POLITICS
08/29/2014 05:44 pm ET | Updated Aug 29, 2014

HUFFPOLLSTER: Political Scientists Weigh In With Midterm Predictions

David Clapp via Getty Images

Political scientists gathered in DC to argue over just how good the midterms will be for the GOP. We weigh in with our own revamped Senate model. And that's it for us until next Tuesday, when we switch over to a morning newsletter. This is HuffPollster for Friday, August 29, 2014.

POLITICAL SCIENTISTS PREDICT GOP GAINS, BUT HOW BIG? - At the annual conference of the American Political Science Association (APSA) in Washington D.C. on Friday, four political scientists predicted gains for the Republicans in both the U.S. House and Senate. In presenting results of their forecasting models on the outcome of 2014 congressional races, however, they disagreed on the magnitude of Republican gains in the House and the potential for a G.O.P. takeover of the Senate.

-James Campbell of SUNY Buffalo predicted a "mini-wave" election, with the Republicans gaining 16 additional seats in the House for the largest U.S. House majority since 1931, and eight seats and majority status in the Senate. Campbell's model is based on a calculation of "seats in peril" for each party based on the late August assessments of the Cook Political Report

-Alan Abramowitz of Emory University was less bullish on Republican prospects. His model, which depends in part on results from the generic congressional ballot question in national polls, predicts a Republican gain of five to seven seats in the House and four to eight seats in the Senate. Based on recent polls showing a 1 to 2 point Democratic lead on the generic congressional ballot, Abramowitz's model predicts a four-seat Republican gain in the House and a 5.5 seat gain in the Senate.

-Robert Erikson of Columbia University presented a forecast of a 14-seat gain for Republicans in the House, based on a model and paper co-authored with Joe Bafumi of Dartmouth College and Chris Wlezien of the University of Texas. Their model also depends heavily on national polling on U.S. House vote preferences.

-James King of University of Wyoming presented results from a different sort of statistical model that allows for the possibility, as he described it, that elections "are not necessarily a referendum on the president." Some are referenda, others may be dominated by "positive events" that might benefit the president's party. Unfortunately for Democrats, King's model sees 2014 as a referendum election and predicts a massive 39-seat gain for Republicans in the House.

The two political scientists who attempted to predict the outcome of the Senate contests warned about the difficulty in such forecasts. Although Abramowitz noted that "the Senate is where action is this year...where party control very much in question," he warned that his Senate model is not as historically accurate as his U.S. House model, "but it does a reasonably good job.

It’s just more difficult to predict Senate elections because they’re lumpier," Campbell agreed. "They’re very competitive and they can move more quickly during the campaign season than House seats."

Campbell -- who, it should be said, once worked briefly as a Republican staffer on Capitol Hill -- concluded with a friendly jibe at Abramowitz -- who, it should be said, was a donor to President Obama's reelection campaign in 2012. "The best indicator of just how bad 2014 is likely to be for the Democrats," Campbell joked, "is that I’ve received no annoyingly gloating emails [this year] from Alan."

WE'VE UPDATED OUR POLLING MODEL - HuffPollster, with Natalie Jackson: "Until our latest changes, the HuffPost Pollster charts also noted the probability that the leading candidate is really ahead -- a number that describes the model's statistical confidence that the colored bands around each trend line do not overlap. Unfortunately, that statistic has proved to be misleading. Given the way the model combines the samples from multiple polls, relatively small margins can quickly produce a very high level of confidence in a statistically significant lead. However, a small lead in the polling averages today may be fleeting. Voter preferences can change over the course of a campaign and, of course, polls themselves can be wrong. So beginning on Aug. 29, the HuffPost Pollster [Senate] charts instead provide a 'win probability' statistic that aims to take into account the potential for shifts in candidate support and the potential for error or statistical bias in polling estimates of candidate support…We compute the win probability not because we believe polls can provide perfect predictions of the future. They cannot (although on the eve of an election, they should come very close). Rather, since so many political observers treat poll results as a forecast, our aim is to quantify the polls’ lack of precision in foretelling election outcomes, to put a real-world "margin of error" around the current polling snapshot. [HuffPost]

And we've added a new house effect correction - The house effect correction that is an integral part of our polling model already pulled results toward the industry average for all pollsters. But the growth of partisan polls and surveys using nontraditional methodologies has created a far greater potential for bias in that average. We have therefore added an additional house effect adjustment. The goal here is simple: to minimize the effects of partisan polls and pollsters with questionable reliability. So we calibrate the model's trend lines to better match estimates from the group of nonpartisan pollsters who were on average within plus-or-minus one percentage point of the trend lines generated by the HuffPost 2012 model after adjusting the model to the actual election results. The average of the polls conducted by these pollsters is calculated, and then the rest of the model is adjusted as if this average were the true average of all the polls." [HuffPost]

SUPPORT FOR AIR STRIKES IN SYRIA - Emily Swanson: "A majority of Americans think the United States should expand its military campaign against insurgents in Iraq into Syria as well, a new HuffPost/YouGov poll shows. According to the poll, 60 percent of Americans now support airstrikes against insurgents in Syria, while 20 percent are opposed. That level of support approached the 64 percent of Americans in the survey who said they support the current airstrike campaign against Iraq. Fifty-six percent of Democrats, 54 percent of independents and 79 percent of Republicans said they support airstrikes in Syria. Support for intervening in Syria has grown dramatically in the past year. A HuffPost/YouGov poll conducted last September found that only 13 percent of Americans thought the U.S. should use airstrikes, while 62 percent said it should not. " [HuffPost]

MOST AMERICANS BELIEVE EMPLOYERS SHOULD GIVE PAID VACATIONS - Dave Jamison and Emily Swanson: "Americans overwhelmingly support the idea of requiring large U.S. employers to provide their workers with at least some paid vacation time, according to a new HuffPost/YouGov poll. In a poll conducted ahead of Labor Day, 75 percent of respondents said they believe in placing such a mandate upon the business community. A mere 17 percent said they oppose it. The support crossed party lines to include 87 percent of self-identified Democrats and 65 percent of Republicans." [HuffPost]

SURVEYS HIGHLIGHT HISTORICAL OPINIONS OF UNIONS - Kathleen Weldon: "On May 14, 1882, unionized workers in New York City held a parade and picnic, and the seeds of the Labor Day holiday were planted. About 50 years later, pollsters began asking Americans to share their opinions on unions. The results highlight how views of the labor movement have -- and haven't -- changed since the 1930s. Some insights from the Roper Center for Public Opinion Research archives: Gallup has been asking the public about their approval of labor unions since 1936. Support for unions has drifted slowly downward since its early peaks. However majorities (54%) still express approval [for labor unions] in the most recent poll….Over 1950s, 1990s, and early 2000s, the public was asked by Gallup where their sympathies generally fell when they heard about a strike. Over this broad time span, the number inclined to side with the union has stayed fairly consistent, ranging from 40% to 52%. Those who tend to side with the company has ranged between 29% and 37%, while the remainder say neither or both….In 1937, 11% of all adults reported belonging to one. By the early fifties, when the proportion of the employed in unions was reaching its peak, membership was at 18%. Labor membership suffered a long decline over the late 20th century, and by 2014, again one in ten said they were union members." [HuffPost]

HUFFPOLLSTER VIA EMAIL! - You can receive this daily update every weekday via email! Just click here, enter your email address, and and click "sign up." That's all there is to it (and you can unsubscribe anytime).

FRIDAY'S 'OUTLIERS' - Links to the best of news at the intersection of polling, politics and political data:

-Harry Enten and Nate Silver argue that migration isn't turning red states blue. [538]

-Gallup finds the average American employed full time works more than 40 hours each week. [Gallup]

-Robert Morris University finds Americans split on the importance of unions. [RMU]

-A Russian pollster finds little opposition to Putin on the situation in Ukraine. [Pew]

-Political scientist Maurice Cunningham takes a look at whether Martha Coakley's well publicized campaign gaffes really hurt her in the 2009 special election loss to Scott Brown in Massachusetts. [MassPoliticsProfs]

-Americans are Team Reclining On Airplanes. [HuffPost]

Subscribe to the Politics email.
How will Trump’s administration impact you?

CONVERSATIONS