The numbers game: behind recent positive GOP polls, a more complicated picture

Rep. Jim GerlachJames (Jim) GerlachFormer reps: Increase support to Ukraine to deter Russia With Trump and GOP Congress, job creators can go on offense Big names free to lobby in 2016 MORE (R-Pa.) is leading his Democrat opponent Lois Murphy by 11 percentage points and has a 50 percent favorability rating, while 60 percent of voters viewed Murphy as a liberal, according to a poll released last week and conducted by the Republican polling firm Public Opinion Strategies.

But in late July, Murphy released a poll conducted by Garin Hart Yang Research, a Democratic firm, showing the race tied and that 17 percent of the voters had not made up their minds on whom to support. Murphy’s poll showed that Gerlach’s unfavorable rating was 46 percent and that 66 percent believed the country was headed in the wrong direction.

The positive results for GOP candidates in the more recent poll appears to defy conventional wisdom that this is the Democrats’ year to win a majority in the House, and that the race between Gerlach and Murphy should be one of the closest in the country. In 2004, Gerlach defeated Murphy by just 6,371 votes and Sen. John KerryJohn Forbes KerryTrump's dangerous Guantánamo fixation will fuel fire for terrorists Tech beefs up lobbying amid Russia scrutiny Overnight Tech: Senate Dems want FCC chief recused from Sinclair merger | Tech rallies on Capitol Hill for DACA | Facebook beefs up lobbying ranks MORE (D-Mass.) defeated President Bush by 51 percent to 48 percent.

To be sure, Democrats have their own polls showing more positive trends for them. A recent poll conducted in late August by RT Strategies and Constituent Dynamics showed Murphy leading Gerlach 50 to 45 percent, and Democrats noted that the National Republican Congressional Committee (NRCC) did not release the results of a poll conducted in July by the Tarrance Group, a GOP firm.

But if polls are useful in providing snapshots of a moving picture, what is really going on in Pennsylvania’s 6th District and other congressional districts with hotly contested races? Even if the more recent polls were internal, the GOP firms that conducted them are well respected. So if there is a Democratic wave this year, why are these polls showing once-troubled incumbents with commanding leads?

The answer is complicated. Several pollsters and political strategists said that polls cannot forecast who will win or lose on Election Day; polls are useful primarily for showing how a race looks at a particular point in time. These same operatives say candidates should not look at the head-to-head matchups and panic if they are trailing.

“Polls are good to show trends, but it is very hard to predict outcomes given [that some races are] so close,” said David Axelrod, a Chicago-based media consultant whose clients include Sen. Barack ObamaBarack Hussein ObamaReport: FCC chair to push for complete repeal of net neutrality Right way and wrong way Keystone XL pipeline clears major hurdle despite recent leak MORE (D-Ill.).

“The horse race is always interesting and clients always want to see it. But it’s not helpful in developing a strategy,” said Craig Smith, a Democratic strategist with Penn, Schoen and Berland.

A weakness in some polls is the placement of the head-to-head horserace question. Political strategists and academics said that the “uninformed ballot” question should come early in a poll. Then, once that number is determined, a pollster can ask questions about the incumbent and challenger’s messages to see how a voter would respond. 

Looking past the horserace number, strategists said that candidates should look first at name identification. If a small percentage of the electorate knows the challenger, then there is an opportunity to define his or her candidacy. But the incumbent also has the chance to define the challenger.

Next, a candidate’s favorable and unfavorable ratings are better predictors of where a race stands and its outcome than the simple matchup.

“I’d look at the unfavorable/favorable ratio,” said a GOP pollster. “If these guys maintain a positive ratio, they’re fine. If it’s inverted, they’re toast.”

For that reason, many pollsters and campaigns only will release the horse-race number. Rep. Curt Weldon (R-Pa.) released a poll Monday conducted by Public Opinion Strategies showing him beating Ret. Admiral Joe Sestak by a 52 percent to 33 percent margin. The campaign would not release any other data from the poll.

But that’s not surprising because campaigns don’t like to broadcast their strategies in public.

“I poll and poll extensively … and I want to keep [the results] to ourselves,” NRCC chairman Tom Reynolds (R-N.Y.) told reporters yesterday.

Carter Eskew, a Democratic strategist with the Glover Park Group, said a candidate’s favorable and unfavorable ratings, and answers to questions on which party is better on national security, for example, could help candidates assess where a race stands.

“I would be encouraged by whether or not [a Democratic candidate’s] numbers had stayed the same, assuming they were good to start, after the tremendous push by President Bush,” Eskew said, referring to Bush’s efforts to make the election about terrorism.

“How well do you stand up to a punch? If I were a Democrat, I would be worried whether the dynamics of the race were changing,” Eskew said. “If I saw no impact from that assault or that my position had grown stronger, then maybe this year is different” from 2002 and 2004.

The biggest factor adding to uncertainty about a race’s outcome is the number of undecided voters. Undecided typically break the challenger’s way, except in 2004, when they voted for Bush.

So has the Gerlach-Murphy race really changed that much in two months? 

“The campaigns are in a different place because between July and September a campaign has been going on,” said the GOP pollster. “It’s not just spending, but message too.”

But Murphy’s spokeswoman dismissed the poll, saying, “I can only assume that in this case ‘NRCC’ stands for ‘Not Really Credible Committee’ because these numbers are totally unreliable.”

Several elements add credibility to polls, said Doug Lonnstrom, the director of the Siena Research Institute. First, better surveys are taken over the course of several days. Second, larger sample sizes reduce the margin of error. Polls with sample sizes of 400, 600 and 1000 interviews have margins of error of five, four and three percent, respectively. In the best polls, the demographics and voter affiliation of those polled match the district’s voter registration pattern. 

There are other more subtle factors that pollsters need to keep in mind.

“You need to look at the historical nature of the district,” Axelrod said. “If the district has never elected a Democrat, but the race is even, then that Republican would be nervous. This is not where you’re supposed to be two weeks after Labor Day.”