Who are the targets of electoral coercion? Evidence from Turkey

Emre Toros & Sarah Birch

The Research Question

  • Although electoral violence and electoral coercion have gained increasing attention, most studies of this topic rely on macro-level data to discern overall causal patterns.
  • This is an important object of research, yet under-researched, as the fear or experience of coercion can fundamentally alter citizen orientation toward electoral processes.
  • Who gets targeted for electoral coercion?

The Literature

Political Composition

Swing and weakly-aligned voters (Robinson and Torvik 2009; Bratton 2008; Collier and Vicente 2012)

Solid party supporters (Gutierrez-Romero 2014; Gutierrez-Romero and LeBas 2018; Hoglund and Piryathne 2009)

Demographics

Poor and rural voters (Gonzales Ocantos et al. 2017)

Urban voters (Dercon and Gutiérrez-Romero 2012)

Least educated (Bratton 2008: 624)

Young and poor (Mares and Young 2016: 282)

Hypotheses

The studies cited above reply on direct reports by citizens to survey researchers, and answers to direct questions about experience of coercion may themselves be affected by distrust and fear.

It is for this reason that we employed a list experiment to tap concerns about electoral coercion, as list experiments enable researchers to gauge respondent concern indirectly and unobtrusively.

Hypothesis 1

  1. Voters with weaker partisanship levels are more likely to experience coercion designed to make them vote for a particular party

Hypothesis 2

 2. Socio economic factors like regional differences, educational levels, gender and income shape voter coercion

The 2018 Turkish Context

  • Electoral storm:

    • 3 parliamentery, 1 local, 1 constitutional referendum, 1 presidential, since 2014: 6 in total, still one to come in 2019

  • Turkey has regularly held competitive multiparty elections, yet more recently, concerns have regularly been raised about electoral integrity.

  • Although allegations of electoral violence and fraud started to become more visible during the 1990s, The decades, starting with the 2002 elections were marked by unprecedented number of accusations of electoral fraud and violence.

  • Thus, though Turkey has a history of elections which have created legitimate governments based on the popular will, recent developments have raised eyebrows about the integrity of electoral processes. This makes the Turkish case a relevant context in which to probe the research questions set out above.

Data and Methods

  • Data is collected in July 2018, just after the June 2018 presidential and parliamentary elections. Countrywide Turkey representative, 1232 face to face interviews.

  • List experiment (Imai 2011; Blair and Imai, 2012)

    • Direct question:
      • Did someone close, like somebody from work or family, force you to vote for a specific party or candidate?

    • List experiment question:

      • The choices of voters for voting a particular party or candidate are shaped by a number of factors. Now I will read some of these factors. We are curious about which factors actually shaped your decision. Please do not tell me which ones, just tell me only how many.

      I voted for the particular party or candidate because …

  1. Control Group
    1 It is time to change the administration
    2 I liked the pledges of party and/or candidate
    3 I want stability in administration

  2. Treatment Group
    1 It is time to change the administration
    2 I liked the pledges of party and/or candidate
    3 I want stability in administration
    4 Someone close forced me to vote for that specific party or candidate

Results - Descriptive

. Asked.Directly List.Experiment
Control NA 1.78
Treatment NA 1.98
Estimated(%) 7.22 20.00
  • There is a 12.78 percent increase when asked the sensitive question by a list experiment and the 20 percent difference between means of control and treatment which is significant at 0.01 level

plot of chunk unnamed-chunk-3

Results - Models on H1: The AKP

  • Analysis with the R package “list: Multivariate Statistical Analysis for the Item Count Technique”
  • H1: Voters with weaker partisanship levels are more likely to experience coercion designed to make them vote for a particular party

plot of chunk unnamed-chunk-4

  1. The coefficient sign for weak partisanship changed from insignificant-negative to significant-positive with the list experiment tool, indicating a strong coercion on weak AKP partisans
  2. Weak AKP partisans are more subjected to electoral coercion compared to strong AKP partisans, controlled for education, income, region and strong partisanship.

Results - Models on H1: The CHP

  • Analysis with the R package “list: Multivariate Statistical Analysis for the Item Count Technique”
  • H1: Voters with weaker partisanship levels are more likely to experience coercion designed to make them vote for a particular party

plot of chunk unnamed-chunk-5

  1. Although the coefficient sign for weak partisanship has changed from negative to positive, it is not significant neither for weak nor for strong affiliated CHP voters

Results - Models on H2

  • H2: Socio economic factors like regional differences, educational levels, gender and income shape voter coercion.

plot of chunk unnamed-chunk-6

  1. Contrary to the general belief in public, there is no statistically significant difference between males and females.
  2. High educated voters are less likely to be subjected to electoral coercion compared to low educated ones.
  3. Voters living in East and South East Anatolia are more likely to be subjected to electoral coercion compared to voters living other regions of Turkey.

Discussion and Conclusion

  • First scientific effort on electoral coercion with micro data for the Turkish context.
  • Once neutralized by the list experiment, it appears that one fifth of voters experience electoral coercion.
  • Damage on the normative appeals of democracy: equal rights, autonomy, criminalization of politics; Schaffer (2007).