Exit polls are also used to collect demographic data about voters and to find out why they voted as they did. Since actual votes are cast anonymously, polling is the only way of collecting this information.
Exit polls have historically and throughout the world been used as a check against and rough indicator of the degree of election fraud. Some examples of this include the Venezuelan recall referendum, 2004, the Ukrainian presidential election, 2004, and the U.S. presidential election, 2004.
Like all opinion polls, exit polls by nature do include a margin of error. A famous example of exit poll error occurred in the 1992 UK General Election, when two exit polls predicted a hung parliament. The actual vote revealed that Conservative Party Government under John Major held their position, though with a significantly reduced majority.
A more fundamental statistical problem with exit polls is selection bias. The polls, though typically much larger than regular opinion polls, still sample only a small fraction of voters. In a heterogeneous population, sloppy selection of the sample can tilt results to any direction. This pitfall can be avoided if the polling organization is competent enough. However, there are problems more inherent to the nature of exit polls. Since the clients (the media) want to publicize results as soon as the real polls close, exit polls must close a few hours earlier. Therefore, late-hour voters are not sampled at all. Some constituents tend to vote early, for example the elderly and stay-at-home mothers, and are oversampled in the exit polls. Other constituents tend to vote late and are under-sampled. This may be the explanation for the 2004 US presidential election discrepancy between early exit polls indicating a Kerry victory, and the final outcome. Additionally, voters may be more or less willing to participate in the exit polls, or more or less willing to sabotage the poll by providing a false vote, depending on their political tendency. This is known as nonresponse and response bias, respectively.
Widespread criticism of exit polling has occurred in cases, especially in the United States of America, where exit-poll results have appeared and/or have provided a basis for projecting winners before all real polls have closed, thereby possibly influencing election results. In the 1980 U.S. presidential election, NBC predicted a victory for Ronald Reagan at 8:15 pm EST, based on exit polls of 20,000 voters. It was 5:15 pm on the West Coast, and the polls were still open. There was speculation that voters stayed away after hearing the results. Thereafter, television networks voluntarily adopted a course of not projecting the presidential victor until after polls closed in the West, Hawaii and Alaska excluded. In the 2000 U.S. Presidential election it was alleged that media organizations released exit poll results for Florida before the polls closed in the Florida panhandle.
Leaks of early exit poll figures for the 2004 presidential election, mainly via the Internet, appeared to indicate a victory for John Kerry. The discrepancies between the exit poll data and the vote count that were outside of the margin of error, coupled with irregularities in the election which seem to explain the discrepancies and what many perceive as evasive tactics by the polling companies, have shed doubt on the legitimacy of that election amongst political activists and some government officials. (See 2004 United States election voting controversies for more detail.) Effective from the 2006 election cycle, in order to minimise controversies created by exit polls, point persons of the different networks are to be secluded from outside contact until such a time it is determined it is safe to release the polls.
Some countries, such as the United Kingdom or Germany, have made it a criminal offence to release exit poll figures before the polling stations have closed, while others, such as New Zealand and Singapore, have banned them altogether. In some instances, problems with exit polls have encouraged polling groups to pool data in hopes of increased accuracy. This proved successful during the 2005 UK General Election, when the BBC and ITV merged their data to show an exit poll giving Labour a majority of 66 seats, which turned out to be the exact figure. This method was also successful in the 2007 Australian Federal Election, where the collaboration of Sky News, Channel 7 and Auspoll provided an almost exact 53 percent two party-preferred victory to Labor over the ruling Coalition.