Pollsters Fail Again as Erdogan Landslide Surprises TurkeyBy and
All surveys underestimated AK Party vote in Sunday's election
Similar flaws from UK to Greece lead to review of poll methods
The fact that Turkey’s AK Party came in first at Sunday’s parliamentary election wasn’t a surprise. Its landslide victory was, upending assumptions about Turkey’s political outlook and marking the latest failure by pollsters to gauge voter sentiment around the world.
Turkey’s survey companies’ missed a dramatic swing over the past five months, as AKP roared back into public favor after losing ground in an inconclusive June vote. The average of five surveys in the month before balloting had the party winning 41.5 percent of votes. In the event, it got almost 50 percent, adding to a series of high-profile polling misfires in the past year from Greece to Israel and the UK.
The bust had implications for markets: While part of the Monday rally in the currency and stocks was a bet on stable government, it also reflected the fact that investors weren’t expecting such a decisive win. The miss sparked debate about whether surveys should be trusted at all, forcing the industry to launch an inquiry to figure out what exactly went wrong.
“Unusually large error margins of the electoral polls in Turkey seem to be following a bizarre global trend,” Yilmaz Esmer, a professor of political science at Istanbul’s Bahcesehir University, said by e-mail Tuesday. “The most important sources of error in any survey are bad samples, high non-response rates, response errors, interviewer errors.”
It’s still “so puzzling” that all surveys managed to underestimate the AK Party’s votes with no publicly available poll overshooting, he said.
Faruk Acar, the head of research firm ANDY-AR, said Turkey’s pollsters will probably pool their efforts in an inquiry. Part of the solution may be to split the country into a higher number of polling districts to have a better sample of respondents, he said.
Ozer Sencar, the chairman of MetroPOLL in Ankara, who made the most accurate prediction in the previous election in June, put the blame on swing voters who didn’t declare their true intentions at surveys. That’s a version of the “Shy Tory” phenomenon identified by British opinion polling companies as a reason why polls underestimated the Conservative Party’s support in the U.K.
“They had just cast their ballots for other parties in June, and didn’t want to admit that they were wrong and returning to the AK Party,” said Sencar, whose latest prediction for AKP votes was 43.7 percent.
Most surveys may have also missed the impact of escalating violence, as the war between the army and Kurdish rebels resumed in Turkey’s southeast, and Islamic State staged a series of suicide attacks. The polls may well have been accurate at the time, but too much happened to sway voters’ opinion too close to the election for them to capture, said Michael Gunter, a professor of political science at Tennessee Technological University.
The violence and political deadlock after the June vote may have persuaded even strong opponents of the Islamist-rooted AK Party and its founder, President Recep Tayyip Erdogan, to silently shift their political allegiances, Sencar said by phone from Istanbul on Tuesday.
ANDY-AR’s Acar said that in the increasingly polarized atmosphere of Turkish politics, some polling companies were liable to hide their own findings if they diverged from the pack.
In 2007, Konda pollster Tarhan Erdem famously predicted days before an election that the AK Party would win 48 percent. That was way more than any Turkish party had won for decades, and far above other predictions. Erdem was viciously attacked by opposition supporters for being a stooge of the AK Party -- then vindicated by the vote result.
Acar said he came up with an AKP vote of 49.7 percent -- almost exactly the eventual outcome -- in a survey shortly before Sunday’s election, but decided not to publish because the result was such an outlier.
“What if it hadn’t been accurate?” he said. “We are under real public pressure here.”
To continue reading this article you must be a Bloomberg Professional Service Subscriber.
If you believe that you may have received this message in error please let us know.