A Year of Decision for Pollsters
Pity the political pollsters, under fire from all sides and confounded by the proliferation of mobile phones and the rise of Donald Trump.
Karlyn Bowman, a polling expert at the American Enterprise Institute, contends that 2016 “will be the last cycle for polling as we know it.” The venerable Gallup organization has abandoned horse-race polling in favor of surveys on issues. Tim Storey, a respected analyst with the National Association of State Legislatures, says flatly that he no longer believes in horse-race polls.
Pollsters are also on the defensive globally. In 2015 they missed the mark in elections in Britain, Greece and Israel. Closer to home they blew the Kentucky gubernatorial election in which surveys consistently gave the lead to the Democrat, Jack Conway. Republican Matt Bevin won by a nine-percentage-point margin. The day after the elections the Lexington Herald-Leader dumped Survey USA as its pollster.
The Kentucky fiasco continued a 2014 trend in which polls understated Republican strength in several Senate races as well as the governor’s race in Wisconsin.
The pollsters’ plight was cogently summarized by Alan Greenblatt in Governing magazine: “Fewer and fewer people have landlines. Buying lists of cellphone numbers…is more expensive for pollsters. Getting people to answer the phone and stay on the line to answer questions is increasingly difficult. Meanwhile, pollsters haven’t yet figured out how best to reliably guide opinions through online surveys.”
Nevertheless, polls are proliferating. They’re essential for candidates in securing donors and indispensable to the media in coverage of this year’s Hydra-headed Republican presidential race. Like it or not, they’re also of value in “helping direct and guide what elected officials do,” in the words of Frank Newport, the editor in chief of Gallup.
The oft-cited RealClearPolitics average of polls has become a benchmark for politicians and the media. Averaging often compensates for small samples or vagaries in particular polls. But averages are no help when all the pollsters have it wrong, as was the case in this year’s Kentucky election.
On a national level, much of the recent skepticism about the polls has focused on their measurement of Trump, who was arguably underestimated by much of the media early in the campaign. Political pundits predicted that Trump would fade because of derogatory remarks about Mexicans, Muslims and Sen. John McCain and his insults of rivals and reporters. But in the latest RCP average of polls Trump has a commanding lead in the GOP presidential race with support of 35 percent compared to 19.5 percent for the runner-up, Sen. Ted Cruz.
Many analysts believe the national horse-race polls are irrelevant. Presidential nominees are selected state by state, and Trump trails Cruz slightly in Iowa, which holds its caucuses on Feb. 1. According to the RCP average, however, Trump has a big lead in New Hampshire, which holds the nation’s first primary a week later.
Ironically, those who dismiss Trump’s standings in the national horse-race polls are relying in many cases on these same surveys for their analysis. The polls show that Trump is strongest among independents, who in many states cannot participate in Republican primaries, and among younger voters, who tend not to vote at all.
By this analysis, Trump is weak among likely voters, as he well may be. But as Republican pollster Kristen Soltis Anderson points out, Barack Obama redefined likely voters in 2008, in part by turning out significant numbers of younger voters.
Other analysts say that Trump’s standing is understated because of what pollsters call the “mode effect.” He fares better in online surveys than in polls conducted by telephone supposedly because there are Trump supporters who are unwilling to declare their preference to a pollster.
The notion that there are secret supporters of a polarizing candidate is a hallowed one in American politics.
In 1948, leftist candidate Henry Wallace, the nominee of the Progressive Party, was shown in polls garnering up to 6 percent of the vote, which almost certainly would have cost President Harry Truman key swing states. Many thought that Wallace would do even better than his poll numbers. But he received only 2.4 percent of the vote and did not cost Truman any state. In a great upset, Truman defeated the heavily favored Republican, Tom Dewey.
(That election, by the way, changed polling practices. Dewey had a big lead that was dwindling as Election Day approached. Gallup did not poll the final week. Ever since 1948, pollsters have taken surveys until the final day of the campaign.)
In 1964, when Barry Goldwater was the Republican nominee and trailing badly in the polls, some supporters nurtured a fantasy that a hidden conservative vote would emerge on Election Day and carry him to victory over President Lyndon Johnson. Instead, Johnson, who portrayed Goldwater as a warmonger while saying nothing about his own plans to widen U.S. military involvement in Vietnam, won in one of the biggest landslides in U.S. history.
In 1968, it was widely believed that there were voters unwilling to tell pollsters of their support for independent candidate George Wallace, an outspoken racist. This, too, was fantasy. Wallace polled 14 percent in the last Gallup Poll and received 13.5 percent of the vote.
So with history as a guide, I doubt the existence of a substantial hidden Trump vote. I also doubt Trump will do well in Iowa, where he seems to lack the organizational ground game of previous caucus winners, notably Obama in 2008 and, on the Republican side, Mike Huckabee in 2008 and Rick Santorum in 2012.
On the other hand, if Trump maintains anything remotely close to his current level of support, he could wind up with a batch of delegates at the Republican National Convention. Under party rules for 2016, the more than a score of states that hold primaries or caucuses before March 15 must award delegates proportionally. From March 15 onward, delegates are awarded on a winner-take-all basis.
On balance, the much-maligned polls have performed useful functions in the presidential race. They’ve pointed out Trump’s appeal, identified constituencies that back the various candidates and, for those who follow Gallup, shed light on the issues that motivate voters.
When it comes to the horse race, to use a distinction articulated by Kristen Soltis Anderson, the polls have been valid rather than predictive. Many Americans are uninterested in politics in non-election years; their opinions of candidates often change once they begin considering their choices.
Finally, this personal note. I was a reporter for The Washington Post for 26 years and covered many political campaigns. Once I participated in an academic event where the principal speaker denounced the media focus on horse-race reporting. When he left the podium to loud applause, he sat down beside me and asked who would win the election.
For better or worse, the polls are here to stay.
(Correction: This article originally misstated the RCP Average in 2012.)