The devil in the details: Influence of group characteristics on voting behavior in the Slovak National Council as seen through multiplex social network analysis

• Thesis title in Czech: Ďábel se skrývá v detailech: Vliv skupinových charakteristik na volební chování v Národní radě Slovenské republiky z pohledu multiplexní analýzy sociálních sítí
• Thesis title in English: The devil in the details: Influence of group characteristics on voting behavior in the Slovak National Council as seen through multiplex social network analysis
• Key words: multiplexní graf|znaménkový graf|detekce komunit|volební chování|analýza sociálních sítí
• English key words: multiplex graph|signed graph|community detection|voting behavior|social network analysis
• Academic year of topic announcement: 2022/2023
• Thesis type: Bachelor's thesis
• Thesis language: angličtina
• Department: Department of Sociology (21-KSOC)
• Supervisor: Mgr. Matouš Pilnáček
• Author: hidden - assigned and confirmed by the Study Dept.
• Date of registration: 25.11.2022
• Date of assignment: 25.11.2022
• Administrator's approval: approved
• Confirmed by Study dept. on: 26.11.2022
• Date and time of defence: 13.06.2023 09:00
• Date of electronic submission: 09.05.2023
• Date of proceeded defence: 13.06.2023
• Submitted/finalized: committed by student and finalized
• Opponents: doc. PhDr. Lukáš Linek, Ph.D.

Guidelines

Research question: How do sociodemographic variables relate to the voting behavior of National Council members?
Background:
My thesis has the main goal of combining one methodological and one conceptual approach to explore an innovative social research method as well as verify existing knowledge about parliamentary voting behavior.
From a methodological standpoint, this thesis is an attempt at exploring the use of „multiple multiplex network analysis“ for analyzing the behavior of Slovak National Council MPs. Multiplex network analysis is an offshoot of network analysis, which combines various networks atop each other with the purpose of more complex modeling of all interactions where a single network expression of the voting behavior would result in significant data loss. Arinik et al. (2020) focus on recognizing clusters in multiple multiplex networks of European Parliament roll-calls (voting in the deliberative assembly context). These authors point out the fact that this method of modeling roll-calls losses less information than aggregated networks, which collapses roll-calls into some averaged value. Multiplex network analysis is data-intensive and therefore relatively new and hence why there is as of now space to replicate their methodology for other deliberative assemblies. This is why I consider the application of their research methods on the Slovak parliament as a potentially useful advanced research method for other social network analysis research.
I shall use this innovative research method as a tool to research the question of whether sociodemographic variables are significantly related to specific patterns of the voting behavior of MPs. As per Tunkis (2017), group sociodemographic characteristics different from party affiliation have a significant impact on subsequent voting behavior with the rest of the party members. However, Tunkis uses a measure of voting loyalty as the dependent variable. With the help of the above-mentioned multiplex network, one could recognize the influence of group characteristics also on groups of MPs that are affiliated with opposition parties. Thus, we could not only verify Tunkis’s findings but also find discrete “latent groups” of MPs across the whole parliament. This method has the potential of being used for all sorts of deliberative assemblies to look for latent voting cliques unified by something other than party affiliation.

Aims:
The main aim is two-fold. First is the attempt at replication of Arinik et al. (2020) research method on local deliberative assemblies as well as a broader exploration of this method as a useful innovative tool to be used for the research of group sociology in general.
Secondly, the thesis aims at an empirical replication of the results of existing research, which says that group sociodemographic characteristics are related to the voting behavior of MPs. Even though Tunkis (2017), as well as Škvrňák (2020), don't use social network analysis as their research method, notwithstanding I consider it useful to presume that their conclusions shall be reflected in some way in my multiplex social network analysis results.

Methods:
Methodically, my thesis shall attempt the procedure of Arinik et al. (2020) with two major changes. First is that the thesis will focus on the Slovak national legislative assembly (National Council of the Slovak Republic) and its roll-calls between 2016-2020, as opposed to the European Parliament of Arinik et al. (2020). Secondly, the thesis will expand the categorization of MPs above just party affiliation. Besides this, the thesis shall include information about the MP’s gender, level of education, age, and ethnic identity.
More precisely, a multiplex social network shall be used to represent all the roll calls over the given period. MPs’ votes shall be recorded onto an adjacency matrix of each roll call. This shall be repeated over all roll-calls of the given period. Next, signed graph partitioning and similarity measures of all possible behavior patterns shall be performed. These with subsequently undergo characteristic cluster analysis, from which voting behavior patterns that have the highest explanatory power of the network shall be produced. This process shall be repeated not only for party affiliations, but also age, education, gender, and ethnic identity. Its result shall be patterns of behavior of MPs with patterns determined by the degree to which different categories will be represented in each pattern. These patterns shall be used finally to compare the findings of existing literature with my results.

Hypotheses:
It can be presumed that at the level of party affiliations, at least two clusters shall be present: the coalition parties cluster and the opposition parties cluster. Furthermore, we can presume that MPs shall converge toward the majority decision of their party affiliation. This presupposition can be expressed as two hypotheses:
H1: Being a member of a party influences the voting behavior of the said party member in favor of the party's majority decision.
H2: Being a member of a coalition party influences the voting behavior of the said party member in favor of the coalition majority decision.
Regarding education, gender, age, ethnicity, and other sociodemographic variables, there is little evidence to explicitly presume some relationship between any single variable and voting behavior. Notwithstanding, the aforementioned existing literature talks of group characteristics' influence on voting behavior. Thus, while I shall presume there is no effect of these variables on voting behavior, I nonetheless aim at looking into whether this is indeed the case by testing these hypotheses on the multiplex social network, as there is a potential of finding counterintuitive results. These presumptions shall be formulated into these hypotheses.
H3: Highest achieved education does not influence the voting behavior of an MP in favor of other MPs with the same education level.
H4: Gender does not influence the voting behavior of an MP in favor of other MPs of the same gender.
H5: Age does not influence the voting behavior of an MP in favor of other MPs of similar age.
H6: Ethnicity does not influence the voting behavior of an MP in favor of other MPs with the same ethnicity.

Characterization of the conclusion:
My first expected conclusion is the evaluation of multiplex social network analysis as a generalizable tool for research of deliberative assemblies and other social groups. This shall be evaluated by the ability to describe voting behavior within an assembly as well as the ability to infer higher-level information like behavior patterns from the multiplex social network.
My second expected conclusion is the verification of hypotheses, which partly refering to the existing literature on parliamentary voting behavior as well as the influence of group characteristics on said behavior. In case the replication of the results shall be unsuccessful, we will be able to say that these group characteristics cannot influence the voting behavior of MPs.

References

Arinik, N., Figueiredo, R., & Labatut, V. (2020). Multiple partitioning of multiplex signed networks: Application to European parliament votes. Social Network
Michael Škvrňák (2021) You’ll never rule alone: how football clubs and party membership affect coalition formation, Local Government Studies, 47:2, 312-330, DOI: 10.1080/03003930.2020.1787167
Peter J. Tunkis (2017) The Ties that Bind: Do Group Associations among Legislators Matter for Political Parties?, Problems of Post-Communism, 64:2, 79-93, DOI: 10.1080/10758216.2016.1172237