The Vote, Leader personality, Negativity and Affective Polarization Link: Evidence from the Turkish Case

Hi guys!

As you can see I am workin’. :)

First thing first, I am really very excited to work on this one with you. I hope we can make it and publish it in the special issue after the workshop.

Enough chit-chat!

Below I have provided the structure of the paper that I have shared with you. Surely this is not definetive we can alter it in any way you want.

The theory of the paper is quite simple: The link between vote choice and leader personality is conditoned a) by negativity used in electoral cycle and b) affective polarization level of the voter. I think this has not been done before although there are studies focusing on the link between vote and leader traits, vote and affective polarization and vote and negativity during the electoral cycle, I do not know a study which checks the link as we propose here.

I have formulated 3 hypotheses:

  1. As the perceived big5 character traits become more introvert, less agreeable, show lack of consciousness, become emotionally unstable and more close the candidate become less preferable to outgroup members

  2. As affective polarization level increases, the candidate become less preferable to outgroup members

  3. As the candidate goes more negative during the electoral cycle, s/he become less preferable to outgroup members

The data for the analysis is the post-electoral survey data are drawn from the Electoral Integrity in Turkey (EI-T) survey, carried out on a nationally representative random sample of 1,232 Turkish voters between 8 and 20 July, stratified by region. The fieldwork was carried by a professional public opinion company and involved face-to-face interviews in 12 NUTS-1 statistical areas, 30 provinces, and 98 districts; respondents were selected using a multi-stage, stratified, clustered random sampling procedure without replacement.

The dependent variable is the presidental vote choice. 4 candidates ran for office in 2018 elections and we have the big5 data only for Erdoğan. This will force us to use a multinominal model with, utilizing Erdoğan as the base category. The independent variables are: big5 character traits, affective polarization level, negativity level, partisanship and controls. Big5 was measured with the usual scale (Alex you would remember we had a small problem with the scale used in questionnaire, but I believe we will be fine). Affective polarization is measured by like-dislike level diffenece for every respondent. Negativity used by the candidate is a 10 level likert scale, where low levels mean high negativity. Partisanship is measured a 4 level scale of proximity to a certain party. Controls are age aducation gender and region (the esea variable: if the respondent lives in east or southeast part of Turkey where the Kurdih population is dense.) I did not work on the directions of the questions, but the results came out as expected.

The base model

Here is the base model:

a <- multinom(
  president_vote ~
    extraversion +
    agreeableness +
    consciousness +
    emotional.stability +
    openness,
  model_data 
)
## # weights:  28 (18 variable)
## initial  value 1359.954768 
## iter  10 value 526.566816
## iter  20 value 407.401295
## iter  30 value 400.147909
## final  value 400.125393 
## converged
tab_model(a, 
          title = "RTE Vote (Base)",
          dv.labels = "",
          show.intercept = F, 
          show.ci = F, 
          collapse.se = T, 
          p.style = "stars",
          p.threshold = c(0.1, 0.05, 0.01, 0.001))
RTE Vote (Base)
Predictors Odds Ratios Response
extraversion 0.92
(0.21) 		Meral Aksener - IYI
agreeableness 2.86 ***
(0.75) 		Meral Aksener - IYI
consciousness 0.37 ***
(0.08) 		Meral Aksener - IYI
emotional.stability 1.91 ***
(0.44) 		Meral Aksener - IYI
openness 0.69 **
(0.12) 		Meral Aksener - IYI
extraversion 0.54 ***
(0.10) 		Muharrem Ince - CHP
agreeableness 3.49 ***
(0.74) 		Muharrem Ince - CHP
consciousness 0.60 ***
(0.09) 		Muharrem Ince - CHP
emotional.stability 1.84 ***
(0.35) 		Muharrem Ince - CHP
openness 0.62 ***
(0.09) 		Muharrem Ince - CHP
extraversion 0.84
(0.32) 		Selahattin Demirtas - HDP
agreeableness 3.52 ***
(1.55) 		Selahattin Demirtas - HDP
consciousness 0.88
(0.29) 		Selahattin Demirtas - HDP
emotional.stability 2.68 **
(1.07) 		Selahattin Demirtas - HDP
openness 0.93
(0.29) 		Selahattin Demirtas - HDP
Observations 981
R2 Nagelkerke 0.734
  • p<0.1   ** p<0.05   *** p<0.01

Reminder for big5 when reading the tables: high scores high extraversy

high scores low agreeablness

high scores high consciousness

high scores low emotional stability

high scores high openness

Results are as expected: An example: as perceived agreeableness of Erdoğan decreases, outgroup voters prefer alternative(read as their) candidates.

the rest of the models include the other IVs.

  1. Base models

  2. Models with campaign negativity

  3. Models with affective polarization

  4. Models with campaign negativity and affective polarization combined

  5. Models with AKP proximity

  6. Full models

Some findings:

On negativity models (models b and d, read the neg_level variable): As Erdoğan goes negative voters tend to vote Akşener compared to Erdoğan (same for Ince but not for Demirtaş)

On affective polarization models (models c and d, read the aff_pol variable): Findings on those bothers me: Although there is no significant reşationship on its own (models c and d), when combined with partisanship variable it becomes significant in the expected way. Would like to hear your comments on this

The full model also provides interesting findings that we can comment on. For example the insignificant esea variable for Demirtaş, is inline with the argument that he was napproved by non Kurdish voters as well.

About how to proceed: I think this will be enough to jump in to the workshop in November. I will be formulating an abstract and share it with you. If you have any objections, updates and/or recommendations, we can circulate it among us via email.

Thats all for now. Best E.