Nonprofit, nonpartisan journalism. Supported by readers.


Why there are so few statewide presidential polls in Minnesota (and almost everywhere else) this year

REUTERS/Jonathan Ernst
Through Sept. 28, pollsters had released the results of just seven Minnesota-only surveys of voters' preferences in the faceoff between Donald Trump and Hillary Clinton.

As this year's unorthodox presidential campaign lurches into the home stretch, national polls on how Hillary Clinton and Donald Trump are faring continue to cascade through the national media. That's given rise to an impression that we're getting more of these polls than ever.

But the polls that matter most are those taken in the states, given that because of the Electoral College we choose our presidents by majorities of voters in each of the 50 states and the District of Columbia. And in Minnesota, there have been notably fewer statewide presidential polls so far this year than in the 2012 campaign. That's also the case in many other states.

There's a reason for that: Political polling in the states has become more expensive, so less of it is being done. Another concern for pollsters: Survey experts warn that the unusual uncertainties inherent in the 2016 campaign election and growing challenges to traditional polling practices could make the process more dicey this time.

We charted polling patterns by examining the archives of RealClearPolitics (RCP), a site that has aggregated presidential polls in each of the 50 states and the District of Columbia since 2004. Through Sept. 28, pollsters had released the results of just seven Minnesota-only surveys of voters' preferences in the face-off between Clinton and Trump. The state saw 12 such polls during the comparable stretch of the 2012 race between President Barack Obama and challenger Mitt Romney.

Minnesota poll update
Through Sept. 28, there have been only seven Minnesota polls matching up Hillary Clinton and Donald Trump -- just three since April. By comparison, there had been 12 Obama-Romney polls by this date in 2012.
Poll *DateClintonTrump
KSTP/Survey USA9/16/16 – 9/20/164639
Star Tribune 9/12/16 – 9/14/164438
Star Tribune 4/25/16 – 4/27/164835
Star Tribune 1/18/16 – 1/20/164338
KSTP/Survey USA10/29/15 – 11/2/154245
PPP7/30/15 – 8/2/154439
* Excludes a national online poll of each of the 50 states taken Aug. 9-Sept. 1 by the Washington Post and SurveyMonkey. That poll showed Clinton leading in Minnesota 49 to 40.

Not swinging anymore?

Among the reasons for lessening frequency of presidential polling so far this year in Minnesota:

First, just as the candidates and their campaigns flock to the swing states, so do the pollsters. In 2008, when Minnesota was widely seen as a key battleground, 13 pollsters did 48 McCain-Obama matchups. The prevailing sense then was that the state was evenly divided; despite the fact that a Republican nominee had not won Minnesota since the state went for Nixon in 1972, it was ready to flip. That belief influenced the GOP to bring its national convention to St. Paul in the summer of 2008. Barack Obama and John McCain visited the state numerous times. But then Obama beat McCain, and later Romney, by comfortable margins. Minnesota's two Democratic senators easily won re-election. Thus the new wisdom heading into this year's campaign: swing state no more.

Second, pollsters' interest in Minnesota could be down this year because no statewide contests are on the ballot other than the presidential race.

But if those were the only reasons for the decline in polling, it follows that pollsters would remain as active this year as in 2012 in closely contested states such as Florida and Ohio. Generally, that isn't happening. Ohio saw 27 Clinton-Trump polls through Sept. 21 vs. 51 for the Obama-Romney contest. In Florida, polls fell to 40 from 64. Similar patterns are unfolding in most of the other swing states.

Expenses soaring

So what else is going on? Rob Daves, the Minneapolis survey consultant who directed the Star Tribune's Minnesota Poll from 1987 to 2007, points to the soaring cost of polling. "It's the 700-pound gorilla," he said.

Response rates to pollsters' calls were above 50 percent in the 1990s, Daves said. "Now if you get 10 percent, you're lucky." In part, that's because the rising numbers of calls from marketers have led people to screen out more calls. This means pollsters must put out more calls to get credible sample sizes. Thus their expenses climb due to the labor and time it takes to make additional calls.

Also, the inescapable popularity of cellphones means more overhead for polling firms, since they need to do more to get samples that reflect the electorate. Daves advises his clients that cellphone replies should account for as much as half of all responses. Steve Wagner, who heads St. Cloud State University's polling effort, recommends a higher cellphone share, 60 percent, in Minnesota. Scott Keeter, director of survey research at the Pew Research Center, thinks that share should be even higher — 75 percent. The higher the cellphone share, the greater the cost. As Wagner notes, lists of mobile phone users are more expensive than lists of traditional landlines because of the need to match cell numbers with Minnesota addresses. Simply calling a 612 area code doesn't guarantee that the person with that number lives in Minnesota. 

If that's not enough, consider the cost implications of legal issues. Recently, the Federal Communications Commission (FCC) issued an administrative ruling that makes it easier for plaintiffs to sue researchers and telemarketers who call cellphones. Guidelines can help the pollsters avoid litigation, but only if they make significant investments in hardware and software. The result: higher costs for call centers, which pollsters must either swallow or pass on to their clients.

The FCC, which allows computerized call dialing to landlines, has banned this practice in dialing cellphones. The agency requires real people to dial — thus again driving up costs.

Wagner said the cost of buying a good cellphone sample has nearly tripled over the last four years. Overall, he estimated that a 20-question, 1,000-call sample for a presidential poll in Minnesota would cost $46,000 today; in 2012, it would have cost $34,000. Scott Keeter said that generally, the cost of political polling has doubled since 2012.

Fewer pollsters active here this year

In Minnesota, just three of the six pollsters who did head-to-head presidential polls in the state in 2012 have been active so far in this year's campaign: Public Policy Polling (PPP), one poll; KSTP/Survey USA, two; and Star Tribune/Mason Dixon, three. The other three pollsters, who could yet survey in coming weeks, are St. Cloud State, Rasmussen and NMB Research/AFF. The latter two have been almost totally absent from state presidential polling this time. St. Cloud State is planning a poll in October. PPP did seven Obama vs. Romney matchups in Minnesota in 2012. 

KSTP/Survey USA did eight Obama-Romney polls in 2012; so far this time, two Clinton-Trump polls. Tom Hauser, KSTP's chief political reporter, cited the lack of statewide races for the U.S. Senate and governor as the principal reason KSTP is doing less presidential polling this time. "Everybody's trying to cut costs," Hauser said, noting that KSTP is the only Twin Cities television station doing presidential polling.

But Hauser said the station is gearing up for separate polls in mid-October that will be a significant expense: surveys in three hard-fought congressional races, in the state's second, third and eighth districts.

The Star Tribune has done three Minnesota presidential polls so far this time and could do at least one more. It did only two in 2012. But despite that, signs of budget pressures are evident there. The newspaper's polling sample in its latest poll, taken Sept. 12-14, was easily the smallest for any of the 10 presidential polls it has done since the 2008 campaign. The paper has lowered its share of landline replies this year, to 70 percent from 80 percent in 2012. Still, that share remains well above the lower level recommended by polling experts. Asked about that, Star Tribune Editor Rene Sanchez said in an email, "We're comfortable with where we've ended up on sample size and the percentage of cell phone users because over the last few years our polls have never been outliers and have ended up foreshadowing final results well. But, polls are only ever a snapshot in time, of course, so you can only put so much faith in them, especially in a volatile election cycle. And we will continue to evaluate things like cell phone use."

Sanchez also said:

We view polling as a vital part of our election-year coverage. Not just the matchup results, but the important stories that can be found from poll questions that focus on the state of mind of Minnesota voters and how their views can vary based on gender, age, income, geography. ...

But, yes they are very expensive, especially if you enlist a reputable, nonpartisan polling firm, as we do. We steer clear of automated polls. Two things drive up the costs, aside from having a good sample size of voters that keeps the poll's margin of error low.

The first is the rising need to track down a substantial percentage of cell phone users. The second is the interest we always have to call back and interview some of the Minnesotans who take the poll, which we believe brings more depth to our stories.

Our expenses on polling are higher this time than last. Initially this year we had plans to do only two big, wide-ranging polls. But, given how extraordinary the presidential race is, we decided this summer to add a third big poll, which we recently published. It's doubtful we will poll again before the election, but I haven't ruled it out yet.

Battleground again?

More Clinton-Trump polls are expected in Minnesota in coming weeks. In addition to the St. Cloud State poll, KSTP/Survey USA plans at least one more. The Star Tribune, as Sanchez noted, could also do one more. Tom Hauser suggested that if the race appears to be tightening in Minnesota, other pollsters could come in. Still, it would take twice as many as were done through Sept. 21 to match the 22 done in the 2012 campaign.

The RCP site now contends that Minnesota has once again become a battleground state. On the weekend before the first presidential debate, the site posted a new poll, from Gravis Marketing and Breitbart, an "alt right" site widely seen as friendly to Trump. The Gravis/Breitbart poll showed Clinton and Trump in a dead heat in Minnesota — 43-43. Yet there are reasons to question the reliability of that poll. It appeared to oversample Republicans. Some of the questions seemed slanted to encourage Trump-favorable responses. There was no clear indication that it met the recommended number of cell phones, since the poll did not reveal this statistic.

Nonetheless, the RCP site lumped the results of that poll in with new Minnesota surveys from KSTP and the Star Tribune, which showed Clinton ahead of Trump by seven points and six points, respectively. The effect was to slash Clinton's lead to four points. That led RCP to shift Minnesota into its "toss-up states" column from "leans Clinton" in its closely watched daily aggregation of state polls.

Cost cutters target polls

As polls have become more costly, fewer media outlets can afford to do them. Newspapers and other traditional news services have seen declining readership or viewing audiences. Nationally, larger news organizations such as the Los Angeles Times have cut polling directors and staff. Smaller news organizations have also been facing similar if not greater pressure to maintain profit levels by cutting costs.

In St. Paul, the Pioneer Press left the presidential polling game after the 2004 election.

"I wasn't involved in the decision to get out of that field, but I certainly haven't lobbied to get back in," Mike Burbach, the paper's editor since 2011, said in an email. Among the reasons: "Polling is expensive. I'd rather spend the money elsewhere."

We found at least 13 dailies that sponsored state presidential polls in 2012 or 2008 missing in action so far this time.

KSTP partner Survey USA, founded by journalist Jay Leve in 1990, has done political polling for many television stations across the country for years. In an email, Leve noted that many media outlets can no longer afford the expense of such polls. "Traditional media companies are shrinking," Leve said. "They have less money to spend with each passing day. If they have less money, they spend less on research. It's that simple."

A churning business

For perspective on Minnesota's presidential polling, we compared trends in the state with patterns in all of the 50 states and D.C. Overall, these presidential polls fell from 685 for the Obama-Romney matchup to 416 for Clinton-Trump — a 39 percent decline.

Amid all of this, the make-up of the pollster army has changed considerably this time. Two highly active firms in 2012, PPP and Rasmussen, have pulled back considerably in the current campaign. 

On the other hand, colleges and universities, mostly smaller institutions, have become big players in the state presidential polling game. Quinnipiac University in Connecticut has become by far the most active of the college pollsters in statewide presidential matchups this year. Its polling consistently garners national attention for the school.

In the Upper Midwest, Marquette University has become the dominant presidential pollster in Wisconsin. At St. Cloud State, Steve Wagner notes that schools, seeking to raise their profiles through widely publicized surveys, subsidize their polling in many ways. Frequently, they have faculty members proficient in polling techniques and students available to do much of the work. That helps them deal with the rising costs of polling.

But in the end, all of these pollsters are on uneasy ground this year. Consider this recent warning from Alan Dershowitz, a prominent constitutional lawyer who follows the political tides closely: "The reality is that polling is incapable of accurately predicting elections such as this one, where so many voters are angry, resentful, emotional, negative and frightened."

Rob Daves said this may be the most difficult presidential election in memory for pollsters. "Each election presents its own challenges, but much higher costs for phone surveys, lower response rates — the two are linked — and the difficulty of identifying a likely electorate are big challenges in measuring candidate support. Couple those things with the high negative feelings people have about both major party candidates and you've got a really tough set of challenges for researchers to cut through.

"It's also difficult for poll watchers as well. As the cost of traditional telephone polling goes up — using real interviewers, not touch-tone dials or the like — many researchers have moved to online surveys, which is a mode that usually doesn't have a scientific sample and is still in its infancy with a lot to be learned."

As Tom Hauser puts it: "No question about it. The conventional wisdom is useless in this case."

David Schultz, co-editor of the 2015 book "Presidential Swing States: Why Only Ten Matter," is a political science professor at Hamline University. Dave Beal, a former business editor and columnist at the Pioneer Press, is a freelance journalist.

Comments (4)

  1. Submitted by Hiram Foster on 10/07/2016 - 09:51 am.


    I have a lot of problem with polls. Maybe one of them relevant here is that polls purchased by news organizations aren’t really news in the sense that they are something that happens that newspapers and other media cover. Instead, they are stories that new media create for themselves and then publish. This creates all sorts of ethical dilemmas for the media. For example, the Star Tribune typically publish information about their polls in a headlined story on the front page. This is done, not because any sort of editorial judgment about the significance of the story, rather it’s simply something they do to satisfy the business side of the paper. In what other instance is the editorial judgment of the paper subordinated to it’s business interests. Note also, that there seems to be some pressure to vouch for the accuracy and validity of polls purchased by the newspaper despite the fact that polls are seen as increasingly unreliable for some of the reasons alluded to in this article.

  2. Submitted by Jeff Michaels on 10/07/2016 - 02:44 pm.

    The fascinating history of the Minnesota Poll

    Seeking a Star Tribune perspective on opinion polls is comparable to having Donald Trump lecture on the evils of arrogance. The newspaper has a long and questionable history regarding opinion polls.

    Take, for example, the 1969 mayor’s race among Minneapolis police captain Charlie Stenvig and Minneapolis city council members Republican Dan Cohen and DFLer Gerry Hegstrom. City elections were held in May then. Back then, the Minneapolis Tribune usually ran results of opinion polling the Sunday before the mayoral election. But it did not that year because its polling showed law-and-order candidate Stenvig was going to win, which he did. I knew people connected to the Cohen campaign and they thought it was a good political move . . . for them. (That was first and last time the local newspaper did anything for the feisty Mr. Cohen.)

    In the 1978 race for governor, the Minnesota Poll final pre-election poll had DFLer Rudy Perpich winning by four points. He lost to Republican Al Quie by seven percentage points. That poll had an interesting twist in that the polling company fired the newspaper because of continuing disagreement over initial polling results. In fact, the episode was so embarrassing, the paper discontinued polling for several years.

    In 1998, the final Minnesota Poll predicted DFLer Skip Humphrey would be the new governor with 35 percent of the vote. And Jesse Ventura would finish third. Anyone remember how that one turned out?

    In 2000, the final presidential Minnesota Poll predicted George Bush would receive only 37 percent of the Minnesota vote and Democrat Al Gore 47 percent. Gore got 47 percent but, as usual, the Republican, George Bush, fared much better with 42 percent of the vote.

    In 2002, the final Minnesota Poll had DFLer Walter Mondale with a five-point lead over Norm Coleman in the U. S. senate race. Interestingly, the St. Paul Pioneer Press Poll had Coleman winning, which he did.

    If people check, they will also find initial polling invariably has the liberal DFLer far out in front, which is designed to increase financial contributions and attract campaign workers. The numbers tend to even out at the end with the final poll because that is the one by which polling outfits are judged.

    If history is a guide, the aforementioned Mr. Trump will do much better in the final Minnesota Poll compared to what initial polls may show.

  3. Submitted by Hiram Foster on 10/07/2016 - 04:01 pm.

    If history is a guide, the aforementioned Mr. Trump will do much better in the final Minnesota Poll compared to what initial polls may show

    I think the Star Tribune switches polling companies making year to year comparisons difficult.

    I do expect this to be a difficult year for pollsters because Trump will bring out a different set of voters. That’s what made the Ventura election so difficult to poll.

  4. Submitted by Hiram Foster on 10/08/2016 - 06:10 am.

    Consipiracy theories

    There is, out there, a view that the Star Tribune, currently owned by billionaire businessman Glen Taylor, manipulates it’s poll results to benefit the DFL. I have never given much credence to these theories. For one thing, I have never seen what advantage would be gained by taking such a risk. For the most part you don’t need polls to tell when the political ground is moving under your feet. A previous commentator cited the Ventura election. Well that was a fairly late breaking phenomenon that polls had a hard time keeping up with, but to some extent they did. Bear in mind that polls are a reflection of what’s happening now, now what is going to happen at some time in the future, even only weeks or days in the future.

    One of the frustrating things about polls is that they can be counted on to fail you when you most need them. Polls do what stock prospectuses tell you never to do, use past results to predict future performance. They work best when things are steady, unchanging. They predict well when things are easily predictable, telling us for example that a popular governor will reelection. But when the electorate changes. when the samples no longer reflect who is actually voting, as they did in the Ventura election, suddenly polls become nothing more than someone’s best guess dressed up in numbers.

Leave a Reply