A fortnight before Turkey votes, Yes votes soar high

2017 referendum

The Yes campaign in Turkey’s executive presidency referendum is cruising to a comfortable victory next month, a survey released to this website says.

The poll, by mobile research company Qriously, found more than three-fifths (61 percent) of Turkish voters will support plans to abolish the position of prime minister and grant sweeping executive powers to Recep Tayyip Erdoğan.

But it also found nearly one-third of likely voters (29 percent) had still not up their minds – a remarkably high proportion, considering the referendum is barely two weeks away.

The survey of 3,418 adults in Turkey was conducted in Turkish between Tuesday 28 – Thursday 30 March 2017 using Qriously’s platform for smartphone app surveys.

The survey also found:

  • The split in the MHP is evenly balanced, with nationalist voters not opposing the executive presidency proposals in sufficient numbers to deny the Yes camp victory;
  • The “AK Party generation” – voters aged 18-44 – overwhelmingly backs a Yes vote out of loyalty to President Erdoğan;
  • Reflecting the city’s traditional position as an electoral microcosm of Turkey, voters in Istanbul strongly back Yes;
  • No voters were more likely to say they understood the constitutional reform in detail, while those who knew only the broad ideas were likelier to vote Yes.

Undecided uncertainty

The headline percentages of the poll were YES (43.6), NO (27.4), UNDECIDED (29.0).

These findings will come as a surprise to those, this website included, who track publicly-available Turkish polling. There are very few polls out there that have produced a YES result above 60 percent: all of those were from pro-government pollsters, and none were published this month.

 

You could also take a dim view of the very high proportion of undecided voters, because it creates uncertainty over the final result. But Qriously says its experience from previous polls suggests the Turkish undecideds are likely to break in a similar fashion to the rest of the population – and that there was a similar level of indecision before last year’s Italian referendum.

That produces a YES (61.4), NO (38.6) result if the Turkish referendum was held yesterday, the day the fieldwork was completed. There are plenty of provisos on relying on a single poll and I discuss these at the end of this analysis.

Evenly divided MHP

This poll found that, by and large, party loyalty reflects voters’ referendum choice. The Yes camp is politically homogenous: a crushing majority (82.1 percent) of Yes voters would support the AK Party in a general election.

 

The No camp has more of a cross party appeal, although not in proportions significant enough to deliver it victory. The key party is the MHP, which has been bitterly divided by leader Devlet Bahçeli’s decision to endorse the Yes camp.

Some commentators have suggested as much as four-fifths of grassroots supporters back a No vote in spite of their leader – which will be crucial to that side’s victory – but this poll suggests the MHP divide is much more evenly balanced between the two referendum choices.

Youngsters out in force

Young people and those approaching middle age are the likeliest to support a Yes vote, this survey suggests, with the 35-44 age bracket most strongly in favour.

Support for No is higher among older people. That’s anyone born before 1972, which just happens to be anyone in living memory of the last successful military coup.

No-one in Turkey below the age of 36 has voted in a national election that the AK Party did not win. For all the talk of government authoritarianism and crackdown on the opposition, it’s easy to forget that the party has delivered a massive increase in living standards and prosperity since it came to power at the turn of the century and, for many voters, this is what still matters.

It is perhaps no coincidence that the 35-44 age bracket, which came out stronger in this poll for Yes than any other, is the one that will have reached political maturity during Turkey’s political and economic turmoil of the 1990s.

Many of them will have welcomed the relative stability delivered by AK.

Regional differences

“He who takes Istanbul takes the country” is the sage mantra of any Turkish election, and this poll suggests Yes are on track to do precisely that.

The city appears to be headed for YES (59.8), NO (40.2), which explains why CHP leader Kemal Kılıçdaroğlu told CNN Turk last week that they will spend many of the campaign’s final days canvassing the city’s districts.

The most striking regional result is southeast Turkey – YES (68.4), NO (31.6) – where this poll suggests reports of the pro-Kurdish HDP’s dwindling influence appear to be true. Other regional differences are less striking: western Aegean and Thrace, both solid CHP areas, are firmly in the NO camp, while some of Yes’s strongest results come from the pro-AK Party Black Sea.

Qriously gauges a respondent’s location through their smartphone’s GPS reading, so it is entirely possible some users were not situated in the part of the country they will vote.

Education

A voter’s level of education is directly linked to their preference in this referendum, Qriously found.

Less educated voters – those who completed compulsory primary education or high school, but nothing more – were overwhelmingly in favour of Yes, while support for No was higher among university graduates.

As this website has previously written, the proposed constitutional changes are very complicated. No voters were also more likely to say they understood them in detail, whereas Yes supporters were more likely to say they only knew the broad ideas of the reforms.

A thought on Qriously

Many readers of this website will not have heard of Qriously before. I helped them with the wording of their questions for this poll, but did not run the fieldwork for them. Their methodology is best explained by themselves here.

This poll has a clear advantage over most publicly-available Turkish polling in that its methods are at least transparent. Qriously polls smartphone users by replacing an advertisement in an app with a series of questions. It’s an opt-in survey, meaning users dismiss it and carry on using their app if they wish.

As with all polls there are drawbacks. I tapped out opinion research and communications consultant Christine Quirk on this one, and she told me a major drawback in this case is in the demographics that cannot be targeted: internet users using a conventional desktop computer; anyone who has a cheap, “non-smart” mobile phone; and anyone who doesn’t own a mobile phone at all cannot be reached.

Mobile phone penetration is high in Turkey – 96.9 percent of households have at least one, according to government data – but the proportion of smartphone users is lower, at around 50 percent.

All these are factors that could weigh against a robust poll result.

Nonetheless, the company is confident they have a sample that broadly reflects Turkey’s demographic make-up. Qriously would argue its referendum methodology is tried and tested, having successfully called the UK’s Brexit and Renzi’s Italian referendum last year, as well as this month’s Dutch election. I leave the judgment to you.

A final note: this is not a one-off poll. There will be two further waves of polling between now and referendum day on 16 April, giving me – and you – ample opportunity for comparison.

Qriously worked with journalist Michael Daventry to develop the survey questions and gave his website James in Turkey exclusive access to the weighted results.

The poll surveyed 3,418 Turkish-language smartphone users based in Turkey between Tuesday 28 – Thursday 30 March 2017. Results were weighted on gender, age and region to be nationally representative of the Turkish adult population based on the latest available census data.

The cost of the poll was met entirely by Qriously.


Corrections: This article was updated on 31 March to credit Christy Quirk for the caveats on the poll’s reliability and to correct Graphic 2: Evenly divided MHP, where the vote proportions for HDP, “Other party” and “None” were accidentally shown as the same figure.

Last modified: Saturday 1 April 2017