★ ★ ★ ★ ★ ★ ★ ★ ★ ★ ★ ★

The battleforcongress

The battle for congress

On November 4, 2014, Americans will go to the polls for this year’s midterm elections. Here’s Vox’s roundup of the most recent predictions from leading election forecasters, and a closer look at the key races that could determine control of Congress.

Senate Chance of party control
Senate Most likely outcome
House Most likely outcome

The main event in the midterms this year is the Republican effort to win the Senate. The map hugely favors them, since Democrats are defending 21 seats — 6 of them in deep-red states — while Republicans are only defending 15. This page shows the latest projections from forecasters of the most likely outcome. As you can see, they don’t vary very much — each party is projected to win about half the seats. But an outcome of 51 Republicans and 49 Democrats will tip the Senate to the GOP, while a 50-50 result will keep the chamber in the Democrats’ hands (because of Vice President Joe Biden’s tie-breaking vote). That’s why forecasting this year’s Senate elections is “a game of inches,” as Eric McGhee, co-creator of the Washington Post’s model, put it to Vox.

Vox's guide to the different models and how they work

Some of the election models we're tracking — Princeton Election Consortium (created by Sam Wang), HuffPostPollster (model created by Simon Jackman, updated and run by Mark Blumenthal and Natalie Jackson), and Daily Kos Poll Explorer (created by Drew Linzer) — rely purely on polls of individual Senate races to make their forecasts. Others — the Upshot (created by Amanda Cox and Josh Katz), FiveThirtyEight (created by Nate Silver), and the The Washington Post (created by John Sides, Ben Highton, and Eric McGhee)— balance out their poll forecasts with a set of "fundamental" factors. These include:

  • The Upshot: Polls of each race, the state’s 2012 presidential result, the political experience of each candidate, candidates’ fundraising, whether it’s a midterm year, the incumbent’s approval rating, and national polls on the public mood.
  • The Washington Post: The state’s 2012 presidential result, the political experience of each candidate, candidates’ fundraising, whether it’s a midterm year, presidential approval nationally, change in GDP, the state’s 2006 Senate result, and a bonus for incumbents.
  • FiveThirtyEight: Silver uses the generic ballot, Congressional approval ratings, fundraising numbers, political experience of candidates, margin of victory, outcome of most recent Senate elections, candidate ideology, and state partisanship.

These fundamentals, though, have already grown much less important, because every forecaster gives polls more and more weight as the election gets closer. For instance, the Post's model is already "mainly polls," Eric McGhee told Vox.

Right now, another set of differences in the models is also significant — they vary in how they estimate and account for uncertainty in what the polls currently show. Right now, when candidates have small leads or the polling data seems of low quality, FiveThirtyEight and HuffPostPollster treat these results with some skepticism. In contrast, Princeton and the Washington Post offer more confident forecasts, particularly in certain races. It's also noteworthy that the Princeton model also uses the median of polls, rather than the trendlines that some other models use. For Vox’s longer analysis of these differences, click here.

For comparison, we’ve also included estimates here from two other forecasters. RealClearPolitics calculates a simple average of several of the most recent polls from each pollster in each race. And Alan Abramowitz, an Emory University political scientist, bases his forecast on the “generic ballot” — national polls that ask which party should control Congress — combined with how many seats each party has up in a particular year.

This page shows each forecaster’s estimate of the likelihood of a GOP Senate takeover. The differences shown here shouldn’t be overstated — in a probabilistic forecast, 60-40 looks a lot like an ordinary coin flip, as Amanda Cox of the Upshot explained at a recent panel. Most of the forecasters are divided over whether to give a very slight edge to the Democrats or Republicans. Still, it’s worth understanding why the models are coming up with these small differences.

Vox's guide to the different models and how they work

Some of the election models we're tracking — Princeton Election Consortium (created by Sam Wang), HuffPostPollster (model created by Simon Jackman, updated and run by Mark Blumenthal and Natalie Jackson), and Daily Kos Poll Explorer (created by Drew Linzer) — rely purely on polls of individual Senate races to make their forecasts. Others — the Upshot (created by Amanda Cox and Josh Katz), FiveThirtyEight (created by Nate Silver), and the The Washington Post (created by John Sides, Ben Highton, and Eric McGhee)— balance out their poll forecasts with a set of "fundamental" factors. These include:

  • The Upshot: Polls of each race, the state’s 2012 presidential result, the political experience of each candidate, candidates’ fundraising, whether it’s a midterm year, the incumbent’s approval rating, and national polls on the public mood.
  • The Washington Post: The state’s 2012 presidential result, the political experience of each candidate, candidates’ fundraising, whether it’s a midterm year, presidential approval nationally, change in GDP, the state’s 2006 Senate result, and a bonus for incumbents.
  • FiveThirtyEight: Silver uses the generic ballot, Congressional approval ratings, fundraising numbers, political experience of candidates, margin of victory, outcome of most recent Senate elections, candidate ideology, and state partisanship.

These fundamentals, though, have already grown much less important, because every forecaster gives polls more and more weight as the election gets closer. For instance, the Post's model is already "mainly polls," Eric McGhee told Vox.

Right now, another set of differences in the models is also significant — they vary in how they estimate and account for uncertainty in what the polls currently show. Right now, when candidates have small leads or the polling data seems of low quality, FiveThirtyEight and HuffPostPollster treat these results with some skepticism. In contrast, Princeton and the Washington Post offer more confident forecasts, particularly in certain races. It's also noteworthy that the Princeton model also uses the median of polls, rather than the trendlines that some other models use. For Vox’s longer analysis of these differences, click here.

For comparison, we’ve also included estimates here from two other forecasters. RealClearPolitics calculates a simple average of several of the most recent polls from each pollster in each race. And Alan Abramowitz, an Emory University political scientist, bases his forecast on the “generic ballot” — national polls that ask which party should control Congress — combined with how many seats each party has up in a particular year.

This year, the House of Representatives is overwhelmingly likely to remain in the GOP’s hands. Democrats would need a net gain of 17 seats to regain control. Unfortunately for them, they face several built-in disadvantages because of the natural geography of House districts, the difficulty in dislodging incumbents, and pro-GOP gerrymandering. To retake the chamber, they’d probably need a wave election — one that’s unlikely to happen with an unpopular Democratic president in office.

Vox's guide to the different models and how they work

There are fewer forecasters making specific House predictions, because there are many fewer polls of individual House races — but we’re currently tracking three of them. Washington Post is using a model that combines polls with “fundamental” factors based on the electoral environment this year. RealClearPolitics and the Cook Political Report use a qualitative ranking judging the competitiveness of individual races. (For each, we’ve allotted races rated as toss-ups equally between the parties.) But the GOP is all but certain to hold on to the House this year. Democrats face three built-in disadvantages — geography, incumbency, and partisan gerrymandering, — that will make it very difficult for them to retake the House this decade without a wave election, as Vox explained here.

  • GOP
  • DEM
  • FiveThirtyEight
    52 48
  • Upshot
    52 48
  • Washington Post
    53 47
  • Princeton Election Consortium
    52 48
  • GOP
  • DEM
  • HuffPost Pollster
    52 48
  • Daily Kos
    53 47
  • RealClearPolitics
    52 48
  • Alan Abramowitz Generic Ballot
    51 49
Forecast average

52 48 seats

  • GOP
  • DEM
  • FiveThirtyEight
    76 24
  • Upshot
    70 30
  • Washington Post
    98 2
  • GOP
  • DEM
  • Princeton Election Consortium
    65 35
  • HuffPost Pollster
    79 21
  • Daily Kos
    90 10
Average chance of GOP takeover

80 percent

  • GOP
  • DEM
  • WaPo
    243 192
  • RealClearPolitics average
    241 194
  • Cook Political Report
    241 194
Forecast average

242 193 seats

Election news

Latest polls make GOP Senate landslide look quite possible

Races to watch

Kansas

Kansas — sen

  • GOP
  • IND
  • 46
    54
  • 47
    53
  • 98
    2
  • GOP
  • IND
  • 36
    64
  • 58
    42
  • 87
    13
Iowa

Iowa — sen

  • GOP
  • DEM
  • 66
    34
  • 70
    30
  • 89
    11
  • GOP
  • DEM
  • 64
    36
  • 61
    39
  • 75
    25
Colorado

Colorado — sen

  • GOP
  • DEM
  • 80
    20
  • 72
    28
  • 98
    2
  • GOP
  • DEM
  • 75
    25
  • 61
    39
  • 79
    21
Georgia

Georgia — sen

  • GOP
  • DEM
  • 60
    40
  • 75
    25
  • 79
    21
  • GOP
  • DEM
  • 83
    17
  • 71
    29
  • 94
    6
North Carolina

North Carolina — sen

  • GOP
  • DEM
  • 27
    73
  • 31
    69
  • 24
    76
  • GOP
  • DEM
  • 36
    64
  • 47
    53
  • 44
    56
Alaska

Alaska — sen

  • GOP
  • DEM
  • 65
    35
  • 74
    26
  • 81
    19
  • GOP
  • DEM
  • 62
    38
  • 68
    32
  • 78
    22
New Hampshire

New Hampshire — sen

  • GOP
  • DEM
  • 25
    75
  • 21
    79
  • 3
    97
  • GOP
  • DEM
  • 25
    75
  • 43
    57
  • 40
    60
Read more on this race
MSNBC: Scott Brown running on ISIS
Arkansas

Arkansas — sen

  • GOP
  • DEM
  • 89
    11
  • 96
    4
  • 99
    1
  • GOP
  • DEM
  • 96
    4
  • 85
    15
  • 98
    2
Louisiana

Louisiana — sen

  • GOP
  • DEM
  • 84
    16
  • 81
    19
  • 98
    2
  • GOP
  • DEM
  • 89
    11
  • 69
    31
  • 86
    14
Kentucky

Kentucky — sen

  • GOP
  • DEM
  • 97
    3
  • 91
    9
  • 99
    1
  • GOP
  • DEM
  • 97
    3
  • 82
    18
  • 98
    2
Michigan

Michigan — sen

  • GOP
  • DEM
  • 4
    96
  • 1
    99
  • 1
    99
  • GOP
  • DEM
  • 0
    100
  • 2
    98
  • 1
    99
South Dakota

South Dakota — sen

  • GOP
  • DEM
  • 99
    1
  • 99
    1
  • 99
    1
  • GOP
  • DEM
  • 100
    0
  • 98
    2
  • 99
    1