SBSQ #31: Trump is super unpopular. So why don’t Democrats have a bigger lead?
A deeper look at the generic ballot. Plus, the game theory of the automated strike zone. And more models coming soon!
This is SBSQ #31, the Greg Maddux edition, named in honor of one of the most cerebral athletes of all time. So even though the headline item is about politics, of course we’re also going to take a nerdy question about baseball. You can leave questions for the Magic Johnson edition (SBSQ #32) in the comments below. But for today, let’s answer these three:
Why aren’t Democrats further ahead on the generic ballot?
Trump’s second term and the forest for the trees, a.k.a. what happened to the 113 Trump predictions follow-up article?
The game theory of baseball’s new automated strike zone
But before we get to the questions, I have an announcement to make:
More models coming soon!
I’ve teased at this in the past, but I can now confirm we’re going to have a Silver Bulletin World Cup model. Not only that, but we’ve also developed a soccer rating system, PELE1, which will serve as the basis for those projections.
This isn’t my first soccer rodeo. I first designed a soccer model, Soccer Power Index, for ESPN in 2010, and then we had SPI 2.0 at FiveThirtyEight, one of our most popular products. But I wanted to start fresh on this. There’s more and better data available now. And I’ve become more experienced at building sports models, having figured out some insights while working on COOPER and ELWAY recently. Furthermore, AI coding tools tend to help with the parts of the job I’m mediocre at (i.e., programming) in a way that frees up bandwidth for thinking about model architecture — international soccer is quite tricky for a number of reasons2. Though I think the LLMs are a long way from being able to build a good model on their own, look for a post about that soon.
In any event, I finished PELE last night. (About a week or so ahead of schedule; that doesn’t usually happen.) We’re going to take a couple of weeks and work on the graphics and presentation before we launch it sometime in early-mid May. And then we’ll publish the World Cup model in late May or early June and update it every day after matches are played. The World Cup itself starts on June 11. We’ll keep PELE running even after the World Cup is done — the international soccer “season” never truly ends. Most of the code stack can also be incorporated into club soccer projections, but that’s (much) further down the line.
The World Cup stuff being on track also means I’m feeling less stressed about getting our midterms model launched reasonably soon also. I don’t think it’s prudent to launch it before early June. Candidates aren’t locked in yet in some key races like Maine and Texas. That we could work around, but districts aren’t even locked yet, either. Long story short: look for the midterms model at some point in June or July.3
Still, I will need to steal some time here and there to finish up these models. So posting volume may be slightly lighter in the weeks between now and roughly Memorial Day. (As it also was for the past week or so, frankly.) We’ll make it up to you with what we hope is a very busy June through November. Onto the questions.
Why aren’t Democrats further ahead on the generic ballot?
Alec Schneider asks:
SBSQ #31, two polling questions: 1) You’ve mentioned Dems will receive a boost when pollers switch from RV to LV screens however Trump approval among LV is higher than RV according to the SB Trump Approval Dashboard, could you square these? 2) What do you make of Trump approval cratering while there is little movement in the SB Generic Ballot? Is it just that a couple highly rated polls from before or early in the Iran conflict prior to Trump approval decline (TIPP and Noble) are buoying the R top line number?
Let’s take a quick snapshot of where things stand, Alec. Trump’s approval rating actually again hit a new second-term low in our tracking today, at 39.3 percent. But Democrats are ahead by “only” 5.7 points in our generic ballot average. I put “only” in scare quotes because that’s probably enough for Democrats to take the House anyway, and make the Senate competitive. But Democrats will probably need a couple more points than that to give themselves even or better odds in the upper chamber. (Or at least, that’s what I assume our model will say. Sometimes it surprises me.)
Trump is the second-most-unpopular president out of the past nine midterms: he’s actually more unpopular now than he was in 2018, when Democrats gained 40 House seats.4 But at this point in 2018, Democrats led by 7.6 points, not 5.7: not a huge difference, but exactly the sort of difference that might tip the balance of the Senate. Moreover, the only president more unpopular than Trump 2.0 was George W. Bush in 2006, and Democrats had a whopping 11.4-point lead as of Apr. 21 that year.
So let’s put this data on a scatterplot. Is this year some sort of outlier?