Bookmark and ShareShare
Tuesday, January 19, 2010

The Role of Polls in Massachusetts

The run-up to today’s Massachusetts special Senate election has been dominated by pre-election polls. The race is being held to fill the late Edward Kennedy’s Senate seat. (Although Republican candidate Scott Brown is quick to point out that it is not Kennedy’s seat, but the “people’s” seat”.) Kennedy was a Democrat. No Republican has been elected to the Senate from the Bay State since 1972. Therefore, it was assumed that the Democratic candidate, Martha Coakley, would sail to victory.

Initial pre-election polls backed up that assumption.

Then things changed. Polls began to show a tightening race. Brown, the Republican, was doing better. The polls generated increased attention being paid to the election. Brown began to pull even with and then move ahead of Coakley. Democrats began to worry. President Obama’s political team began to worry. Losing a Democratic Senate seat meant complications in the drive to pass healthcare reform. Obama’s team made the decision to send the president to Massachusetts this past Sunday to bolster Coakley’s chances. Obama lent his voice to get-out-the-vote telephone calls. The race, and the chance that Brown could win, became headline news. The eventual outcome suddenly became vested with extreme importance. Returns are being followed with a fervor usually reserved for presidential elections.



All of this occurred without a single vote being cast. The sturm and drang evolved directly out of pre-election polls monitoring the status of the race.

Throughout the history of modern polls, some have argued that this type of impact of pre-election polls is a not a good thing.

In fact, various countries around the world at various times have attempted to ban the publication of pre-election polls in the days leading up to an election. Many still do. One of these is Greece. In fact, just this past fall, the American Association for Public Opinion Research (I am a member of AAPOR’s Executive Council) issued a statement decrying the Greek government’s law banning the publication of election survey results in the final 15 days before an election.

A research document published by the World Association for Public Opinion Research a few years ago -- also supporting the value of pre-election polls -- noted “...there is a tendency for politicians to propose restrictions on the publication of opinion polls by the media - particularly in the run-up to major political elections.”

The WAPOR research found that:

In 2003, 30 countries - nearly half of those surveyed - have some kind of restriction on the publication of polls. And this may be just the tip of the iceberg. We have no evidence for many countries and these blind spots are precisely where the political situation makes such restrictions even more likely.

Examples cited by WAPOR (in addition to Greece): Switzerland has a 10-day before Election Day embargo on pre-election polls; Spain a 5-day embargo, Turkey a 7-day embargo, Canada a 2-day embargo, Poland a 1-day embargo, etc.

What's behind the antipathy towards pre-election polls?

The AAPOR statement pointed out that “attempts around the globe to limit the release of survey research” are based on two assumptions.

  • One: That election poll results influence voters.
  • Two: Voters can be misled by false forecasts.


AAPOR called both of these assumptions “false.”

The first of these refers to the classic bandwagon or anti-bandwagon effect. If polls show one candidate ahead, voters could decide to support that candidate solely because of their conformist desires to be one of the crowd (“to jump on the bandwagon”). Or idiosyncratic, deviant voters could decide to vote against the leader solely to make a nonconformist statement. The AAPOR statement concluded that such bandwagon effects have not been supported by available scientific evidence.

A more sophisticated version of this argument deals not with voter preference, but with voter turnout. Pre-election polls can dramatically affect efforts to get certain types of voters out to the polls.


There is little question that turnout is being affected by pre-election polls in Massachusetts. Millions are being spent on get-out-the-vote (GOTV) efforts. These dollars probably would not have been spent absent polls showing the tightening race. Democrats may end up voting in higher numbers than they would have, given the prospect of a Republican victory. Republicans may end up voting in higher numbers than they would have, given the prospect of a Republican victory.

My position on both of these "voters are affected by pre-election polls” arguments (see here for an expanded discussion) is straightforward. I don't see the problem. If a voter makes up his or her mind based on polls, so what? If a voter is energized to turn out because of pre-election polls, so what? Pre-election polls do nothing more that summarize what one's neighbors are thinking. Knowing how one’s fellow citizens are going to vote is a perfectly legitimate input into one's own decision making process. I don't think we would want to ban asking one’s neighbors about the candidates.

There is a great deal of election information, noise, and flotsam floating about in an election environment. All of it, in theory, could affect voter preference and turnout. Pre-election polls are one of these elements. In many ways taking into account poll data is preferable to being swayed by a random comment made by a coworker. Or the content of a 30-second attack ad.
In short, polls are a legitimate source of voter information. One that voters should be able to take into account if they so choose.


The second assumption noted by AAPOR is that voters will be misled by bad polls. This is certainly a possibility. The majority of polls are accurate. But occasionally the portrait of a race raised by polls is not accurate. It’s not frequent, but it can happen. That’s the chance one takes in a democracy. And that’s the chance one takes in a free speech environment. Bad polls can occur.

Other types of bad or misleading information can also be out there in the pre-election environment. Voters can be misled by 30-second ads and soundbites, by inaccurate word-of-mouth assertions overhead at restaurants or at work, and by all sorts of less-than-perfect input.

Voters are most likely sensitive to the fact that not all information they encounter in an election environment is accurate. Voters are, no doubt, these days more sophisticated about this possibility than they have been in the past. Among other things, I believe voters know that pre-election polls are not infallible. They almost certainly know that every word in a 30-second attack ad is not to be taken seriously. Voters need to be cautious and careful. But banning certain types of information is not the best way to make this happen.

I have also heard arguments that an emphasis on horse race polls takes attention away from candidates’ positions on vital issues. That headlines and news leads focusing on the horse race push out headlines and news focusing on candidates' positions on the issues.

That may be the case. This, it would seem to be, is more the fault of news gatekeepers than polls themselves. If one doesn’t like what media outlets emphasize in their news coverage, one should lobby these gatekeepers to change. Not attempt to outlaw various sources of information.

Certainly much information about where candidates stand on the issues is widely available. I think gatekeepers headline polls so frequently because they assume that’s what readers and viewers are most interested in. One way to change that is to change what voters are interested in. Not to ban polls.

The current election in Massachusetts is an example of what I have called a “continuous” election. As opposed to a “surprise” election -- which would occur if no one had the slightest idea what was going on or who would win prior to Election Day.


What would happen if we did indeed have a “surprise” election? One with no pre-election polls. And no way of knowing who was ahead or any indication of what voters were thinking about the candidates?

For one thing, there would be attempts to characterize the race, regardless. Long before modern polling, journalists and others concocted straw polls to try to get a handle on the status of the race. Absent polls, journalists would stop people in the streets, attempt to judge relative sizes of crowds, query political insiders, and use whatever other techniques they could come up with to characterize the race. None of which would be reliable or valid.

I sometimes think there is a genetic basis for humans' strong desire to know who is ahead in an election. This drive is not going to be obviated by outlawing pre-election polls. Journalists and others will attempt to characterize the status of a race regardless. Better to do this with neutral scientific polls than with guesses.

Also. Regardless of the status of public polls, campaigns are going to conduct internal polls. Which gives candidates a serious advantage over average citizens if there are no public polls. Campaigns in this scenario would be making shifts and changes based on reasons unknown to the public. Better in the current situation in Massachusetts that the entire world is aware of the status of the race than just the insiders who have the wherewithal to conduct their own private polls.

Final point. The objective of elections is to elect the best possible candidates. Monitoring how voters are viewing the candidates as the election proceeds, I believe, helps maximize that process. Voters who are aware of who is ahead and who is behind -- before Election Day -- will adjust their voting intentions accordingly. That's usually a good thing. If Scott Brown does win the Massachusetts Senate race (I am writing this before election results are known), it will not be a surprise to most Massachusetts voters. The voting is taking place in an environment in which most voters know that Brown has a real chance of winning the election. Their vote, therefore, has ratcheted up to a new level of consequence. That's a good thing. Certainly better than if voters were totally in the dark. As would be the case if this were a surprise election.

1 comments:

Anonymous said...
January 19, 2010 at 1:47 PM  

In a column in this morning's Washington Post E.J. Dionne pointed out that the last Politico showed Brown up 52 percent to 43 per cent, getting 26 percent of the black vote, 77 per cent of the Hispanic vote, and having a two-to-margin with voters under 30. That contradicts what we know about the past preferences of those groups. If those numbers prove to be dramatically off, it would seem to bring the quality of the polling into question, and the possibility that voters were mislead, does it not?

Post a Comment

Comments are moderated by Gallup and may not appear on this blog until they have been reviewed and deemed appropriate for posting.

Copyright © 2010 Gallup, Inc. All rights reserved. | Terms of Use | Privacy Statement