Electoral processes, being the principal mechanism of civil expression, are the object of thorough observation. Several methods, all backed up by the power of statistical inference, have been developed to monitor elections and to contrast their end results in order to detect anomalies. Among the most commonly used methods we find exit polls, quick counts and parallel vote tabulation.
Exit polls, a mechanism that holds a privileged status for the media, political parties and international watchdog organizations, have been used and are still used in most countries around the world. Despite that, in theory, they allow for a quick and precise estimate of the results of an election, experience shows that they’re not always reliable.
First, as with every poll, they have an error margin that varies according to the statistical sample used. That’s why, depending on how close an electoral race is, this error margin can become a serious hindrance. A good example is what took place during the UK parliamentary elections of 1992. Two exit polls, carried out by networks BBC and ITV, predicted a victory of the Labour Party over John Mayor’s Tories. After the official count, it was shown that these polls were wrong, and that the difference of votes between the parties was within the margin of error, which generated confusion and a lot of tension.
Another problem with these polls during elections is the selection of a sample. Selecting a sample that is small enough for the poll to be viable, but that at the same time is representative of a heterogeneous society, is an art in itself. At the end of the 2004 Venezuelan recall referendum, the NGO SÚMATE questioned the results presented by the electoral body, having as their only proof an exit poll carried out by their volunteers, one that showed a 30% difference from the official results. International watchdog organizations such as the Carter Center and the OAS played a key role legitimizing the official result and discarding that exit poll. Further studies proved that the mistake was due to the sample used, and the over representation of polling stations where the voters were more willing to voice their opinion.
During the 2004 United States presidential election this phenomenon came into play as well, where voters who favoured one option were more prone to divulge their preference, thus skewing the statistics. As the Election Day went on, several polls showed a slight advantage for the Democratic candidate John Kerry. When voting stopped, and to the surprise of many, it was shown that George W. Bush had been re-elected. In spite of serious allegations of fraud, the main problem was that those voters who favoured then-president Bush answered less often to pollsters than those who favoured Kerry. This mistake is known as no answer.
A frequent criticism to exit polls is that, since they are meant to put a finger on the pulse of events and offer projections before the day is done, they don’t take into account those voters who go to their polling stations by the end of the day. Therefore, many electors may not have yet cast their vote when the poll is taking place. This can become a problem when there are supporters of a given choice that have a strong tendency to vote in the afternoon. Besides, the early publication of results may influence these people, discouraging them or driving them to vote for the winning trend. In the 1980 United States presidential election, NBC broadcast estimates that had Reagan as the winner at 8:15 Eastern Time; it is speculated that this dissuaded voters on the west coast to go and vote, even when the polling stations were still open.
These inconveniences have made that countries like the UK, Venezuela and Germany forbid the publication of exit polls before official results are announced.
Although we don’t mean to diminish the power of statistics, and this is a subject that the authors of this blog are not very well-versed, we believe that exit polls have proved in numerous occasions how susceptible they are to error. Therefore, mentions or broadcasts made by the media about the exit polls must include at least the technical sheet and their use must be rigorously regulated by the electoral authorities so that the Election Day can unfold without problems.