Nate Silver's Blog, page 81

Live Politics Podcast: Will Trump Fire Rosenstein? Withdraw Kavanaugh?

FiveThirtyEight












 












More: Apple Podcasts |
ESPN App |
RSS
| Embed


Embed Code

In a live taping of the FiveThirtyEight Politics podcast, the crew discusses the future of both Deputy Attorney General Rod Rosenstein and Supreme Court nominee Brett Kavanaugh. The team also tries to make sense of the rumors that swirled around Rosenstein’s employment status on Monday and debates the politics at stake in whether or not Kavanaugh’s nomination is pulled.

You can listen to the episode by clicking the “play” button in the audio player above or by downloading it in iTunes , the ESPN App or your favorite podcast platform. If you are new to podcasts, learn how to listen .

The FiveThirtyEight Politics podcast publishes Monday evenings, with occasional special episodes throughout the week. Help new listeners discover the show by leaving us a rating and review on iTunes . Have a comment, question or suggestion for “good polling vs. bad polling”? Get in touch by email, on Twitter or in the comments.

The GOP’s Least-Worst Option Is If Kavanaugh Withdraws — And Soon

Brett Kavanaugh has never been a popular Supreme Court nominee — and he’s probably becoming more unpopular still following allegations earlier this month by Christine Blasey Ford that Kavanaugh had sexually assaulted her when they both were in high school. No one this unpopular has ever been confirmed to the Supreme Court; the only previous nominees who polled as poorly as Kavanaugh either had their names withdrawn (Harriet Miers) or lost their confirmation vote (Robert Bork). And all of this polling was taken before at least two other accusations surfaced of potential sexual misconduct involving Kavanaugh1 — and before Ford and Kavanaugh’s testimony before the Senate Judiciary Committee, which is scheduled for Thursday.

President Trump and Congressional Republicans are not afraid to take unpopular actions in pursuit of their ideological goals. Last year, they spent many months trying and failing to pass a repeal of Obamacare, even though those efforts were extremely unpopular. And they passed a tax bill that was highly unpopular at the time of its passage, although its numbers have since improved some. The Supreme Court is at least as much of a priority for Republicans.

The difference on Kavanaugh is that there are several other conservative nominees who could potentially replace him — and who may have been better picks in the first place. In other words, you would think Trump and Senate Majority Leader Mitch McConnell have better options than rolling the dice with Kavanaugh. Amy Coney Barrett, for example, a judge on the 7th Circuit and one of Trump’s reported finalists when Kavanaugh was chosen, has several advantages from the GOP’s point of view. She’d potentially be more conservative than Kavanaugh, at least on issues such as abortion; she’s already been confirmed (to her circuit seat) by the current Senate; and it might not hurt Republicans to choose a woman when the four conservatives on the current Supreme Court are all men.

Barrett also isn’t facing several accusations of sexual misconduct, as Kavanaugh is.

But there’s a midterm coming up in just six weeks. And there’s about a 3 in 10 chance that Republicans lose the Senate, according to the FiveThirtyEight forecast. Could Republicans really get Barrett or another nominee confirmed before then? And if not, could they confirm her in the so-called lame-duck session after the midterms but before the new Congress meets on Jan. 3.

The answers are “possibly” and “probably” — but the timing is getting dicier by the day. As of Tuesday morning, we’ll be 42 days away from the Nov. 6 midterms, and exactly 100 days away from when the new Congress convenes. The eight current members of the Supreme Court variously took between 50 and 99 days to be confirmed:

How long does it take to confirm a Supreme Court justice?

Days from nomination to confirmation for current members of the Supreme Court

Justice

Days from nomination to confirmation




Ruth Bader Ginsburg
50

–



John Roberts*
62

–



Neil Gorsuch
65

–



Sonia Sotomayor
66

–



Stephen Breyer
73

–



Samuel Alito
82

–



Elena Kagan
87

–



Clarence Thomas
99

–



Days until midterms
42

–



Days until next Congress
100

–

*Roberts was initially nominated for associate justice and then withdrawn and re-nominated for chief justice; our count of his confirmation time includes the combined time from both nominations.

Source: WIKIPEDIA

That makes the timing awfully interesting (and makes Republican complaints about Democratic delays to the process a little easier to understand). If Kavanaugh were to withdraw his name today, and Trump were to nominate someone else in his place tomorrow, the GOP might be able to confirm the replacement before the midterms — but the timing would be tight and would require a faster confirmation process than for any current member of the Supreme Court.

The lame-duck session would be a safer bet, but it’s not without risk for the GOP. One problem is that they might lose the Senate — to repeat ourselves, there’s about a 30 percent chance of this. Because the Senate is a much heavier lift for Democrats than the House, in the scenarios where the GOP loses the Senate, they’d probably also lose the House by wide margin; in our simulations, Republicans lose an average of about 50 (!!) House seats in scenarios where they also lose the Senate. The House doesn’t have any say in the Supreme Court nomination process, but would Republicans really want to push forward a nomination after losing by such a landslide margin?

My guess is probably yes — a Supreme Court seat really is that important to them. But the politics are uncertain; there aren’t really a lot of recent precedents for a party taking such significant action during the lame-duck session. And several Republican senators, after just having seen their colleagues take a drubbing in the 2018 midterms, might be skittish about what such a vote would mean for their survival in 2020, when the Senate map is a fairly tough one for the GOP.

In addition, there’s the chance the next nominee could have vetting problems, too. Historically, about 25 percent of Supreme Court nominations lapse, are voted down or are withdrawn.

Here’s the thing, though. The longer the GOP takes to replace Kavanaugh, the worse the timing problems become for them. If, say, the confirmation process on Kavanaugh drags out for another two weeks before he’s voted down or withdrawn, and then Trump takes another two weeks to choose a replacement because the overall process has become such a mess, then confirmation before the midterms would be extremely challenging. There also might not be enough time to seriously vet the new nominee before the lame-duck session, giving Republicans less margin for error then, too.

So why not just “plow right through” and vote to confirm Kavanaugh anyway, allegations and everything else aside? Although there’s a good chance McConnell is bluffing, that seems to be the current plan, with McConnell having promised a vote in the “near future” on Kavanaugh and no accusers other than Ford set to testify.

The problem is that this is an extremely live news story; with several new accusations having come out against Kavanaugh over the weekend and debates about the credibility of Ford’s allegations still ongoing. It’s hard to know what would happen to Kavanaugh if more accusations came out after he’d already been confirmed to the Supreme Court, but the possibilities include impeachment and serious long-term damage to the Court’s reputation — along with whatever additional price the GOP had to pay at the midterms. Even if the GOP were able to confirm Kavanaugh before the midterms this year, a landslide election could put the GOP in a considerably worse position to hold the Senate when other Supreme Court nominations come up in 2019 through 2024.

Put another way, there are huge risks to the GOP in both rushing to confirm Kavanaugh and in letting the process play out for several more weeks — which means that encouraging Kavanaugh to withdraw now, however painful it might be, is probably their least-worst option.

There is one other possibility, which is that McConnell — who reportedly didn’t want Kavanaugh to be chosen in the first place — could be rushing through the process in the hopes that Kavanaugh will be voted down (or forced to withdraw once it becomes clear that McConnell doesn’t have the votes). Back when Ford was Kavanaugh’s only accuser, this had seemed like a fairly likely exit strategy: The hearings would be engineered to allow Kavanaugh to save face, and perhaps to allow Republicans to stoke some grievances with their base. But wavering GOP senators such as Susan Collins and Jeff Flake would find some excuse to oppose his nomination and his nomination would be pulled. This scenario still seems like a distinct possibility — but the fact that the Kavanaugh story is developing so rapidly, with the stakes continuously increasing with every news cycle, could mean that McConnell is now pot-committed to the bluff even if he’d been hoping to keep his options open before.

I haven’t said much about the potential electoral upsides to the GOP of the confirmation process, such as possibly increasing base turnout, and putting vulnerable Democratic senators such as North Dakota’s Heidi Heitkamp to a tough vote on Kavanaugh or another nominee. That’s because I’m a little bit skeptical of them. Contrary to the conventional wisdom, it’s not clear that rank-and-file voters care about the Supreme Court as much as party activists and other “elites” do. And despite predictions that Anthony Kennedy’s retirement would help the GOP, Republicans’ electoral outlook has only gotten worse since then (and they’ve had especially poor polling in the past week or two).

For all that said, the Kavanaugh story has become unpredictable enough that its electoral effects are fairly uncertain, even if they’re weighted toward the downside for the GOP. If I were a Republican member of Congress facing reelection in 2018 or 2020, I’d just much rather take my chances with Barrett than with Kavanaugh.

Politics Podcast: Our Forecast Is 7,000 Lines Of Code

FiveThirtyEight












 












More: Apple Podcasts |
ESPN App |
RSS
| Embed


Embed Code

In the latest installment of the FiveThirtyEight Politics podcast’s “Model Talk,” Nate Silver discusses how unexpected events, like a troubled Supreme Court nomination, are processed by the forecast model. He also answers listener questions about how long the model’s code is and where it’s stored.

You can listen to the episode by clicking the “play” button above or by downloading it in iTunes , the ESPN App or your favorite podcast platform. If you are new to podcasts, learn how to listen .

The FiveThirtyEight Politics podcast publishes Monday evenings, with occasional special episodes throughout the week. Help new listeners discover the show by leaving us a rating and review on iTunes . Have a comment, question or suggestion for “good polling vs. bad polling”? Get in touch by email, on Twitter or in the comments.

Can A Democrat Really Win A Senate Race In Tennessee? Or Mississippi?!?!

Republicans don’t have a lot of exposure in the Senate this year — but they’re doing what they can to help Democrats make the most of it. The GOP entered this election cycle with only eight of their own seats up for grabs. Republican incumbents retired in three of those seats, however, and while Democrats don’t really stand a chance in Utah — where Mitt Romney will almost certainly succeed Orrin Hatch — the races to replace Sen. Jeff Flake in Arizona and Sen. Bob Corker in Tennessee are highly competitive.

In addition, Mississippi Sen. Thad Cochran’s early retirement this spring triggered a special election that will add to the Nov. 6 docket and that also gives Democrats a plausible chance at a pickup. Just how plausible? Read on. We’re covering each of these races — along with the Minnesota special election, the lone Democratic retirement of the cycle following Sen. Al Franken’s resignation — in today’s installment of POLLS vs. FUNDAMENTALS, the extremely dorky series of articles in which I evaluate the conflicting perspectives that polls and non-polling factors (“fundamentals”) provide on the Senate this year. In contrast to races featuring Democratic incumbents, where including fundamentals in our forecast generally helped Democrats, it helps Republicans in states such Mississippi and Tennessee:

A big gap between polls and fundamentals

Forecasted margin of victory or defeat in open-seat races and races with appointed incumbents

Republican’s forecasted margin of victory or defeat


Race
Incumbent
Fundamentals
Adjusted Polls




Arizona
—
-1.4
-3.8


Minnesota special
Smith (D)
-15.8
-8.2


Mississippi special
Hyde-Smith (R)
13.1
5.9*


Tennessee
—
13.8
-1.1


Utah
—
26.8
32.5

*Mississippi will hold a special election on Nov. 6, with candidates from all parties listed on the same ballot. If no candidate gets a majority of the votes, a runoff between the top two vote-getters will be held on Nov. 27. In the table, Hyde-Smith’s 5.9-point advantage reflects her lead over Democrat Mike Espy in polls of their prospective runoff. Another Republican candidate, Chris McDaniel, trails Espy by 17.6 points in runoff polling. In polls of the nonpartisan blanket primary scheduled for Nov. 6, Republicans lead Democrats in aggregate by 20.1 points in our polling average.

But let’s start in Arizona, which is a more straightforward case (and first in alphabetical order). In contrast to the relatively complex fundamentals calculation that our model makes for races featuring incumbents, the one for open-seat races is more straightforward: It accounts for only state partisanship, the generic congressional ballot, candidate experience, fundraising and a variable indicating whether or not a candidate is undergoing a scandal. In Arizona, Democratic nominee Kyrsten Sinema and Republican Martha McSally are both current U.S. representatives, meaning that the experience variable is a tie. And neither is caught up in a scandal. So the questions are who’s raised more money and whether the blue lean of the generic ballot is enough to offset the red lean of Arizona.

And the answer is … an open-seat race in Arizona ought to be pretty damned close in a political climate like this one. Hillary Clinton lost Arizona by only 4 percentage points in 2016, but the state has been more Republican-leaning than that in the past, and it’s been more Republican in statewide races than in federal ones. (Our partisanship variable accounts for state legislature results as well as presidential voting.) Sinema has a slight lead in fundraising, however, so the fundamentals calculation tips the race ever-so-slightly toward her, projecting her to win by nearly 2 percentage points. Sinema currently leads by slightly more than that, 4 percentage points, in the polling average. Nonetheless, this is a good example of how races tend to gravitate toward the fundamentals: Sinema’s polling lead had averaged 7 or 8 points before McSally won a contentious GOP primary last month.

FiveThirtyEight treats races featuring appointed incumbents (as opposed to elected incumbents) as tantamount to open-seat races, both because some of the variables we use to evaluate incumbents aren’t available in the case of appointees and because appointed incumbents don’t have a very good track record at winning re-election. As a case in point, the Minnesota and Mississippi special elections still have the potential to create headaches for Democrats and Republicans, respectively, as the appointed incumbents in those states (Sen. Tina Smith in Minnesota and Sen. Cindy Hyde-Smith in Mississippi) aren’t performing especially well in polls.

Even so, it’s probably too ambitious to think that Republicans will win the Minnesota special election given that elected Democratic incumbent Amy Klobuchar is poised to win by a landslide margin in Minnesota’s other Senate race; so-called double-barrel Senate elections (in which both of a state’s Senate seats are on the ballot in the same year) are almost always won by the same party. But whereas the fundamentals calculation projects Smith to win by 15 points — she’s raised more money and has held the higher elected office1 — Republican nominee Karin Housley has held Smith’s margin to the high single digits in most polls. It could be that the Housley name has a little extra currency in hockey-obsessed Minnesota (Housley’s husband, Phil Housley, is a member of the Hockey Hall of Fame and the current coach of the Buffalo Sabers). Still, Smith has about a 9 in 10 chance of winning — odds that Democrats ought to be reasonably happy with, considering the unpredictable race that they might have faced in 2020 if Franken had remained on the ballot.

The Mississippi special election is the most complicated race on the November ballot — and one of the hardest to forecast. If you take the polls there more at face value, it looks like an underrated opportunity for Democrats; based on fundamentals, however, it’s more of a long shot.

Here’s how it works. On Nov. 6, Mississippi will hold a nonpartisan blanket primary featuring multiple candidates from each party. If no one receives a majority of the vote, the top two finishers will advance to a runoff on Nov. 27. (Basically, this is the same thing that Louisiana does with its congressional races every year.) There are three major candidates in the race: Hyde-Smith, the appointed Republican incumbent; Chris McDaniel, the controversial tea party-backed Republican who nearly defeated Cochran in the GOP primary four years ago; and Mike Espy, an African-American Democrat who served as a U.S. representative in the 1980s and 1990s and then as Bill Clinton’s first Secretary of Agriculture. (There’s also a second Democrat, Tobey Bartee, but he has only about 2 percent of the vote in polls.)

Polls of the blanket primary show Espy and Hyde-Smith roughly tied for first with about 30 percent of the vote each, and McDaniel some ways behind in the high teens, with a substantial number of undecided voters. None of those numbers are too surprising; President Trump endorsed Hyde-Smith, although she was ahead of McDaniel even before the endorsement. So there’s likely to be a runoff and it’s likely to feature Espy against Hyde-Smith.

There’s also polling of potential runoff matchups, however, and in those polls, Espy does surprisingly well — at least as compared to how the fundamentals might expect him to do. On average, he trails Hyde-Smith by only about 6 percentage points in head-to-head polls and leads McDaniel in a potential runoff by about 18 points. So the Lite version of our model, which forecasts the runoff based on polls only, gives Espy a decent shot of beating Hyde-Smith (and assumes he’d crush McDaniel on the off-chance McDaniel made the runoff instead). The Classic and Deluxe versions, by contrast, which account for fundamentals, consider that Mississippi is a red state and that the two Republicans are likely to get significantly more votes combined than the two Democrats (Espy and Bartee) on Nov. 6, which is historically a good predictor of runoff results. Thus, they see Espy as a fairly heavy underdog against Hyde-Smith, and also think he could have a tough time with McDaniel despite leading him in polls.

Every scenario in the special Senate election in Mississippi

Probability of occurring, based on 50,000 simulations as of Sept. 19, 2018

Scenario
Lite
Classic
Deluxe




Espy wins a majority on Nov. 6
5.4%
5.9%
6.2%


Espy beats Hyde-Smith in runoff
17.4
8.9
9.3


Espy beats McDaniel in runoff
0.6
0.4
0.5


Hyde-Smith wins a majority on Nov. 6
2.7
2.1
1.8


Hyde-Smith beats Espy in runoff
71.8
80.5
80.4


Hyde-Smith beats McDaniel in runoff
1.5
1.4
1.2


McDaniel wins a majority on Nov. 6





McDaniel beats Espy in runoff

0.3
0.3


McDaniel beats Hyde-Smith in runoff
0.5
0.4
0.3

Needless to say, all of this is pretty complicated. We’re doing the best we can to model these possibilities based on data from past races in Louisiana, as well as from California and Washington state, which use a somewhat similar top-two format, but there’s inherently a fair amount of uncertainty here. The bet the fundamentals would place is simply that a runoff held three weeks after Election Day in a deep-red state like Mississippi — one that could possibly determine control of the Senate — is highly likely to favor Republicans. But polls of the runoff are more equivocal.

In comparison to Mississippi, there’s nothing especially tricky about Tennessee — it’s just a case where polls and fundamentals have totally different perspectives on the race. Polls show roughly a toss-up there between Democrat Phil Bredesen, the former governor, and Republican U.S. Rep. Marsha Blackburn. (If anything, Bredesen has just the slightest lead, as he tends to be doing better in the higher-quality surveys.) The fundamentals, by contrast, project a Republican win of 14 percentage points. Tennessee is very red, having voted for Trump by 26 points. Unlike some other Southern states, it’s also quite red in statewide elections; Republicans have a 28-5 advantage in the Tennessee Senate, for example. Blackburn has also slightly outraised Bredesen so far, making it among the only competitive Senate races where the GOP has the fundraising lead. The model does give Bredesen credit for being a former governor, but he has a lot going against him.

Rather than parse Tennessee itself any further, let’s instead see what has happened in the past when polls and fundamentals clash. In the chart below, I’ve listed Senate races since 1990 where there was at least a 10-point gap between the polls and the fundamentals with 60 days to go in the campaign, based on backtested results from the FiveThirtyEight model. I’ve limited the analysis to races deemed to be competitive by the Cook Political Report and where there was an adequate amount of polling. For example, the Indiana Senate race in 2016 meets all of those criteria; Democrat Evan Bayh, a former governor and senator, was well ahead in polls, but the fundamentals calculation regarded Republican Todd Young as the favorite. (Young eventually won handily.)

What happens when polls and fundamentals clash?

Competitive U.S. Senate races since 1990 featuring at least a 10-point gap between polls and fundamentals with 60 days until the election

Democratic Candidate’s Margin 60 Days Before Election





State
Year
Polling Average
Fund-amentals
Actual Result
Fundamentals more accurate than polls?
Race moved in direction of fundamentals?




IL
1990
19.4
7.1
30.1




IL
1992
34.2
12.2
10.2
✓
✓


CA
1994
16.7
0.9
1.9
✓
✓


NJ
1994
18.5
5.2
3.3
✓
✓


OH
1994
-17.9
-7.9
-14.2

✓


SD
1996
-0.7
-13.7
2.6




CA
1998
0.6
10.7
10.0
✓
✓


IL
1998
-4.7
10.2
-2.9

✓


NY
2000
-0.0
17.0
12.3
✓
✓


TX
2002
-1.8
-16.5
-12.0
✓
✓


LA
2004
-20.0
3.9
-21.7




NC
2004
8.3
-9.1
-4.6
✓
✓


SD
2004
3.2
15.7
-1.2




MD
2006
6.6
21.3
10.0

✓


NJ
2006
-1.2
19.9
9.0

✓


PA
2006
8.1
-3.4
17.4




TN
2006
-3.3
12.4
-2.7

✓


VA
2006
-1.4
-18.1
0.4




WA
2006
10.4
24.1
16.9

✓


AK
2008
2.5
-28.2
1.2

✓


VA
2008
22.7
5.7
31.3




AR
2010
-27.8
4.1
-21.0

✓


NC
2010
-6.4
-17.1
-11.8
✓
✓


NV
2010
1.8
13.1
5.7

✓


CT
2012
-1.6
23.0
11.8
✓
✓


SD
2014
-13.7
-30.8
-20.9

✓


GA
2016
-13.1
-24.3
-13.8

✓


IN
2016
9.9
-6.1
-9.7
✓
✓

Races are listed if: (i) They had at least a 10-point gap between polls and fundamentals with 60 days to go until the election; (ii) they were rated as competitive rather than “safe” by the Cook Political Report; (iii) they had an adequate amount of polling (equivalent to a cumulative polling weight of at least 2); and (iv) they did not feature a viable third-party candidate.

What happened in these races? Polls came closer to the final margin about two-thirds of the time (in 18 of 28 cases). So if you had to choose between polls and fundamentals, you’d pick polls. However, the race moved in the direction of the fundamentals three-quarters of the time (in 21 of 28 cases). That is, if the Republican was doing better according to the fundamentals analysis than according to the polls, the Republican tended to gain ground 75 percent of the time, and likewise for the Democrat.

So the best forecast comes from taking a blend of (mostly) polls and (some) fundamentals. Exactly how much our model weights each component depends on the amount of polling and the amount of time left until Election Day. Essentially, the fundamentals calculation is treated as the equivalent of 1 or 2 recent, high-quality polls. So if there are 10 or 15 recent polls of a state, the fundamentals calculation has little influence. In states such as North Dakota where polling is fairly sparse, they can have more sway.

What that means for Tennessee is that any poll showing Bredesen tied or ahead — and perhaps even behind by 1-2 percentage points — is good news for Bredesen, because the model expects the race to revert toward Blackburn based on the fundamentals. With every new poll, it weights the fundamentals less and less. With that said, Bredesen isn’t out of the woods yet; that Indiana race took a long time to tip toward Young in 2016 before he surged ahead on Election Day.

Finally, we come to Utah, where there isn’t much suspense: Romney is a massive favorite according to both polls and fundamentals, and the only real question is whether he’ll be a reliable vote for Trump or a thorn in the president’s side.

CLARIFICATION (Sept. 20, 2018, 1:35 p.m.): A previous version of this article gave the impression that former Sen. Al Franken would have been up for re-election this year had he not resigned. His seat would normally not be up for election again until 2020.

Will Democratic Senators Lose Despite The ‘Blue Wave’?

Our congressional forecasts reflect a blend of several different methods of prediction. But for the most part, those methods tell a fairly consistent story. In the House, for instance, district-by-district polls, the generic congressional ballot and historical trends in midterm elections all point toward Democrats winning the national popular vote by somewhere in the range of 8 to 10 percentage points, which would very probably be enough for them to take control of the chamber.1 So do what we call the “fundamentals,” non-polling indicators that have empirically been useful predictors of races for Congress, such as fundraising totals, past margins of victory and several factors related to incumbency.2

Likewise in the Senate, the different versions of our model, which blend these methods together in different ways, tell a similar overall story to one another. It’s a reasonably happy story for Republicans because the Senate map, which consists overwhelmingly of Democratic-held seats, is highly favorable for the GOP this cycle. The polling-driven Lite version of our Senate forecast has the GOP finishing with 51.3 Senate seats and having about a 5 in 7 chance (more precisely, 71 percent) of keeping control of the chamber.3 The Classic version, which incorporates fundamentals, has them with 50.8 seats and about a 2 in 3 chance (68 percent) of maintaining control instead.4 If you go down to the decimal point, the Classic forecast is ever-so-slightly better for Democrats than the Lite forecast — which implies that the fundamentals are ever-so-slightly better for them than polls — but it isn’t a big overall difference.

The word “overall” is doing a lot of work in the previous sentence, however, because while the top-line prognosis on who might control the Senate is similar in all three versions of our model, the forecasts differ quite a bit from race to race. In particular:

The fundamentals are more bullish than polls for Democrats in several states with Democratic incumbents — most importantly, in Florida, Missouri and North Dakota.
But fundamentals are more bearish for Democrats in two important open-seat races: Tennessee and the Mississippi special election.
Polls and fundamentals are fairly consistent with one another in states with Republican incumbents — including in Texas, where fundamentals support the notion of a competitive race between Sen. Ted Cruz and Rep. Beto O’Rourke.

In this article, I’m going to focus on the first category only: races featuring Democratic incumbents. We’ll cover the other two categories in an upcoming piece.

The ‘fundamentals’ say it should be really hard for Democratic incumbents to lose

(Warning: Sports analogy ahead.) In soccer, the most obvious advantage of having a man advantage after an opposing player has been sent off on a red card is that it’s much easier to score goals. But what’s nearly as important is that it becomes very difficult for the other team to score a goal; defense becomes a cinch.

So it also goes in congressional races when your party is having a wave election: You’ll win plenty of your opponent’s seats, but — not to be overlooked — you’ll usually also lose almost none of your own seats. The GOP didn’t lose any of its own Senate seats in 1994, 2010 or 2014, for example. Democrats didn’t lose any of their own seats in 2006 or 2008. Between these five election cycles, the “waving” party went undefeated5 in defending its Senate seats.

This year, however, we have a seeming contradiction: The polls are pointing toward a wave in the House, with an average projected gain of 35 to 40 seats for Democrats and a popular vote win of 8 to 10 points.6 And yet, Democrats are at risk of losing several of their own Senate seats, which could offset any gains they may make among GOP-held seats and make it much harder for Democrats to take control of the Senate.

One way out of the dilemma for Democrats is if their incumbents in the Senate aren’t in quite as much trouble as the polls show — and our fundamentals calculation suggests that could be the case. Below is a comparison of polls and fundamentals in each seat held be an elected Democratic incumbent. Note that elected incumbent excludes appointed incumbents such as Tina Smith in the Minnesota special election; we’ll deal with her race in the next installment.

Most Democratic incumbents are doing worse than the ‘fundamentals’ would predict

Forecasted margin of victory for Democratic senators who are running for re-election, according to FiveThirtyEight’s fundamentals and adjusted polls

Forecasted margin of victory


Race
Incumbent
Fundamentals
Adjusted Polls




Florida
Nelson
+7.7
-0.1


North Dakota
Heitkamp
+7.9
-2.4


Missouri
McCaskill
+8.3
+1.0


West Virginia
Manchin
+9.0
+10.4


Ohio
Brown
+10.4
+13.5


Pennsylvania
Casey
+11.6
+13.1


Indiana
Donnelly
+11.6
+2.8


Montana
Tester
+13.4
+5.7


Wisconsin
Baldwin
+17.9
+10.7


Michigan
Stabenow
+18.8
+18.5


Virginia
Kaine
+21.3
+15.0


New Jersey
Menendez
+23.3
+6.9


New Mexico
Heinrich
+25.1
+20.1


Minnesota
Klobuchar
+27.6
+23.3


Washington
Cantwell
+29.3
+16.3


Maine
King*
+30.7
+30.6


Connecticut
Murphy
+33.2
+23.3


Delaware
Carper
+36.6
+9.9


Massachusetts
Warren
+37.4
+28.0


Rhode Island
Whitehouse
+38.9
+23.5


New York
Gillibrand
+39.9
+30.9


Maryland
Cardin
+40.6
—


Vermont
Sanders*
+45.7
—


Hawaii
Hirono
+51.8
—

*King and Sanders are independents who caucus with the Democrats.

The Minnesota special election is not listed because Tina Smith is an appointed rather than an elected incumbent and our model treats races with appointed incumbents as open-seat races. California is not listed because the race features two Democrats and no Republicans. There is no polling of the Senate races in Maryland, Vermont and Hawaii.

Adjusted polling as of Sept. 17; fundamentals as of Sept. 16.

A handful of Democrats, such as West Virginia’s Joe Manchin, Ohio’s Sherrod Brown and Pennsylvania’s Bob Casey, are running slightly ahead of their fundamentals, but most Democrats are underperforming them. Florida, North Dakota and Missouri, where polls show near toss-ups in races that the fundamentals suggest should be Democratic-leaning, are the most important cases. Montana’s Jon Tester and Indiana’s Joe Donnelly, although they lead in most polls, also fall into this category, as fundamentals suggest they should have a slightly clearer advantage. Polls also show New Jersey’s Bob Menendez in a somewhat competitive race, when fundamentals imply Republicans should have no business competing in New Jersey in such a Democratic-leaning year — even accounting for Menendez’s corruption scandal.

Before we go further, it might be helpful to break fundraising out from the other factors we include in the fundamentals. Why? It isn’t necessarily because fundraising is the most important or predictive factor, although, it’s one of the most predictive ones. Rather, it’s because we tend to think of fundamentals as background conditions, whereas fundraising is in part a reaction to those conditions. Voter enthusiasm, which by most indications is on Democrats’ side this year, can translate to higher fundraising totals, for instance.

Democrats have a fundraising advantage in every single Senate race with a Democratic incumbent. (Our evaluation of fundraising is based on individual contributions only, not money from PACs or parties or self-financing from the candidates.) This isn’t a huge surprise — the incumbent candidate usually raises more than the challenger.7 But the Democratic fundraising edge is quite emphatic, with the median Democratic incumbent in the Senate having almost a 6-1 fundraising advantage over his or her Republican opponent. This makes a pretty big difference. As you can see in the next chart — which breaks out how much fundraising contributes to the fundamentals calculation in each race — there are several races where the Democratic incumbent’s edge would otherwise be fairly marginal, but where fundraising tips the scales toward them.

Democrats’ fundraising advantage helps buoy their lead

Forecasted margin of victory for Democratic senators who are running for re-election, according to various aspects of FiveThirtyEight’s fundamentals

Forecasted margins by fundamentals inputs


Race
Incumbent
fundamentals Forecasted margin of victory
Fundraising
Other Factors




Florida
Nelson
+7.7
+1.9
+5.8


North Dakota
Heitkamp
+7.9
+4.3
+3.7


Missouri
McCaskill
+8.3
+5.0
+3.3


West Virginia
Manchin
+9.0
+2.4
+6.6


Ohio
Brown
+10.4
+7.7
+2.6


Pennsylvania
Casey
+11.6
+4.5
+7.0


Indiana
Donnelly
+11.6
+4.4
+7.2


Montana
Tester
+13.4
+5.7
+7.7


Wisconsin
Baldwin
+17.9
+7.0
+10.9


Michigan
Stabenow
+18.8
+2.3
+16.5


Virginia
Kaine
+21.3
+6.2
+15.2


New Jersey
Menendez
+23.3
+4.5
+18.8


New Mexico
Heinrich
+25.1
+5.0
+20.1


Minnesota
Klobuchar
+27.6
+8.1
+19.5


Washington
Cantwell
+29.3
+7.5
+21.8


Maine
King*
+30.7
+6.1
+24.6


Connecticut
Murphy
+33.2
+8.5
+24.6


Delaware
Carper
+36.6
+7.1
+29.5


Massachusetts
Warren
+37.4
+8.5
+29.0


Rhode Island
Whitehouse
+38.9
+5.4
+33.6


New York
Gillibrand
+39.9
+7.9
+32.0


Maryland
Cardin
+40.6
+7.9
+32.7


Vermont
Sanders*
+45.7
+7.7
+38.1


Hawaii
Hirono
+51.8
+7.8
+43.9

*King and Sanders are independents who caucus with the Democrats.

The Minnesota special election is not listed because Tina Smith is an appointed rather than an elected incumbent and our model treats races with appointed incumbents as open-seat races. California is not listed because the race features two Democrats and no Republicans. There is no polling of the Senate races in Maryland, Vermont or Hawaii.

Adjusted polling as of Sept. 17; fundamentals as of Sept. 16.

It’s also worth describing how our model “thinks” about incumbency — and how the incumbency advantage is a bit more complex than people might assume. For example, were it not for an overwhelming fundraising advantage over his Republican opponent Rep. Jim Renacci, the fundamentals model would actually have Ohio’s Brown as being more vulnerable than West Virginia’s Manchin.

How can that be? Isn’t West Virginia much redder than Ohio? Well, sure.8 (Although Ohio is getting pretty red too.)

But the most important fact about incumbents is that incumbents, by definition,9 won their previous election. So it can help to look at the incumbent’s previous margin of victory for clues about his or her strength. For instance, West Virginia is much redder than Ohio now, but that was also true in 2012 — and in 2012, Manchin won his race by 24 points, while Brown won his by just 6. That’s a hint that the incumbency advantage isn’t one-size-fits-all and that there are some characteristics — either relating to Manchin himself or to his state, West Virginia — that made him a stronger incumbent than Brown.

One clear trend in the data, for instance, is that incumbents from smaller-population states such as West Virginia tend to be harder to beat than those from more populous ones such as Ohio. Look up the historical rosters of senators from Hawaii or Delaware, for instance, and you’ll tend to see the same incumbents getting re-elected every six years until they finally retire or pass away; less so in larger states such as Florida. I wasn’t surprised when we discovered this trend in the data; there are several fairly straightforward reasons why you’d expect it to be the case.10

Another pattern is that more demographically and politically idiosyncratic states tend to be associated with larger incumbency advantages. We measure each state’s similarity to other states by means of its CANTOR score, which evaluates how comparable it is to other states based on a variety of demographic, political and geographic variables. We found that states that have fewer good comparables — say, West Virginia or Vermont, or even California — tend to re-elect their incumbents more often than states whose demographics more closely reflect those of the country as a whole, such as Ohio or Missouri. This also ought to be fairly intuitive; it’s easier for an incumbent to be a uniquely good fit for his or her state if that state is more unique.

Manchin also has a considerably more moderate voting record than Brown, another factor that our model considers. He’s voted with the Democratic Party only 63 percent of the time over the past three Congresses, according to our version of a party unity score, as compared with 89 percent of the time for Brown. Our model finds that incumbents who break ranks with their parties more often tend to do better than those who always vote the party line, other factors held equal.

With that additional background, let me quickly go through the model’s logic on some of the key races with Democratic incumbents.

Florida. It might seem surprising that the fundamentals calculation regards Florida’s Bill Nelson as the most vulnerable Democratic incumbent, since Florida is quite purple and there are Democrats up for re-election in some genuinely red states. But Florida is a populous, diverse state — and as I described, those states tend to be associated with a smaller incumbency advantage. Moreover, Nelson has a very good challenger in Florida Gov. Rick Scott; one way our model accounts for candidate quality is by looking at the highest elected office the opponent has held, with races against current or former governors or senators falling into the top category. Scott has also kept a relatively even pace with Nelson in terms of public fundraising — and that’s not even accounting for all the money he’s kicked into the campaign himself. That said, the fundamentals calculation still thinks Nelson ought to be ahead in such a blue-leaning year.
North Dakota. Part of why a party’s incumbents are so tough to beat when it’s having a wave election is that they probably won their races under tougher conditions before. For instance, Democratic incumbent Heidi Heitkamp barely won her race in 2012, and North Dakota has grown redder since then. Nonetheless, Democrats are poised to win the overall popular vote for the House by around 9 percentage points this year, a significant blue shift from 2012, when they won it by just 1 point. Also helping Heidkamp is that she has a moderate voting record, that North Dakota is a small state, and that so far she has a significant fundraising advantage over her Republican opponent, Rep. Kevin Cramer.
Missouri. Democrat Claire McCaskill would otherwise be regarded as quite vulnerable by the fundamentals calculation — she has a more liberal voting record than colleagues like Heitkamp, and Missouri is a larger and more diverse state. But she’s been prepared for a competitive race since she won re-election in 2012, and she has a 4:1 fundraising advantage so far over Republican opponent Josh Hawley. One factor our model doesn’t account for is who the incumbent beat in their last election; in McCaskill’s case, it was former U.S. Rep. Todd Akin, whose comments about “legitimate rape” undermined his candidacy. That potentially leads the model to overrate McCaskill’s chances slightly; it gives her full credit for her 16-point margin of victory against Akin, when the results would likely have been much closer against another Republican opponent.
West Virginia. We’ve already discussed Manchin’s case. He’s a throwback to an older, independent-minded type of incumbent who’s a uniquely good “fit” for his state in a way that usually defies partisan trends. In a red-leaning national environment, Manchin might nonetheless be in trouble — but in a blue year, it shouldn’t be surprising that he’s leading Republican opponent Patrick Morrisey in polls.
Ohio. In theory, Brown might be vulnerable, but he hasn’t drawn a very tough challenger — Renacci has raised relatively little money and hasn’t shown much of an independent streak in Congress.
Pennsylvania. Casey’s story is similar to Brown’s: Pennsylvania has become redder, but it’s still not an easy pickup for Republicans in a blue year, and Casey’s opponent, Rep. Lou Barletta, is probably not who you’d pick to maximize your chances of winning.
Indiana. Donnelly benefits in our analysis from a moderate voting record and from facing a relatively inexperienced opponent in businessman and former state representative Mike Braun. As is the case with McCaskill, however, our model may slightly overrate Donnelly because it doesn’t account for the fact that he faced a problematic opponent, Richard Mourdock, when he won the seat in 2012.
Montana. Tester is potentially Manchin-like in being a good fit for a quirky, low-population Montana. He has a considerably more liberal voting record than Manchin or Heitkamp do, but that’s partly offset by Montana being slightly more purple than North Dakota or West Virginia.
Michigan and Wisconsin. Michigan’s Debbie Stabenow has had an emphatic lead in polls, while Wisconsin’s Tammy Baldwin’s advantage has varied more from survey to survey (although with an average lead of around of 11 points in our adjusted polling average). The fundamentals calculation is skeptical that either race will become more competitive. Indeed, the four Rust Belt states that flipped to Trump in 2016 (Ohio, Pennsylvania, Michigan and Wisconsin) all seemed poised to easily re-elect their Democratic incumbents this year.
New Jersey. Our model actually does account for scandals, and Menendez’s scandal is a significant one — although the federal case against was him dismissed after a mistrial, with most jurors prepared to say he was not guilty. But New Jersey is a blue state, and the scandal penalty to Menendez’s margin would have to be enormous to cost the incumbent his seat in a blue year. This is a risk Democrats could have avoided entirely by nominating another candidate, but probably not one that will come back to bite them.

One final, important note: While several of the Democratic incumbents, such as Heitkamp and Menendez, are underperforming where they “should” be in the polls, our model will give them only so much time to catch up to its expectations. As time passes and as more and more polls come in, the effect of the fundamentals will diminish in our forecast — to the point where they really only serve as a tiebreaker in Senate races that are otherwise too close to call. We’ll cover GOP incumbents and open-seat races, where the fundamentals calculation sometimes helps Republicans, in another article soon.

Politics Podcast: What Manafort’s Plea Means For Trump

FiveThirtyEight












 












More: Apple Podcasts |
ESPN App |
RSS
| Embed


Embed Code

The FiveThirtyEight Politics podcast is recording a live podcast in New York City on Sept. 24. You can come! Get details and tickets here.

President Trump’s former campaign chairman, Paul Manafort, entered a plea agreement on Friday in which he agreed to cooperate with the Mueller investigation into Russian meddling in the 2016 election. In this emergency installment of the FiveThirtyEight Politics podcast, the crew explains what his cooperation could look like and debates how much of a risk this poses for Trump.

You can listen to the episode by clicking the “play” button above or by downloading it in iTunes , the ESPN App or your favorite podcast platform. If you are new to podcasts, learn how to listen .

The FiveThirtyEight Politics podcast publishes Monday evenings, with occasional special episodes throughout the week. Help new listeners discover the show by leaving us a rating and review on iTunes . Have a comment, question or suggestion for “good polling vs. bad polling”? Get in touch by email, on Twitter or in the comments.

Politics Podcast: The Battle For The Senate Is Full Of Interesting Races

 

By Nate Silver and Galen Druke and Nate Silver and Galen Druke












 












More: Apple Podcasts |
ESPN App |
RSS
| Embed


Embed Code

Note: The FiveThirtyEight Politics podcast is recording a live podcast in New York City on Monday, September 24th. Find details and tickets here.

Our Senate forecast is live and in the latest installment of the FiveThirtyEight Politics podcast‘s “Model Talk,” Nate Silver discusses why Republicans are favored to keep the Senate and which races promise to be the most interesting. As usual, he also answers listener questions on topics ranging from the model’s coding language to whether polls affect public opinion.

You can listen to the episode by clicking the “play” button above or by downloading it in iTunes , the ESPN App or your favorite podcast platform. If you are new to podcasts, learn how to listen .

The FiveThirtyEight Politics podcast publishes Monday evenings, with occasional special episodes throughout the week. Help new listeners discover the show by leaving us a rating and review on iTunes . Have a comment, question or suggestion for “good polling vs. bad polling”? Get in touch by email, on Twitter or in the comments.

Republicans Are Favorites In The Senate, But Democrats Have Two Paths To An Upset

Many of the individual race forecasts in the FiveThirtyEight Senate model, which launched on Wednesday, look pretty optimistic for Democrats. The model shows Sen. Joe Manchin in a strong position to retain his seat in West Virginia, for instance. It has Democrats as ever-so-slight favorites to win the GOP-held Senate seats in Nevada and Arizona. It thinks Democratic incumbents like Missouri’s Claire McCaskill and North Dakota’s Heidi Heitkamp could close well down the stretch. It even gives Democrat Beto O’Rourke a credible shot in Texas — although it has Sen. Ted Cruz as the favorite in the race.

But despite that, the model has Democrats as reasonably clear underdogs to take control of the Senate. Even though it’s more optimistic than the consensus about Democrats’ chances in several individual races — and even though the model is generated by the same program that gives Democrats around a 5 in 6 chance of winning the House — it nevertheless says Republicans have somewhere between a 2 in 3 and 7 in 10 chance to hold the Senate, depending on which version of our model you look at.

This isn’t any sort of paradox: The Senate map is simply very, very daunting for Democrats. In fact, it’s about as unfavorable a map as any party has faced in the Senate, ever. Democrats have 26 seats up for election in November; Republicans have just nine. Moreover, 10 of those 26 seats are in states that President Trump won in 2016, including five states (West Virginia, North Dakota, Indiana, Montana, Missouri) where he won overwhelmingly.

Just how uphill a task is that for Democrats? We currently forecast Democrats to win the popular vote for the U.S. House by 8 to 9 percentage points (similar to their advantage on the generic congressional ballot) — a margin that by almost any definition would qualify as a “wave election.” As a point of comparison, Republicans won the House popular vote by 7 percentage points in 1994 and in 2010, and Democrats won it by 8 points in 2006, all of which are usually considered wave years. But our model thinks that even an 8- or 9-point advantage would probably not be enough for Democrats to win the Senate. Instead, they would need around an 11-point advantage in the House popular vote before becoming favorites to claim the Senate, our model estimates.

But just as Republicans are far from doomed in the House, they are far from safe in the Senate. Democrats need to gain only a net of two seats to take control of the Senate, and they have five plausible opportunities: Arizona, Nevada, Tennessee, Texas and (most debatably) the Mississippi special election, which involves a nonpartisan blanket primary on Nov. 6 with a potential runoff three weeks later. Meanwhile, Republicans have three very good opportunities to pick off Democratic incumbents — those are McCaskill in Missouri, Heitkamp in North Dakota and (surprisingly) Bill Nelson in Florida — and there are several other states where they’re still in the running, such as Indiana. But no one of those races is a sure thing for Republicans. In fact, the Classic and Deluxe versions of our model have Republicans as slight underdogs in all of the Democratic-held seats, although the polling-driven Lite version of the forecast has them favored in North Dakota and Florida.

In essence, there are two ways by which Democrats might win the Senate: a macro path and a micro path.

First, the macro path. As I mentioned, Democrats would become favorites to take the Senate if they won the overall popular vote for the House by more than about 11 points. At that point, the tailwind would simply be so strong that Democrats would probably find a way to win all or almost all of the toss-up races. (Keep in mind that Senate races are not truly independent of one another; instead the outcomes are correlated to a meaningful extent.) We wouldn’t bet on Democrats winning the House popular vote by 11 points, which would be an overwhelming margin — the most lopsided margin for either party since 1982. But considering that they’re already ahead by 8 or 9 points right now, it can hardly be ruled out.

Or, the Democrats could win by means of the micro path and just have the coin come up heads in a lot of the toss-up races, even if the overall political environment isn’t any better for them than we’re currently projecting. In our Classic forecast, there are 11 seats that each party has at least a 10 percent chance of winning: Arizona, Florida, Indiana, Missouri, Mississippi, Montana, North Dakota, Nevada, Tennessee, Texas and West Virginia. Democrats need to win eight of those 11 to take the Senate. If each of these races were truly a coin flip — meaning, a 50-50 proposition where the outcomes were independent from one another — then Democrats would have to do the equivalent of coming up with eight heads in 11 tries, the chances of which are about 11 percent.

Overall, the macro path is probably the easier one for Democrats. That is to say, if you woke me up on Nov. 7 and told me that the Democrats had won the Senate, I’d guess it had been because the overall political environment had been even better for them than we’d expected — and not because they’d navigated their way through the thicket of individual races.

With that said, almost all the Senate races are fascinating on an individual level. They generally feature high-quality candidates; Democrats are competing almost everywhere they can compete, and Republicans have mostly avoided nominating the Todd Akins and Christine O’Donnells of the world, instead having chosen more traditional nominees. In quite a few races, moreover, there’s a clash between polling and “fundamentals” — non-polling factors like fundraising and a state’s past voting history. The model is skeptical that Democrats have truly made Tennessee a toss-up, for instance, even though that’s what polls show there. On the flip side, it’s surprised that Nelson, McCaskill and Heitkamp aren’t polling better given that parties rarely lose many of their own seats in wave elections. (We’ll go into more detail on these themes in an upcoming Election Update.)

Finally, a couple of notes on how the Senate and the House forecasts interact with each other. As we explain in our methodology post, the forecasts are literally generated by the same model, so data from the Senate forecasts can affect the House numbers and vice versa.

And one scenario we can almost rule out is Democrats winning the Senate but failing to take the House. There’s less than a 2 percent chance of that, according to our model, simply because the Senate is such a heavy lift for Democrats that they’ll almost certainly have won the House by the time they get there.

How our House and Senate forecasts interact

According to FiveThirtyEight’s 2018 midterm forecasts, as of Sept. 12

Scenario
Probability of scenario occurring


Senate
House
Lite
Classic
Deluxe




Democratic
Democratic
28%
34%
33%


Democratic
Republican
2




Republican
Democratic
45
49
45


Republican
Republican
26
17
21

By the same measure, the election is in a relatively precarious balance. Sure, right now, the Democrats are favored to take the House and the Republicans to hold the Senate. But even a relatively modest shift in the Democrats’ direction would make them favorites to take the Senate as well, and likewise for Republicans and the House. In fact, we’d bet against both of our forecasts being right simultaneously! There’s a greater than 50 percent chance that either Republicans win the House or Democrats win the Senate1 by the time we get to Election Day.

Check out all the polls we’ve been collecting ahead of the 2018 midterms.

How FiveThirtyEight’s House And Senate Models Work

References

2016 presidential forecast / 2016 and 2014 Senate forecast / 2012 presidential forecast

Cook Political Report ratings / Inside Elections ratings / Sabato’s Crystal Ball ratings / elasticity scores / generic congressional ballot / pollster ratings / scandals database

The Details

The principles behind our House and Senate forecasts should be familiar to our longtime readers. They take lots of polls, perform various types of adjustments to them, and then blend them with other kinds of empirically useful indicators (what we sometimes call “the fundamentals”) to forecast each race. Then they account for the uncertainty in the forecast and simulate the election thousands of times. Our models are probabilistic in nature; we do a lot of thinking about these probabilities, and the goal is to develop probabilistic estimates that hold up well under real-world conditions. For instance, when we launched the 2018 House forecast, Democrats’ chances of winning the House were about 7 in 10 — right about what Hillary Clinton’s chances were on election night in 2016! So ignore those probabilities at your peril.

The methods behind our House and Senate forecasts are nearly identical. In fact, they’re literally generated by the same program, with data from House races informing the Senate forecasts and vice versa. However, there are a few differences between how House and Senate races are handled, which we’ll describe as they arise below. Additionally, the congressional models have a different flavor to our presidential forecasts in two important respects:

Races for Congress are more localized than presidential races, and this is reflected in the design of the models. In presidential elections, the same candidates are on the ballot everywhere, and outcomes are extremely correlated from one state to the next. It wasn’t a surprise that President Trump won Michigan in 2016 given that he also won demographically similar states such as Wisconsin and Pennsylvania, for instance. Sometimes that sort of thing happens in congressional elections too, with one party winning almost all of the toss-up races. Nonetheless, about three-quarters of the uncertainty in the House and Senate forecasts comes from local, race-by-race factors. If the presidential model is laser-focused on how the polls are changing from day to day and what they say about the Electoral College, the goal of the congressional models is to shine some light into the darker corners of the electoral landscape.
As compared with the presidential model, the House model is less polling-centric. (This is less the case for the Senate model.) Both the House and Senate models use a broad variety of indicators in addition to polling. With 435 separate House races every other year — and typically 33 to 35 Senate races — it’s possible to make robust empirical assessments of which factors really predict congressional races well and which ones don’t. Nonetheless, our models default toward using polling once there’s a lot of high-quality polling in a particular state or district. In the Senate, most races have abundant polling. But this is less true in the House, where districts are polled sporadically and polling can be an adventure because of small sample sizes and the demographic peculiarities of each district.

Three versions of the models: Lite, Classic, Deluxe

In 2016, we published what we described as two different election models: “polls-only” and “polls-plus.”2 But now we’re running what we think of as three different versions of the same model, which we call Lite, Classic and Deluxe. I realize that’s a subtle distinction — different models versus different versions of the same model.

But the Lite, Classic and Deluxe versions of our models somewhat literally build on top of one another, like different layers of toppings on an increasingly fancy burger. I’ll describe these methods in more detail in the sections below. First, a high-level overview of what the different versions account for.

The layers in FiveThirtyEight’s House forecast





Which versions use it?



Layer
Description
Lite
Classic
Deluxe




1a
Polling
District-by-district polling, adjusted for house effects and other factors.
✓
✓
✓


1b
CANTOR
A system which infers results for districts with little or no polling from comparable districts that do have polling.
✓
✓
✓


2
Fundamentals
Non-polling factors such as fundraising and past election results that historically help in predicting congressional races.

✓
✓


3
Expert forecasts
Ratings of each race published by the Cook Political Report, Inside Elections and Sabato’s Crystal Ball


✓

Lite is as close as you get to a “polls-only” version of our forecast — except, the problem is that a lot of congressional districts have little or no polling. So we use a system we created called CANTOR3 to fill in the blanks. It uses polls from states and districts that have polling, as well as national generic congressional ballot polls, to infer what the polls would say in those that don’t or that have only minimal polling. The Lite forecast phases out CANTOR and becomes truly “polls-only” in districts that have a sufficient amount of polling.

The Classic version also uses local polls4 but layers a bunch of non-polling factors on top of it, the most important of which are incumbency, past voting history in the state or district, fundraising and the generic ballot. These are the “fundamentals.” The more polling in a race, the more heavily Classic relies on the polls as opposed to the “fundamentals.” Although Lite isn’t quite as simple as it sounds, the Classic model is definitely toward the complex side of the spectrum. With that said, it should theoretically increase accuracy. In the training data,5 Classic miscalled 3.3 percent of House races, compared with 3.8 percent for Lite.6 You should think of Classic as the preferred or default version of FiveThirtyEight’s forecast unless we otherwise specify.

Finally, there’s the Deluxe flavor of the model, which takes everything in Classic and sprinkles in one more key ingredient: expert ratings. Specifically, Deluxe uses the race ratings from the Cook Political Report, Nathan Gonzales’s Inside Elections and Sabato’s Crystal Ball, all of which have published forecasts for many years and have an impressive track record of accuracy.7

Within-sample accuracy of forecasting methods

Share of races called correctly based on House elections from 1998 to 2016

Forecast
100 Days Before Election
Election Day




Lite model (poll-driven)
94.2%
96.2%


Fundamentals alone
95.4
95.7


Classic model (Lite model fundamentals)
95.4
96.7


Expert ratings alone*
94.8
96.6


Deluxe model (Classic model expert ratings)
95.7
96.9

* Based on the average ratings from Cook Political Report, Inside Elections/The Rothenberg Political Report, Sabato’s Crystal Ball and CQ Politics. Where the expert rating averages out to an exact toss-up, the experts are given credit for half a win.

So if we expect the Deluxe forecast to be (slightly) more accurate, why do we consider Classic to be our preferred version, as I described above? Basically, because we think it’s kind of cheating to borrow other people’s forecasts and make them part of our own. Some of the fun of doing this is in seeing how our rigid but rigorous algorithm stacks up against more open-ended but subjective ways of forecasting the races. If our lives depended on calling the maximum number of races correctly, however, we’d go with Deluxe.

Collecting, weighting and adjusting polls

Our House and Senate forecasts use almost all the polls we can find, including partisan polls put out by campaigns or other interested parties. We had not traditionally used partisan polls in our forecasts, but they are a necessary evil for the House, where much of the polling is partisan. Having developed a system we like for handling partisan polls in our House forecasts, we’re also using them for our Senate forecast.

However, as polling has gotten more complex, including attempts to create fake polls, there are a handful of circumstances under which we won’t use a poll:

We don’t use polls if we have significant concerns about their veracity or if the pollster is known to have faked polls before.
We don’t use DIY polls commissioned by nonprofessional hobbyists on online platforms such as Google Surveys. (This is a change in policy since 2016. Professional or campaign polls using these platforms are still fine.)
We don’t treat subsamples of multistate polls as individual “polls” unless certain conditions are met.8
We don’t use “polls” that blend or smooth their data using methods such as MRP. These can be perfectly fine techniques — but if you implement them, you’re really running a model rather than a poll. We want to do the blending and smoothing ourselves rather than inputting other people’s models into ours.

Polls are weighted based on their sample size, their recency and their pollster rating (which in turn is based on the past accuracy of the pollster, as well as its methodology). These weights are determined by algorithm; we aren’t sticking our fingers in the wind and rating polls on a case-by-case basis. Also, the algorithm emphasizes the diversity of polls more than it has in the past; in any particular race, it will insist on constructing an average of polls from at least two or three distinct polling firms even if some of the polls are less recent.

There are also three types of adjustments to each poll:

First, a likely voter adjustment takes the results of polls of registered voters or all adults and attempts to translate them to a likely-voter basis. Traditionally, Republican candidates gain ground in likely voter polls relative to registered voter ones, but the gains are smaller in midterms with a Republican president. The likely voter adjustment is dynamic; it starts with a prior that likely voter polls slightly help Republicans, but this prior is updated as pollsters publish polls that directly compare likely and registered voter results. (If you’re a pollster, please follow Monmouth University’s lead and do this!).
Second, a timeline adjustment adjusts for the timing of the poll, based on changes in the generic congressional ballot. For instance, if Democrats have gained a net of 5 percentage points on the generic ballot since a certain district was polled, the model will adjust the poll upward toward the Democratic candidate (but not by the full 5 points; instead, by roughly two-thirds that amount — around 3.5 points — depending on the elasticity score9 of the district). As compared with the timeline adjustment in our presidential model, which can be notoriously aggressive, the one in the House and Senate models is pretty conservative.
A house effects adjustment corrects for persistent statistical bias from a pollster. For instance, if a polling firm consistently shows results that are 2 points more favorable for Democrats than other polls of the same district, the adjustment will shift the poll part of the way back toward Republicans.10

The House and Senate model use partisan and campaign polls, which typically make up something like half of the overall sample of congressional district polling.11 Partisanship is determined by who sponsors the poll, rather than who conducts it. Polls are considered partisan if they’re conducted on behalf of a candidate, party, campaign committee, or PAC, super PAC, 501(c)(4), 501(c)(5) or 501(c)(6) organization that conducts a large majority of its political activity on behalf of one political party.

Partisan polls are subject to somewhat different treatment than nonpartisan polls in the model. They receive a lower weight, as partisan-sponsored polls are historically less accurate. And the house effects adjustment starts out with a prior that assumes these polls are biased by about 4 percentage points toward their preferred candidate or party. If a pollster publishing ostensibly partisan polls consistently has results that are similar to nonpartisan polls of the same districts, the prior will eventually be overridden.

CANTOR: Analysis of polls in similar states and districts

CANTOR is essentially PECOTA or CARMELO (the baseball and basketball player forecasting systems we designed) for congressional districts. It uses a k-nearest neighbors algorithm to identify similar congressional districts and states based on a variety of demographic,12 geographic13 and political14 factors. For instance, the district where I was born, Michigan 8, is most comparable to other upper-middle-income Midwestern districts such as Ohio 12, Indiana 5 and Minnesota 2 that similarly contain a sprawling mix of suburbs, exurbs and small towns. Districts can be compared to states,15 so data from House races informs the CANTOR forecasts for Senate races, and vice versa.

The goal of CANTOR is to impute what polling would say in unpolled or lightly polled states and districts, given what it says in similar states and districts. It attempts to accomplish this goal in two stages. First, it comes up with an initial guesstimate of what the polls would say based solely on FiveThirtyEight’s partisan lean metric (FiveThirtyEight’s version of a partisan voting index, which is compiled based on voting for president and state legislature) and incumbency. For instance, if Republican incumbents are polling poorly in the districts where we have polling, it will assume that Republican incumbents in unpolled districts are vulnerable as well. Then, it adjusts the initial estimate based on the district-by-district similarity scores.

All of this sounds pretty cool, but there’s one big drawback. Namely, there’s a lot of selection bias in which races are polled. A House district usually gets surveyed only if one of the campaigns or a media organization has reason to think the race is close — so unpolled districts are less competitive than you’d infer from demographically similar districts that do have polls. CANTOR projections are adjusted to account for this.

Overall, CANTOR is an interesting method that heavily leans into district polling and gets as close as possible to a “polls-only” view of the race. However, in terms of accuracy, it is generally inferior to using …

The fundamentals

The data-rich environment in House and Senate elections — 435 individual House races every other year, compared with just one race every four years for the presidency — is most beneficial when it comes to identifying reliable non-polling factors for forecasting races. There’s enough data, in fact, that rather than using all districts to determine which factors were most predictive, I instead focused the analysis on competitive races (using a fairly broad definition of “competitive”). In competitive House districts with incumbents, the following factors have historically best predicted election results, in roughly declining order of importance:

The incumbent’s margin of victory in his or her previous election, adjusted for the national political environment and whom the candidate was running against in the prior election.
The generic congressional ballot.
Fundraising, based on the share of individual contributions for the incumbent and the challenger as of the most recent filing period.16
FiveThirtyEight partisan lean, which is based on how a district voted in the past two presidential elections and (in a new twist) state legislative elections. In our partisan lean formula, 50 percent of the weight is given to the 2016 presidential election, 25 percent to the 2012 presidential election and 25 percent to state legislative elections.
Congressional approval ratings, which are a measure of the overall attitude toward incumbents.17
Whether either the incumbent or the challenger was involved in a scandal.18
The incumbent’s roll call voting record — specifically, how often the incumbent voted with his or her party in the past three Congresses. “Maverick-y” incumbents who break party ranks more often outperform those who don’t.
Finally, the political experience level of the challenger, based on whether the challenger has held elected office before.

In addition, in Pennsylvania, which underwent redistricting in 2018, the model accounts for the degree of population overlap between the incumbent’s old and new district. And in California and Washington state, it accounts for the results of those states’ top-two primaries.

The Senate model uses almost all the same factors for incumbents, but there are some subtle differences given that senators face election once every six years instead of once every other year. For instance, previous victory margin is less reliable in Senate races; that’s because a longer time has passed since the previous election was held. In addition, the Senate model uses more sophisticated data in calculating the effects of incumbency. Candidates in smaller,19 more demographically distinct states20 tend to have larger incumbency advantages. The Senate model also accounts for changes in the state’s partisan orientation since the seat was last contested. Finally, the Senate model uses a more advanced method of calculating candidate experience.21

In open-seat races, the model uses the factors from the list above that aren’t dependant on incumbency, namely the generic ballot, fundraising, FiveThirtyEight partisan lean, scandals, experience and (where applicable) top-two primary results. It also uses the results of the previous congressional election in the state or district, but this is a much less reliable indicator than when an incumbent is running for re-election.

But wait — there’s more! In addition to combining polls and fundamentals, the Classic and Deluxe models compare their current estimate of the national political climate to a prior based on the results of congressional elections since 1946, accounting for historic swings in midterms years and presidential approval ratings. The prior is designed in such a way that it phases it out completely by Election Day.

Incorporating expert ratings

Compared with the other steps, incorporating expert ratings and creating the Deluxe version of the model is fairly straightforward. We have a comprehensive database of ratings from Cook and other groups in House races since 1998 and in Senate races since 1990, so we can look up how a given rating corresponded, on average, with a margin of victory. For instance, going into the 2018 midterms, candidates who were deemed to be “likely” winners in their House races won by an average of about 12 points:

What do ratings like “lean Republican” really mean?




Expert Rating
Average margin of victory




Toss-up
0 points


“Tilts” toward candidate
4 points


“Leans” toward candidate
7 points


“Likely” for candidate
12 points


“Solid” or “safe” for candidate
34 points

Based on House races since 1998.

But, of course, there are complications. One is that there’s an awful lot of territory covered by the “solid” and “safe” categories: everything from races that could almost be considered competitive to others where the incumbent wins every year by a 70-point margin. Therefore, the Deluxe forecast doesn’t adjust its projections much when it encounters “solid” or “safe” ratings from the experts, except in cases where the rating comes as a surprise because other factors indicate that the race should be competitive.

Also, although the expert raters are really quite outstanding at identifying “micro” conditions on the ground, including factors that might otherwise be hard to measure, they tend to be lagging indicators of the macro political environment. Several of the expert raters shifted their projections sharply toward the Democrats in early 2018, for instance, even though the generic ballot was fairly steady over that period. Thus, the Deluxe forecast tries to blend the relative order of races implied by the expert ratings with the Classic model’s data-driven estimate of national political conditions. Deluxe and Classic will usually produce relatively similar forecasts of the overall number of seats gained or lost by a party, therefore, even though they may have sharp disagreements on individual races.

Simulating the election and accounting for uncertainty

Sometimes what seem like incredibly pedantic questions can turn out to be important. For years, we’ve tried to design models that account for the complicated, correlated structure of error and uncertainty in election forecasting. Specifically, that if a candidate or a party overperforms the polls in one swing state, they’re also likely to do so in other states, especially if those states are demographically similar. Understanding this principle was key to understanding why Clinton’s lead wasn’t nearly as safe as it seemed in 2016.

Fortunately, this is less of a problem in constructing a congressional forecast; there are different candidates on the ballot in every state and district, instead of just one presidential race, and the model relies on a variety of inputs, instead of depending so heavily on polls. Nonetheless, the model accounts for four potential types of error in an attempt to self-diagnose the various ways in which it could go off the rails:

First, there’s local error — that is, error pertaining to individual states and districts. Forecasts are more error-prone in districts where there’s less polling or in districts where the various indicators disagree with one another. Some states and districts are also “swingier” (or more elastic) than others; conditions tend to change fairly quickly in New Hampshire, for instance, but more slowly in the South, where electorates are often bifurcated between very liberal and very conservative voters.
Second, there’s error based on regional or demographic characteristics. For instance, it’s possible that Democrats will systematically underperform expectations in areas with large numbers of Hispanic voters or overperform them in the rural Midwest. The model uses CANTOR similarity scores to simulate these possibilities.
Third, there can be error driven by incumbency status. In some past elections, polls have systematically underestimated Republican incumbents, for example, even if they were fairly accurate in open-seat races. The models account for this possibility as well.
Fourth and finally, the model accounts for the possibility of a uniform national swing — i.e., when the polls are systematically off in one party’s direction in almost every race.

Error becomes smaller as Election Day approaches. In particular, there’s less possibility of a sharp national swing as you get nearer to the election because there’s less time for news events to intervene.

Nonetheless, you shouldn’t expect pinpoint precision in a House forecast, and models that purport to provide it are either fooling you or fooling themselves. Even if you knew exactly what national conditions were, there would still be a lot of uncertainty based on how individual races play out. And individual Senate races are, at best, only slightly more predictable, as they can be highly candidate-driven.

Odds and ends

OK, that’s almost everything. Just a few final notes:

Our models also project the turnout in each race, based on factors such as the eligible voter population and turnout in past midterms and presidential races. Competitive races tend to produce higher turnout than noncompetitive ones. Projecting turnout is important in understanding the relationship between the national popular vote and the number of seats that each party might gain or lose.
I should emphasize that we do not make ad-hoc adjustments to the forecasts in individual races. They’re all done strictly by algorithm.

Editor’s note: This article is adapted from a previous article about how our election forecasts work.

Model Creator

Nate Silver FiveThirtyEight’s founder and editor in chief. | @NateSilver538

Version History

1.3 Added Senate forecast.Sept. 12, 2018

1.2 Added individual race pages.Aug. 26, 2018

1.1 Improvements to account for top-two primary results and to better handle redistricting in Pennsylvania.Aug. 20, 2018

1.0 House model launched.Aug. 16, 2018

Related Articles
2018 House forecast
2018 Senate forecast

2018 Senate Forecast

ELECTION 2018

FiveThirtyEight

House forecast

Latest polls

Generic ballot

Trump approval

Midterms coverage

Search for a race or candidate

Search

How do you like your Senate forecast?

Lite

Keep it simple, please — give me the best forecast you can based on what local and national polls say

Classic

I’ll take the polls, plus all the “fundamentals”: fundraising, past voting in the state, historical trends and more

Deluxe

Gimme the works — the Classic forecasts plus experts’ ratings

Forecasting the race for the Senate

Updated Aug. 29, 2018, at 6:24 PM

1 in 2

Chance Democrats win control (49.7%)

1 in 2

Chance Republicans keep control (50.3%)

AVERAGE

MEDIAN

CURRENTBREAKDOWN

CURRENTBREAKDOWN

MAJORITY

MAJORITY

Breakdown of seats byparty

↑

Higher

probability

55 D45 R

54 D46 R

53 D47 R

52 D48 R

51 D49 R

50 R50 D

51 R49 D

52 R48 D

53 R47 D

54 R46 D

80% chance outcome inthis range

80% chance outcome inthis range

10% chance TK gain TKthan -2 seats

10% chance TK gain TKthan -2 seats

10% chance TK gain TKthan 4 seats

10% chance TK gain TKthan 4 seats

+4

+-2 Democratic seats

AVG. GAIN

+-2

In the event of a split Senate, the vice president determines the majority.

Our forecast for every Senate seat

The chance of each candidate winning in all 35 Senate seats up for election in 2018, as well as the party controlling the 65 seats not on the ballot this cycle

Not up for election

Up for election

Party flip

>50% non-incumbent party

The balance of power

23 Dem. seats notup for election

MAJORITY

35 seats up for election

42 Rep. seats not up for election

10

Solid R

>95[b]%

2

Likely R

>75%

1

Lean R

>65[c]%

0

Toss-up

65[e]%

1

Likely D

>75%

21

Solid D

>95%

Every seat on the map[a]

AK

CA

CO

ID

IL

IA

KY

LA

MD

MN

MO

NY

OR

TN

TX

VA

WI

AL

AZ

AR

IN

KS

ME

CT

DE

GA

HI

SC

SD

MA

MI

MS

NE

NV

NH

NJ

NM

NC

ND

RI

OH

OK

PA

FL

MT

UT

VT

WA

WV

WY

Our latest coverage

5 HOURS AGO

Just About Everyone In Florida Has Already Decided Who To Vote For

By Dhrumil Mehta and Janie Velencia

21 HOURS AGO

Election Update: The Most (And Least) Elastic States And Districts

By Nathaniel Rakich and Nate Silver

1 DAY AGO

Here’s A New, Less Volatile Version Of Our Generic Ballot Tracker

By Nate Silver

How the forecast has changed

We’ll be updating our forecasts every time new data is available, every day through Nov. 6

Chance of controlling the Senate

1 in 10

1 in 4

1 in 2

3 in 4

9 in 10

49.7%

49.7%

50.3%

50.3%

NOV. 6

ELECTION DAY

NOV. 8, 2016

Seats controlled by each party

60-40

55-45

EVEN

55-45

60-40

FORECAST LAUNCH

47-53

47-53

KEY

AVERAGE

80% CHANCE OF FALLING IN RANGE

How the Senate has swung historically

Our forecasted seat breakdown in the Senate for 2018 and the change in the breakdown for every Senate election since 1924

Net advantage

Swing

0

20

40

60

80 seats

0

20

40

60

80

1926

1930

1934

1938

1942

1946

1950

1954

1958

1962

1966

1970

1974

1978

1982

1986

1990

1994

1998

2002

2006

2010

2014

2018

◄ More Democrats | More Republicans ►

COOLIDGE

COOLIDGE

HOOVER

HOOVER

ROOSEVELT

ROOSEVELT

TRUMAN

TRUMAN

EISENHOWER

EISENHOWER

KENNEDY

KENNEDY

JOHNSON

JOHNSON

NIXON

NIXON

FORD

FORD

CARTER

CARTER

REAGAN

REAGAN

H.W. BUSH

H.W. BUSH

CLINTON

CLINTON

W. BUSH

W. BUSH

OBAMA

OBAMA

TRUMP

TRUMP

Average

Average

How this forecast works

Nate Silver explains the methodology behind our 2018 midterms forecast. Read more …[f]

The third-party candidates listed represent our best approximation of who will appear on each state’s general election ballot. The candidates listed will update as each race is finalized; some listed candidates may not ultimately qualify for the general election.

This analysis treats currently vacant seats as being held by the party that previously controlled them.

[g]

Forecast models by Nate Silver. Design and development by Jay Boice, Aaron Bycoffe, Rachael Dottle, Ritchie King, Ella Koeze, Andrei Scheinkman, Gus Wezerek and Julia Wolfe. Research by Andrea Jones-Rooy, Dhrumil Mehta, Mai Nguyen and Nathaniel Rakich. Notice any bugs? Send us an email.

RELATED STORIES

We Gave You 3 Million Russian Troll Tweets. Here’s What You’ve Found So Far.

We Gave You 3 Million Russian Troll Tweets. Here’s What You’ve Found So Far.

Why We’re Sharing 3 Million Russian Troll Tweets

Why We’re Sharing 3 Million Russian Troll Tweets

The 500-Page Inspector General’s Report In 900 Words

The 500-Page Inspector General’s Report In 900 Words

The Polls Are All Right

The Polls Are All Right

Why Obama vs. Trump Is Such A Blowout

Why Obama vs. Trump Is Such A Blowout

Get more FiveThirtyEight

Newsletter

Videos

Podcasts

Twitter

Facebook

Data

RSS

Contact

Jobs

Masthead

Privacy and Terms of Service

About Nielsen Measurement