2014 Senate Forecasts Combined


Share on: TwitterTw ▪ FacebookFb ▪ Google+G+

The number of organizations forecasting election outcomes has increased rapidly in recent years.

The New York Times presents its forecast alongside five others, which is admirable at a moment when people are learning more about the applications of data and forecasting for politics and beyond.

This table takes the next step: a combination of all six forecasts.

Aug 27
Aug 27
Aug 4
Aug 22
Aug 7
Aug 26
Aug 27
45New HampshireN.H.89.2899190LeanLikelyLikely99
48North CarolinaN.C.58.5594950Toss-upTiltToss-up92
55West VirginiaW.Va.12.012110LeanLeanLikely1

Together these forecasts indicate that control of the Senate is still up for grabs.

In three states, the Democratic candidate’s aggregate chance of winning is between 45% and 55% – these races are incredibly close. With wins in two of these states (plus wins where their chances are currently higher than 55%), Democrats would hold 50 seats in the Senate, and Vice President Biden would break the tie.

To be clear, Republicans maintain a slight edge. Currently their combined chance of winning is greater than 50% in a sufficient number of states to secure a 51 seat majority. Additionally, Republicans have fewer vulnerabilities than Democrats. There are seven states where the aggregate chance of either party winning is currently between 25% and 75%. Of these, Democrats must win five, while Republicans need only win three to gain a majority.

In contrast to the House of Representatives where the Republicans are almost certain to maintain their majority, control of the Senate will probably not be decided until every vote is cast, counted, and possibly recounted. The performance of individual candidates and their campaigns could easily make the difference, so it’s going to be a good fall to volunteer, donate, or at least tune in to the debates.

Six smart teams disagree

Among the competitive races above, the forecasters disagree most about the races in Michigan and North Carolina. The Democrat Gary Peters is generally favored in Michigan, though the Cook Political Report calls the race a toss-up, while the Washington Post Election Lab forecast gives him a 99% chance of winning. Similarly in North Carolina, the New York Times gives Democratic Senator Kay Hagan a 49% chance of winning reelection, while the Washington Post forecast is 92%.

The forecasters don’t disagree about everything – there is consensus that the races in Louisiana and Alaska are very close.

Bias or something else?

While there is disagreement among the six forecasters, this does not appear to be the result of partisan bias. Each organization provides a forecast for all 36 Senate races, so an average score can be calculated for each forecaster. These six averages are all within a 2.6 percentage point range, so bias does not explain the disparities.

Organizations do vary in how they view the competitive landscape generally. The New York Times and the Washington Post place fewer races in the middle “toss-up” percentiles; in contrast, the Cook Political Report rates nearly all the competitive races above as toss-ups.

Each of the six organizations have different approaches to forecasting. The New York Times and the Washington Post use models they’ve developed to relate data and outcomes from past campaigns to the ones taking place now. Cook, Rothenberg, and Sabato have extensive experience and rely more on qualitative factors. FiveThirtyEight has yet to release forecasts for 2014 Senate races based on a formal model or algorithm, but according to Nate Silver, their forecasts are, “based on an assessment of the same major factors that our algorithm uses.”

Of the six organizations, the Washington Post’s forecasts generally differ most significantly from the combined state-by-state forecasts. It will be interesting to evaluate their model against the eventual outcomes, or to see if all the forecasters come into greater alignment as November 4th approaches.

The forecasters

The people and organizations who develop these forecasts are awesome. Developing the models and conducting analysis takes a real investment by an organization. The individuals involved are incredibly bright, and every one of them could gain personally if they did forecasting for a private purpose instead. So, say hi to them sometime:


The combined percentage in the table above represents the Democratic Party Candidate's chance of winning the Senate seat for each state. This percentage is calculated as a simple average of the six forecasts listed. For organizations providing qualitative forecasts, they are coded as follows:

COOK ROTH SABATO Chance Dem wins
(above or below 50)
Safe Solid Currently Safe Safe 49
Likely Likely Favored Likely 40
Lean Lean Lean Leans 25
Tilt Toss-up/Tilt 10
Toss-up Toss Up Pure Toss-up Toss-up 0

Share on: TwitterTw ▪ FacebookFb ▪ Google+G+