Files in this item



application/pdfZHONG-DISSERTATION-2021.pdf (795kB)Restricted to U of Illinois
(no description provided)PDF


Title:Candidates, the economy and voting behavior
Author(s):Zhong, Wei
Director of Research:Gaines, Brian
Doctoral Committee Chair(s):Gaines, Brian
Doctoral Committee Member(s):Winters, Matthew S.; Bowers, Jake; Cho, Wendy K. Tam
Department / Program:Political Science
Discipline:Political Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Negative partisanship
Economic voting
Political sophistication
Abstract:This dissertation includes three independent but related essays on political behavior. The first essay, “Candidate Evaluation and Negative Partisanship: Evidence from the 2016 U.S. Presidential Election", builds on existing literature arguing that candidates are playing more and more important roles in elections, but at the same time affective polarization is intense, and negative partisanship is critical to vote choice. Negative affection to the opposition party leads voters to be more loyal to their own party. A natural question then arises, that is, do candidates matter to negative partisanship? Do they promote strong oppositions by emphasizing negative advertising and campaigning? Are candidates’ traits, issue positions, or ideologies more likely to affect party loyalty in voting than the other? Therefore, in the first essay, I examine candidate effects in negative partisanship using 2016 U.S. presidential election data to explore whether and how candidates matter for party loyalty in voting. What does it mean to say that they might or might not? I exploit the largely ignored open-ended questions asking for evaluations of political candidates in the American National Election Studies. These rich data offer the best clues as to how opinions about parties, candidates’ traits and personalities, and issues inter-mix in voting decisions. Given such large text, I applied natural-language processing techniques including tokenization, lemmatization, and doc-term matrix, to open-ended responses to construct a matrix of relevant (repeated) terms, and train a machine learning classifier that classifies vote choice from those data. I show that the more voters dislike the opposing party’s candidate in terms of personal traits and characteristics, the more loyal they are to their own party’s candidates. Voting, in this sense, is driven more by negative reactions than by positive ones. The second essay, “Economic Information in Retrospective Voting", focuses on voters’ differences in understanding of the economy, and how this variation leads to their evaluation and interpretation of the government’s economic policies and the incumbent’s performance. I focus on the research question: what types of economic information do voters use to form the concept of the national or aggregate economy? What does the economy mean to them? How do these choices influence the formation of their broader economic retrospections and political support to incumbent? To answer these question, I use individual-level panel survey datasets and show differences of choices of economic indicators do contribute to different retrospective economic assessment. For example, in 2019, people who are highly attuned to either job or investment indicators are more likely to report higher and positive retrospective economic assessment, whereas groups who pay attention to living costs and tax information are more likely to say economic conditions are worse under Trump administration. The choice of these economic indicators is affected by individuals’ broad economic experiences, including individual own economic structure, financial insecurity and partisan attachment. Specifically, citizens are more likely to choose economic indicators that are highly correlated with their economic well-being, such as unemployed workers preferring job information, and citizens with investments preferring to consider stock market prices when thinking about the economy. Likewise, strong financial insecurity increases citizens’ willingness to learn about all kinds of information, in particular living-cost information. These resulting differences in perceptions are also politically important: individuals who seek out particular economic information will strongly react to the incumbent’ s relevant economic policy and hold incumbents to account during elections, compared to individuals who care less about the same information. The result contributes to the general literature on economic voting by clarifying how voters use economic information and form concepts of the economy, and use these to assess the incumbent. It also contributes to the literature on heterogeneity in voters’ economic evaluations by drawing attention to the distinction and motivation between economic information and resulting voting behaviors. In the third essay, I re-examine and unpack the role of political sophistication on a survey-experiment dataset, in explanations of heterogeneity on discerning credible and less credible information sources when voting. I use data from Argentina, a survey experiment, which tests if voters can discern credible and less credible information sources about incumbent’s corruption. To explore which of the various components of sophistication, including political knowledge, education and political discussion, multiplicatively contribute to discerning the information sources, I employ a causal forest model. I show that the heterogeneity is mainly driven by an individual’s region and age as well as relevant political sophistication factors including education and political knowledge. Moreover, investigating the interactive relationship between treatment effects and these two sophistication factors. I observe that educational attainment and political knowledge can be substitutes for each other. I also find that groups with a mixed combination of education and political knowledge are not distinctive from each other in discerning more credible from less credible information. Although political discussion itself does not contribute to the heterogeneous treatment effects, along with education or political knowledge, discussions promote the differences between the least and most educated or knowledgeable group in distinguishing between more credible and less credible information. This also highlights the heterogeneity of citizens when processing information relevant to voting behavior. The paper contributes to the operationalization of sophistication. Researchers often omit a group that might have mixed education and knowledge (e.g., high education and low political knowledge, or vice versa). As a result, we know comparatively more about the multiplicative effects of the various components of sophistication in voting behavior.
Issue Date:2021-07-12
Rights Information:Copyright 2021 Wei Zhong
Date Available in IDEALS:2022-01-12
Date Deposited:2021-08

This item appears in the following Collection(s)

Item Statistics