28 October 2015

Turkey's Elections and Opinion Polls

On 1 November, that is this coming Sunday, Turkey will be holding a re-run of the June elections.

Opinion polls generally indicate that the same hung Parliament will return. I already posted my views on that outcome.

This post is about the percentages and the results of various opinion polling firms.

Below is a confusing table summarizing the current predictions of leading companies. You do not need to pay attention to it. I will explain its relevance in a minute.

Poll Date
Company
Sample Size
Other
24–25 October
Gezici
4.864
43,0
26,1
14,9
12,2
3,8
21–24 October
ORC
2.650
43,3
27,4
14,0
12,2
3,1
17–21 October
Kurd-Tek
2.592
39,10
28,14
14,36
14,21
4,19
16–20 October
AKAM
24.540
40,45
28,22
14,33
14,17
2,83
(Published 21 October)
KONDA
 ?
40,9
30,4
14,3
11,8
2,6
16–20 October
Betimar
2.000
44,30
25,48
13,58
12,97
3,67
13–19 October
Konsensus
2.002
43,0
28,6
13,1
11,9
3,4
17–18 October
Gezici
4.864
41,3
27,0
15,6
12,5
3,6
10–16 October
Perspektif
1.255
42,3
27,6
14,7
12,8
2,6
12–16 October
Argetus
4.472
43,1
26,2
15,1
12,6
3,0
12–17 October
MAK
 ?
43,5
27,8
13,6
12,1
3,0
10–15 October
Denge
 ?
43,4
25,9
13,7
13,6
3,4


A Turkish social scientist studied 88 recent polls [link in Turkish] to determine the standard deviation for each political party. Since standard deviation is a measure of how dispersed the data are, he was trying to see how far off poll results were and their overall distribution.

The findings are startling. The standard deviation for AKP is between 4 and 12 percent. An overwhelming majority of polling firms predicted much bigger AKP majorities.

Similarly, the results were quite skewed for HDP. Except it was the other way around. Most polling companies predicted dismal showings.

What is interesting is the fact that for CHP and MHP the standard deviation was 1 and 2 percent (a standard deviation of 0 would indicate a perfect distance from the mean value).

In other words, since their methodology was accurate and reliable for two of the parties, it is hard to blame technical error.

Why am I talking about this?

Well, for the first time in Turkish history, a Kurdish polling firm (out of Erbil, no less) conducted an opinion poll on Turkish elections. That is the company on the third row, Kurd-Tek.

Notice the AKP and HDP results?

Theirs is the only prediction placing AKP below 40 percent mark. And it is one of the only two putting HDP above 14 percent.

If, on 1 November, AKP stays below 40 percent and HDP above 14 percent, there will be many polling firms with a lot of 'splaning' to do.

No comments:

Post a Comment