Explaining Exit Polls
An exit poll in the election context is a survey based on interviews with voters as they leave (or exit) their balloting locations. An exit poll can also be used for other purposes, such as surveying library users or museum visitors.
To estimate the outcome of an election in a particular constituency, a sample of its voting precincts is drawn and at least one interviewer is sent to each sampled precinct. On a pre-selected and systematic basis, the interviewer intercepts people who have already voted in order to obtain an interview, say every 4th or 10th person who comes out of the voting place. The interviewer usually hands the voter a questionnaire on a clipboard and asks him or her to fill it out, fold it up, and deposit it in a survey ballot box. In this sense, an exit-poll interviewer uses self-administered questionnaires, one at a time.
There is a regular series of national exit polls conducted for major national news organizations like the television networks and the Associated Press. The national exit polls were originally developed as separate network data collection operations, beginning in the 1970’s, but then the separate operations were consolidated into a single data collection enterprise in order to save money. The sponsors are the five major networks (ABC, CBS, CNN, NBC, and Fox) and the Associated Press. This began as the Voter Research Service (VRS) and then became the Voter News Service. Since 2003, it has been called the National Election Pool (NEP).
The typical exit poll questionnaire gathers 3 types of data:
- who the respondent voted for in the day’s key elections in a particular jurisdiction (a state or city)
- a variety of attitudes held by the voter
- the demographic characteristics of the voter
The latter 2 sets of information can be used to explain why they voted as they did and what kinds of people voted for each candidate. These exit poll questionnaires are relatively short (typically less than 25 questions) and take less than 5 minutes to complete.
Election night projections of the outcome of key races are based only in part on exit polls. Interviewing people as they leave their voting place overcomes a lot of the problem of respondents’ misreporting whether they voted or not when they are interviewed on the telephone. Analysts have sophisticated statistical models that use a lot of additional data beyond that gathered in the exit polls to make their projections, including incorporating historical voting data such as past turnout and the partisan division of the vote from the sample precincts in which the interviews were taken. Computer models use this information to evaluate exit-poll results as they become available, looking at whether the turnout is much higher or lower than usual and whether the vote is more or less Democratic or Republican than in the past.
In close races, the models also employ raw vote totals, first at the sample precinct level as they become available, and then at the county level for all counties in a state as they become available. However, even under these highly rigorous conditions, rare mistakes have occurred.
What is important to note is that at the close of Election Day, exit poll results are weighted to reflect the actual election outcomes. It is in this way that the final exit poll data can be used for its primary and most important purpose – to shed light on why the election turned out the way it did. That is, exit polls are just as important for the information they gather about the voters’ demographics and attitudinal predispositions towards the candidates and the campaign issues as they are for making the projections reported by news organizations on Election Night.
It is this data that provides powerful explanations, to be examined in election postmortems, for why people voted the way they did - telling us which key demographic groups voted for which candidates or which propositions and why. It is in this way that the so-called “mandate” of the election can be measured (and reported) accurately without relying on the partisan “spin” that the candidates, their campaign staff, and political pundits typically put on interpreting the election outcome.
Finally, exit polls do something pre-election polls cannot do: capture the voting intent of last minute deciders. In 2004, 5% of voters made up their minds on who to vote for on Election Day, according to exit polls. 11% decided who to vote for in the last week before the election.
Exit pollsters face many challenges and operational issues conducting these surveys especially given the changing nature of how and when people can vote (for example there has been an increase in the share of all votes collected through absentee ballots).
Challenges facing exit pollsters
- Early leaks
An exit poll sample is not representative of the entire electorate until the survey is completed at the end of the day. Different types of voters turn out at different times of the day.
- Selective refusal
Selective refusal of voters as they leave the polling place. In recent national and state elections, Republicans have declined to fill out an exit poll questionnaire at a higher rate than Democratic voters, producing a slight Democratic skew.
- Changes in the way people vote
Such as the trend toward absentee voting, voting by mail, and other early voting.
- Operational costs
Training and supervising the hundreds of interviewers, programmers and other support people that are needed to conduct interviewing at precincts around the country make for considerable operational issues facing exit pollsters.
This information was developed by AAPOR as part of a comprehensive online journalism polling course created in partnership with NewsU, a project of the Poynter Institute and funded by the Knight Foundation. The course launched September 2007.
Back to top