YOU would be excused if you thought that opinion polls were utterly useless, following the election results last Sunday.
Not only did they get the order of the candidates wrong, most failed to predict the winner.
No published poll before February 17 predicted DIKO candidate President Tassos Papadopoulos to finish third, while only one major poll actually showed DISY candidate Ioannis Kasoulides first.
Pollsters are in disagreement as to why they got last Sunday’s results so wrong.
Actually, they disagree about whether they really did get it wrong, and what the limitations of polls are.
Olympios Toumazos of RAI Consultants, the company which ran polls for Phileleftheros and Mega TV, had this to say:
“The basic attribute of polls is that they show trends, and I think that the trends shown throughout the election period were correct.
“These were: Papadopoulos being in the decline ever since Christofias announced his candidacy, the subsequent constant rise of Christofias, and the meteoric rise of Kasoulides’ percentages in the two weeks before the first Sunday.”
“However, where polls have failed is in predicting the absolute percentages of the candidates,” he said.
This does not carry any further repercussions for the validity of polls, he insists.
Instead, “this is a classic case of the ‘exception which proves the rule’,” Tomazos opined.
So what produced last Sunday’s result?
“Two things: the so-called ‘flying vote’ and ‘vote concealing’,” he said.
“The former was not part of the polls and most people flown over voted either Kasoulides or Christofias, therefore leaving Papadopoulos on the back foot.
“Vote concealment is something that we have learned to expect from voters from the Left, and thus accommodated it in our results.
“However, this year voters from the Right concealed their votes and this had a deleterious effect on the percentages of Papadopoulos.”
Toumazos modestly admitted that, “the experience has made us wiser.”
Christina Kokkalou of Insights Market Research, which had conducted three polls per week for Simerini, gave her own take on the results.
“During the last week prior to the elections, we conducted daily polls for Kypros Chrysostomides’ Front for the Restructuring of the Centre.
“These showed Papadopoulos’ percentages declining since Wednesday February 13; this continued into the election weekend,” she averred.
“This was because of the greater momentum created by party loyalty in favour of Christofias and Kasoulides.”
Party loyalty was harder to come by in the parties which had backed Papadopoulos.
The lagging mobilisation rates of EDEK and the European Party was not reversed, with a quarter of the voting strength of each of these parties opting for choices other than Papadopoulos.
Loizos Constantinou, the analyst in charge of polls at Kypros Chrysostomides’ legal office, claimed that, “since Thursday February 7, the mobilisation rates of these two parties were in freefall, while DIKO’s mobilisation froze on Monday February 11.”
Kokkalou added that, “the abstention from the over 60s with which Papadopoulos was popular with (half of the over 60s who were going to vote for Papadopoulos did not go to the polling booths), and the new voters swung the result against the incumbent.”
“Finally, Papadopoulos got 83 per cent of DIKO voters on his side, while nine per cent chose Christofias and five per cent went for Kasoulides.”
When the result is too close to call, are polls rendered ineffective?
Toumazos thought so: “polls give a range within the margin of error.
“When ranges overlap, then predictions are hard to make.”
A spokesperson from Noverna, who runs polls for Politis, said the same: “polls yield results which are within a particular confidence level.
“If confidence levels overlap, then there is no more that can be done.”
Kokkalou was adamant that pollsters could go ahead with an accurate analysis of their gathered data.
“Polls can be accurate, and margins or error can themselves be analysed and minimised according to the capabilities of particular polling companies”, she said.
Constantinou pointed out another twist in the tale: “When there are not enough undecided voters, as was the case in these elections, the elections become a zero sum game: my loss is your gain.”
For example, if a candidate falls from 32 per cent to 30 per cent, then his rival may rise from 30 per cent to 32 per cent, accordingly.
Exit polls also seemed to make a hash of it on Sunday.
None could agree either on the order or who would win, while their results were obfuscated by margins of error.
Kokkalou, who ran Sigma TV’s exit poll, claimed that the discord shown in the results of the various exit polls, “was no light matter and those responsible should be held accountable.”
“Why did Sigma’s exit poll which had a sample of 960 people get it right, while CyBC’s touted 9,000 strong sample get it wrong?” she wondered.
Constantinou said that it was the quality of the sample which mattered, and not the size of it.
“Getting the correct sample is crucial: it must be balanced and truly representative of the electorate”, he argued.
Constantinou added that, “the advantage that exit polls have over previous polls, besides their privileged timing, is that those sampled are inclined to be more honest.”
Toumazos was hopeful that it would be easier to approach the actual results this Sunday.
“This is because we have only two candidates, we can weigh the results based on last Sunday’s votes and we have also learnt a lot from last Sunday.”
Toumazos maintained, however, that, “predictions will still be difficult.”
Electoral analyst Christoforos Christoforou ventured an explanation as to why Papadopoulos’ candidacy failed to gain any last-minute momentum.
He argued that, “various proactive measures taken by the President acted as a boomerang, since people perceived them as an indication of nervousness and electioneering, more than anything.
“The array of social benefits, the petrol cuts and the 142 last ditch promotions in the National Guard instigated by Papadopoulos did more harm than good to his election prospects.”
