Though the results on Election Night up and down the ticket were ultimately different than the predictions from experts, pundits and analysts, a factual look back shows that the Democratic Congressional Campaign Committee’s (DCCC) internal polling clearly and accurately predicted the rise, fall, and ultimate landing for House Democratic candidates.
{mosads}These in-house polls – conducted using all live callers or a combination of live calls to cell phones and automated calls to landlines – showed the momentum behind many Democratic challengers and their ability to beat House Republicans, but also a notable drop in support in the final week of the campaign, due at least in part to FBI Director James Comey’s letter.
Ultimately, the DCCC’s final internal data sets closely reflected the final results of dozens of House races, demonstrating the accuracy and effectiveness of the polls overall.
Across 38 competitive House races, Republicans ultimately won by an average final margin of 7 points, with Democrats receiving 45 percent of the vote to the Republicans’ 52 percent. By comparison, the DCCC’s final polling in these districts showed Democratic candidates with a similar average deficit of 5 points, 41 percent to 46 percent. The support levels, of course, change once undecided voters picked a side.
Across much of the DCCC’s district-specific polling, the final internal polls accurately predicted the final margin within two points. In many of those districts, data also showed a clear decline thanks to the impact of Comey’s letter.
In Illinois’ 10th District, the DCCC poll had Brad Schneider up by 5 points on Nov. 3. The final margin was 5 points. The Oct. 22 to 23 poll had Schneider up by 11 points.
In New Jersey’s 5th District, our poll had Josh Gottheimer up by 4 points on Nov. 3. The final margin was 4 points. The Oct. 10 poll had Gottheimer up by 6 points.
In Virginia’s 10th District, the Nov. 1 poll showed LuAnn Bennett down by 5 points. The final margin was 6 points. The Oct. 18 to 19 poll showed Bennett and Republican contender Barbara Comstock tied.
In Nebraska’s 2nd District, our Nov. 1 poll showed Brad Ashford and Don Bacon tied. The final margin was 2 points. The Oct. 6 to 7 poll showed Ashford up by 3 points.
In California’s 25th District, the Nov. 1 poll showed Bryan Caforio trailing by 6 points. The final margin was 7 points. The Oct. 26 poll showed him trailing Steve Knight by only 2 points.
In California’s 49th District, our Nov. 1 poll showed Doug Applegate trailing by 2 points. The final margin was 2 points.
In Colorado’s 6th District, the Nov. 2 poll had Morgan Carroll down by 8 points. The final margin was 8 points. The Oct. 29 poll showed the race tied.
In Nevada’s 4th District, our Oct. 29 poll showed Ruben Kihuen leading Cresent Hardy by 5 points. The final margin was 4 points.
These examples clearly demonstrate the ability of the DCCC’s internal polling to accurately capture the state of the race. Comparing the final election results to polls released weeks and months prior to Election Day is as unhelpful as it is inaccurate.
Polls provide a snapshot in time and are used to inform strategic decisions, capture trend lines, and provide insight into the dynamics of a race. To suggest we cook up “junk science” just to release it publicly and drive the media narrative is not only an insult to the profession, it’s wildly untrue.
In total, the DCCC conducted 226 general election in-house polls in the 2016 cycle, and only thirteen of them were released to the press. Any external review of the DCCC’s in-house polling program is only considering 6 percent of the sample, which any pollster would realize is not the full picture.
Further, the DCCC has a Data and Analytics team of 13 experts, with decades of experience in the field of polling and analytics. Every in-house poll was either done with all live callers or a combination of live and automated callers, and all contained a minimum of 100 cell phone interviews, following best practices within the field. These in-house polls never utilized mix-mode online panels.
We also use voter turnout probability scores as the basis for sampling and weighting, with voters’ demographics and past turnout history – in all recent elections, not just 2012 – to identify the likely composition of voters in each district. In our targeted districts, we projected turnout to be 68.2 percent of registered voters; thus far, turnout in these districts has been 66.5 percent, with ballots still being counted.
The DCCC polling team is proud to have conducted accurate polling and provided critical data sets to inform the important work being done internally at the DCCC.
Ultimately, thanks in part to the information provided by our polling team, House Democrats outperformed the national environment and managed to gain a net of six seats despite strong headwinds against us.
Amanda Hoey is the Director of Polling and Modeling for the Democratic Congressional Campaign Committee. She previously worked at The Feldman Group, a Democratic polling firm, and is a graduate of the University of Connecticut.