Are Polls Accurate Predictors of Election Results?

Last updated: February 23, 2017

Polls have become an indispensable way for us to try and determine public mood. However, they’re not always accurate. They’ve been wrong many times in the past, most recently by predicting Obama would win primaries Clinton ended up taking. So do they tell us who’s going to win, or are they nothing more than an educated guess?

Are Polls Accurate Predictors of Election Results?
Yes because...

Research companies worldwide, such as YouGov, which provide data and statistics for election polls provide a fair representation of the population, and therefore a fair representation of public opinion.

Assuming that the company in question is a reputable one, hence the example listed above, it is possible to find out how many people YouGov assign to their panels. In the UK this number totals 200,000. A sizable portion of the voting population, taking into account low turnout numbers in the UK. A wide sample, including different locations and ages, means that the resulting poll will provide a certain indicator of public opinion, and though this may not be exact it is still valuable nonetheless. It would be impossible to say if a resulting poll in America said that Obama led by 4% in the polls, however, if it said that he led by 98% across a nationwide sample of a suitable number in relation to America's population, then this would signal a massive problem for the opposing candidate.
No because...
We already know they are inaccurate. The past five polls have significantly exaggerated Labour support - most notably in 1992 when they consistently, and right up to polling day, predicted a defeat for John Major who went on to win the election with the largest number of votes ever polled by a party in a UK election. The suggestion that polls are accurate because they attempt to cover a fair representation of the population is a non-sequitur. The sample may be representative, but even if it is 100 percent so (which seems unlikely) that doesn't mean it will accurately reflect how the sample will vote (particularly in an election with such a large number of undecided voters).

Are Polls Accurate Predictors of Election Results?
Yes because...

They are a good indicator of the closeness of the race

The current polls in America show a close race between Republicans and Democrats. Voters can be assured that their vote will count, which is possibly the most important result that these polls have. Inversely, this may make people turnout who were possibly thinking about abstaining.
No because...

Are Polls Accurate Predictors of Election Results?
No because...

The only accurate poll is the exit poll, as evidenced through history

In 1948, the American polls and the American newspapers had branded Harry S. Truman a loser. The Chicago Tribune even ran the story 'Dewey defeats Truman'. All of this was the result of inaccurate polls. The only accurate polls in America are the exit polls, and even then these can come in early, as they did in the Gore Bush standoff over Florida. Polls do not take into account swing voters, and assume that the answers received by participants are solid and firm. Therefore, poll results can be taken for granted by those parties that the polls favour, which may be the reason why parties that originally lead in the poll lose, because they alienate their original supporters through negligence or overconfidence.
Yes because...

Are Polls Accurate Predictors of Election Results?
No because...

Voting is not a predetermined decision

Despite many American citizens being party members, seen in the high turnout of the Caucuses, there are still many registered voters who have no concrete affiliation to one party or the other. The voter, not voting on the party line, has up until the moment of punching their ticket to decide who they will vote for. Obviously, a poll conducted a week before an election is immediately out of date and loses any relevance as an indicator of victory, especially in the final week of elections, when PR and spin are at their highest in order to sway swing voters.
Yes because...

Continue the Debate - Leave a Comment

4 Comment threads
0 Thread replies
Most reacted comment
Hottest comment thread
4 Comment authors
Notify of
Daniel O\'Leary

Speaking from Canada, as a Canadian, the polls never had Prime Minister Justin Trudeau winning the election in October 2015, never mind winning a majority as he did.
Polls are amazingly inaccurate, and politicians, most politicians , refer to polls when they are favourable to them but are quick to dismiss them when they show them behind or losing.
Polls IMO should be abolished, they are of no consequence.

mike mclean

all that these polls do is to tell us what the people in the sample likely think, assuming they are being honest. With response rates of 8 percent on average, it is implausible that the tiny samples these pollsters extract represent the views of the population as a whole. Some of the methodologies these pollsters publish are so flawed to make their data meaningless. Its sobering that on the morning of the Brexit vote, an “influential” poll predicted a “remain” win by 12%. Those who are interested should study the Major vs Kinnoch pre-election polling and the subsequent election result in the UK. The response of the pollsters was that those polled must have been telling lies :)


For the Polls in the United States, if you try to see where or what the data is that led to the poll outcome you often find the sample spaces are far smaller than the required “n” value for a 95% confidence interval based on the variation. It amazes me that most American polls are collected and run by Journalism companies not using actuary scientists (statistical scientists) to determine a correct outcome. Although, as any intelligent viewer can tell, American news agencies all have their slant (cnn far left, fox far right). It is probably why I tend to read American news on the BBC.