Since 1997, FairVote has released projections for U.S. House elections using a model that relies only on previous election results. These projections have been remarkably accurate. Months before the 2014 election, we projected outcomes in 368 U.S. House races, and were correct in 367. Two days after the 2014 election, we made projections for 373 2016 U.S. House races, and expect a similar level of success.
Our 2016 projections reveal:
While the model’s accuracy could likely be improved by incorporating data from public opinion polls, on the candidates, or on their campaigns, the accuracy of our projections in the absence of such information illustrates an important truth about elections under our current system: In the vast majority of cases, the particulars of candidates and campaigns have little impact on the end result. Uncompetitive races mean that outcomes are essentially predetermined, leaving voters without meaningful choices or a compelling reason to go to the polls.
The full projections can be found in the spreadsheets linked above. Users also can also use these sheets to make their own projections, by adjusting the values for the national preference between the major parties (we use a 50-50 projection), and the average advantage earned by incumbents (we give an average "incumbency bump" of 2.77%).