Talk:Opinion polling for the June 2015 Turkish general election

Distribution of undecided votes

In December KamuAR poll and October Metropoll poll, the references indicate that some voters are undecided (8.6 and 14.5 respectively) and poll results with undecided votes distributed to the parties are not available. In the article, the results are given distributed. Is it your own calculation, T.C. Ataturkiye, and if so, what formula is used here? It seems that the votes are not distributed linearly proportional with each party's vote.

Further, I have just noticed that the percentages do not add up to 100, but to 106.2 in the first one, so I guess there is an error with the distribution of the KamuAR poll. Also, other parties are only 0.7 in the MetroPoll poll, so they should not become 6.0 even after all 14.5 undecided/invalid votes are distributed, so I imagine something is wrong with it as well.--Cfsenel (talk) 13:46, 31 December 2014 (UTC)

Hi. The distribution of undecided votes is my calculation, and you were right to point out that there was a mistake in the KamuAR poll. I've now rectified it, thanks for letting me know. In the MetroPoll poll, the 'others' also takes into account the 3.9% who said they would cast a protest vote as well as the 0.9% who said they would vote for the Felicity Party, which gives a total of 5.5% when the undecided votes are not distributed.

The method I use to distribute undecided voters is to first find out how many people said they would vote for each party. For example, if the sample is 4,117 like the KamuAR poll and the AKP is found to have 35.2% without distributing the undecided voters, I do 0.352*4117 in order to find out how many people said they would vote for the AKP. I use the same method for all the other parties. I then use this method to find out the number of undecided voters (which in KamuAR is 8.6%), so I do 0.086*4117, and then subtract this number from 4117. This gives us the number of people polled minus the undecided voters. I then divide the voters for each party by this new number, then times it by 100% to get the new percentage. If you notice any flaws then please tell me. T.C. Ataturkiye (talk) 16:15, 31 December 2014 (UTC)

It should be correct, I calculated the same thing. But in the MetroPoll poll, I think casting a protest vote does not mean they will vote for other parties, the poll already has an 'others' section. I don't know that it means in other countries, but in the Turkish context it means casting an empty ballot, or an invalid one by marking multiple parties, writing on the ballot, casting the envelope without the ballot paper, tearing the ballot etc. (In the 2011 election, they constituted 2.22% of the total vote, and in 2014 local elections it was 4.39%, although this also includes people who cast an invalid vote by mistake.) So in effect, their votes are distributed among other parties as well. Considering this, the numbers I calculated were AKP 47.5; CHP 26.7; MHP 15.7; HDP 8.3; Others 1.8.--Cfsenel (talk) 21:14, 31 December 2014 (UTC)

See this from the same company, for example. When asked what they voted in the previous election, some said 'others' and some said 'protest vote', which is different.--Cfsenel (talk) 21:24, 31 December 2014 (UTC)

Hi, I think you're right. The invalid/blank ballots are discounted when the results are announced, so the most sensible option would be to discard them rather than considering them as 'other'. T.C. Ataturkiye (talk) 23:19, 1 January 2015 (UTC)

Unreliable sources

Sadly, most of the poll results are given with references to websites like secimanketi.tv, secimanketleri.org, etc, none of which are reliable sources, not any more than a personal blog. See tr:Tartışma:2015 Türkiye genel seçim anketleri for details, some current mistakes those caused in the article are: Pollmark poll claimed to be dated 29 Dec in the article ([1][2]) (which they of course don't bother to give some crucial information, like sample size) is actually from 10-14 Oct ([3]). They presented it as a new poll again, to get more clicks and ad revenue. The 22 Aug ORC poll ([4]), which is actually dated 11-13 Aug, is deliberately misrepresented by secimanketi.tv by claiming both MHP and HDP are at 10.2 (actually MHP is 14.2, (http://www.orcarastirma.com.tr/wp-content/uploads/2014/08/orc.agustosgundem.2014.pdf) secimanketi.tv took 4 points and put it to the 'others' part). These sources should not be used at all. In most cases the sources are available in the official website of the polling company, if not, a reputable newspaper article may be used as a source. The Turkish article has better sources for most of them, with the errors corrected.--Cfsenel (talk) 11:42, 5 January 2015 (UTC)

This is certainly not good news. If I'm honest, I'm not sure any newspaper or polling company are completely viable as a source. For example, PollMark's website has been under construction since at least the local elections (most likely for long before that as well) and the company is supposedly owned by an advisor to Tayyip Erdogan. It sounds like they send out some made-up polling results to different journalists at regular intervals showing the AKP at high numbers. I'm not sure what sort of real and respectable polling company would have their website under construction for at least a year.

As far as the two election poll websites, I suggest we either delete the polls they release for which we can't find additional sources, or we put a specific notice beside these polls stating that they are more than likely to be unreliable for the reasons you have given above. T.C. Ataturkiye (talk) 23:05, 12 January 2015 (UTC)

In the Turkish Wikipedia, I found additional sources for almost all of them, and removed one that I could not find a second source. It also has some of the poll dates corrected now. But you are right, T.C. Ataturkiye, polling is in deplorable state in Turkey right now. For example, 29 Dec 2014 Pollmark results were actually announced on Facebook! They are seriously claiming that their website was blocked by the government to prevent their results from being announced to the public. What an excuse! At least they seem to have a little sense of professionalism in preparing and posting their results on Facebook. In the case of the 2 Jan 2015 PCR poll, I doubt even such an organization exists, the whole thing seems to be a complete farce. For all I know, a bored person may have created it on Excel (and not put a lot of effort to it, I would do a much better job creating mock poll graphs), sent the screenshot from the Excel graph to secimanketi.tv and secimanketleri.org, who then published it. I was going to remove it from the Turkish article, but a respectable (!) Turkish newspaper (Sabah) picked it up and published it on their website, probably because it showed AKP at a high percentage. Just pathetic.--Cfsenel (talk) 23:15, 16 January 2015 (UTC)

Thanks for finding alternative sources, Cfsenel. As far as I know, one of Pollmark's founders is a close advisor to Erdoğan so I would hardly believe them if they actually claimed that their site had been blocked by the government. I initially thought PCR was a foreign polling firm and hoped that this might give it a bit of credibility, but again you are right. A simple google search yields absolutely no trace of such a polling form. Furthermore, in the image of their excel poll results graph posted here, they spell their name 'Politic's Communication and Research'. There are no grammatical circumstances in which there should be an apostrophe between the 'c' and the 's' in 'politics'. I think the inability to make up a name properly, as well as the general unprofessionalism in their results graphic, shows that whoever made up these results simply sent some random numbers to secimanketi in order to publish it in the media. Who knows, it could have been the secimanketi site itself making up polls to increase site views. While I have no evidence to support this, I think that many of these bogus poll firms are 'owned' by individuals who are close to the government (as in the case of Pollmark) and thus have good access to various journalists. This, I would argue, gives them the ability to send out some random poll results they make up to different newspapers in order to show the government at a high percentage. If you remember the opinion polling for the presidential elections, the projected results were so different to the actual results that KONDA actually had to issue an apology. While I have no evidence to back up my claims (although the appallingly unprofessional polling graphic that PCR made can probably count as evidence), I seriously think that we should make it known somewhere that these polls are definitely not as reliable as those in other countries. We can maybe do this on this article or maybe Opinion polls in Turkey page. For example, in Opinion polling for the 2015 United Kingdom general election article, each poll result is accompanied with a huge PDF report by the polling company on the polling results and techniques. Whereas in Turkey, we count ourselves lucky to find a poll that bothers telling us the sample size. This, along with the more general concerns about the government's influence over the media and also the large difference between SONAR/Gezici and others such as Pollmark I think need to be briefly explained or at least noted down somewhere to avoid confusion. T.C. Ataturkiye (talk) 00:41, 17 January 2015 (UTC)

You are right, I remembered the company names wrong, such an excuse would not make sense especially for Pollmark. I actually meant KamuAR when I was talking about announcing results on Facebook (here is their page: https://tr-tr.facebook.com/Kamuar ). Apparently they have posted a new poll result, but it would be bizarre to show a Facebook page as the source when adding it to Wikipedia.--Cfsenel (talk) 05:25, 25 January 2015 (UTC)

GENAR poll

I have removed the GENAR poll results. We have very little expectations about the information provided given the state of polling in Turkey, but I don't think our current information about their latest poll even meets those. GENAR is certainly a known company that should be on this page, but this is not a poll result announcement. When a company's president says on a TV program that according to their polls "MHP is around 14", I don't think this constitutes encyclopedic data. I don't mean to say we need the decimal points (indeed, they are mostly unnecessary given the uncertainty range) but the language is so vague here that (e.g. one party's votes are given with high precision while others are only mentioned approximately) I believe we should add the poll results when and if the results are actually announced by the company.

I also am not sure whether PCR and Emax polls should be here while even the company's existence is doubtful, but of course one could say that it is not for us to decide and it is sufficient to mention their questionability it in the notes. At least they are full announcements, even though probably fake.--Cfsenel (talk) 02:09, 10 February 2015 (UTC)

Online polls

I don't think online polls have any place on an encyclopedia at all. That is, an online poll by a polling company which designs determines a sample and contacts them online could be mentioned (although it is hard to see how that could work given the percentage of people who do not use internet), but an open-to-all online poll on a website has zero encyclopedic value.--Cfsenel (talk) 05:13, 9 May 2015 (UTC)

Censorship of the turkish equivalent article

So it appears the turkish equivalent of this article is being censored along with some other articles in Turkey since 6th of June. The turkish wiki has placed a banner on their Wikipedia warning about this. See The Signpost about it.

Here's the warning message they placed as a banner on their Wikipedia: tr:Türkçe_Vikipedi'deki_maddelerin_erişime_engellenmesi. They are referring to all other articles as "disabled" with the exception of this one which they refer to as "partially disabled".
And here's a news report by Hürriyet Daily News about it: http://www.hurriyetdailynews.com/wikipedia-releases-warning-on-turkeys-censorship-monitoring.aspx?PageID=238&NID=84255
Also see: http://www.birgun.net/haber-detay/vikipedi-sansure-isyan-etti-83076.html

So shouldn't this also be part of this article (even though it's a bit meta)? Please add it.

--Fixuture (talk) 22:47, 26 June 2015 (UTC)