The election is on Tuesday, and we’ll publish as much content as we can between now and then.
asserted
we → publish → content
But it’s also Halloween1, so let’s have a little fun today.
asserted
’s → ’ → fun
This is the long-awaited (??) revival of Silver Bulletin Subscriber Questions (SBSQ).
asserted
This → await → Questions
More specifically, it’s edition #32, which is obviously the Magic Johnson Edition.
asserted
which → ’ → ?
Most of the questions are election-related, but we’ll take a couple of sports ones at the end.
asserted
we → relate → end
But it’s also the SuperMegaJumboDeluxe Edition.
asserted
it → ’ → ?
We’re in lightning-round mode here; I’ll give you a couple of relatively off-the-cuff paragraphs on each topic.
asserted
I → ’re → topic
You can leave questions for SBSQ #33 (the Larry Bird Edition) in the comments below.
asserted
You → leave → comments
What’s up with the governors’ races in Alaska and Vermont?
What if Republicans win the House?
What if the polls froze?
Why didn’t 2018’s blue wave reach the Senate?
asserted
wave → ’ → Senate
What should we make of big registered vs. likely voter gaps?
What can we learn from undecided voters?
asserted
we → make → voters
Does FLIPR account for voter suppression or election interference?
asserted
FLIPR → account → suppression
The conservative golden age, revisited
Should models account for near-misses in sports?
asserted
models → revisit → sports
Sophie asks:
Why is the Silver Bulletin mascot a wolf and not a dolphin (for FLIPR)?
Sylver, who is indeed a wolf 🐺, was recommended by Fivey Fox, and drawn by the same graphic artist as Fivey Fox.
asserted
who → ask → Fox
I think you’d agree that it would be pretty weird for a fox to be friends with a dolphin?
asserted
fox → think → dolphin
The reason we update the forecast only once per day is because we have to go find him in the East River — he’s never quite in the same location — and bribe him with sushi-grade tuna.
asserted
he → update → tuna
Fivey, meanwhile, is living out his dreams by having become a ranger at Acadia National Park, where he never has to explain that 30 percent chances are supposed to happen 30 percent of the time.
asserted
chances → live → time
What’s up with the governors’ races in Alaska and Vermont?
asserted
What → ’ → Alaska
Jeremy asks:
The paid subscribers want to know what is happening with your gubernatorial forecast, specifically with Alaska and Vermont.
asserted
what → ask → Alaska
Those are probably the forecasts where FLIPR is most going out on a limb.
asserted
FLIPR → go → limb
V. M. Kozma asks:
There is a large gap between FLIPR and prediction markets on the Alaska senate election that completely goes against the overall trend of prediction markets being less positive on Democrats’ chances.
asserted
markets → ask → chances
An obvious explanation would be that the model fails to handle ranked choice voting.
asserted
model → fail → voting
We covered the Alaska Senate race in last Sunday’s post.
asserted
We → cover → post
Since then, we’ve also added a scandal flag for Mary Peltola.
asserted
we → add → Peltola
(Because of her alleged abuse of campaign staffers; it’s more ambiguous whether the efforts to recruit other Dan Sullivans to the ballot would constitute a “scandal” by our standards.)
asserted
efforts → allege → standards
It’s one major case where we’re lower on Democrats’ odds than the consensus.
asserted
we → ’ → consensus
But that’s also partly because of ranked-choice voting.
asserted
that → ’ → voting
It’s not because we’re failing to account for RCV; rather, I’d argue that it’s because we are handling RCV explicitly whereas other models are treating it as a rounding error:
asserted
models → ’ → error
Basically, the story would be that people are looking at the topline difference between Peltola and the “real” Dan Sullivan, where she’s ahead by a couple of points.
asserted
she → look → points
But if Peltola doesn’t hit 50 percent in the first round and Sullivan picks up votes from the “fake” Dan Sullivan and the other Republican on the ballot, he could gain a point or so under ranked-choice voting reallocations
In the governor’s race, FLIPR has been very bullish on the Democrat, Jonathan Kreiss-Tomkins, although we’re now aligned with the conventional wisdom there and it looks like Kreiss-Tomkins could win in the first round outright.
uncertain
Tomkins → hit → round
That’s also because of RCV, although for a slightly different reason.
asserted
That → ’ → reason
One mistake, as people might be making in Alaska’s Senate race, is to ignore RCV.
uncertain
people → make → RCV
But you can also err in the opposite direction by just summing up the cumulative vote for each party and assuming all votes transfer cleanly.
asserted
votes → err → party
In fact: (a) there is a fair bit of dropoff voting, i.e. voters don’t complete all options on their ballot; (b) maybe ~15 percent of voters switch parties with their second choices.
asserted
percent → be → choices
So if Kreiss-Tomkins is at 47 percent or something after the first round, it’s going to be very hard for any Republican opponent to catch up.
asserted
opponent → go → round
Right now, we have the first round projected as follows:
asserted
round → have → ?
Kreiss-Tomkins (D): 47.1 percent
Bernadette Wilson (R): 27.5 percent
Dave Bronson (R): 14.8 percent
Treg Taylor (R): 10.6 percent.
Nominally, the three Republicans have 52.9 percent of the projected first-round vote combined.
asserted
Republicans → have → vote
But say 15 percent of Bronson and Taylor voters drop off, and another 15 percent flip to JKT.
asserted
percent → say → JKT
Then he winds up winning easily.
asserted
he → wind → ?
On top of that, some polls ask about various runoff matchups, and we use that data directly where we can.
asserted
we → ask → data
But look, this is tricky, and AK-Gov is certainly one of our more assumption-driven forecasts.
asserted
Gov → look → forecasts
…and 26 more, not listed.
The Aporia
