). So this data could change. If anything, I now expect the 2 way and multiparty data to mostly say the same things. 

Right now, it gives Harris a 64.6% chance of winning, and Trump a 35.4% chance. But again, I'm kind of reluctant to even include third party data since RFK dropping out probably invalidates all of this. 

It's kind of frustrating having RCP separate the two types of data. I was thinking of switching swing states over to multiparty data at some point if it is as reliable as 2 way, but there just hasn't been a good time this election cycle. Maybe some time in october, idk. I might even merge the two tables into one near election day for my final prediction. 

Anyway, here's the map for the multiparty data. It's 282-240 Harris with NC being a true tossup. 

Looks a lot more bullish for Harris, but again, this is probably because of RFK remaining in the race. With him dropping out, I'd defer to the two party data/map.

Senate

This senate map is getting awfully divided and rigid around the 49-51R, where I don't really see it being anything else. You would need a 4-5 point shift either way just to flip any more states. 

I mean, I'm just gonna call this for the republicans. They have an 89.4% chance of winning the senate. The democrats are still at 7.7%, and the chance of a tie is 2.9%. 

A little simulation fun

So, as we all know, I've been messing with election simulators. I built a rudimentary one but I struggle to improve it. Anyway, 538 put up their model for the election day with their own simulator, and I want to do my own simulations with my own simulator. I'll do 100 simulations for each of the three areas of prediction and post the range of results.

Presidential 2 way

Harris: 59%

Trump: 38%

Tie: 3%

This is actually very similar to what 538 got, but with me getting more 269-269 outcomes. It's possible this specific sample of 100 skewed, I did get more harris outcomes than i would have predicted, but rules are rules, i generated 100 outcomes, and these are the outcomes it gave me (actually i did run it again and got a similar outcome). For reference, the simulator generates random outcomes for each state independent of the other states. This is different than my trend model, which assumes that the states are interconnected, and if one flips others flip the same way. My formal prediction is based on trends, this is just a more random model I feel like employing from time to time.

Presidential 5 way

Harris: 69%

Trump: 27%

Tie: 4%

This is a little closer to what I'd expect given I give Harris a 65% chance. It's a little bullish for Harris, but I'd normally expect it to be more tilted toward the favored candidate the further away we get from it being 50-50. 

Senate

Dem control: 1%

Tie: 11%

Rep control: 88%

I expected it to be a bit more tilted than the presidential given how rigid the map is. However, the results on the individual seat distributions are weird. I'll post those two:

47-53R- 6%

48-52R- 26%

49-51R- 56%

50-50T- 11%

51-49D- 1%

Given how many senators were in the "likely' category, the chance of a dem losing a senator was much greater than the GOP, meaning if there's deviation from 49-51R, it's likely in the republican's favor.

Realistically speaking, a 50-50 Tie is the best we can hope for as dems, with the 51-49D outcome only happening 1% of the time. To be fair, I trust the "trend" model of elections more than these simulations, but these simulations produce some interesting outcomes. 

Conclusion

Not much has actually changed from last week, considering how much I've just talked about things. The presidential race is still tightly competitive at roughly 50-50 (48-52 in Trump's favor), although my simulation data gives Harris a larger edge.I think that might be due to Florida and Texas being in play now. 

The third party data has it more in Harris's favor, but with RFK dropping out, I assume the 2 way data is more realistic and accurate as of now. 

The senate looks bleak for dems, I think that republican control, with some flavor of 49-51R is the most likely outcome. 

And yeah. That's my prediction this week.