Election coverage has become almost synonymous with showcasing the latest poll. During an election, national media outlets are overrun with polls showing which candidate is ahead and what topics people are most concerned about.
Polls provide a representation of a population’s current beliefs, attitudes and voting behaviors. They offer a snapshot of the race and are used to show citizens’ views on major issues. Without polls, it can be hard to predict or measure the country’s wants and needs.
Both voters and candidates use poll information to make election decisions, relying on them for information on campaigns. However, not all polls are created equal.
“When and *if* they are representative of a given population, they provide insight to the public and individual campaigns about the probability of various outcomes,” said Coye Cheshire, UC Berkeley School of Information professor, in an email. “Polls can affect voting behavior (such as turnout), and this can occur regardless of whether the polls are accurate or not.”
‘Significant impacts’ on campaigns, less influence on voters
Polls are much more likely to influence a candidate’s campaign decisions than a voter’s behavior, according to campus political science and public policy professor Henry Brady.
Candidates will use polls to determine which issues to emphasize in campaigns and which populations to prioritize in their outreach, added Laura Stoker, campus political science professor emeritus.
“(Polls) play an important role; they inform candidates and campaigns on the effectiveness of their messaging and strategy,” said Jennifer Shanoski, Merritt College chemistry professor and Berkeley Unified School District, or BUSD, school board candidate, in an email. “(They) can influence how the campaign needs to pivot, if at all, in order to better influence the outcome.”
These types of polls are typically internal polls that are never publicly released, according to Brady. He added, however, that public polls may also be used as an incentive for fundraising: They can show a candidate has a chance, but needs more donations to beat his opponents.
Jonah Gottlieb, campaign manager for the Right to Housing Slate for Berkeley Rent Board, noted the importance of candidates only using poll data to reframe their positions on issues to sound more appealing to voters, rather than them changing positions based on what is “politically popular.” The slate includes Soli Alpert, Vanessa Danielle Marrero, Ida Martinac, Nathan Mizell and Negeene Mosaed.
“Many people also wonder if people change their votes based on polling predictions,” Cheshire said in an email. “I do not believe that there is consistent, strong empirical evidence that people *change* their voting decisions based on polling. However, there are a few interesting psychological effects from polling that *do* appear to influence voter behaviors.”
These effects may include higher voter turnout after polls show a close race and an increase in “bandwagon voting” when exit polls show the favorable candidate winning, Cheshire added. He noted people like to say they helped accomplish a successful outcome, so they are more likely to go out and vote if their candidate is winning.
Polls are also typically more impactful in primary races, rather than “first-past-the-post winner-take-all” races between two candidates, according to Brady.
“Someone who liked A over B over C might vote for candidate B instead of A because A has no chance (according to the polls) of beating C, but B does,” Brady said in an email. “In this case they vote for their second choice strategically to make sure that their least preferred alternative (C) does not win.”
Many Berkeley candidates, however, are not paying attention to polls, as they can cost about $10,000, according to District 1 candidate Elisa Mikiten. Gottlieb agreed, saying polls are not a strategic use of time and resources for local grassroots campaigns. Instead, candidates focus on mobilizing residents to vote through door-to-door campaigning and personal connections.
According to Shanoski, however, professional polls can get a more robust representation of voters’ beliefs and priorities. She added she would like to see more polling for down ballot races.
“Polling is usually done for important races. One reason for polling is to lift up down ballot races,” Shanoski said. “I would love to see more attention paid to small local races, which have significant impacts on the communities of everyday working people.”
Polls do not always match the real world
Most assume the information shared through polls are accurate and representative of the larger population. While this is often the case and pollsters work hard to create a truly representative sample of a given population, there are still sources of error that can influence poll results.
“All of the effects discussed above tend to implicitly assume that polls are actually representative samples of a given population,” Cheshire said in the email. “A poll, just like any survey of a population, is only as good as its sample.”
The biggest problem in creating an accurate poll is nonresponse from those surveyed, according to Institute of Governmental Studies co-director Eric Schickler.
Additionally, polling may not reach marginalized groups, further influencing survey conclusions, BUSD board candidate Tatiana Guerreio Ramos added. Cheshire noted this is especially true for polls conducted via the internet or phone.
“It is probably obvious to most people that if you want to understand the likely voting outcomes in a given town, you should not just poll a group of people at a rally for a specific candidate,” Cheshire said in the email. “While no legitimate polling agency would do something that biased, the reality is that obtaining a random, non-biased, representative sample of a given population is incredibly difficult.”
As a result, pollsters will weight responses depending on the demographics of those who respond to better match the composition of the people who are expected to vote, Brady added.
However, Brady noted two problems with this approach: Weighting survey responses to count for a whole demographic may ignore qualities about those who respond that are not shared by the rest of the demographic population and determining what the weights should match.
“We might want the weights to match the composition of those who will go to the polls, but we don’t know before the election who will go to the polls,” Brady said in the email. “Hence, we must use questions on the survey (Do you intend to vote?) to figure out what the composition of the voting electorate might be, but those questions will not necessarily reliably identify the right people.”
Schickler warned against relying on only one poll for information, as the margin of error can sometimes be large enough to influence results enough to make them meaningless. He noted, however, that if multiple polls — ideally by different pollsters — show the same result, it is an indication that something is happening.
Poll aggregators, such as FiveThirtyEight, or 538, and Real Clear Politics, combine reliable polls to minimize errors and provide more accurate information. As a result, a poll that is “really off the mark” is balanced by other polls in the aggregate, Brady added.
“Polls can be wrong for many reasons, and they have a large margin of error,” Brady said in the email. “Right now, 538 says that Democrats have about 60% chance of winning the Senate and Republicans have a 40% chance — that means that the outcome is pretty close to a coin toss.”
Despite the 20-point spread, polls have skewed Democratic since 2015, according to Stoker. She noted this may be due to discrepancies between survey responders, who tend to be more Democratic-leaning, and people who show up to vote.
This creates a two-point bias towards Democrats, Stoker said. Taking these two points into account will create a more representative poll.
“People are less likely to vote than they say they are,” Stoker said. “People who come into surveys tend to be more democratic and engaged than the categories they are representing.”
Prioritizing the ‘horserace’ over electorate concerns
The prominence of polls in election coverage is likely because they are easy stories to write and enhance the media outlet’s prestige, according to campus journalism professor Edward Wasserman.
However, Wasserman said coverage can be careless, as reporters don’t often go back to the way poll questions are posed, which may introduce bias into the results.
“My overall take on polls is that we want to go back to what polls are for. The reason for polls is to enable candidates to calibrate messaging to better manipulate voters,” Wasserman said. “They are meant to be tools of the campaigns and giving them the prominence that the news media gives them has some serious drawbacks.”
He noted that ideally, the media should provide voters with information on candidates’ positions on issues the electorate cares about.
Polls are “certainly beneficial” in identifying the types of issues that are animating the electorate, according to Brady.
“Indeed, they can often tell us that the electorate has some concerns that we had not recognized,” Brady said in the email. “This has occurred recently with abortion where it has become a much bigger issue than we expected.”
However, most polls focus on the “horserace,” or who is going to win, noted Wasserman.
In addition to creating an “enormous distraction” from information needed to make an informed decision, horserace coverage can also have a depressing effect on voter turnout, Wasserman said. If voters think the election is a foregone conclusion, many people will not want to waste their time voting if they believe it will not have a significant impact based on polls.
“(Polls) focus attention on the relative position of candidates in a campaign, which relegates voters to be a spectator in the campaign,” Wasserman said. “The media focuses too much on who is winning vs who ought to win.”