Saturday, December 3, 2016

Polls and Politics: Five Things You Need to Know

By Gail Bambrick

They make the news every day in election season, but how much is too much?

Jeffrey Berry

“Where polling does have an impact in general elections is in fundraising,” says Jeffrey Berry. “If your candidate is way behind, you are less likely to give money.” Photo: Melody Ko

On the eve of national elections, political polls make the news almost every day as the jockeying becomes more intense and pollster websites like Nate Silver’s fivethirtyeight.com become stalking grounds for political junkies.

In the first of a continuing series, “Five Things You Need to Know,” Jeffrey M. Berry, the Skuse Professor of Political Science in the School of Arts and Sciences, gives us the lowdown on polls: how influential they are, who uses them and why we need to be skeptical of the results. Berry, who has focused his research on American politics and political behavior, says that we need to be especially careful when reading about poll results in the media, whose coverage of them is often “illusory.”

1) Polls do not sway voter behavior in general elections.

They tend not to be highly influential because there are only two candidates, and if you are a Republican, you are not going to vote for the Democrat just because the Democrat is ahead. There is no evidence that independents are susceptible to a bandwagon effect generated by polls.

Where polling does have an impact in general elections is in fundraising. If your candidate is way behind, you are less likely to give money.

2) Polls influence primaries, but not issues.

People may like all the candidates, but use the polls to see who has a better chance against an opposing party incumbent. But it can be the reverse: Christine O’Donnell was trailing in the polls throughout the Republican primary for the special election to fill Joe Biden’s Senate seat in Delaware, but then won a decisive victory over [nine-term U.S. Rep.] Mike Castle. So clearly Delaware voters did not give up on her just because she trailed all the way.

The impact of polls on how voters view issues is indirect, but it can affect legislators who may, for example, be more likely to vote to eliminate “Don’t Ask, Don’t Tell” if they think the public has become more tolerant toward gays and lesbians.

3) Campaign polls should always be disregarded.

Campaign polls are private, and candidates leak them to the press only when they have good news. But they are rarely published, so you never get to see how the questions are asked or how the sample was drawn.

Most other polls are sponsored by media outlets or by vendors that sell their goods to media outlets. The media are trying to generate a story, or in some cases, there are independent pollsters trying to create a name for themselves. Quinnipiac University in Connecticut has a large polling unit and the news generated by its polls has enhanced the school’s visibility.

4) Automation has made polls cheap to produce and more prevalent—but not more relevant.

Some companies do automated polling so they don’t have to pay people to phone respondents. Rasmussen and Survey USA are the leading vendors. More polling is being done than ever before because of automated polls, so it has become a bigger facet of political life. They use very small samples—500 people—so the error margin is pretty high. There is also polling over the Internet.

The accuracy is pretty good, as long as you account for error margin—a statistic that’s always published but commonly ignored. Journalists are particularly guilty on this count because they like to emphasize the changing nature of the horse race. Yet often the change they write about is within the margin of error—4 to 5 percentage points. A lot of times the change they write about is illusory.

5) Polls can and have been abused.

The push poll is the most common example. Someone calls and instead of saying “are you for Jones or for Smith,” they will say “are you aware that Smith backs killing babies,” then ask “who are you for, Jones or Smith?” This is disguising campaign calls as polling. It’s clearly against the code of ethics of the pollsters’ trade association, the American Association for Public Opinion Research. Founded in 1947, it is an organization of approximately 1,900 survey research professionals from academia, nonprofit organizations, polling firms and government.

A good resource for unbiased information is www.pollster.com, run by political scientists with links to articles about polls and polling. It is now hosted on the Huffington Post and is being called HuffPost Pollster. But it maintains its integrity, and here the results of several polls over time are graphed on issues and elections. This wider view is more accurate and isolates the idiosyncrasies that may occur in a single poll.

Gail Bambrick can be reached at gail.bambrick@tufts.edu.

Have a suggestion for “Five Things You Need to Know”? Email the editor, Taylor McNeil, at taylor.mcneil@tufts.edu.

Posted October 13, 2010