An exit poll is a survey conducted with voters as they leave polling places after casting their votes, designed to estimate how different groups voted and to investigate the characteristics, motivations and attitudes associated with voting behaviour. Unlike a pre election poll, which asks people before voting what they intend to do, an exit poll asks people after they have voted what they say they actually did.
Because the election has already occurred for the individual respondent, an exit poll can examine reported vote choice alongside demographic characteristics, issue preferences and other information that is difficult to obtain from the official vote count alone. Exit polls are therefore principally research instruments for analysing voter behaviour and providing early estimates or explanatory information, rather than ordinary surveys of general political opinion. Their timing makes them fundamentally different from conventional pre election polling.
The sampling design of an exit poll is specialised because respondents are encountered at or near polling places rather than being selected from a general population telephone list or online panel. Pollsters select polling locations or other voting sites according to a sampling design and then approach selected voters according to specified procedures. The objective is to obtain a sample that can be used to estimate the behaviour of the broader electorate.
The design must account for geographical variation, turnout patterns, different types of polling locations and the possibility that voters at different locations or times of day may differ systematically. The people willing to complete an exit questionnaire may also differ from those who decline. Consequently, an exit poll remains a sample survey rather than a complete record of every voter.
A major analytical value of an exit poll is its ability to connect vote choice with respondent characteristics and reported attitudes. Official election results normally establish how many votes each candidate or party received in a particular area, but they do not necessarily explain which demographic groups supported each option or which issues were associated with those choices.
An exit survey can ask respondents about age, education, occupation, gender, race or ethnicity where appropriate, political identification, important issues, perceptions of candidates and other characteristics. Analysts can then examine relationships between these variables and reported vote choice. Such relationships are estimates derived from the sampled respondents and should not be interpreted as perfectly precise measurements of every individual voter.
Exit polls can also be used to investigate why people say they voted as they did. Questions may ask respondents which issue mattered most, whether they approve of the government's performance, whether they believe the country is moving in the right direction, or whether particular political concerns influenced their decision. These answers allow researchers to study associations between attitudes and reported electoral behaviour. They do not necessarily establish causation.
If people who support a particular candidate also report concern about an issue, the data can show that the characteristics occur together within the sample, but it does not by itself prove that the issue caused the vote. Questionnaire design, response options, timing and the wording of questions remain important sources of measurement variation.
Exit polling has an unusual relationship with official election results because the official count eventually provides the authoritative numerical record of votes actually counted, whereas an exit poll is an estimate based on respondents. Early exit poll estimates can therefore change as more votes are counted and as statistical models incorporate additional information. Differences between an exit poll and the final official result do not automatically demonstrate that voters lied or that the poll was improperly conducted.
Sampling error, nonresponse, coverage of voting locations, weighting, questionnaire design and other methodological factors can all contribute to discrepancies. The final election count and the survey answer are also different kinds of information, because one records counted ballots while the other records responses from a sample of voters.
Exit polls are particularly useful for studying subgroup patterns because they can estimate how voting behaviour varies across categories of voters. For example, analysts can compare reported support across age groups or examine whether voters who identified a particular issue as important differed in their reported vote choice. However, subgroup estimates generally involve smaller numbers of observations than the full sample and therefore have greater statistical uncertainty.
AAPOR warns that margins of sampling error for subgroups can be larger than the margin for the full survey and that very small subgroups should be treated cautiously. This principle applies directly to exit polling because detailed cross tabulations can quickly divide the original sample into relatively small groups.
Exit polls should not be confused with election result projections, voter surveys conducted before election day, or official vote counts. A projection may combine partial vote counts with statistical information to estimate an eventual result, whereas an exit poll is a survey of voters. A pre election poll measures intention or expectation before voting, whereas an exit poll asks after voting. An official result is based on ballots counted under the applicable electoral rules, whereas an exit poll is based on sampled respondents and statistical estimation. These distinctions are important because the same headline percentage can have very different evidential meanings depending on how it was produced.
The main limitation of an exit poll is that it still depends on people being reachable, willing to respond and accurately reporting the requested information. Some voters may decline to participate, some may misunderstand questions, and some may give inaccurate answers. Pollsters therefore use sampling procedures, weighting and statistical adjustments to reduce known problems, but no adjustment can guarantee perfect reconstruction of the entire electorate.
Exit polls are most informative when their methodology is transparent and their estimates are considered alongside official results and other evidence. Their principal contribution is not simply to guess the final vote total but to provide structured information about who voted, what they reported voting for and which attitudes were associated with their choices.
