So, as we have clearly seen, my own model for political races is a lot different from 538's. I estimated that there's a 69% chance of the dems taking the senate, but they assume a 52-55% chance of the GOP taking it, looking at it throughout today.
Most models I'm also seeing seem to assume the dems are just going to lose the house, without really justifying it well. They basically base their opinions on models, and models are well...garbage in garbage out. They're only as valid as what you put into them are.
My own general methodology for political races is to let the polls do the talking. I look at all races within 8 points, because I assume a general 4 point margin of error, and then I calculate the Z score based on the polling average, and use a Z score chart or calculator to spit out the probabilities.
I feel like this has generally been a good methodology. I've been using it since 2016, and honestly? I BEAT Nate's model doing this in terms of accuracy. Nate assumed like a 70% chance that Hillary would win. I was signalling the warning bells on election day eve that yeah, all of this hubris for hillary winning was misplaced and that Hillary's actual chances were 56%. And ultimately? Trump won.
Now, polls can never be 100% accurate. That's why margin of errors exist. They are going to be potentially off by a few points, sometimes as many as 8, but you generally can spitball it. I mean, I come close. I decide these states are in play, I crunch the numbers, and I spit out probabilities. And even when I'm wrong, honestly, my predictions are vague enough where they can fall within the probabilistic range of things.
I learned the hard way, trying to predict 2008 and 2012, that using other methodologies like past trends can often be wrong. For example, in 2008, I thought McCain would win NC. I had no justification for this, other than it was a southern state that traditionally went republican. But then Obama won it. The polls were right, and my opinion based on trends was wrong. I made the same mistake in 2012 with Virginia. VA was within a point or so polling wise, and could've flipped either way, but honestly, I once again went with trends, and trends were...wrong.
The thing is, voters change from election cycle to election cycle. And if the voters say they're going a certain way, I'm inclined to believe them.
I will admit, in 2016 and 2020 things got messy, with both elections having a systemic polling error that ran in the democrats' favor, but in 2016 we were talking an average error of around 2-3 points, and around 4 points in 2020. Still, that's largely within one standard deviation, and is not that unexpected. In 2020 in particular I ended up throwing out a lot of pro trump leaning polls on the basis of bad methodology, but this time, i'm just going to go with the actual polling averages.
So, that brings me to 538. 538 rose to fame because in 2012, they got every state right, and in 2016, they got it more right than any of the other big guys, because they admitted Hillary had a CHANCE of losing, while the rest of the country and the media echo chamber was drunk on democratic hubris.
And let's be honest, they tend to be a decent site with showing polls and their own predictions. They generally do a better job than even RCP in collecting polling averages, and creating models for politics.
But....sometimes I feel like they inject too many of their personal feelings in their models. Like I remember their hacky anti Sanders coverage in 2020, and how they thought harris was the best positioned candidate to win the primary. Why? Because they came up with a model of 5 different electoral groups and Harris appealed to all 5. She could do it all. Appeal to centrists. Appeal to progressives. Appeal to various minority demographics.
Yeah. She did worse than andrew yang. No one liked her. Because she was as fake as a $3 bill. Idk, sometimes these guys have awful political instincts.
Honestly, I like to leave my personal politics out of my models. Sure, I might comment on stuff, but generally speaking, I prefer to let the data do the talking.
So, 538. 538, on their website, offers 3 models. The default one is their "deluxe" model. But, in order to understand what that is, we must first go through the others. Their "lite" model is "What Election Day looks like based on polls alone". When we apply this model, they have a 57% chance of the dems taking the senate, basically spouting back the obvious fact my own model pointed out. That dems are favored in the senate. I think 538 calculates stuff differently, where they tend to have supercomputers spitting out up to 40,000 scenarios based on data and calculating based on that. I tend to be a bit more simplistic, because I dont have fancy computer programs to crunch that many numbers and run scenarios. I admit I have fooled around with random number generators in 2020 to give me a range of outcomes based on the data, and if anything it just solidified already existing trends. Basically based on this model, the dems won 57% of scenarios. In the house, the dems' odds went up to 22%. Now, to be fair, they tend to have a wider range of outcomes I assume. I only looked at 7 races for the senate, and wouldn't even delve into the house. Again, I dont have the means to do complex stuff like 538 pulls off. But, their model for the senate ranges from the dems holding 57 seats, to the GOP holding 60. Honestly, I just assume anything with a less than 2% chance of flipping is a foregone conclusion and focus exclusively on closer races. Seriously, Gretchen Whitmer is going to win with her 25.5% polling advantage. Even if the polls are off by 8 points, which has happened in Michigan before, she's going to win. There's no reason to even consider this a toss up.
So, their "classic" model. This is "What Election Day looks like based on polls, fundraising, past voting patterns and more". This is where you start getting into the screwery I tried to pull off in 2008. Looking at past elections. uh...how much did that help in the rust belt in 2016? Or virginia in 2012? Or North Carolina in 2008? Again, if a random sample of voters says they're gonna do something, and the data is legitimate, there's no reason to question that. Do people make their minds up last minute? Sure. Is there a shy trump voter effect? Maybe. There are factors that no pollster can count on messing with results. But unless the race is close, there is no point in considering those factors. They're random noise that no one can predict. Honestly, fundraising I dont think means anything. Again, go back to 2020. Look at people like harris and bloomberg. They had huge warchests. Harris had a lot of media support. The dems were pushing her hard. But she lost. Because all the money and institutional advantages in the world don't matter if people don't like you. Institutional advantages are fingers on the scale, but they're not necessarily foregone conclusions. And again, past voting patterns. Eh, voters change. I mean, I would rather trust existing data based on polls than "these guys voted this way for the past 20 years therefore they'll continue to vote this way."
As for what their classic model spits out? Dems still win 56% of the time, but the house forecast drops like a brick to 12% dems. This still seems fairly valid. Again, the senate forecast favors the dems in my opinion. I'm not saying that because I am biased toward them and want to make them look good. I'm the dude for the past 1.5 years saying "hey the dems are in trouble if the election were held today we're talking apocalpytic red wave". That's being proven wrong. That's just how it is.
So the deluxe model, "What Election Day looks like when we add experts’ ratings to the Classic forecast". So basically, at this point, this is their opinions. You're putting peoples' subjective opinions into a model. And that's how suddenly the GOP is favored in the senate in a 55-45 way. I don't trust this as far as I can throw it. Yes, this is supposed to be a red year. Dems are in power. They control everything, they're very unpopular, but then the GOP is doing stupid crap like taking away abortion rights. If you wanna fire the dems up, you're doing a good job doing that. Thanks for the electoral gift GOP. And of course, January 6th hearings have been on the past few weeks and the GOP is STILL a trumpist cult. So that's gonna hurt them.
The fact is, even though this is a very bad year for the dems, with inflation running rampant and the dems literally not doing anything, the GOP is imploding so hard in their own way it's actually kind of evening things out. This isn't 2010 at this point. I remember 2010. I was a republican then. They had energy. While the dems didn't. The dems have more energy right now than they did then. And the GOP is very divided.I'm not saying things are perfect. The dems should lose in theory. The Biden presidency is as underwhelming as I knew it would be. if not worse. It's going full Jimmy Carter mode if anything. But the GOP just doesn't have that magic either. And given the nature of people voting for lesser evils, while the GOP may make some gains in the house, they seem to be flopping in the senate and governorships, only seemingly doing well in North Carolina. Ya know, the state I didn't expect Obama to win but he did.
But, 538 knows this, once you turn off the model with their "expert" opinions and just rely on the raw data. And they've even discussed the situation themselves. They point out that in practically every election, the party in power tends to lose seats in elections. They point out the fact that Biden is going full Jimmy Carter and how the senate favors rural environments while the dems are primarily an urban party. But at the same time, much like I just said, they mentioned that SCOTUS going full activist mode is actually helping the dems. As is January 6th.
This doesn't really explain too much for why the dems are favored to win the senate but not the house though. I mean, shouldnt these trends affect both? It seems like they're trying too hard on "well the president's party almost always loses seats in mid terms". Yeah but if the president's party is fired up over the GOP doing stupid crap that alienates people, maybe they will do better this time.
Idk, I mean, the house is weird. It technically does favor republicans, simply because of the urban/rural divide and how districts are drawn. The GOP wins rural areas, the dems cities, and the suburbs are up for grabs. Of course, I would argue that we shouldnt ignore cities if youre a republican or rurals areas if you're a dem. That's a huge reason our politics is screwed. keep in mind, you can have 30% of rural people voting dem or 30% of city dwellers voting GOP. I would argue the shift toward trump in PA is because the dems literally made the rural areas redder in order to win narrow bands of suburbanires, so the gains are cancelled out by losses elsewhere, accelerating these geographical trends.
But, in terms of districts, as long as those trends stand, any district that has a major urban area in it is going to often go blue. And many district mostly rural is gonna go red. And a lot of the time, you got these districts with cities that have like say, 50000-100000 people in them, but then are little islands of blue in a larger sea of red. And the areas around them collectively have more people than the cities.
So if the dems don't favor rural people at all, well, they're gonna struggle in the house. Because you actually need some rural people to be able to carry districts. And since districts are based in part on population, but are often random parcels of land with weird shapes, a lot of districts end up either being very blue, or very red. And it's hard to make enough districts that favor the blue parties. Even worse, on the state level a lot of gerrymandering happens to break up blue areas even more, so maps end up favoring the GOP even more.
In some ways the dems are digging themselves into this hole by forgoing working class people in favor of just going purely by "well we can win cities and we should focus on suburbs". It's why the dems end up becoming a mess of contradictions as they end up having to be economically left to appeal to cities, but then economically right to appeal to suburbs, which are often where affluent people live. It also means that they shift culturally left, while the GOP can be culturally right and win over all the disaffected people.
I mean, honestly? It's weird. Like, in PA, you can have it generally be slightly blue leaning overall, but it ends up voting pretty red in legislative districts simply because most districts end up being red due to them being based on a certain land area. If your district is primarily urban you can make a blue district, but look at a freaking map. Like in PA, you only have two major cities. Pittsburgh and Philly. And those districts are very blue. And maybe you can make a couple blue districts out of the lehigh valley and the wilkes barre/scranton corridor. Okay. Well....what about say, Reading, or Lancaster, or York, or Harrisburg, or Altoona? Uh...okay, you have a city with 40-80k people who are blue voters...but then what? All the areas around those cities are rural as fudge. Yeah. Welcome to why dems lose elections. There are 17 congressional districts in PA, and the dems only have 6 safe ones, vs the GOP having 8. Simply because of how land works. We voted Biden and have a democratic governor by the way. Oh, and the three competitive districts are fairly urban and suburban. The lehigh valley I mentioned is one of those districts. Bucks county is another. These are very urbanized areas but they still go center because of all the people in between those areas.
So I can kind of understand why in a state like PA, the dems stand to lose house seats, but maintain the senate and governorship. Sure, if you take the collective vote of everyone in those cities, you will outvote the rural people.
But then when you have districts, you'll have a very gerrymandered environment where you'll have a few house districts that are very blue, as in, blue by insane 50 point margins, but then the rest are red. And even a lot of purpleish ones can lean red. Because you're taking like one or two cities with 50k people each, and weighing them against the masses of Pennsyltucky. Keep in mind, outside of major cities, PA quickly turns into Alabama with snow. We even have weirdos flying confederate flags even though we were never in the confederacy. It's weird. One minute you can be in a super urban area....15 minutes later you're in some rural area where there's amish driving buggies everywhere and Trumpers are waving signs and flags at your local sheetz station. It's weird.
I didn't intend to go on this long, but it's interesting when i really dig into it. Honestly, I wish we could jsut have a system of proportional representation. But I'm not sure that system existed in the 1700s so we're stuck with this mess. While I like a lot of aspects of our constitution, I have to admit, it IS horribly dated by modern standards and that further causes major political dysfunctions involving the obvious majority party consistently losing elections to the more rural minority party. It's a crapshow.
No comments:
Post a Comment