Three essays on the determinants of voting behavior

Resumen

This collection of essays focuses on the study and analysis of electoral behavior in its relation to the political, economic, social and geographical context. Thus, this thesis is an interdisciplinary work that brings in new theoretical and methodological approaches to contribute to interesting questions that are not yet completely understood. The study of voting behavior is particularly relevant in the current context of events in which we do not fully understand yet the causes of new political dynamics such as the rise of radical and populist parties or the high persistence of electoral behavior; as well as the political consequences of these phenomena and others for left-behind areas, for the dynamics of political change, the current decline of established parties in many countries, or the growing disaffection of voters with the democratic system around the world. The issues that this thesis analyzes are also relevant to those interested in using new methods to generate causal answers to research questions. This thesis uses research designs and techniques that rely on a variety of natural experiments or quasi-experiments that provide our results with strong internal validity. While the use of quasi-experimental techniques per se is not a novel contribution of the thesis itself, the application of such techniques to the particular questions that I study is a contribution to the fields of political economy and political behavior. The main theoretical idea of the thesis is, then, that voting is highly endogenous, with interrelated determinants that evolve over time. That is, voting is a continuous behavior that is modified by new social, economic, or political factors. These factors are the new social cleavages reported by (Ford & Jennings, 2020). For example, place of residence, education level, or age are new social cleavages that can alter the continuum of voting behavior over one’s life. For this reason, the answers provided by much of the previous literature are not sufficient to explain the new political issues. Therefore, it is relevant to conduct new research to better understand the new determinants of these patterns. The main thesis put forward here is that voting is a continuous and endogenous behavior, where the greatest explanatory weight falls on the individual’s past behavior, but with such weight evolving over time as society changes. Thus, the specific goal of the thesis is to find out what influences voting behavior, whether it is one’s own experience, the evolution of the environment, or geographic location, in order to find out what are the new determinants of this behavior. To this end, it focuses on examining new political phenomena that are not yet fully understood. Specifically, one of the phenomena that I study is the decline of mainstream parties and the rise of new or challenger parties in recent years, such as Movimento Cinque Stelle in Italy, Podemos or Ciudadanos in Spain or En Marche in France (see De Vries & E. Hobolt, 2020). Similarly, the thesis also contributes to our understanding of a second recent phenomenon such as the rise of radical right-wing populist parties such as Fratelli in Italy, Vox in Spain, Chega! in Portugal, Golden Dawn in Greece, Front National in France or AfD in Germany. To explore these questions, the thesis develops two converging but distinguishable lines of research. The first line of research examines how the election itself, that is, the experience of casting a vote for the first time (also called voting shock or voting experience), influences voting behavior. This first line of research is related to the endogeneity and continuity of voting behavior, and to political theories that describe the act of voting as a habit and analyze how the trajectory of voting over people’s life changes with accumulated experience with elections. This line of research aims to verify whether there is a persistence of voting behavior, its evolution across different social contexts, and the timing of the effects (short- and long-term). In this regard, the first chapter of the thesis addresses the short-term effects of having had a first experience with voting, while the second chapter analyzes the long-term effects. The third chapter deals with a second but related line of research that examines how location determines voting. This line of research addresses political geography as a divide, examining how unequal access to public goods and services affects people’s voting behavior. In developing this line of research, I seek to determine the impact of place on voting and how the political system’s decisions about the location of public services is a critical factor in the emergence of new parties. The high endogeneity of voting motivates the use of advanced causal estimation techniques. Theoretically, one of the reasons why endogeneity occurs is that current behavior may be, at least partially, explained by past behavior. How to deal with the presence of this type of endogeneity is a very salient issue in the social sciences because there is always limited information and it is very difficult to account for explanatory variables that come from the past. Econometrically, endogeneity occurs when relevant explanatory variables are omitted from the estimation which makes the unobserved factor of the estimate (error term) to be correlated with the dependent variable. There are several ways in which one could address endogeneity. For my purposes, two relevant approaches are based on identifying treatment and control groups that may be equally affected by unobservables that could create endogeneity: 1) A first approach is a fully experimental design, in which groups of equals are randomly assigned into a treatment or a control group; 2) A second approach is a quasi-experimental approach using natural experiments. This second approach uses observational data. A quasi-experimental approach is based on the idea that there are certain situations in which a distinction between groups of equals arises through the natural and quasi-randomized assignment of a variable that acts as a treatment. As a result of this quasi-random partitioning between equals, the effect of the treatment variable can be estimated. By exploiting this natural exogenous scenario with causal inference techniques such as synthetic control, propensity score matching, difference-in-differences, or regression discontinuity, among others, it is possible to avoid estimation biases. In the dissertation, I use a variety of quasi-experiments and techniques. For example, in the first chapter I use the minimum voting age as an exogenous variable that causes a difference in the first-voting experience and combine it with a difference-in-differences approach. In the second chapter, I also leverage from the quasi-random differences created by voting age thresholds, but use a regression discontinuity design to analyze long-run effects of voting. In the third chapter, I exploit the fact that there are arbitrary/fortuitous regional boundaries that lead to differences in access to public goods and services. Those boundaries create quasi-random variation in access to public services that can be exploited using a geographical regression discontinuity design. In addition to the use of quasi-experimental methods, in recent decades there has been a great advance in political geography, especially in the application of geographical techniques applied to the study of political or social issues. We can find Rodden (2010) and W. T. Cho & Gimpel (2012) as precursors of the use of Geographic Information Systems (GIS) in political science. The advancement of geographical software along with the universalization of its use through data analysis software such as R, Stata or Python has expanded the possibilities of using GIS. The main contribution of applied geography is the possibility of studying the context surrounding political phenomena (Ethington & McDaniel, 2007). Knowing the context makes it possible to discover the endogenous factors related to the environment. The potential of spatial analysis as a research tool was already demonstrated in early works such as Snow (1948), who relied on geographical differences to determine the causes of cholera. In the case of political science, the potential of applied geography as a tool to study relevant research questions is also large, making it possible to study questions that we could not study otherwise. In this regard, in the last decades there have been a number of very relevant papers that have used geographical analyses to answer questions that were extremely difficult to answer before, such as for example Peisakhin & Rozenas (2018); Albertus (2020); Monogan et al. (2017),or Nall (2015). The third chapter of this thesis is also an example of how applied geography is a great tool for political science, as it exploits the link between geography and natural experiments through the application of GIS and Geographical Regression Discontinuity to the study of the effects of differences in public services on the vote. Although I use Spain as a case study in the three chapters, this thesis is not about Spain, but about voting behavior and its new determinants, broadly understood. Spain, due to its historical, geographical, social and political characteristics, is an ideal scenario to conduct advanced quantitative research on electoral behavior. The internal validity of my estimates can then be supported with further analyses of other countries in which similar political problems arise. In this way, I believe that the conclusions drawn here for the case of Spain can be of general interest for people interested in current issues in political science. Regarding experience, I find that electoral experience leads to positive reinforcement of past behavior, i.e., contributing to create a habit of voting (see Green & Shachar, 2000; Shachar, 2003). In the first chapter of the dissertation, I find, using differences in voting exposure due to the natural experiment provided by eligibility thresholds, that the role of the first experience with voting depends strongly on the social context in which the first experience occurs. I find that the social context makes a difference in whether and how experience has an effect on future electoral behavior. Those who had a prior voting experience remain more loyal to mainstream parties in the new anti-mainstream context that aroused after the Great Recession. The main contribution of this result lies in the fact that, to my knowledge, this is one of the first studies to estimate the conditional role of social and economic context in strengthening or weakening the effect of electoral experience. Electoral exposure or experience is not only relevant for future voting behavior in the short-term. I find that the effects of first-voting experiences persist in the long term. The second chapter of the thesis shows that the effect of the initial voting experience lasts for at least four decades. I am able to isolate the effect of having had an experience with voting from the effects of other posterior experiences by using the natural experiment provided by the interruption of democracy in Spain. I find that people with previous experience with voting (e.g. people that were eligible to vote in 1936) turned to the polls at a higher rate than inexperienced voters in the first democratic elections after the return to democracy in Spain in 1977. To my knowledge, the findings of this chapter are novel because they show that the effects of previous experiences with democracy last for much longer than previously believed. Additionally, the effects I find have been obtained using research strategies with strong internal validity. This fact allows me to provide a strong causal interpretation. Moreover, in the second chapter I also analyze the potential existence of heterogeneity in these long-term effects of experience with voting across different population groups. I find that the effect of experience is not homogeneous: voters’ socio-demographic characteristics determine whether the effect exists or not and its magnitude. Another contribution of the dissertation is to analyze to what extent the differences in the provision of public goods and services affects voting behavior. I find that voting behavior changes due to differences in this provision in a context in which differences are created by decentralization processes. In order to reach this conclusion, I leverage the fact that political boundaries do not always follow natural geographic boundaries. Arbitrary political boundaries create cases in which geographically similar municipalities are artificially split into two different political regions, creating inequalities in access to public services for otherwise equal citizens. Inequalities in the access to public services lead to dissatisfaction with the system, which is then reflected in subsequent voting behavior. Those voters that are artificially placed into regions with worse access to public services are more likely to vote for anti-establishment parties and for parties with strong preferences for a more centralized provision of public services. These conclusions have several implications. One is that voting experience plays a critical role in both party choice and turnout. I find that in a situation of contextual economic and political change that creates a window of opportunity for challenger parties, experience serves as a shield against change. Voters with previous experience with voting are less likely to vote for challenger parties, slowing the process of transformation of the party system. Related to this, electoral experience itself is, then, an explanatory mechanism for why party systems are so stable over time. Thus, the high stability of the party systems found by Lipset & Rokkan (1967) is consistent with my findings. It is difficult to change the political system, even in an environment that is susceptible to change, because the fact that elections are repeated every few years strengthens the loyalty of the electorate to previous voting options. These findings underscore the importance of repeated elections for the stability of democratic systems. The status quo seems to be better preserved as more citizens are previous voters, acquire electoral experience and create loyalties to certain actors in the system. In contrast, in circumstances where the number of voters who are excluded from participating in past elections is large, it is more likely that challenger parties will grow in support. The dissatisfaction with the system can also occur geographically. Sectors that feel "left behind" because they suffer the negative consequences of the geographically unequal distribution of public goods and services may manifest this dissatisfaction into voting for challenger parties. The changes in political behavior that these differences in access to public services produce are a side effect of decentralized political systems that require further analysis. The conclusions of the thesis also have policy implications, as they inform our knowledge of which factors make democracy more stable and citizens less radical. Implementing mechanisms to favor participation in elections is key for improving the functioning of democracy. The results also suggest that developing mechanisms for collaboration among different levels of government in the distribution of public goods and services between regions so as to reduce inequalities is a way of avoiding the feeling of abandonment of citizens and the political reactions that come with it. It is also worth mentioning that this is not a dissertation on Spain, but a study conducted with evidence coming from Spain. This makes it possible to draw conclusions about Spain and its democracy that can be applied to other countries. In this regard, the second chapter of the thesis explores the legacy of experience with democracy during the Second Republic when democracy came back forty years later. As democratic interruptions have also occurred in other countries, it is relevant to find that electoral experience in the past creates a tie with democracy that survives even after long periods of dictatorship. Finally, while the results of this thesis are relevant, they also leave many questions unanswered. One clear limitation of my findings is that given the identification strategies that I use, results have strong internal validity but limited external validity. There is, thus, room to contribute to similar research questions in the future by analyzing similar topics in other countries and contexts in order to generalize the conclusions obtained in this thesis.