Are the 2018 midterm election polls accurate?

A woman arrives at a polling station in Lark Community Center as the early voting for midterm elections started in Texas, in McAllen, Texas, U.S., October 22, 2018. REUTERS/Carlos Barria - RC16F40C5550
Editor's note:

Visit our 2018 Midterms page for Brookings experts’ research and analysis related to the upcoming midterm elections, and subscribe to the Brookings Creative Lab YouTube channel to stay up to date on our latest videos.

After the surprising outcome of the 2016 election, many people lost faith in the results of polls and surveys, believing that they had been inaccurate in their predictions of how people would vote in an election. Brookings Senior Fellow William Galston explains how the 2016 election actually showcases the use of polls, discusses the usefulness of polls to candidates, and describes how polls can go wrong.

What you need to know:

  • In a representative democracy, candidates always want to know what the electorate is thinking.
  • For the past 70 years or so, most candidates have used surveys and public opinion polls to supplement what they can get from talking with their constituents.
  • 2016 is widely taken to be an election that invalidated the polls and revealed how inaccurate they are. In fact, the opposite is the case. The national surveys were closer to an accurate prediction of the national election outcome in 2016 than they were in 2012.
  • The problem in 2016 was not at the national level but at the state level. In many states, there weren’t any polls or any good polls.
  • Hillary Clinton’s campaign did not conduct standard surveys in a number of states that turned out to be crucial in the three weeks before the election.
  • Polls most frequently go wrong because they don’t predict turnout correctly. In 2016, for example, the standard models of who was going to turn out did not take into account the kind of mobilization of less frequent voters, such as blue collar voters, that Donald Trump was able to achieve in key states and key parts of those states.
  • In 2018, the big question is “Who is going to show up?”
  • Survey researchers make a distinction between registered voters and likely voters. If they get that model wrong, then their survey will not be accurate, even if it is technically impeccable in every other respect.
  • Candidates conduct polls to see which issues are salient and resonate, and what kinds of arguments for and against specific issues are likely to be most effective. They want to find out how people will respond to issues and arguments.
  • Both major political parties are locked in on the issues that they feel will drive their supporters to the polls. The Republican Party, under President Trump’s determined leadership, is focusing on immigration and the federal judiciary. On the Democratic side, survey research has determined that health care, and in particular coverage for pre-existing conditions, is the top concern on the minds of many voters. They are also hitting women’s issues hard because survey research has determined that women are far more discontented than men with not only the policies of the Trump administration but also the personal conduct of the president.