1

Party Preferences and Economic Voting in Turkey

(Now That the Crisis is Over)

Prepared for presentation at the EcoMod Conference “Middle East and North African Economies:
Past Perspectives and Future Challenges”, Brussels, Belgium, June 2-4, 2005

Cem Başlevent
Istanbul Bilgi University
Department of Economics
Istanbul, TURKEY
[email protected]

Abstract

The purpose of this study is to re-examine the factors that shape party
preferences in Turkey by estimating an individual vote intention
function. The economic variables in the empirical model are items that
can be used to test the conventional ‘economic voting’ hypotheses, i.e.
whether individuals’ economic evaluations about the past or the near
future affect their party choice. In an earlier paper, based on data from
2002, evidence was found in favor of these hypotheses. Those who had
been affected adversely by the economic crisis of 2001 were found to
be very unlikely to vote for the incumbent parties of the time. In the
present paper, we focus on the comparison of the characteristics of the
intended voters of the currently-ruling - and so far successful - Justice
and Development Party (AKP) with those of other parties. According to
multinomial logit estimates, those who make the more optimistic
evaluations about the state of the economy are more likely to vote for
the AKP. This finding is at variance with the dominant empirical
finding of what is known as ‘grievance asymmetry’, i.e. that voters may
react more to negative changes than to corresponding positive ones.

1. Introduction
The general elections of November 2002 were considered as a political earthquake in
Turkey. With thirty-four percent of the votes, the ‘moderate-Islamist’ Justice and
Development Party (Adalet ve Kalkınma Partisi, AKP) won nearly two-thirds of the
seats in the parliament owing to the election system that imposes a ten percent
national threshold for representation. The AKP has been the recipient of a substantial
amount of protest votes by large masses that were adversely affected by the dismal
economic conditions that prevailed in Turkey after the former ruling coalition led the
country into its worst ever economic crisis. While the center-left Republican People’s
 2

Party became the only party other than the AKP to enter the parliament, members of
the former coalition suffered the heaviest losses as their combined vote share dropped
by about 39 percentage points (to 14.7 percent) within the three and a half years
following the April 1999 elections. This was the first time in Turkey that ruling
parties were totally wiped out of the parliament.
The purpose of this study is to analyze party preferences in Turkey and
determine the characteristics of the electoral base of the AKP with a particular focus
on the role of economic evaluations. This issue has previously been dealt with in our
earlier paper (Başlevent, Kirmanoğlu, and Şenatalar, 2005). The empirical work of
that paper was based on an April 2002 electorate tendency survey conducted at a time
when the effects of the economic crisis were still strongly felt. The empirical model
performed well in predicting that the incumbent parties would be punished by those
who were badly affected by the economic conditions. The survey utilized in the
present paper was conducted in December 2003 at a time when the economy had
began its recovery, most noticeably in the form of a significant decline in the inflation
rate and a more stable exchange rate against the euro and the U.S. dollar.
Empirical findings from a survey conducted one year after the 2002 elections
could be informative for two main reasons. The first one is that, after one year in
power, the AKP was regarded more as a ‘conservative-democrat’ party, rather than
pro-Islamist. It would be interesting to see if that perception translated into political
support to the party from the more liberal/modern segments of the society. Another
reason is that the model allows us to test - in a multivariate setting - the role of
economic evaluations when the ruling party is successful in creating a stable
economic environment. This is not a result to be taken for granted since Lewis-Beck
and Paldam (2000) report that there is dominant empirical finding on what is known
as the ‘grievance asymmetry’, which asserts that voters may react more to negative
changes than to corresponding positive ones.

2. Economic Voting
In examining party preferences, we estimate an individual vote intention function that
includes both economic and non-economic factors. The economic factors considered
are the variants of the two main approaches to economic voting theory, i.e. the
retrospective and prospective pocketbook/sociotropic evaluations of individuals.
Economic voting is a general concept that relates electoral support for incumbent
parties to economic performance built upon the hypothesis that voters hold the
government responsible for economic events.1 Central to the economic voting
argument is the idea that uncertainty about the future and about the sincerity of the
candidates (parties) make the past action the best guide for assessing the future
utilities of candidates (parties). Empirical analysis involving economic voting may be
carried out at the individual level to discover the determinants of voters’ party choice
or at the aggregate level to analyze election outcomes. At the individual level,
sociotropic and egotropic (or pocketbook) voting are two approaches to modeling
economic voting behavior and so are the retrospective and prospective voting
hypotheses.
Retrospective pocketbook voting hypothesis is the notion that party choice is
influenced by individuals’ assessments of their personal (household) economic
1
For more on economic voting, see the special issue of Electoral Studies: Economics and
Elections (Volume 19, Number 2/3, June/September 2000).
 3

situations in the recent past. If an individual feels that her economic situation
worsened (improved) under a given government, she will be less (more) likely to vote
for the ruling party in the next election. Retrospective sociotropic voting refers to the
notion that voters’ consider the state of the national economy when they decide which
party to vote for. Taking into consideration the general economic situation does not
necessarily reflect an altruistic thinking of the voter for the economic needs of others,
but may be based upon the notion that a better macroeconomic situation benefits the
individual.
Both types of retrospective voting imply that votes are cast on the basis of
economic performance, rather than economic policy proposals and promises.
Hypotheses of prospective voting, on the other hand, argue that party choice is based
on the assessment of future economic prospects rather than the past. Prospective
pocketbook voting refers to the consideration of individual economic expectations for
the future while individuals who consider the larger national prospects are said to vote
sociotropic prospectively.
Examination of the relationship between economic performance and electoral
success in Turkish politics goes as far back as Bulutay and Yıldırım (1969) and
Bulutay (1970) that make use of aggregate data. Among the more recent studies,
Çarkoğlu (1997) develops an aggregate vote function for Turkey that relates
variations in electoral support for the incumbents in 21 elections during the 1950-
1995 period to macroeconomic conditions, namely the percentage changes in
consumer price index and real per capita GNP and the unemployment rate. The
author finds that “rising unemployment and inflation rates lead to declining electoral
support for the incumbents whereas higher per capita GNP growth rates lead to higher
levels of support” (p.90). Kalaycıoğlu (1999) estimates a model of party preference to
determine the factors that explain the voting behavior in Turkey. Discriminant
analysis on individual level data reveals that voting behavior is motivated more by
cultural and ideological factors than by socio-economic factors. In assessing the role
of economics on voting, Kalaycıoğlu employs an economic self-satisfaction variable
based on the current level of satisfaction of the respondents’. As such, this variable
represents voters’ pocketbook evaluations at the time of survey.
Taking a different approach, Akarca and Tansel (2002) use province level data
from the 1991 and 1995 general elections to examine the association between the vote
shares of political parties and economic growth. Their finding is that the major
incumbent party benefits from a good performance at the expense of extremist parties
in the opposition while the junior party in the coalition and the centrist opposition
parties are unaffected by the economic conditions. Finally, Esmer (2002) carries out
binary logit analysis to examine voter behavior in Turkey. He concludes that the most
important determinant of the party choice is the left-right ideology whereas indicators
of economic well-being are not good predictors.

3. Data and Model Specification

In the empirical work, we use data drawn from the Electorate Tendency Survey
conducted in December 2003 by Veri Araştırma, a private research company. The
survey was conducted in 26 (out of 81) provinces of Turkey on a nationally
representative sample.2 The original sample included 1,806 individuals at or above
2
The 26 provinces are Adana, Ankara, Bursa, İstanbul, İzmir, Konya, İçel, Hatay, Antalya,
Aydın, Tekirdağ, Manisa, Kütahya, Isparta, Sakarya, Karabük, Çorum, Trabzon, Samsun,
 4

the voting age of 18. The respondents were asked which party they would vote for if
general elections were to be held on that same day. Of the respondents, 8.9 percent
gave no answer, 6.9 percent stated that they were undecided about their votes, and 5.4
percent stated that they would not vote for any of the parties.3 Put differently, 78.7
percent of the respondents stated an explicit party preference. The AKP is the leading
party in the sample with 62.4 percent of explicit party preferences.
In the empirical investigations, we work on a sample of 1,520 individuals that
excludes the undecided and the no-answer cases. Thus, we examine the choice
between the five major parties in Turkey plus the minor parties lumped as one party
and the “no vote” decision. The reason we include the non-voters in the analysis is
that a non-negligible share of voters seem to have lost belief in the political system.
In the November 2002 elections, 8.6 million out of the 41.4 million registered voters
failed to turn up at the ballot-box. While this was the lowest turn-out rate (79.1%) in
the last 7 elections, it amounted to a drop of nearly 9 percentage points from the 1999
general elections. By including the non-voters as a separate category, we hope to be
able to gain some insights as to what kind of people these are.
Among the four parties other than the AKP, the Republican People’s Party’s
(Cumhuriyet Halk Partisi, CHP) ideological position would be best described as
center-left. The True Path Party (Doğru Yol Partisi, DYP) is on the center-right; the
Nationalist Action Party (MHP) gets its support from Turkish ultra-nationalists while
the Democratic People’s Party (Demokratik Halk Partisi, DEHAP) is the
representative of Kurdish nationalists.
The vote intention function we estimate is made up of economic and non-
economic factors, and the specification is similar to the one used in Başlevent et al.
(2005). The multinomial logit estimates for the earlier paper are given in the
Appendix. The economic factors included in the present model are items that
correspond to the four variants of the economic voting theory discussed earlier. The
survey questions on the evaluation of economic changes in Turkey and in
respondent’s household during the 12 months preceding the survey are responded to
on a 5-point scale such that larger values correspond to more negative economic
evaluations: 1 = much better, 2 = better, 3 = the same, 4 = worse, and 5 = much
worse.
As reflected by the distribution of the variables in Figure 1a, the 2002 survey
was conducted at a time when the effects of the economic crisis of 2001 were still
strongly felt by the majority of the population. In each case, the share of those who
respond negatively is more than half. By the end of 2003, the picture that emerges
from the same distribution has changed dramatically (see Figure 1b). Positive
evaluations are the majority except in the ‘retrospective pocketbook’ case. Evidently,
there are many individuals who are hopeful about the economy’s future even though
their personal condition has not improved during the past year.

Erzurum, Kayseri, Sivas, Adıyaman, K.Maraş, Diyarbakır, and Van. These were chosen after
a stratification of provinces according to geographical location. Within these provinces, both
urban and rural locations were represented in the sample in proportion with the number of
voters they contain.
3
The last figure, which is down from 16.9 percent in April 2002, could be an indication that,
during this period, a restoration of Turkish people’s trust in the political system has taken
place.
 5

Figure 1a: Distribution of economic voting variables (April 2002)

Prospective pocketbook 4% 55% 37%

Retrospective pocketbook 21% 43% 20%

Prospective sociotropic 28% 43% 15%

Retrospective sociotropic 10% 59% 29%

Much better Better The same Worse Much worse

Figure 1b: Distribution of economic voting variables (Dec. 2003)

Prospective pocketbook 51% 35% 10%

Retrospective pocketbook 31% 40% 25%

Prospective sociotropic 66% 16% 10%

Retrospective sociotropic 61% 18% 14%

Much better Better The same Worse Much worse

In fact, comparison of the pocketbook and sociotropic variables provides a

way of observing whether individuals feel that they are economically better or worse
off compared with the society in general. In our sample, about 44 percent of the
respondents have answered the retrospective pocketbook and sociotropic questions in
the same way, while the corresponding figure for the prospective questions is 58%
(See Table 1). The majority of the remaining individuals feel that they are worse off
than others. Since the ‘average’ person should in reality be doing just as well as the
rest of the society, it seems that people tend to overrate other people’s welfare, or
underplay their own. It may also be interesting to know whether people’s
retrospective evaluations coincide with their prospective ones. That happens to be the
case 65% of the time for the sociotropic variables and 55% of the time for the
pocketbook variables. The majority of the rest have more positive evaluations for the
next 12 months.

Table 1: Comparison of economic voting variables (in %)

A: Retro. A: Pros. A: Pros. A: Pros.
pocketbook pocketbook sociotropic pocketbook
B: Retro. B: Pros. B: Retro. B: Retro.
sociotropic sociotropic sociotropic pocketbook
A better than B 9.7 12.4 21.8 38.8
A equal to B 43.5 57.7 64.8 54.5
A worse than B 46.8 29.9 13.4 6.7
 6

The non-economic factors included in the model are standard socio-

demographic variables (age, education, gender, the within-province location of the
place of settlement, and religiosity) as well as other factors specific to Turkey that
might affect voting behavior. These are operationalized through dummy variables for
(i) belonging to the ‘Alevi’ sect of Islam (7.6% in sample), (ii) being an ethnic ‘Kurd’
(10.1% in sample), and (iii) the opinions on the issue of Turkey’s membership in the
European Union. A similar question on the abolition of the death penalty present in
the April 2002 survey is missing in the December 2003 survey since the death penalty
was already abolished by that time.
Education is entered in the model as a single continuous variable that takes on
its values depending on the years spent in school. The information is gathered from
the survey question on the highest degree completed. We assigned 3 years to those
who have - at most - completed elementary school (71% of sample), 8 to middle
school (9%), 11 to high school (15%), and 15 to university graduates (5%). We
expect this variable to perform well as a predictor of the party choice since Esmer
(2002) reports that the level of education is repeatedly found to be positively related
to leftist orientation in earlier work by the same author.
As in all predominantly Muslim countries, Islam plays an important role in
Turkish politics, and a self-identified degree of religiosity has previously been found
to be a significant factor of voters’ preferences (Esmer, 1995; Kalaycıoğlu, 1999;
Çarkoğlu and Toprak, 2000). The survey question on whether the respondent is a
believer and how closely he or she follows the rules of Islam is responded to on a 5-
point scale. While the overwhelming majority of the respondents claim to be
believers, about two-thirds consider themselves “religious”. In an attempt to
incorporate what’s known as ‘religious voting’ into the model, we include this
variable linearly such that larger values correspond to higher degrees of religiosity.
Even though Turkey is far from becoming a full member of the European
Union (EU), the issue of whether Turkey should become a member is a very
contentious one. In our sample, around seventy percent of the respondents state that
they support the EU membership of Turkey while 15 percent are against membership,
and the rest have their reservations or are undecided. In the model, we treat the
against-membership group as the reference category and include dummies for the
remaining two groups (Cond-EU, for “conditional”, and Pro-EU). We expect this
variable to capture the pro vs. anti-Western division among Turkish people.
The population of Turkey includes a very small share of non-Muslims while
two sects of Islam combine for nearly 99 percent of the population. While most of the
muslims are Sunni’s, there’s also a smaller population of Alevi’s. The share of
Alevi’s, who are ethnically divided among Turks, Kurds, and Arabs in Turkey’s
population, is estimated at nearly 20 percent (Güneş-Ayata and Ayata, 2002). The
reason Alevi’s are of interest politically is that they are known to be supporters of the
democratic-secular state, and they usually vote for center-left parties. Kurds, on the
other hand, have a separate ethnic origin, and ethnically aware Kurds are known to be
more likely to vote for the DEHAP, which they view as their representative. Güneş-
Ayata and Ayata (2002) estimate the share of ethnic Kurds who vote for the DEHAP
between one-quarter and one-third. Although the share of ethnic Kurds in Turkey’s
population is estimated at anywhere between 12 and 20 percent (Andrews, 1989), the
DEHAP and its predecessors have never been able to pass the 10 percent nationwide
 7

threshold in any election except in 1991 when they formed a pre-election coalition
with another party.4

3. Empirical Results
Multinomial logit estimation results are reported in Table 2. Since the model yields
too many results to discuss here, we focus on the differences between the
characteristics of the AKP voters and the voters of other parties. The coefficients
reported are exponentiated coefficient estimates, otherwise known as odds-ratios.
They measure the ratio of the probability of voting for a party to the probability of
voting for the reference category, which in our case is the AKP. For example, a
coefficient of 2 (along with a p-value of less than 10%) on the gender dummy for
party X means that for two individuals who are otherwise identical, a female is two
times more likely than a male to prefer party X to the AKP. Conversely, coefficients
less than 1 indicate a negative association between the value of the explanatory
variable and the probability of choosing the relevant party over the AKP, that is if
statistical significance (p-value<.1) is present.
Leaving the discussion on the economic voting variables to the end, we first
focus on the rest of the explanatory variables in the model. As far as the socio-
demographic variables are concerned, there isn’t strong evidence that age is a
determinant of the choice between the AKP and other parties with the exception of the
CHP. People who intend to vote for the AKP are younger than those who plan to vote
for the CHP. However, since the coefficients on the age variable across the six
estimated equations are jointly significant, it is appropriate to include the age variable
(and all of the rest of the variables for the same reason) in the model. The education
variable turns out to be a good predictor of the choice between the AKP and other
parties. With the exception of the MHP and HADEP, individuals who fall into the
other categories are more educated than AKP voters.
Coefficient estimates on the gender dummy indicate that females are more
likely to choose the AKP over the MHP and DEHAP. Regarding the place of
settlement dummies, the coefficient estimates confirm the general opinion that the
AKP is strong in largely populated areas. As expected, the degree of religiosity is
strongly positively associated with the intention to vote for the AKP. The DYP
constitutes an exception here since there is no statistically significant difference
between the AKP and DYP voters with respect to the religiosity variable.
As for the issues variables, estimation results indicate that the intention to vote
for the AKP is not positively associated with being pro-EU membership. Even though
statistical significance of the coefficients is generally absent, the magnitudes of the
coefficients suggest that the intended voters of CHP, DYP and DEHAP are more pro-
EU than AKP voters. On the other hand, MHP voters and those who intend to vote
for minor parties are the two groups who are more strongly against EU membership
than those who plan to vote for the AKP. It is worth mentioning that in the study
based on the April 2002 survey, the finding was that AKP voters are statistically
significantly less pro-EU than many of the other groups. The reason for the change
must be that, as reflected by the vote share in the present data set, the AKP is getting
more support from the modern/liberal segments of the society.

4
Incidentally, the HADEP ended up entering the elections under another name (DEHAP) in
the face of the possibility of being closed down by the Constitutional Court.
 8

Table 2: Multinomial Logit Results of the Vote Intention Function (Dec. 2003)
No
DYP CHP MHP DEHAP Others
vote
Subsample 48 202 62 40 122 98
size (%) (3.2) (13.3) (4.1) (2.6) (8.0) (6.5)
Demographic variables
1.016 1.017 0.985 0.993 0.995 1.006
Age
0.145 0.017 0.162 0.645 0.592 0.493
1.087 1.102 1.037 0.890 1.144 1.112
Education
0.061 0.000 0.367 0.148 0.000 0.001
0.928 0.817 0.440 0.470 1.029 0.641
Female
0.818 0.332 0.007 0.076 0.902 0.083
1.007 1.155 1.994 1.603 1.219 3.222
District
0.986 0.602 0.068 0.408 0.570 0.017
0.434 0.561 0.486 2.200 0.882 3.324
City center
0.052 0.038 0.094 0.152 0.710 0.011
0.920 0.440 0.645 0.418 0.577 0.517
Religiosity
0.716 0.000 0.038 0.001 0.000 0.000
Economic voting variables
2.246 2.404 1.865 1.053 2.265 2.005
Retro. socio.
0.000 0.000 0.002 0.853 0.000 0.000
2.285 2.762 3.338 2.683 2.865 2.969
Pros. socio.
0.000 0.000 0.000 0.000 0.000 0.000
1.397 1.240 0.869 2.162 1.747 1.337
Retro. pocket.
0.126 0.133 0.492 0.006 0.000 0.093
1.490 1.459 1.198 1.507 1.678 2.011
Pros. pocket.
0.096 0.018 0.456 0.165 0.003 0.000
EU dummy variables
1.273 1.006 0.835 2.089 0.180 0.954
Conditional
0.745 0.988 0.693 0.472 0.000 0.908
2.768 1.533 0.512 2.333 0.493 0.658
Pro-EU
0.102 0.182 0.061 0.328 0.011 0.204
Identity variables
1.353 25.760 5.652 4.982 6.446 8.905
Alevi
0.779 0.000 0.007 0.025 0.000 0.000
0.721 0.732 0.408 80.404 0.883 1.634
Kurd
0.603 0.425 0.237 0.000 0.778 0.234
Notes: Reference category is the AKP, with 948 votes (62.4%). Sample size: 1,520.
Table entries in each cell are exponentiated coefficient estimates (top) and p-values
(bottom). Restricted log likelihood: –1941.36. Log likelihood of model: –1390.78.
Likelihood ratio (84 d.f.): 1101.16. Pseudo R2: 0.2836.
 9

Estimation results concerning the identity variables indicate that Alevis are
highly unlikely to vote for the AKP while the CHP is confirmed as the most popular
party among this group of voters. Kurds, on the other hand, are extremely more likely
to vote for the HADEP rather than the AKP as well as the rest of the parties. There’s
also some evidence that the AKP is relatively more popular among Kurds as it has
traditionally emphasized the notion of the “brotherhood of Islam” rather than the
Turkish identity as the basis of unity in the society.
As far as the economic voting variables are concerned, there is strong evidence
that voters with positive economic evaluations are planning to support the AKP
government in the next elections. Interestingly, the coefficients on the sociotropic
voting variables are statistically more significant than those on the pocketbook
variables. In other words, people who believe that the state of the nationwide
economy is improving are more likely to support the AKP than those who feel their
personal situation has improved. The only exception here is that people who thought
that the national economy worsened are as much likely to vote for the DEHAP as the
AKP. This result does not come as a surprise since the two parties are known to be
the most popular among the economically disgruntled masses.

4. Conclusion
Multinomial logit estimates of the vote intention function we estimated
confirm our earlier finding that relatively younger, more religious, and less educated
people, especially males, constitute the electoral base for the AKP. Those who have
positive evaluations of the AKP’s performance are also likely to vote for the AKP.
We have thus found evidence that economic evaluations play a significant role in the
party choice of Turkish voters alongside non-economic factors. We concluded that
‘economic voting’ is present in Turkey in the sense that successful incumbents are
rewarded just as unsuccessful ones are punished.

References
Akarca, A. T. and A. Tansel (2002) “Economic Performance and Political Outcomes:
An Analysis of the 1995 Turkish Parliamentary Election Results” Paper
presented at the Ninth Annual Conference of the Economic Research Forum
held in Sharjah, UAE, October 2002.

Andrews, P.A. (1989) Ethnic Groups in the Republic of Turkey. Weisbaden: L.

Reichert.

Başlevent, C., H. Kirmanoğlu, and B. Şenatalar (2005) “Empirical Investigation of

Party Preferences and Economic Voting in Turkey”, European Journal of
Political Research, 44 (3), forthcoming.

Bulutay, T. (1970) “Türk Toplumsal Hayatında İktisadi ve Siyasal Gelişmeler”

(Economic and Political Developments in Turkey’s Social Life) SBF Dergisi,
25, 3, pp.79-119.

Bulutay, T. and N. Yıldırım (1969) “Türk Seçmenlerinin Oy Verme Eğilimlerinde

İktisadi Sebeplerin Üzerinde Bir Deneme” (An Essay on Economic Reasons
 10

Behind Voting Tendencies of the Turkish Electorate) SBF Dergisi, 22, 4, pp.7-
39.

Çarkoğlu, A. (1997) “Macro Economic Determinants of Electoral Support For

Incumbents in Turkey, 1950-1995” New Perspectives on Turkey, Fall, 17, pp.
75-96.

Çarkoğlu, A. and B. Toprak (2000) Türkiye’de Din, Toplum ve Siyaset (Religion,

Society and Politics in Turkey), in Turkish, Turkish Economic and Social Studies
Foundation (TESEV) Publications. Istanbul.

Esmer, Y. (1995) “Parties and the Electorate: A Comparative Analysis of Voter

Profiles of Turkish Political Parties” in Ç. Balım, E. Kalaycıoğlu, C. Karataş, G.
Winrow, and F. Yasamee (eds.) Turkey: Political, Social and Economic
Challenges in the 1990s. Leiden, New York, Köln: E. J. Brill.

Esmer, Y. (2002) “At the Ballot Box: Determinants of Voting Behaviour in Turkey”,
in Y. Esmer and Sabri Sayarı (eds.) Politics, Parties and Elections in Turkey.
Lynn Rienner.

Güneş-Ayata, A. and S. Ayata (2002) “Ethnic and Religious Bases of Voting”, in Y.

Esmer and Sabri Sayarı (eds.) Politics, Parties and Elections in Turkey. Lynn
Rienner.

Kalaycıoğlu, E. (1999) “The Shaping of Party Preferences in Turkey: Coping with the
Post-Cold War Era” New Perspectives on Turkey, Spring, 20, pp. 47-76.

Lewis-Beck, M.S. and M. Paldam (2000) “Economic Voting: An Introduction”

Electoral Studies, 19, 2/3, pp. 113-121.
 11

Appendix: Multinomial Logit Results of the Vote Intention Function (April 2002)
No
DYP CHP HADEP MHP ANAP SP DSP
vote
Subsample 148 134 101 94 85 50 36 306
size (%) (11.4) (10.3) (7.8) (7.2) (6.5) (3.8) (2.8) (23.5)
Demographic variables
1.026 1.017 1.019 0.971 1.026 1.012 1.024 1.016
Age
0.001 0.060 0.171 0.006 0.005 0.329 0.071 0.017
0.960 1.007 0.871 0.972 1.012 1.002 0.974 1.057
Education
0.323 0.866 0.033 0.535 0.795 0.978 0.691 0.060
1.929 1.477 1.703 1.431 3.568 1.497 1.495 2.177
Female
0.003 0.120 0.147 0.158 0.000 0.220 0.302 0.000
0.619 1.165 0.885 0.765 1.024 0.929 0.373 0.906
Urban
0.046 0.609 0.762 0.338 0.939 0.845 0.043 0.631
0.447 1.504 0.420 0.507 1.400 1.697 0.599 1.711
Metropolitan
0.011 0.212 0.139 0.065 0.324 0.195 0.286 0.019
0.467 0.255 0.202 0.397 0.691 0.852 0.255 0.376
Religiosity
0.000 0.000 0.000 0.000 0.068 0.524 0.000 0.000
Economic voting variables
1.503 0.870 0.572 1.175 0.753 0.993 0.743 1.176
Retro. socio.
0.020 0.440 0.072 0.374 0.132 0.979 0.191 0.242
0.921 0.974 1.780 0.665 1.028 0.938 0.729 0.937
Pros. socio.
0.500 0.851 0.026 0.003 0.859 0.735 0.144 0.518
0.717 0.926 0.982 0.858 0.686 0.801 0.757 0.886
Retro. pocket.
0.013 0.628 0.948 0.324 0.022 0.260 0.254 0.273
0.743 0.934 1.044 0.613 0.941 1.083 0.582 1.121
Pros. pocket.
0.073 0.724 0.906 0.009 0.774 0.767 0.029 0.438
Issues variables
3.265 2.052 2.411 1.582 2.901 2.300 3.501 1.824
Cond-EU
0.000 0.138 0.257 0.269 0.038 0.036 0.050 0.026
2.588 4.046 3.259 1.491 4.783 0.578 2.862 1.369
Pro-EU
0.000 0.000 0.051 0.146 0.000 0.145 0.033 0.105
2.266 2.016 1.074 1.297 1.601 0.837 2.554 1.430
Cond-abol
0.000 0.022 0.941 0.344 0.138 0.642 0.020 0.081
0.846 3.273 22.936 0.303 3.006 0.660 0.754 1.965
Pro-abol
0.612 0.000 0.000 0.020 0.001 0.390 0.646 0.004
Identity variables
4.954 90.083 8.783 1.391 14.239 3.827 10.358 15.817
Alevi
0.064 0.000 0.018 0.793 0.001 0.281 0.016 0.000
0.890 0.253 25.696 0.515 0.443 3.075 0.000 0.472
Kurd
0.784 0.011 0.000 0.305 0.132 0.015 1.000 0.037
Notes: Reference category is the AKP, with 347 votes (26.7%). Sample size: 1301.
Table entries in each cell are exponentiated coefficient estimates (top) and p-values
(bottom). Restricted log likelihood: –2556.87. Log likelihood of model: –1988.29.
Likelihood ratio (128 d.f.): 1137.15. Pseudo R2: 0.2224.