Can a poll really tell us who's going to win next year's election?

Can a poll really tell us who's going to win next year's election?
© Getty Images

Many credit Saint Bernard of Clairvaux (1091-1153) with being the first to philosophize that “the road to hell is paved with good polls.” Let’s consider this in light of the coming presidential election. What can this summer’s polls — six months before the first votes get cast in Iowa, New Hampshire and South Carolina — tell us about who’s going to win and what will matter most to Americans in 2020?

First, it’s really early. I’m old enough to remember when Howard Dean was going to win the Democratic nomination in the summer before the 2004 primaries kicked off. Hillary Clinton was considered unbeatable in August 2007 against Barack Obama. Remember when Newt Gingrich was going to be the Republican nominee a year before the 2012 primary elections began, and when Jeb Bush was guaranteed to be the 2016 GOP standard-bearer? (Let’s also not forget those pre-election polls in 2016 that put Hillary Clinton into the White House.

So, how much confidence do you now have in current polls showing former vice president Joe BidenJoe BidenSarah Huckabee Sanders becomes Fox News contributor Poll shows Biden, Warren tied with Trump in Arizona Giuliani says he discussed Biden with Ukrainian official MORE winning the 2020 Democratic nomination? About Sen. Bernie SandersBernie Sanders2020 candidates have the chance to embrace smarter education policies Bernie Sanders Adviser talks criminal justice reform proposal, 'Medicare for All' plan Poll shows Biden, Warren tied with Trump in Arizona MORE (I-Vt.) being primed to again win the progressive protest vote? About polls that show Donald TrumpDonald John TrumpSarah Huckabee Sanders becomes Fox News contributor The US-Iranian scuffle over a ship is a sideshow to events in the Gulf South Korea: US, North Korea to resume nuclear talks 'soon' MORE losing to just about any of the two dozen Democrats?

ADVERTISEMENT

Let’s stipulate that horse race polls are super fun to talk about. Who’s got momentum? Will the latest gaffe or gotcha zinger hurt? Will that amazing debate performance last week move her from 17th in the polls to 11th? But horse race surveys tell us almost nothing about what’s happening with voters on values and issues, and polls that do ask substantive questions are asking Americans who mostly have not yet focused on the coming elections. Expecting this summer’s polls to give us a solid view into next year is like debating which neighborhood cat would be fastest in a sprint; it might be interesting, but it’s kind of pointless.  

Second, current polls are not as good as past electoral truths and trends at predicting what matters and who will win. For example, presidential elections have been really close in the past quarter-century and are likely to be so again in 2020. It’s throwaway analysis to say the country is closely divided, but it nevertheless portends another close one in November 2020. Past is often prologue. 

Since the 1992 Clinton-Bush election, the winning presidential candidate has won by an average of only 4 points. For example, Obama won 51-47 during his 2012 re-election and George Bush’s 2004 re-election was the second closest for an incumbent in U.S. history. Even the best polling cannot adequately measure these small differences. The usual “plus or minus 3 percent margin of error” means that polls generally predict within a range of six points (and six is usually less than four).

Strong party-line voting is another inherited trend that better predicts elections than this summer’s polls. Partisanship — whether a voter identifies as a Democrat or Republican, liberal or conservative — is one of the two strongest predictors of voting choice. No one needs a poll to know that most Americans are going to support one party or another. (The other best vote predictor is religion, like how many times a week someone goes to church.)  

Party loyalty is historically high, partly because we have for the first time in more than a century one consistently liberal party in the U.S. — the Democrats — and one consistently conservative party — the Republicans. In fact, more than 90 percent of Republicans and Democrats are to the right or left of the other party’s median voter. Team and tribe are driving the debate and elections, often without heed to the headlines and issues. Even the anomaly of Trump and the once conservative GOP supports this. In that instance, Trumpian group identity appears to be offsetting ideology.  

ADVERTISEMENT

Third, today’s polls ignore the influence of tomorrow’s leaders. One of my favorite bumper stickers is, “If the People Lead, the Leaders Will Follow.” But this is not always true, especially in presidential elections. In 2016, for example, Trump moved immigration up as an election priority simply by the force of his arguments, not because it was a top issue for voters before he became a candidate. Especially in the Republican primary, voters followed him, not the other way around.

For an example from farther back, look to George H.W. Bush’s 1992 re-election effort. In the spring of 1991, Bush had a whopping approval rating in the high 80 percent range following a successful first Iraq war. Spinners of conventional wisdom predicted certain victory in 1992. Needless to say, Bill Clinton got the keys to the White House, partly because of his unrelenting reminder that, “It’s the economy, stupid!” 

No doubt, you questioned whether the Saint Bernard quote above was accurate. His road to hell was paved with good intentions, not good polls. Nevertheless, if you intend to predict the 2020 presidential nominating and general elections by looking at this summer’s polls, you’re likely to have a hellish time getting it right. It’s not that political polls are inaccurate, because they’re not. It’s that this summer’s polls are not enough and they’re coming at us too early to be helpful. We’ve got a long way to go, and there is much more at play than what voters are (occasionally) thinking about now.

Donnie Fowler is an adjunct professor at the University of San Francisco. He is CEO and executive director of Tech4America and co-founder of Democracy Labs. He has worked on the presidential campaigns of Dick Gephardt, Jesse Jackson, Bill Clinton, Al Gore, Wes Clark, John Kerry, Barack Obama and Hillary Clinton, and was a Clinton White House staffer as a liaison to Congress for presidential appointments. Follow him on Twitter @fowlerdonnie.