Let’s talk shop about every Senate race

Today’s newsletter was originally conceived of as something slightly different, a sort of very meta column about FLIPR vs. prediction markets. But I’m feeling a little meta-d out this weekend — maybe it was seeing You Can See Everything on Friday.1 And I realized I had a lot of spare thoughts about individual Senate races, including some we haven’t discussed much at Silver Bulletin. So let’s keep this more tangible and I’ll go race-by-race through the map.

The headline calls this “shop talk,” but what it absolutely isn’t is any sort of insider speculation. That’s not my thing. Rather, it’s how I’d speak to a friend who knows the territory, where we share enough assumptions in common that I can be slightly unguarded. I do this a lot on Still Counting or in external media appearances, but it’s also squarely within the territory of Model Talk.

Because, yes, I have thoughts and intuitions beyond what’s reflected in FLIPR. My jumping-off point for these questions are races where we differ from other models, prediction markets, or the conventional wisdom. What are they seeing that I’m missing, or vice versa? Are there other reasonable modeling choices I might have made that would have produced a meaningfully different forecast?

I guess that’s a way of saying “where do I trust FLIPR and where don’t I?” although I don’t think that’s quite the right framing. For one thing, I mostly do trust FLIPR. For another, the whole point of building a model, in some sense, is to set clear rules ahead of time so you don’t have to trust your gut or make ad hoc assumptions in midstream. Now that we’re in “silly season,” it sometimes feels like everyone from the expert ratings agencies to the pollsters themselves are hedging their bets or sometimes herding. FLIPR might be wrong, but it gives you an independent opinion every day, based on rules we locked into place months ago. Still, I keep a mental list each cycle of things we might change in our models the next time around.

I promise: we’ll get into the individual Senate races shortly. But for context, let’s quickly look at the topline numbers.

Models and markets have converged, but FLIPR remains a bit more confident in Democrats’ Senate chances

In the House, there’s little difference between FLIPR and prediction markets. It would take a polling error of epic proportions — bigger than in 2016 — for Democrats to fail to pick up the chamber. Statistically, it’s possible, but you’d start to get into the fat-tailed ends of the distribution where both the polls and a lot of the assumptions we’ve been making about politics, including decades-old trends about weak performance for the president’s party in midterms, are deeply wrong.

In the Senate, there’s a bit more daylight. I wouldn’t call it a major disagreement, and the gap has closed some in recent days, but it’s also more than a “rounding error’s” worth. Polymarket2 gives Democrats a 2-in-3 chance of winning the Senate, whereas FLIPR puts it at 3-in-4.

FLIPR launched on Aug. 11, but I’ve run the Polymarket data series back to Aug. 1. In the markets, there was a reasonably big jump in Democrats’ odds on Aug. 11; I don’t remember that having been an especially good day of news or polls for Democrats, so that could plausibly have been markets reacting to FLIPR itself.

FLIPR and Polymarket then moved basically in tandem with one another through roughly Sept. 22, with both seeing a sharp increase in Democrats’ win probability shortly after Labor Day. (Though if you squint, I think you can say that FLIPR led the markets by a couple of days.) Then our forecast continued to improve for Democrats while their market rally stalled out. But now our forecast has been very flat for the past two weeks — there’s actually been just the slightest downtick for Democrats over the past couple updates — while Polymarket has shifted toward Democrats over the past few days instead.

What’s behind this little shift toward Democrats in prediction markets?

I don’t have a foolproof theory, but a couple of hypotheses. First, markets may be building in some information — or assumptions — based on early voting data, something that FLIPR doesn’t really3 look at. I’m on record as not being a huge fan of this, though friend-of-the-newsletter Joel Wertheimer objects, and says early voting data may be worthwhile if you focus on which party is turning out its unlikely voters, meaning people who have skipped most recent elections. My impression is that Democrats are doing better in that department, or at least that’s what people who more closely follow early voting than I do seem to think.

I’ve also noticed that prediction markets often become considerably more confident as elections approach, shifting away from 50/50 toward whichever side is perceived to have the edge. And that makes a certain amount of sense. The closer we get to Nov. 3, the less time remaining for October Surprises, etc. But I wonder if this principle is being applied slightly incorrectly by traders. Most of the remaining uncertainty in the race isn’t about how polls will change over the final week-and-a-half, but rather about whether there will be another major polling error again — in other words, whether polls were right in the first place. And we won’t know that until we have the results in. Meanwhile, FLIPR’s assumption is that you shouldn’t expect to see a lot of late polling movement in the modern, highly polarized political environment.4 From my vantage point, a satisfying conclusion5 would be if prediction markets and other models catch up to FLIPR by Election Day, so that everyone winds up with basically the same forecast but FLIPR got there a little sooner by correctly detecting earlier that the national environment was very poor for Republicans.

Where do FLIPR and prediction markets differ the most?

The differences between FLIPR and prediction markets are internally consistent. In the Senate, there are roughly a half-dozen races where we’re modestly or materially more bullish on Democrats’ chances, although also 1 or 2 where the difference runs in the other direction and we have better odds than Polymarket does for Republicans.

The following is a rather technical way to think about it. But basically, if you plugged Polymarket’s forecasts for individual races into FLIPR, and relied on FLIPR to model the correlations between races and estimate Democrats’ overall win probability, my guess is that it would replicate Polymarket’s topline number pretty well, and produce a 65-70 percent chance of Democrats winning the Senate instead of FLIPR’s actual 75 percent. In that sense, this isn’t really a macro-level disagreement about the intrinsic amount of uncertainty in the forecast. Rather, the disagreements boil down to individual races. Here’s a scatterplot comparing FLIPR and prediction markets in every Senate race as of today’s forecast.6

Overall, there’s far more agreement than disagreement; the correlation coefficient is .99. Let me also give you the comparison in tabular format to better highlight where there are differences, however.

OK, enough build-up. It’s time for our tour of every race, in descending order of Democrats’ winning odds according to FLIPR — although I’ll lump some of the less competitive races together.

Virginia and all bluer states

I’m not going to have much to say about those races — there are several on the Republican side too — where FLIPR is at 99.8 percent or something and Polymarket is at 98 percent. These are mostly technical differences — it’s not particularly attractive to tie up capital in a somewhat illiquid market to arbitrage a very small ROI.

Georgia and North Carolina

FLIPR and prediction markets are well-aligned in these states, and I don’t have much to say here either. Democrats ought to win purple-ish states where they have strong candidates (in the form of Jon Ossoff and Roy Cooper). The early voting data looks pretty good for Ds in these states too, and Republicans seem to be kind of giving up, particularly in North Carolina. And while Georgia and North Carolina voted Trump in 2024, they actually moved slightly closer to the tipping-point relative to the national popular vote. Democrats only lose these states if there’s some kind of incredible surprise in the polls, in which case they might even lose the House.

New Hampshire and Minnesota

We wrote about these states recently. I take the 11 percent chance for Michelle Tafoya in Minnesota seriously, but a severe underperformance for Peggy Flanagan will probably be excused by some combination of the national environment and Minnesota’s blue lean: it’s a D +13 state in our generic ballot baseline. Whenever there’s a disagreement between polls and “fundamentals,” uncertainty tends to be higher, and FLIPR accounts for this. But that means, while there’s an outside chance of a Tafoya win to the point where Democrats can’t just take the race for granted — Minnesota actually comes up as the tipping-point state 7 percent of the time — it also wouldn’t be a shock to see Flanagan outperform her polls and win by 10 points or something after all.

In New Hampshire, I have less of an instinct to be a contrarian buyer of John Sununu stock, mostly because Chris Pappas has “proof-of-concept” in having been elected in a congressional district that represents half the state with performances that are at-or-above baseline. New Hampshire is one of the most swingy/elastic states, relatively responsive to changes in the national environment and this one might be more interesting in a slightly less blue environment. A Sununu win would nevertheless be less shocking than a GOP victory in Georgia or North Carolina. But I’m not going to sweat the difference between FLIPR’s 93.8 percent and Polymarket’s 92.

Michigan

Yeah, prediction markets are probably reasonable here. I say that because FLIPR doesn’t really have any mechanism to account for El-Sayed’s progressivism, even though I think it would be reasonable to do so. In fact, in Michigan, we have the fundamentals being slightly better for El-Sayed than his polls in part because of his prodigious fundraising. If we accounted for his ideology rating within the fundamentals, I imagine we’d be pretty close to Polymarket. Who knows if there could also be some sort of ethnic or religious bias against him too; no Muslim-American has ever been elected to the U.S. Senate.

At the same time, if El-Sayed holds on, you could look at Michigan as a successful example of modeling. Because Michigan ought to be pretty blue in this environment, we had AES favored even when he slightly trailed Mike Rogers in the polls.

The presence of Maine, Minnesota, Michigan and arguably New Hampshire on the list of competitive races is important from a macro standpoint, though. Not all of these are solid blue states — Michigan voted for Trump twice — but they all ought to be easy wins for Democrats in this political environment given “generic” candidates. One concern for Democrats is if pure partisanship matters more than candidate quality. That could hurt them in races like Texas or Kansas where they have the better candidate but a lot of partisan gravity to overcome. However, it would help them in Michigan, Maine and Minnesota, where they’re vulnerable because of more questionable Democratic candidates and/or relatively strong Republican ones.

In other words, say that FLIPR is underweighting “fundamentals" and overweighing polls. In that world, Democrats win all of the blue and purple states, so they defend Michigan, Minnesota, Georgia and New Hampshire, and flip North Carolina and Maine. Now they have to find two more pickups in states that voted for Trump three times in a row.

But this is not actually so bad for Democrats. In “simple fundamentals” — which is just the generic ballot among likely voters + the state’s partisan lean score — Alaska is only R +0.2, Florida R +1.1, Iowa R +1.2, Ohio R +1.6, and Texas R +1.9. (Then there’s a big gap to Kansas and other states.) This is a relatively straightforward calculation: a D +9.5 generic ballot would reflect roughly a 12-point swing from 2024. So you’d expect this tranche of states, which Trump won by around 12 points in 2024, to be highly competitive.

In other words, if the generic ballot is right, fundamentals and polls aren’t that misaligned. Is there a big enough candidate quality difference in Alaska or Ohio or Texas to make up for an R +0 to R +2 background environment? I’d argue that’s highly feasible; candidates mattering less than they once did should not imply that they don’t matter at all. So to save the Senate, Republicans probably need to beat the point spread on the generic ballot on top of coming up with some clutch performances in individual races.

Ohio

FLIPR is a hair higher on Sherrod Brown than prediction markets. But look, this is a newsletter where I promised some gut-checks. And I actually wonder if FLIPR is too low on Brown.

I just get very anchored to 2018 when Brown won by almost 7 points in an environment that looks pretty similar to this one. That’s just strong proof-of-concept in a way that FLIPR might not fully be capturing. Our “fancy fundamentals” calculation incorporates past winning margins for incumbents, although not for non-incumbents like Brown. And Brown also considerably outperformed Kamala Harris’s margin in Ohio in 2024, losing to Bernie Moreno by just 3.6 points. Indeed, many Ohioans still think Brown is the incumbent, whereas Jon Husted is an unelected incumbent, and our model treats races with unelected incumbents as open-seat races. To me, Ohio looks like the most straightforward D pickup after North Carolina.

Maine

More straightforward than Maine, for instance, which is ground zero for the question of whether polling errors are persistent, since Collins has repeatedly beaten her polls by a lot. According to FLIPR, Troy Jackson is slightly ahead in the polls, way ahead in “simple fundamentals” which account for Maine’s blue lean in presidential races — but a slight underdog in “fancy fundamentals” that account for Collins’s strong re-election margins in the past.

I think FLIPR has some fairly smart heuristics here, including accounting for the fact that uncertainty is high in this race. But it’s not a race that Democrats should feel safe about, and personally I might be closer to Polymarket’s 64 percent chances for Jackson than FLIPR’s 70.5 percent. I also expect Maine to be less correlated with national outcomes than most other states, i.e. I can imagine Jackson winning in a disappointing night for Democrats or Collins winning in an otherwise good night for Dems.

Texas

The streams have somewhat converged here. Ken Paxton has gained a point or so in polls, and FLIPR’s odds for James Talarico have declined slightly as a result — but Talarico’s chances have held steady or slightly improved in prediction markets, possibly because of traders reading tea-leaves on early-voting data. It does feel like there’s a bit more herding here in the polls than in some of other Senate races, though I’m not sure who that favors, exactly.

We’ve written about Texas a lot. It’s a flip that ranks high on “unthinkability” because of Texas’s symbolic importance and the fact that Democrats haven’t won a Senate race there since 1988. While we’re slightly higher on Talarico than the markets, I’m surprised the gap isn’t larger.

Still, the math showing Talarico as a modest favorite checks out. As I wrote under the Michigan comment, a D +9.5 generic ballot would shave Republicans’ partisan edge in Texas down to merely R +1.9. Paxton is basically the definition of an underperforming candidate, and a 3-to-5-point overperformance for Talarico would be well within the reasonable range.

Furthermore, Texas has not particularly been the site of large polling errors. In fact, Beto O’Rourke considerably beat his polls in losing to Ted Cruz by 2.6 points in 2018. Take that margin, add a scandal penalty for Paxton, and you get to FLIPR’s forecast of a narrow Talarico win in line with polls.

What about all the past premature predictions of a purple Texas? Well, basically what happened is that the predicted Democratic gains in the suburbs were (more than) offset by significant improvement for Trump among working-class Latinos in South Texas. But those gains are relatively recent, and polls show them potentially reversing this year.

To be fair, forecasting a Democrat to be favored in Texas is not exactly a forecast that one sleeps easily with, but I’m not sure if that’s an analytical reaction or an emotional one.

Alaska

This is the one race where FLIPR is notably lower on Democrats’ chances than markets are. That strikes me as important in terms of the overall Senate equation because Alaska, up in the Arctic Circle, is one of the least correlated states with the Lower 48. You can imagine Democrats having a somewhat disappointing night but Mary Peltola winning, or maybe the reverse.

What accounts for the difference? It might partly be FLIPR’s heuristics on ranked-choice voting. We put a fair amount of work into these, and that’s the sort of detail that gut impressions might ignore. 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. 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. For what it’s worth, FLIPR has been extremely bullish on Democrats’ chances in the Alaska gubernatorial race as compared with the consensus in another race that reflects our RCV model. So it seems to have some strong opinions about Alaska.

The argument that we’re underrating Peltola is decent too, though. Like Brown, she has strong proof-of-concept, having won Alaska’s at-large seat to the U.S. House in 2022 in a national environment that was considerably less blue than this one. Then she lost in 2024, but by only 2.4 points. I’ll cop out by saying I think FLIPR is a bit too low on Peltola because we aren’t really sufficiently accounting for candidate quality — while prediction markets might be a bit too high if they aren’t thinking enough about RCV.

Iowa and Kansas

These are the two states where the prediction markets are most bullish on GOP chances relative to FLIPR. My topline takeaway is that I think that’s probably reasonable, and these are states where I worry about FLIPR’s forecast a bit.

But despite their geographic and demographic proximity — which implies the possibility of correlated polling error between them — Iowa and Kansas are actually rather different cases. “Simple fundamentals” pegs Iowa as merely an R +1 state, which isn’t much of an edge, obviously. But Ashley Hinson is one of the stronger Republican candidates on the map, even if Josh Turek is a strong candidate too. In Texas, Alaska, and Ohio, all of which have similar partisan lean scores, Democrats have a much clearer candidate quality advantage, I’d argue. In Iowa, they’re sitting a bit more naked in the event of a systematic overestimation of Democrats in red states. And of course, there was infamously a big polling error in Iowa in 2024.

Then again, the diesel prices story makes some sense in Iowa, and regular gasoline prices are ticking up in the state too. My subjective view is that this is about as close to a pure 50/50 toss-up as it gets, so I suppose I’m once again splitting the difference between FLIPR and prediction markets.

Kansas is redder than Iowa, having voted for Trump by 16 points in 2024 as compared to 13 in the Hawkeye State. Our blended fundamentals calculation, even accounting for the blue national environment, has Kansas at R +4.4, so Adam Hamilton would have to outperform by several points. And Kansas hasn’t elected a Democrat to the Senate since 1932! And yet, polls aside, there are some silver linings here for Democrats. Sharice Davids has repeatedly won in the blueish KS-3 district centered around well-educated Kansas City suburbs. Laura Kelly and Kathleen Sebelius won statewide in gubernatorial races, and Democrats absolutely crushed it in the abortion referendum in the state in 2022. And it’s trending a bit blue relative to the national baseline: Hillary Clinton lost it by 21 points in 2016 compared to Harris’s 16.

Furthermore, Roger Marshall is kind of the canonical example of a candidate with significant vulnerabilities who might not have been prepared for a blue wave. But, we’re suffering from a lack of quality polling in Kansas as compared to most of the other key races. Yet again, I’ll boringly suggest that the truth is probably somewhere in between FLIPR and Polymarket.

Nebraska

There’s less gap between FLIPR and prediction markets here, by contrast. For what it’s worth, FLIPR assumes that Dan Osborn, an independent, would caucus with the Democrats if he wins, and it makes the same assumption for Seth Bodnar in Montana. I’m not sure what assumption is baked into prediction markets, but Democrats cleared way for Osborn in Nebraska which is presumably a strong hint.

Still, for whatever reason, I’m intuitively a bit skeptical that the “trick” of running quasi-Democrats as independents is quite as clever as Democrats seem to think it is. Nebraska, like Kansas, is a state where Osborn would have to overperform the fundamentals by quite a lot.

And yet, even a recent Republican internal poll had Pete Ricketts ahead by only 2 points. If there’s the typical 3-to-4 point internal poll bias, that suggests that it really is close to a toss-up.7 Furthermore, Osborn lost in Nebraska by only 7 points against Deb Fischer in 2024, a pretty bad year for Democrats, and we’re projecting the national political environment to shift by more than that in November.

A 1-in-4 chance seems basically reasonable.

South Carolina

Here’s where I feel like we start to get into more speculative territory, even though the difference between FLIPR (27 percent for Annie Andrews) and prediction markets (20 percent) isn’t all that large. I suppose I have a reflexive concern about taking the bull case on Democrats in South Carolina given two factors. First, the much overhyped Jaime Harrison race, where he eventually lost to Lindsey Graham (RIP) by 10 points. And second, the state is relatively inelastic. It has a large Black population, a lot of conservative, evangelical whites, and not many swing voters.

Even according to FLIPR, there’s little chance — 4 percent — that South Carolina will be the tipping-point state. If polls are systematically overestimating Democrats in red states, Andrews’s chances are basically zip even if Darline Graham, Lindsey’s sister, is a weak candidate. If the polls are accurate about red-state performance, Democrats could pick up South Carolina if they’re having a fantastic night, but at that point they’ve probably already won Texas and so forth and have more than enough seats in hand.

Montana

In general, candidate selection decisions have gone well for Democrats in the Senate. Yes, Flanagan and El-Sayed have vulnerabilities, but they’re in states that should be quite blue based on national conditions. Montana is the most notable exception, where there’s been an unresolved stand-off between Alani Bankhead, the Democrat, and Seth Bodnar, the independent, whom we assume would caucus with Democrats if he wins.

A Bodnar internal poll recently showed himself narrowly winning, in fact, and Democrat Jon Tester was re-elected in the state in 2018 in an environment that might be similar to this one. FLIPR builds in some slightly complex assumptions about third-party candidates in cases like this, seeing them as having asymmetric distributions that are a little bit feast-or-famine. Basically, the hope for Bodnar would be that Democrats vote tactically, recognizing that he’s the best/only chance to defeat Kurt Alme.

I think I’m on FLIPR’s side here in terms of Montana being a race that presents a lot of intrinsic uncertainty. As in Nebraska, I suppose I’d expect the state to be polling closely if it were a true head-to-head race. Our forecast shows Alme beating Bodnar by 8 points on average, but with Bankhead getting 22.5 percent of the vote. If that Bankhead vote collapses, there might be a little crack in the door for Bodnar. To be fair, FLIPR Deluxe’s win probability has bounced around a fair bit in Montana, reflecting that this is one of our more assumption-driven forecasts.

Florida

In Florida, though, I think I’d go with Polymarket’s 9 percent for Angie Nixon and not FLIPR’s 14 percent. Granted, that isn’t a huge difference.

According to the fundamentals, Florida ought to be in the same vicinity as Ohio or Iowa — former swing states that have now turned red but could get the gang back together and moonlight as a battleground in such a blue national environment. But Ashley Moody has maintained a comfortable-ish lead in the more credible polls, and it’s easy to see why. New domestic migrants flowing into Florida are quite red, so our partisan lean calculations in the state are always lagging a bit.8 Meanwhile, Democrats are a mess in the state, and it’s one of the few competitive states this year where voter registration stats have continued to tilt more red. On top of that, Nixon is a DSA member in a state that has long been relatively capitalistic. So far, early-voting data in Florida has been much better for Democrats than in recent elections, but the governor’s race and the various competitive House races are probably the ones to watch here. The ~10 percent of the time that Nixon wins will reflect cases where everyone might even wonder WTF is wrong with polls because they missed a truly cataclysmic Democratic tsunami.

Louisiana and Mississippi

Polymarket is more bullish than FLIPR on Democrats’ chances in each state, especially in Mississippi. Neither state has much in the way of reliable polling, and ironically, some of the Mississippi hype may be because one of those fake Wedgewood polls that showed Democrat Scott Colom in a surprisingly close race against Cindy Hyde-Smith.

Hyde-Smith is a mediocre incumbent, having underperformed Mississippi’s partisan baseline in her past Senate races in 2018 and 2020; indeed, her 2018 race required a runoff. And in Louisiana, Julia Letlow is untested as a statewide candidate, although she was elected to Congress from Louisiana’s 5th Congressional District three times by comfortable margins. What makes me really skeptical, though, is that Louisiana and Mississippi are extremely inelastic states, with very few true swing voters. Granted, these are outlier-y states that are tough to model, but a 3-5 percent chance for Democrats instead of 8-13 percent as at Polymarket feels more grounded.

Arkansas and all redder states

These reflect a blend of other highly inelastic Southern states and then South Dakota and Idaho, where independents are running in place of Democrats. Nominally, prediction markets imply that FLIPR is underrating Democratic chances in some of these states. Never say never, I suppose, but like for the blue states at the top of the list, these markets aren’t very liquid, and I suspect reflect technical differences in differentiating “very slim” from “none” as much as any substantive disagreements.

Summing up

I’m slightly hesitant to include this section. But since it’s already reflected in the text above, here are the cases where there’s a difference of at least 5 points of win probability between FLIPR and Polymarket and where I’d break the tie if I were allowed to make subjective adjustments to FLIPR. For many of these states, I’m really in between FLIPR and the markets, but if you forced me to come down on one side, here’s where I’d go:

• Markets higher on Democrats; Nate mostly agrees with markets: Alaska

• Markets higher on Democrats; Nate mostly agrees with FLIPR: Mississippi (and Louisiana)

• FLIPR higher on Democrats; Nate mostly agrees with markets: Maine, Michigan, Iowa, Kansas, South Carolina, Florida

• FLIPR higher on Democrats; Nate mostly agrees with FLIPR: Ohio9, Texas, Montana.

Take all of that for what it’s worth, or what it isn’t. I don’t want this to be treated as some sort of quasi-official “extra Deluxe” Silver Bulletin forecast. FLIPR is what we should be judged by. But these are the states I’ll be watching carefully because of what they imply about how our models might be improved in the future.