I reached out to Harry, a FOTP ("Friend of The Point") to chat more about the model, how we should use it and what it tells us about November.
Our conversation, conducted via email and lightly edited for flow, is below.
Cillizza: Explain what "The Forecast" aims to do. And how it's different than race ratings that CNN does?
Enten: "The Forecast" aims at giving us the best understanding of where the results will end up. That's very different from the race ratings that CNN normally does, which aim to explain the current environment. Now, obviously knowing the current environment is key to knowing the future. Yet, there are a few key differences. We look at a slew of other variables that help us jump from an understanding of the current environment to understanding the Election Day political environment. Stuff like district polls, the generic congressional ballot, fundraising by the different candidates, etc. We know, for example, that race ratings sometimes lag behind the national environment. In years in which Democrats have done well, the race ratings tend to understate the Democratic side.
Two other things "The Forecast" aims to do that are different: We're providing a margin of error on our estimate based on how well it did in the past. (More on that below.) We're also summing up the individual estimates to provide an overall estimate. Now that comes with a margin of error, but we're putting our money where our mouth is.
Cillizza: What are the factors that you are plugging into "The Forecast"? Are any weighted to matter more?
Enten: Well, we mentioned a few of them above, but let's see if we can get a little more in-depth.
On the Senate, it's really the state polling. That gets the most weight by far. Our dataset dates back to 1992. We also control for money raised, the generic ballot, the past voting in the state and the political experience of the candidates.
In the House, we basically have two models. One takes into account the race ratings and the fundamentals. We're taking into account those ratings from four outlets (CNN, Cook Political Report, Inside Elections and Larry Sabato's Crystal Ball). Fundamentals include money raised by the candidates, whether or not the incumbent is running, the previous House vote in the district, how it voted in other elections (state House, state Senate, etc.), whether or not the incumbent running is a freshman. Unlike the Senate, the polls do not get more weight than the other variables.
Cillizza: The House model says that there is an outside chance Democrats gain 89 seats. And a similar chance that Republicans pick up 25. Explain this -- since it seems very weird.
Enten: So I think this is a misunderstanding, and I'm glad you brought it up. We're talking about the final margin between the parties. So a 25-seat majority is 230 for the GOP and 205 for the Democrats. An 89-seat majority is 262 for the Democrats to 173.
Now those may seem wide, but there are a lot of ways the forecast can go wrong. If the national polls are way off, then you have error. Maybe it's districts with an incumbent where the forecast is off.
A good forecast isn't just about nailing the result. It's about understanding where the forecast can go wrong. Perhaps people want something more specific, but I'd rather be honest about how confident we actually feel.
Cillizza: Your Senate model predicts a three-seat GOP gain. That's a little bit counter conventional wisdom to the idea that Democrats have an outside chance at winning the majority. Why?
Enten: Again, I think we were a little unclear. We're thinking Republicans end up with 52 seats to Democrats' 48. It's all about margins. That's better for Republicans than the 51 GOP to 49 Democratic seats right now.
Republicans are benefiting from a few factors. There are just so many Democrats running in red states. Specifically, Heidi Heitkamp in North Dakota is forecast to lose (though that race is in the margin of error).
Democrats are leading in a number of close races like Arizona, Florida and Missouri. They may win all of those, but on average they won't.
I'll point out that Democrats still have a shot. They could pick up four seats and end up with 53 in the best case for them. It's just not likely.
Cillizza: Finish this sentence: "The best way to use 'The Forecast' is _____________." Now, explain.
Enten: " ... as a guide for understanding what will happen in November."
I think we'll get a number of results wrong, especially in the House. We may end up getting 20 seats wrong (out of 435). The hope is that even if we end up getting a number wrong, the errors will cancel out.
Finally, I hope people realize that people will realize the true extent and precision we can get. When you hear pundits on TV say "X is going to be the result," I hope people now know what is the actual margin of error around that estimate.
It's wider than I think people think it is.
- Midterm momentum in the Senate swings to GOP
- Trump laments that 'two maniacs' stopped GOP momentum in midterms
- Animal abuse registry gains momentum
- Trump: Maniacs stopped the momentum
- Swing vote senators summoned to WH to talk Supreme Court
- Democrat: We won 'potential for momentum'
- Can Hulu keep 'Handmaid's Tale' momentum?
- Macy's momentum may be starting to slow
- France's fight against sexual violence gathers momentum
- Senator explains GOP immigration plan