Why Pollsters Fail, Election Surprises From UK To Canada To Argentina
The increasing "unreliability" of voters in a digital age are consistently knocking polling experts off their mark.
What do Argentina, Canada, Israel and the United Kingdom have in common? All have had national elections this year with results considerably at odds with voter poll predictions.
In Canada and the United Kingdom, this could in part be explained by electoral systems: Where there are single-member constituencies, small variations in vote numbers can produce dramatic changes in a party's parliamentary representation.
In the last elections in Britain, the UK Independence Party (UKIP) won 12.6% of ballots, and the Scottish National Party (SNP), 4.7%. Nevertheless UKIP votes were scattered across a large number of constituencies and those of SNP concentrated in Scotland, which finally yielded UKIP one seat in the Westminster parliament, and 56 for SNP.
And yet quirks of electoral systems do not explain everything, since pollsters not only failed to estimate the number of seats respective parties would win, but also the distribution of voting.
There are at least two possible explanations. The first is that while all registered voters are eligible to constitute the sample population used in polling, not all of them will be equally motivated to vote. For that reason in the United States, polls distinguish between answers given by all registered voters, and those of "likely voters."
To establish their predictions, pollsters resort broadly (though no exclusively) to checking questions like, whether or not the respondent knows where he or she must vote, or the number of times they had voted previously.
The number of actual Democratic Party voters for example is generally fewer than its potential voters, as some of the party's support bastions (young people or minorities), tend to vote less often than they say.
Old land lines
This type of difference has apparently been a crucial factor in Argentina's presidential elections. Between primary elections held in August and the first round vote in October, the number of voters increased by two million. The majority of these voted for the center-right candidate Mauricio Macri (this is the case in spite of voting being compulsory by law in Argentine elections).
The second explanation for discrepancies between poll forecasts and election results may be increasing electoral volatility. There are signs in various parts of the world of decreasing party loyalties and the increasing prominence of personalities. That could mean two things, if true: that more and more people will determine their vote in the last days of a campaign, and many will change their minds for strategic reasons.
This seemed to have happened in Canada. The slogan ABC, or Anything But Conservative, may well have caused votes to flow from the New Democratic Party to the Liberal Party, perceived as more likely to defeat the Conservatives.
[rebelmouse-image 27089620 alt="""" original_size="640x481" expand=1]
New Canadian P.M. Justin Trudeau and his wife, Sophie. Photo: Renegade98
All this reminds us that poll firms usually lag behind social and technological changes. Until the late 20th century for example, election polls in the U.S. were carried out using ordinary telephones found in all homes.
Pollsters then noted they were increasingly unable to gauge the youth vote, given the increase in homes inhabited by single young people with no land line, but just a cell phone. When they tried to adapt by calling mobile phones, they found that habitual cell phone users were disinclined to answer their calls, and when they did, less inclined still to answer questions. They also took their numbers (including area codes) to new voting districts if they moved. Calling cellphones is in any case more expensive than the standard telephone.
The bottom line today is that polling firms must be prepared to constantly track and adapt to ever-changing behaviors and conditions. Until then, as we look ahead to the second round of voting to choose Argentina's next president, we may just have to wait until election day.