Our midterms model, which we’re now calling FLIPR 🐬 — Forecast with Leading Indicators, Polls and (Expert) Ratings — launched today!
asserted
we → call → Indicators
We’ll be updating the model nearly every day, and will have lots more analysis to come.
asserted
We → update → analysis
FLIPR is in the lineage of the f/k/a FiveThirtyEight election forecasts1, but we’ve also put a lot of obsessive work into refining and improving it.
asserted
we → put → it
We greatly appreciate your patience as we’ve gotten everything ready.
asserted
everything → appreciate → patience
Some elections fill me with a sense of existential dread.
asserted
elections → fill → dread
Even the most carefully designed models, which dutifully consider not only how to forecast each state but also how to calibrate the degree of uncertainty in the forecast, will have trouble evaluating the impact of a once-in-a-century pandemic like in 2020 or a major party switching candidates midstream like in 2024.
asserted
party → design → 2024
The next tier down from this are elections that seem relatively normal on the surface, but where typically reliable indicators are contradictory.
asserted
indicators → seem → surface
In 2022, President Biden’s deep unpopularity and the long history of the president’s party performing poorly in midterm elections suggested the possibility of a red wave.
uncertain
party → perform → wave
Polls of individual races, however, and special election results, were much more equivocal, perhaps reflecting the degree to which the Supreme Court’s decision to overturn Roe v. Wade had galvanized Democrats.
asserted
Wade → reflect → Democrats
Democrats pared their House losses to just 9 seats and actually gained a seat in the Senate, but it wasn’t an easy election to forecast.
asserted
it → par → Senate
It looks like a pretty normal midterm, to the extent anything can be normal in American politics these days.
asserted
anything → look → politics
And normal midterms mean bad news for the incumbent president’s party, especially when the incumbent is as unpopular as President Trump.
asserted
incumbent → mean → Trump
Polls, historic trends and leading indicators are all fairly consistent in pointing toward Democratic gains.
asserted
Polls → lead → gains
We’re still 84 days out from the election, for one thing.
asserted
We → ’re → thing
The uncertainty is something that inherently probabilistic models like our midterm election forecast, a.k.a. FLIPR, are specifically designed to grapple with.
asserted
models → design → forecast
In the House, Republicans basically won the war in an unprecedentedly aggressive round of mid-decade redistricting, which left the map with a mild Republican bias.
asserted
which → win → bias
That gives Democrats a hurdle to clear.
asserted
That → give → hurdle
Not to be overlooked: extremely partisan maps drawn by both parties have greatly reduced the number of competitive districts in the House.
asserted
maps → overlook → House
Even if Democrats have a great year — say, they win the national popular vote for the U.S. House by 9 percentage points, which would qualify as a landslide by modern standards — they’ll have trouble matching the 41-seat gain that they had in 2018.
asserted
they → have → 2018
The current composition of the Senate is 53 Republicans and 47 Democrats3, and Vice President JD Vance casts the tiebreaking vote.
asserted
Vance → cast → vote
That means Democrats need to gain a net of 4 seats to take control of the chamber.
asserted
Democrats → mean → chamber
In a roughly neutral, presidential-year-type political environment, that would be a nearly impossible task, requiring gains deep into Republican territory.
asserted
that → require → territory
Moreover, it’s plausible that Democrats will have to pick up 5 Republican seats, not 4, because Abdul El-Sayed currently trails Mike Rogers in the polling average in Michigan, a state that Democrats are defending.
asserted
Democrats → ’ → that
But every available indicator suggests that this very much isn’t going to be a neutral political environment.
uncertain
this → suggest → ?
The Democratic Party has a whole host of problems, which we chronicle obsessively here at Silver Bulletin.
asserted
we → have → Bulletin
Turning out their voters in midterms isn’t one of them, however.
asserted
Turning → turn → them
And for midterm election purposes, you’d rather have a little internal disarray than have a president with a 38 percent approval rating on the basis of an unpopular quagmire in the Middle East and understandable voter anxiety about spiking gas prices.
asserted
you → have → prices
Not to mention voter fatigue over an 80-year-old, lame duck president at a time when the incumbency “advantage” has turned into persistent anger all around the world toward the status quo and whoever is in charge of it.
asserted
whoever → mention → it
Let’s get into more specifics about these indicators, and then we’ll discuss why FLIPR is spitting out the probabilities that it does.
asserted
it → let → probabilities
Polling, history and recent elections tell the same story
asserted
Polling → tell → story
Our generic ballot average, which we update every day, currently shows Democrats ahead by 6.7 percentage points.
asserted
Democrats → update → points
Silver Bulletin readers and other election nerds can become obsessed with every little zag in the polls, but our version of the generic ballot has been fairly stable for several months now.
asserted
version → obsess → months
However, that 6.7-point margin might actually lowball Democrats.
uncertain
margin → lowball → Democrats
That’s because most generic ballot polls at this not-exactly-early-but-not-exactly-late point are being conducted among registered voters rather than likely voters — and therefore, don’t account for turnout.
asserted
polls → ’ → turnout
FLIPR makes a likely voter adjustment.
asserted
FLIPR → make → adjustment
In line with our longstanding practices, the adjustment is based mostly on polls that release both likely and registered voter versions of the same survey.
asserted
that → base → survey
Thus, FLIPR’s version of the generic ballot shows Democrats ahead by 8.1 points instead.
asserted
version → show → points
Well, FLIPR has a calculation for that, too.
asserted
FLIPR → have → that
We estimate a “prior” based on all congressional elections since 1946, calibrated based on the president’s approval rating, the result of the previous election, and whether the election was a midterm, and adjusted for ever-increasing partisan polarization.
asserted
election → estimate → polarization
In that calculation, Democrats are favored to win the popular vote for the U.S. House by 8.7 points instead.
asserted
Democrats → favor → points
…and 14 more, not listed.
The Aporia
