SURPRISE TURKISH ELECTION RESULTS PUT SKEPTICAL

UNCLAS SECTION 01 OF 04 ISTANBUL 000180 SIPDIS E.O. 12958: N/A TAGS: PGOV TU SOPN SUBJECT: SURPRISE TURKISH ELECTION RESULTS PUT SKEPTICAL EYE ON POLLING FIRMS REF: 05 ANKARA 4888 ISTANBUL 00000180 001.2 OF 004 Consulate Istanbul and Embassy Ankara jointly authored this cable. ¶1. (U) Summary: The relatively low support for the ruling Justice and Development Party (AKP) in the March 29 local election caught many political watchers off-guard, which calls into question the validity of public opinion surveys in Turkey. Only one published poll * from A&G Research -- correctly calculated support for the ruling Justice and Development Party (AKP) below 40 percent for municipal assemblies nationwide. While many Turkish firms fail to employ best practices employed by established U.S. companies, some Turkish firms are raising the bar. End Summary. -------------- Most Polls Were Wrong on Support for AKP -------------- ¶2. (U) Most published polls overestimated public support for the AKP in the lead-up to the March 29 local elections. Polling companies generally forecasted support for the ruling AKP significantly above -- outside the range of the survey,s sampling error -- the actual support of 38.9 percent for municipal assembly nationwide. Konda, a polling firm that correctly forecasted the 2007 national elections results, overshot support for AKP at 47.9 percent. Other firms had AKP at 45.6 percent (Veritas) and 51.9 percent (Metropoll). Only A&G correctly predicted support for AKP below 40 percent; placing them at 39.8 percent. ¶3. (U) Many polling companies also over-predicted the margin of victory for AKP in the mayoral contests in Istanbul and Ankara. Metropoll and Veritas showed wide double digit margins for the AKP candidate in each city, while the results were much closer at 6-7 percentage points. Veritas widely overestimated support for AKP in many other mayoral races, wrongly calling the AKP candidate in nearly all its reported surveys. Sonar Research also overshot support for AKP for Ankara,s mayoral race. Even A&G overestimated support for AKP in Istanbul, while underestimating support for the party,s candidate in Ankara. While the polls correctly called the winners of the four major races in the local elections -- Ankara, Istanbul, Izmir, and Diyarbakir -- the wide, statistically significant gap between the actual result and the predicted numbers argue against the general quality of polling in Turkey. -------------- Common Polling Problems in Turkey -------------- ¶4. (SBU) Statistical failures of this scale across a number of polling companies suggest that polling methodology in Turkey is weak. Past reporting questioned the quality of Turkish polling, particularly Anar Research and Pollmark; however, one Turkish firm was found to use acceptable survey methods ) Infakto Research (reftel). Emre Erdogan, Chairman of Infakto told us his company deliberately does not conduct many polls, focusing instead on conducting quality surveys. State-INR, the Maryland based World Public Opinion, and the International Republican Institute all have commissioned Infakto to conduct polls in Turkey. Infakto does not do polling for political parties, though it occasionally has done work for individual candidates. ¶5. (SBU) Emre Erdogan, who trained under prominent Turkish survey research academics Ersin Kalaycioglu and Ali Carkoglu, criticized many pollsters for only operating in an election year. These firms often hire inexperienced temporary interviewers. In addition, many firms do not spend resources for quality control and there is little accountability for producing erroneous results, according to Erdogan. ¶6. (SBU) Another typical mistake most Turkish polling firms commit is the use of quota sampling for achieving predetermined targets for the proportion of certain sectors of Turkish society. While this can be an acceptable method, it assumes the polling firm has a perfect understanding of the demographics of Turkish society to set the quota. This is a dubious assumption, given that the Turkish state itself only discovered last year that it had underestimated the number of Turkish voters by several million people. One ISTANBUL 00000180 002.2 OF 004 exception was A&G Research; Chairman Adil Gur told us his firm randomly picked its respondents, without setting out to achieve certain numbers of target groups. ¶7. (SBU) How polling firms manipulate the data is another source of error. The most common mistake is to drop the undecided respondents in a survey, thus proportionally increasing the value of those who indicated support for a party. This assumes undecided voters will act like decided voters at the ballot box, an often erroneous assumption. This mistake can severely distort the accuracy of the poll, since undecided respondents can often account for 20-30 percent of a survey,s total respondents. Other methods of data manipulation can add additional error to a survey; even with the best intentions of the researcher, data weighting can introduce inaccuracies if the fundamental assumptions on which they are based are mistaken. The general lack of transparency of most published polls in Turkey increases the possibility that additional error entered the data after the results were collected. ¶8. (SBU) Despite these problems the general consensus among academics and industry leaders is that polling in Turkey has improved over the past ten years. The growth of foreign direct investment in Turkey has increased the demand for quality survey research. Also, Turkish universities now conduct classes in scientific survey research, compared to the past where students needed to go abroad to learn these methods, according to Emre Erdogan. He also stressed that interviewers can ask more sensitive questions today than they could five years ago. ¶9. (SBU) Still, there is a large gap in quality as some firms continue to conduct shoddy or biased polls. Some industry leaders that we talked with, including from Konda, A&G, Genar, Anar, told us they strive to attain accurate results by expending the resources needed to reach out to all the regions of Turkey, including remote rural areas and poor outer suburban areas surrounding major Turkish cities (Varos). The above firms also hired Kurdish speaking interviewers when conducting polls in predominately Kurdish areas. However, many firms fail to go beyond the major cities or do not strive to randomly select their respondents. ¶10. (SBU) Industry leaders and academics highlighted Konda and A&G as two of the better polling firms in Turkey. Emre Erdogan noted Konda used good methods and was not influenced by politics. During the lead-up to the 2007 national elections, Konda was criticized for producing results that showed strong support for the ruling Justice and Development Party (AKP). Konda,s data proved to be an accurate forecast of the election, as most commentators were caught by surprise by the level of AKP,s victory in 2007. The head of Genar Research, M. Teyfik Goksu told us Konda and A&G both strive to conduct sound polling without a political bias. -------------- Looking under the Hood at Konda and A&G -------------- ¶11. (SBU) Chairman Agir Dir told us Konda has been conducting scientific polls in Turkey for over two decades, following methods used by American polling firms. His firm only conducts face-to-face interviews by trained pollsters. Konda does not focus its business on political polling; instead it conducts numerous regular projects that include marketing and social research. Konda conducted the most comprehensive non-governmental survey of the Turkish people in 2007. The poll, titled &Who are we?8 commissioned by Milliyet newspaper, interviewed 47,000 people in 79 out of 81 provinces in Turkey to gather social and ethnic information on the Turkish people. Konda has since conducted subsequent surveys in conjunction with Milliyet on topics such as religion and ethnicity, two extremely sensitive topics in Turkey. Agirdir told us these surveys could not had been conducted five years ago, noting that Turks are now able to talk more openly on these issues. (NOTE: Some Infakto interviewers conducting a poll regarding Armenian issues were detained by security forces in Eastern Anatolia in 2004. END NOTE.) ¶12. (SBU) Adil Gur, Chairman of A&G, told us his company ISTANBUL 00000180 003.2 OF 004 strives to produce accurate polls because &reputation is the greatest currency for a polling company.8 He stressed A&G does not conduct polling for political parties, noting many foreign investment banks commission surveys with his firm. Gur said his firm only conducts face-to-face polling, arguing that telephone polling fails to get a random sample. Gur also stresses that, for national election polling, his firm uses only trained interviewers that go out to all 81 provinces. Also, he insisted he used Kurdish speaking interviewers in the southeast; however, these people cannot have connections with either AKP or the pro-Kurdish (DTP). -------------- Other polling Companies -------------- ¶13. (U) Other polling companies have not used proper methods and this has affected their ability to forecast election results. For example, Metropoll poorly estimated support for AKP in the 2007 election and the 2009 local election; its latest data showed AKP getting nearly 52 percent of the nationwide vote, 14 percentage points over the actual vote. Metropoll uses telephone polling, a questionable means to achieve randomness because hard line telephone penetration is relatively low in the eastern provinces, according to many polling experts in Turkey. Most other firms opted to conduct the more expensive face-to-face polling for that reason. Furthermore, Metropoll refuses to employ Kurdish speaking interviewers; according to chairman Sencar, Kurdish speakers are unreliable since they are &politically biased8. Reporting from 2005 also found Metropoll,s survey methodology questionable (reftel). ¶14. (SBU) Some polling companies appear to be politically motivated, publishing favorable results in likeminded newspapers. On the final day allowed by law to report opinion polls before the March 2009 election, the Fethullah Gulen supported newspaper, Zaman published a massive poll conducted by VERITAS research, predicting a &crushing victory8 for AKP. This poll uniformly overestimated support for AKP in 16 mayoral contests as well as the overall national vote. ¶15. (SBU) Other polling firms also have clear political sympathies, but their numbers do not always fall according to their prejudices. Sonar research Chairman A. Hakan Bayrakci told us he does not like AKP or Saadet Party (SP) and does not accept commission polls for them because he views them as parties with a demonstrated religious agenda. Sonar,s results before the 2009 local elections were a voice among many other firms that overshot support for AKP. Sonar did underestimate support for AKP in the 2007 national elections, however, Bayrakci failed to provide details of Sonar,s polling methods, despite our questions on the subject. -------------- Polling Companies with Ties to AKP -------------- 16 (SBU) A number of polling firms have close ties with the ruling AKP and/or Prime Minister Erdogan. Both Emre Erdogan and the press reported the Prime Minister has commissioned polls from Ankara Social Research Center (Anar),Pollmark, Denge Research, and Genar. Anar has a long history with the AKP: Interior Minister Besir Atalay used to be Anar,s chairman before entering politics; the current chairman, Erdogan Uslu, has served as an AKP advisor; and Uslu used to share an office in Anar with current President and former AKP PM Abdullah Gul (reftel). Emre Erdogan and the press also reported that Denge,s Chairman, Hasan Basri Yildiz, is the Prime Minister,s cousin. Genar Chairman M. Teyfik Goksu successfully ran as the AKP candidate for mayor for the Istanbul district of Esenler in the local elections. According to Emre Erdogan, Pollmark is close to the AKP leadership. He told us the Prime Minister refuses to use Metropoll ) despite the more religious attitude of the firm -- because Metropoll,s founding members left Pollmark to establish their own completing company. (COMMENT: Despite these firms, ties to AKP, it is unclear if they bias their results in favor of the ruling party. Pollmark, Denge, and Genar, all underestimated support for AKP in their published results before the 2007 national elections. END COMMMENT.) ISTANBUL 00000180 004.2 OF 004 -------------- The opposition,s ambivalence towards polling -------------- ¶17. (SBU) According to Metropoll Chairman Ozer Sencar, other political parties have not shown much interest in using polling data. CHP party chairman Baykal said on a televised interview that he does not trust polls. Sencar told us Baykal chooses candidates based on closeness to the party, despite polling data showing alternative candidates with greater popular appeal. ¶18. (SBU) Comment: Public opinion research has improved in Turkey, yet poorly conducted polling results will continue to flood Turkey,s sensationalistic newspapers. Reliable firms, such as A&G and Konda, appear determined to improve their methods and to start consistently providing accurate forecasts. The demand for accurate poll results ) especially from international businesses -- will likely increase accountability and thus standards for the industry. Political parties that use polling as part of their campaign strategy ) the ruling AKP, for example ) may benefit from the increased quality of survey research. However, as a governing party, AKP could also find itself unable to take hard political decisions that run contrary to prevailing public opinion. Wiener