Uh, hey, Galen. Uh, well, our forecasts are probabilistic. What we can say is that it would be a surprise if Democrats were not to win the House, and the Senate is closer to a toss-up. Our Senate forecast might tilt toward Democrats. We're supposed to put that behind a paywall, but people see that all the same. It gets out on Twitter and everything else. Um, you know, so according to the model, you would slightly rather have Democrats have the hand to play in the Senate than Republicans, but the Senate's going to be a long slog until the bitter end, I would assume.
Um, okay. So, basically, we're going to get a 5D Fox replacement. Galen and Nate were up to something two hours ago that we're not telling anybody about. And we are also going to get specific about the forecast. But first, we're going to talk about something else because some election results actually derailed a couple narratives last night. And of course, I'm talking about the Wisconsin gubernatorial primary where David Crowley came from behind and won by a half a percentage point with more than 95% of the expected vote counted against Democratic Socialist Franchesca Hong. After Franchesca Hong led in all of the — not just recent polling but basically all of the polling going back months in this race and was leading by 20 points at the end. So Nate, I want to get your thoughts out of the gate on this and then, you know, as a former Scotty myself, I got some thoughts.
Scotty? I hadn't heard that term. There are excuses you can make. One problem is you didn't actually have that many late polls. There clearly weren't a lot of late developments in the race. Um, Crowley had dropped out and re-entered, right? Um, Sara Rodriguez, I believe her name is, had dropped out. Uh, Mandela Barnes had dropped out. Um, so you had a lot of late-breaking news. But still, the fundamental problem, Galen, with polls is that they are not actually reaching a representative sample of voters. Um, because to take a poll is an unusual thing to do. In telephone polls now, which are used less and less, you might get a response rate of anywhere between 1% and 10% depending on the quality of what you're trying to do, how many times you try to reach people and so on. Um, randomly dial a number or take a registered voter list and like randomly dial a number. It's much harder to randomly ping someone in the middle of the day on the internet. Um, so polling is challenging for that reason.
In general elections though, there are a lot of things that you can anchor polls down to. Number one is that in general elections, you almost don't need polls. You can just look at what's the national political environment. Of course, polls like the generic ballot are part of that calculation. So, it's a little bit circular, but look at what's the national environment and then where does the state line up in partisanship based on how it voted in the past presidential election. Like that alone will get you like an 80% decent answer most of the time. Maybe you account for incumbency, but even incumbency, that advantage is diminishing also. Um, demographically things are very predictable, right? If I give you a template and I tell you — I just give you, actually, Galen, if I give you somebody's name and their zip code, from that alone you can probably correctly identify who they're going to vote for 75% of the time or more, right? Because from the name you can impute usually gender, a lot about age, actually a fair amount about race. From the zip code, you can infer tons of things about someone's economic circumstances and urban or rural. Those things jointly are quite powerful. And so therefore like you almost don't need like any actual like voter input to have like a good model in some sense of a race.
But the information you get from polls themselves is often biased in some ways. It's often biased in favor of whoever candidates are easier to reach on the phone and or most enthusiastic. Those are not quite the same thing. Um, and you know, traditionally when you had telephone surveys, younger voters were harder to reach on the phone. Um, now with internet surveys, you might have the opposite problem, right? Who are people who don't get online? Well, probably not very much in the country, but it might be older, poorer voters, which in a lot of states is still part of the Democratic base, right? Among white people, old white people, old white working-class people have become very Republican, right? That's not true among older Black and Hispanic voters, which is still a big Democratic base group. And by online surveys, voters who are very loyal to the Democratic Party, who are often somewhat moderate, especially on cultural issues, um, they might not be as enthusiastic as like a DSA voter, and they kind of will disappear from these polls, right? I don't think it's a case of people refusing to say who they're supporting. It's not like a shy Trump phenomenon or shy Crowley or whatever phenomenon. It's like these voters are being lost and approximated by people who are non-representative, right? Maybe you find some Black voters and some older voters, but the mere fact that you found them for your poll, your online survey, whatever else, or web-to-text survey, whatever else, makes that itself — it's like an observer effect, right? You can never completely remove that bias.
