Nonprofit, nonpartisan journalism. Supported by readers.

Donate

Why you should mostly ignore internal polls (but can learn a little bit from them)

“Challenger is now within just a few points of Incumbent,” the article reads. This is big news! Just a few months ago, nobody gave a second thought to Challenger’s upstart campaign; Incumbent had held office for so many years, surely he would cruise to re-election. Yet, now it appears Challenger is within striking distance of a major upset.

“…according to an internal poll released by Challenger’s campaign,” the story continues.

Cue sad trombone.

As a savvy news consumer, you’ve learned to take reports about internal polls — as opposed to public polls like NBC/Wall Street Journal or Quinnipiac, which are done by third parties with transparent methodology — not so much with a grain of salt as with a whole salt mine. Generally, if the public is seeing internal poll results at all, it’s only because there’s something in there that the campaign wants them to see.

So are internal polls totally worthless, just another piece of campaign propaganda? To some degree, yes — but, if you you apply a healthy dose of skepticism to what you read, stories about internal poll findings can still give you some information about a race.

With three hotly contested congressional races in Minnesota this year, you’re sure to see a lot of stories reporting on internal polls between now and November; here are some things to keep in mind as you read them.

Why campaigns — and outside groups — conduct internal polls

The reason political campaigns conduct polls boils down to two things: one is to provide the campaign with critical strategic information and a read on the state of the race — this kind of information is rarely released to the public.

The other reason to commission it is as a bit of spin — to manipulate the process with the goal of convincing donors to open their checkbooks, or to persuade the public and the press that a candidate is viable. This side of the polling coin is what the public usually sees.

Congressional campaigns usually contract with a third party to get polls done, but the idea is, with an internal poll, the candidate’s team gets to decide what questions are asked, how they are asked, and what information they want to find out.

Outside groups, like party-aligned super PACs or partisan advocacy groups, might conduct polls on a race, too, when it’s clear what the upside might be — to lend some weight to a challenger’s campaign, or to steal a challenger’s momentum.

A partisan pollster will work the levers — the content of questions, their wording, their order — in an attempt to get the desired response, whether it’s to jack up an opponent’s negatives or your own favorability, or both, or neither.

Taking the temperature of the electorate

Most congressional campaigns with any money to spend will commission at least one internal poll; incumbents or even marginally viable challengers will do several over the course of the election. (The cost of a comprehensive internal poll for a U.S. House race can be around $30,000, or it can run $45,000 or more, depending on a few things, such as sample size.)

Generally, they’ll start by doing a wide-ranging poll with a large sample size, perhaps in late spring or early summer, to get a comprehensive, initial look at who the likely voters are, what they care about, and what the overall dynamics of the race are like.

The poll — which, for a congressional race, might have 500 or so respondents — has to accurately reflect the district’s age, race, and political makeup, and ask a series of broad questions to get a sense of how enthusiastic respondents are about voting, what top-of-the-ticket coattails may exist, what issues are most important, and, perhaps most importantly, how many voters are undecided.

According to Jim Meffert, a Democrat who mounted an unsuccessful challenge to Rep. Erik Paulsen in 2010, “there’s two reasons to do polls before August. One is to really benchmark the undecideds. It’s name ID, it’s message testing, you’re trying to figure out what’s going to work, what the strategy is going to be, what people think of your opponent.”

Strategically, it’s important for a comprehensive poll to ask respondents their opinions on national political figures — for this cycle, it’ll be people like Donald Trump, Hillary Clinton, and President Obama, and perhaps state leaders like Gov. Mark Dayton — as well as on big issues like Obamacare, national security, and immigration.

Those kinds of questions give campaigns an idea of which issues resonate most when they’re thinking about what to emphasize in mailers and TV advertisements.

As the campaign gets going, if it has the cash, it will go back out and commission polls to test initial assumptions and see how change there has been, particularly with respect to how many people remain undecided.

Getting the results you want

But what about that second goal of internal polling — producing a positive result that shows a candidate is electable, or even ahead?  For example, imagine a challenger who’s running against a relatively strong incumbent, who’s conducting an internal poll to see where the electorate is at — but also maybe to pick up some buzz and a few checks from donors.

In that situation, what the poll might do is read a generic, positive paragraph about the challenger and the incumbent — the kind you might find in their website’s bio section. After polling for initial impressions, the interviewer will then read out information about the candidate’s opponent, usually framed with varying degrees of negativity.

If the opponent is an incumbent, the poll might bring up a key vote that can be interpreted as a stand against something broadly popular, like Social Security or vets’ programs.

They also may try to connect the incumbent with another unpopular member of the party — internal Democratic polls used to link Republican candidates with former Rep. Michele Bachmann — to drive up negatives.

After the respondent is subjected to the barrage of negative info, the idea is that the challenger’s favorability ratings go way up. If it’s enough to put a challenger within striking distance, or even in the lead, the campaign’s case is easy: voters will flock to them once they know what they are really about.

That, in turn, makes for a neat pitch to donors: give the campaign money so they can run ads to get out their own message, attack the incumbent, and — as the polls supposedly prove — give the challenger a chance.

This objective was essentially stated by the Democratic-aligned House Majority PAC in a poll they conducted for Mike Obermueller, who challenged GOP Rep. John Kline in 2014. The outside group’s poll found Obermueller up six points on Kline, but said “Obermueller will only have this chance in 2014 if he has the media dollars to articulate his message.” (He ended up losing to Kline by 17 points.)

A mysterious cake

For the public, internal polls are practically useless without the aforementioned context and caveats. But they’re often reported relatively straight in the press.

A big problem for the public, and the press, is that campaigns rarely release the entire polls, or the methodology of those polls — leaving people in the dark as to what was asked, how it was asked, and who was surveyed. One Republican strategist described it as being like serving a cake at a dinner party, assuring the guests it’s good, but refusing to say what’s in the cake.

Unless there’s a leak, then, the public doesn’t get an idea of the ingredients of the poll, and how the result got cooked. But people are still interested — and many races, particularly congressional ones, have a dearth of this type of horse-race intel — so internal polls often get written up anyway, with campaigns offering reporters dueling spin on their results, which they always certify as legit.

For example, take a pair of recent CD2 polls, written up in the Pioneer Press last week. It cited the Lewis poll, which had him up 12 points on Craig, and found a full quarter of the electorate to be undecided.

Craig’s poll had her up one point over Lewis, but only six percent of voters were undecided.

Michael Brodkorb, the GOP strategist and blogger, went on Twitter to offer his thoughts and a bit of context. “The bulk of polling done by campaigns is not meant to predict the outcome of an election, but rather to understand the electorate,” he tweeted. “But the political reality is that the public assumes all polling by campaigns is done to just predict the outcome of the election.”

According to Meffert, each campaign had a clear motive in releasing their top-line internal numbers — alongside no information on methodology — to the press. “The only reason you say anything about them is fundraising,” he said.

“Lewis needs to make it a viable race, make it seem like there’s an opening, and Craig wants to make it seem close, she needs to get that national money.”

If that were true, it’d be an exception to the rule. In 2012, writing in the New York Times, Nate Silver did an analysis of internal polling by going through past polls released to the public in U.S. House races. He found that the ones made public “were about six points more favorable to their candidate than independent surveys on average — and that they were typically less accurate in the end.”

“The release of internal campaign polls is not that different from any other form of spin in this respect — but that is precisely the point,” Silver wrote. “Internal numbers that a campaign releases to the public should be thought of less as scientific surveys and more as talking points.”

Useful to campaigns, but potentially dangerous

If done right, internal polling can be a useful tool for candidates, even with all the attendant problems.

Of course, there’s money: managing a campaign necessarily means trying to get the most out of finite resources, and polling is an expensive part of running for office.

For a well-funded challenger like Stewart Mills in the 8th District, or an open-seat candidate like Craig with incumbent-like levels of cash on hand, that may be an easy decision.

But for a challenger with little margin for error, like Meffert back in 2010, it’s a big deal. “There was a lot of hand-wringing about it,” he said, “whether we spend the money and whether we should do it.”

He says there are other risks involved, too. “Once you do a poll, every campaign has enough people in the voting universe that when you do a poll, your opponent will know it, there’s no way to hide it,” he said, adding that he knew what attacks Paulsen would throw at him, and vice versa.

“You have to balance that with whether your message and whatever’s gonna come out of it transcends any preemptive strike your opponent does.”

There’s also the issue of exactly how much stock to take in the poll itself. If history is any judge, placing too much faith in them can stir up hopes in the high-pressure bubble environment of the campaign trail.

These days, the poster child for the danger of drinking too much of your own polling Kool-Aid is Mitt Romney.

A big storyline in the aftermath of his defeat to Obama was how much his team believed in their internal polling — and how blind-sided they were when those polls turned out to be wrong. Per Nate Silver, in key swing states, Romney’s internal polling outperformed his actual numbers by an average of 4.7 percent.

Of course, a congressional campaign is different than a presidential campaign. And as Silver notes, it’s not as if Romney campaigned less hard in New Hampshire because his polls had him up. But if the purpose of internal polls is to provide an accurate read on the race to a candidate’s team, then the Romney example serves as a cautionary tale.

What can be learned from internal polls

With all these problems, you might be tempted to ignore internal polls — and if you did, you’d be just fine.

But, by reading between the lines, you can get some valuable information about the dynamics of a race, and even what a campaign may be thinking.

With the Craig and Lewis polls, for example, each poll’s estimation of how many voters are undecided is revealing. Fewer undecided voters limits Lewis’ potential upside in the district; if a quarter of the electorate really is undecided, as Lewis’ poll found, it bolsters the notion that many voters here are independent, non-partisan types — the exact kind of voter that Lewis claims will put him over the top in this open-seat contest.

Or take this poll, released this week by the conservative American Action Network, an advocacy group founded by former Sen. Norm Coleman. It found Paulsen leading DFL State Sen. Terri Bonoff by over 25 points in the 3rd District race. An internal Bonoff poll, meanwhile, showed her tied with the incumbent.

Clearly, Paulsen and his allies have no interest in making the race seem close for fundraising purposes — the Republican is a prolific fundraiser. Rather, the GOP wants to make Bonoff’s challenge appear as flimsy as possible to cut off whatever momentum she’s picking up.

The main thing to keep in mind when considering internal polls? They won’t tell you much about who is going to win or lose the race, but the numbers a campaign chooses to emphasize can give you an idea of what tactics it thinks will lead to victory.

You can also learn about all our free newsletter options.

Comments (2)

  1. Submitted by David Broden on 08/30/2016 - 04:16 pm.

    A More Thoughtful View and Understanding of Poll Differences

    The article keeps wandering across the spectrum of what various polls are intended for and how they are used. Having prepared, conducted, and analyzed polls of many candidates and worked with some excellent and some not so good polling firms the following are some key points to keep in mind. There really are just two types of polls; 1) Polls that ask the answer- those are the polls structured to act as objective polls but in which the questions ensure the answer- today most political candidates want this type so they are show the electrorate that they are in a favorable postion and need support. These polls serve no purpose, have no value and are misrepresentative in many ways yet they are used and quoted. The second is a truly objective poll– these require some serious understanding of grammer, statistics, correlation, and data analysis, The also must ask the same question at least two ways an often 3 or 4 times to track the responder. Inclusion of a benchmark question such as is the nation on the right track or is the economy better or worse must be included as a reference and balance. The few of these around these days continue to serious and useful polls even with the difficulty in obtaining a vaild statistical sample due to changes in citizen communication methods. One national major polling firm with whom I worked began each planning session with a few messages– 1)we will establish a rigorous and valid sample even if it takes two tries to validate the poll, and 20 we will not ever structure a poll to ask the answer someone wants — we want the true opinion of the electorate regarding the candidate and the issues. Multiple questions of the same topic presented in reverser language and in differnent context provided the correlation and validation. Queston today is are we serious about the facts or do we want the ideology to be the answer. Let’s hope objectivity win that is GOOD GOVERNMENT

    Dave Broden

  2. Submitted by Jean Schiebel on 08/30/2016 - 11:53 pm.

    Polling

    I received the poll call..the problem was I don’t live in the 3rd district, since it was a robo call..I just hung up. How accurate can it be if this mistake is being made?

Leave a Reply