It’s no secret that politicians often rely on polls to craft their message, sometimes to a fault. Hence the saying: “I must take a poll to find out where my people are going so I can lead them there.”
Few things are more calming to political pundits than hard numbers. They provide an objective, mathematical anchor in the midst of opinion and conjecture. They form the basis of campaign strategies and marketing tactics. Based on voters’ opinions — and the ages, genders, and ethnic backgrounds of those voters — politicians may shift their stances, harden their tone, or back off a particular issue.
But, heading into the 2016 election, we must face a difficult truth: The accuracy of election polling has been declining, and it’s likely to get worse.
To find inaccurate polling results, one need look no further than the 2014 gubernatorial election in Maine. Just days before the election, the Maine People’s Resource Center predicted that Rep. Michaud would win a narrow victory. Even RealClearPolitics, a poll aggregator, indicated that Gov. LePage was leading by only 1.4%. As it turned out, Gov. LePage won re-election by nearly 5%.
Pollsters similarly underestimated the strength of Bruce Poliquin in Maine’s 2nd Congressional District, who — despite polls suggesting a tight race — won his House seat by over 5% of the vote.
Such miscalculations from local and state polling agencies may be surprising, but even national organizations often blunder. Most pollsters didn’t predict the landslide Republican victory in 2014, Israeli Prime Minister Benjamin Netanyahu’s comfortable re-election earlier this year, or the British Conservative party’s success in the general election held in May.
In a recent article published in The New York Times, Cliff Zukin, professor of political science at Rutgers University and past president of the American Association for Public Policy Opinion Research, explained that political polling is in crisis as statisticians scramble to reconcile the results of recent elections with their polling data and mathematical models.
According to Zukin, two trends are contributing to the increasing unreliability of election polling in the United States: the growth of cellphones and the decline in people willing to answer surveys.
Today, a majority of people are difficult or impossible to reach on landline phones. Pollsters must resort to cell phones, but the Federal Communications Commission has prohibited the practice of calling cellphones using automatic dialers, which pass the call on to a live interviewer only after a person picks up the phone. To complete a 1,000-person survey, it’s not unusual to have to dial more than 20,000 random numbers. Doing so manually takes time and money.
Small news websites often launch a polling department merely to bolster their credibility, and settle for third-rate statistical techniques to reduce costs and speed up the process. Fancy statistical models — which embrace dubious assumptions about demographics and turnout — have replaced large numbers of respondents and methodically random sampling.
The second unsettling trend is rapidly declining response rates. In the late 1970s, pollsters considered an 80 percent response rate acceptable, but by 2014 the response rate had fallen to 8 percent. As the response rate dwindles, the chances that bias will creep into the data grow.
Another issue is poll herding, where a pollster either re-weights an outlying poll to align it with other poll results, or fails to publish it altogether. This practice can be problematic when voters suddenly shift their views dramatically, because pollsters are often unwilling to break from the pack to report the changes. In races like the 2014 Virginia and Iowa Senate contests, herding seems to have shrouded substantial movement as Election Day approached.
We may feel the impact of inaccurate polling long before the 2016 general election or even the primary season. For instance, the entrance criteria for the upcoming Republican presidential debate on August 6th — whereby the ten candidates leading in national polls will be allowed to participate — are unlikely to select the ten most viable candidates.
Zukin cautions: “Our old paradigm has broken down, and we haven’t figured out how to replace it. Political polling has gotten less accurate as a result, and it’s not going to be fixed in time for 2016.”
Remember that the next time you read about a candidate falling behind in Iowa or making gains in New Hampshire.