AMERICAN POLITICAL POLARIZATION IN THE 2010s by NATHANIEL OLDS A THESIS Presented to the Department of Political Science and the Robert D. Clark Honors College in partial fulfillment of the requirements for the degree of Bachelor of Science May 2022 An Abstract of the Thesis of Nathaniel Olds for the degree of Bachelor of Science in the Department of Political Science to be taken June 2022 Title: American Political Polarization in the 2010s Approved: Priscilla L. Southwell, PhD Primary Thesis Advisor It is often claimed that political polarization is becoming more and more rampant in the United States, but researchers in the field of political science debate this. Some support this claim and some argue against it, but many accept that it is more complicated than it is made out to be. This thesis examines some of the competing investigations on the subject and expands on them, using data from the American National Election Studies (as many others have done in the past). Data is pulled from the 2012, 2016, and 2020 ANES Time Series, which poll thousands of people in conjunction with each of these years’ November elections. Although there is no one way to measure this somewhat subjective subject, this thesis focuses on cross- referencing the ANES Party Identification variable with the Feeling Thermometer variable (a measure used to determine respondents’ feelings towards a particular subject). Feeling Thermometers for both the Republican and Democratic Parties are analyzed to view how public opinion towards these parties has changed over the past ten years. In terms of these measures, this thesis finds indications that polarization is relatively high, but inconclusive support for the claim that polarization has consistently and dramatically increased over the past decade. ii Acknowledgements I would like to thank my primary advisor, Professor Priscilla Southwell, for her essential guidance through the research process, as well as my second reader, Professor Neil O’Brian, and my Clark Honors College representative, Professor Carol Paty, for serving on my thesis committee and supporting me throughout the execution of this project. I would also like to thank Cameron Mulder and Taha Mirghorbani for their assistance in determining the best approach to locating and analyzing the data involved in this project. I would especially like to show my appreciation to Gabriele Hayden for her invaluable assistance through the specifics of the data analysis and presentation process. Additionally, I would like to recognize Miriam Alexis Jordan and Alex Brown for their help with formatting and Clark Honors College guidelines. Finally, I am eternally grateful to my family for supporting me through these often-confusing four years. iii Table of Contents Introduction 1 A Polarized Nation? 1 Defining Polarization 2 Differences of Opinion (Does Polarization Exist?) 3 Methods 9 Selecting Variables for Polarization 9 Specifics of Variables and Data 11 Methods of Data Analysis 12 Data and Analysis 15 Conclusion 33 Bibliography 36 iv List of Figures Figure 1: Party ID and Feeling Thermometer: Republican Party, 2012 ANES Time Series, Stacked Column Graph 15 Figure 2: Party ID and Feeling Thermometer: Republican Party, 2012 ANES Time Series, Clustered Column Graph 16 Figure 3: Party ID and Feeling Thermometer: Democratic Party, 2012 ANES Time Series, Stacked Column Graph 18 Figure 4: Party ID and Feeling Thermometer: Democratic Party, 2012 ANES Time Series, Clustered Column Graph 18 Figure 5: Party ID and Feeling Thermometer: Republican Party, 2016 ANES Time Series, Stacked Column Graph 20 Figure 6: Party ID and Feeling Thermometer: Republican Party, 2016 ANES Time Series, Clustered Column Graph 20 Figure 7: Party ID and Feeling Thermometer: Democratic Party, 2016 ANES Time Series, Stacked Column Graph 22 Figure 8: Party ID and Feeling Thermometer: Democratic Party, 2016 ANES Time Series, Clustered Column Graph 22 Figure 9: Party ID and Feeling Thermometer: Republican Party, 2020 ANES Time Series, Stacked Column Graph 23 Figure 10: Party ID and Feeling Thermometer: Republican Party, 2020 ANES Time Series, Clustered Column Graph 24 Figure 11: Party ID and Feeling Thermometer: Democratic Party, 2020 ANES Time Series, Stacked Column Graph 26 Figure 12: Party ID and Feeling Thermometer: Democratic Party, 2020 ANES Time Series, Clustered Column Graph 26 Figure 13: Party ID and Feeling Thermometer: Republican Party, Average Feeling Thermometer Rating Across All Three Years, Line Graph 28 Figure 14: Party ID and Feeling Thermometer: Democratic Party, Average Feeling Thermometer Rating Across All Three Years, Line Graph 28 Figure 15: Party ID and Feeling Thermometer: Republican Party, All Years, Stacked Column Graph 30 Figure 16: Party ID and Feeling Thermometer: Democratic Party, All Years, Stacked Column Graph 30 v List of Tables Table 1: Party ID and Feeling Thermometer: Republican Party, 2012 ANES Time Series 15 Table 2: Party ID and Feeling Thermometer: Democratic Party, 2012 ANES Time Series 17 Table 3: Party ID and Feeling Thermometer: Republican Party, 2016 ANES Time Series 19 Table 4: Party ID and Feeling Thermometer: Democratic Party, 2016 ANES Time Series 21 Table 5: Party ID and Feeling Thermometer: Republican Party, 2020 ANES Time Series 23 Table 6: Party ID and Feeling Thermometer: Democratic Party, 2020 ANES Time Series 25 Table 7: Party ID and Feeling Thermometer: Republican Party, Average Feeling Thermometer Rating Across All Three Years 27 Table 8: Party ID and Feeling Thermometer: Democratic Party, Average Feeling Thermometer Rating Across All Three Years 27 vi Introduction A Polarized Nation? Public figures, journalists, and political observers often declare that the United States is becoming more and more polarized. The sentiment is so prevalent in our current climate that it has become commonly accepted. However, the term “polarization” has been discussed so much in this context that it has become difficult to determine its definition, let alone its extent. One would hope that this amongst scholars of political science, the discourse would be much clearer, but even in that context, there are significant disputes as to the definition and veracity of the term. It is defined by different boundaries, different data points, and even different words within the category of “polarization.” In most forms, the idea of U.S. politics becoming more and more polarized is a very significant shift, one that has become increasingly salient. We need some consensus on what the term means if we hope to address it, but this cannot be achieved in isolation. The goal of this project is not to fully answer the question of whether the U.S. is becoming more polarized, as many scholars with more resources and experience have been unable to do so. My goal is to take a broader look at this debate, propose a unique view on the subject, and hopefully, add a new perspective to the conversation. Ultimately, it seems that a coherent answer to the overall question of U.S. polarization is difficult to articulate, for multiple reasons, but indications can be found within certain types of data that provide evidence of slightly worrying polarization, but also a seemingly less pronounced increase than is often described. Defining Polarization While many of us use the term “polarization” generally to refer to growing conflict between opposing political groups, it is helpful to reexamine this phenomenon, as it is often used to refer to slightly different political occurrences. Often, the term is used to refer to an actual shift in the viewpoints of political parties, such that (in the case of the United States) Republicans and Democrats begin to exhibit views that become more and more opposed to each other over time. In other instances, the term refers to an increase in animosity and tension between the opposing parties. To add another layer to these definitions, there is often a dispute as to whether they refer to party elites (actual politicians and representatives of political parties) or the general public, as some argue that the elites misrepresent the public and take more polarized stances (Fisher et al., 2013, p 88). One could even define polarization as simply a diminishing “middle ground” in politics, with a decrease of moderates and independents (which may be due to either people’s views becoming more extreme or simply a higher number of people deciding to join one of the major political parties). All of these definitions generally refer to an increase in conflict, although they display different aspects of this conflict and have different implications. Multiple definitions will be examined in the forthcoming analysis of previous research, as many of them interact with each other in both supportive and opposing ways. When it came to the purposes of this project, the definition was viewed through the lens of the changing opinions of the public towards political parties over time. However, the varying thoughts on the nature of “polarization” are very important to keep in mind in a general discussion of the subject. 2 Differences of Opinion (Does Polarization Exist?) Previous research on political polarization generally falls into two distinct sets of approaches and conclusions. One argues that the party elites and the electorate of both major U.S. political parties have grown further apart in their ideology over time. The other, however, argues that the phenomenon is more complicated. An important distinction is made between the elites and the electorate, and many argue that the latter is much less polarized than the former. Some argue that polarization as a shift in ideology is not as much of a factor as is an increasing number of people identifying with opposing political parties (Fiorina et al., 2008, p. 578). Layman, Carsey, and Horowitz (2006) provide an example of the first argument, stating that “the Republican Party’s elites, activist base, and electoral coalition have become much more traditionalist, whereas their Democratic counterparts have grown more modernist and secular” (Layman et al., 2006, p. 86). They describe this as mainly a cultural shift, suggesting that issues of morality and race extend the preexisting divisions between the parties and cause them to move further apart. They do express some doubt that public polarization is as pronounced as it is often made out to be, and acknowledge that such polarization may be less than posited. However, they note that “some party identifiers are moving their own attitudes toward the very liberal or very conservative positions of their party’s elites,” indicating some truth to the idea of polarization in the electorate (Layman et al., 2006, p. 94). Central to their argument is an analysis of data from the American National Election Studies (ANES) that seems to show growing divergence between the position of the Republican and Democratic 3 Parties on social welfare, cultural issues, and racial issues (Layman and Carsey 2002, as cited in Layman et al., 2006). Fiorina and Abrams (2008) offer a different take, arguing that recent “polarization,” as it is usually described, is actually “party sorting,” which occurs when “subpopulations…sort themselves out in ways that heighten their differences” (Fiorina & Abrams, 2008, p. 578). Furthermore, they also highlight analyses by of ANES data on welfare, culture, and race from around the same time as Layman et al. They argue that “on some issues there appears to have been little sorting, and on other issues the sorting appears largely limited to one party while the other party remains unchanged or even becomes less well-sorted” (Levendusky, 2006, 2007, as cited in Fiorina & Abrams, 2008, p. 578). This contrast between different conclusions from research working with the same data highlights the difficulty of analyzing this particular subject. The exact definition of “polarization” is difficult to determine, and even when looking at the same data, different scholars can draw different conclusions. It is also worth noting, however, that although they doubt the strength of public polarization, the same authors acknowledge that “there is general agreement among informed observers that American political elites have polarized” (Fiorina et al., 2008, p 565). There is nuance in both of these approaches, and neither provides an easy answer. The first approach to polarization also includes Abramowitz and Saunders (2008), who take issue with a book by Fiorina, Abrams, and Pope (2006), and push back against their claims that polarization is exaggerated (Layman and Carsey critique this same book, showing the ongoing debate between these scholars). By using the ANES data once more and calculating the difference between the number of liberal and 4 conservative positions on repeated questions over the years, the authors found “an increase in ideological polarization since the 1980s,” especially amongst citizens with high political interest and engagement (Abramowitz & Saunders, 2008, p. 544). The authors also utilize data from the Congressional Quarterly’s Guide to U.S. Elections to argue that “states have become much more sharply divided along party lines since the 1960s,” and note an increase in the partisan vote margin and number of uncompetitive states in national elections (Abramowitz & Saunders, 2008, p. 548). They use this to dispute another claim by Fiorina et al. (2006) that geographic polarization between red states and blue states is not as pronounced as it is assumed. To counter this, however, we can look at the work of Ansolabehere, Rodden, and Snyder (2006), who make a case for a “purple America.” Contrary to Abramowitz and Saunders (2008), they argue that when one extends the timeframe over the past century and takes into account more than just presidential elections, one can track a decreasing amount of single-party dominance across the states. They note that state legislatures are much more balanced now than in previous decades. At the time the article was written, it was typical that “neither party holds more than 60 percent of the seats,” and they state that “in no chamber today does one party hold 90 percent of the seats,” whereas such occurrences were common in the early 20th century (Ansolabehere et al., 2006, p. 114). Additionally, they also return to ANES data (along with the General Social Survey) to analyze the distinction between moral and economic polarization. They found that while polarization is apparent on moral issues, economic issues do not show the same trend, and furthermore, “the influence of economic policy preferences—on which the vast majority of Americans are moderates—has dominated 5 the more divisive moral issues in explaining vote choice” (Ansolabehere et al., 2006, p. 109). Intriguing trends can be observed in more recent articles, particularly those that review this older era of research. For instance, Baldassari and Gelman (2008) take a stance in line with Fiorina in their analysis of ANES data. They find insignificant support for ideological polarization, but they do find support for the idea that parties are “sorting voters along ideological lines” (Baldassari and Gelman, 2008, p. 439). However, Kozlowski and Murphy (2021) reevaluate this claim over a decade later. They argue that while Baldassari and Gelman’s analysis was accurate at the time, party sorting has ceased to tell the full story in the following years, and that issue alignment within these parties “is tighter now than at any period in at least 70 years” (Kozlowski and Murphy, 2021, p.11). This issue alignment provides a stronger argument that members of the parties are drifting away from the opposition and becoming unified at the extremes. However, this is surely not the end of the debate, as Fiorina and others continue to publish counternarratives as the back-and-forth continues. A particularly interesting evaluation of the subject is that of Westfall, Van Boven, Chambers, and Judd (2015), which takes a slightly different approach and addresses how we perceive polarization. The authors analyze public opinion of the concept of political polarization itself and determine the individual characteristics that may cause one to believe that the phenomenon is more common than it truly is. Using ANES data once more, they note that “those who perceive the greatest political attitude polarization in the United States—and, hence, those who most exaggerate political polarization—are those who are themselves most polarized” (Westfall et al., 2015, p. 6 155). In other words, having relatively extreme views and being politically active seems to be connected to a belief that polarization is rampant, which could result in an interesting feedback loop. This article serves as a sort of bridge between some of the competing narratives of the other articles: while it does seem to take actual polarization as a given, it also acknowledges that the phenomenon is often exaggerated by the public. Another noteworthy angle is the measure of general aggression and dislike towards opposing parties. Iyengar, Sood, and Lelkes (2012) examine this perspective of polarization through several data sources (including ANES once again). Their analysis argues that “Democrats and Republicans harbor generally negative feelings toward their opponents” and that “stereotypes of party supporters have become increasingly differentiated” by whether one refers to one’s own party or the opposition (Iyengar et al., 2012, p. 421). The authors associate these trends with an increase in polarization and value them over ideology as a measure of the subject. Among all of this, there also remains a final question of whether this seemingly increased animosity is simply a reflection of disillusionment with politics and political parties in general. Groenendyk (2018) argues that while we may see high partisan dislike towards opposing parties, this may be a result of partisans that “feel less enthusiastic about their party,” yet “resist changing their party identity” and justify it by focusing on their “hatred of the opposing party” (Groenendyk, 2018, p. 160). Varying explanations abound, with a relatively solid body of evidence for each one (often stemming from the same data). As the debate over what polarization is and whether it is present in U.S. politics continues, it becomes clear that the topic is incredibly 7 complex. Whichever approach, method, or data set one uses, there is always room for variation and debate. As such, no set of variables will tell the full story, but with some of the previous research on the subject in mind, one can attempt to approach the topic from a unique and useful angle. 8 Methods Selecting Variables for Polarization The analysis of specific issues, such the moral vs. economic dimension discussed by Ansolabehere et al. or the cultural angle brought up by Layman et al., seems to be something of a dead end, as it is quite an impossible task to arrange the vast array of political, social, and economic issues in such a way that accurately displays the country’s partisanship or creates a simple ideological scale. In order to produce anything relatively coherent and digestible, one must simplify to a set of core issues, preferably an attitudinal set that has been consistently measured over several years. However, this attitudinal set and the methods used to analyze it can vary greatly among researchers. Many scholars working with the ANES data have come to vastly different conclusions. As acknowledged before, differing conclusions are quite likely to arise in any effort to answer the question of polarization, but an effort to determine which political issues are the most representative of the nation’s electorate seems especially subjective. Similarly, an analysis based on political elites and their actions produces questions as to the proper approach. Ansolabehere et al. would argue that looking at presidential elections is not enough, and this is likely accurate. However, what would this wider analysis include? The parties to which elected officials belong? Or a more complicated analysis of their actions in office, which may include anything from how they vote to how they speak to each other? Even then, this does not answer the question of the nation as a whole. The “elites vs. public” polarization debate has been continuing for years, and it is commonly held that the former does not necessitate the latter, even 9 by those who argue both are occurring, such as Layman et al. (2006). It is also less of a point of contention that elites are becoming more polarized, so perhaps this ground is well-trodden. The electorate seems to be the best place to focus, and while a large set of ideological stances may be very subjective and hard to measure, citizens’ relationships with political parties are much more straightforward. Everyone makes affective judgements about politics, politicians, and political parties, even if those judgements boil down to not caring one way or the other. These judgements are arguably the measure that is most closely tied to everyday conflict, as it affects the manner with which we engage with politics in our daily lives. Someone who hates the opposing party and loves their own party will act a certain way when engaging in political discussion. An independent who has neutral feelings about both parties will engage in a different manner, and an independent who dislikes both parties will act in yet another way. The degree to which people feel animosity towards political parties is the core of the manner in which they treat people who are members of those parties or hold those beliefs, and it is a central part of polarization in people’s everyday lives. The research provided by Iyengar et al. (2012) seems particularly relevant to me as a result of this. As such, my analysis focused on the measures of public opinion towards the two major political parties in the United States: the Democratic Party and the Republican Party. With the electorate established as the center of the investigation and the parties established as the main variable, we can move to finer details of the data itself. 10 Specifics of Variables and Data As with many other investigations of political polarization, my analysis utilized data from the American National Election Studies. The data sets were pulled from the last three election years: 2012, 2016, and 2020. This decade is often presented as a time when many things changed in politics, in particular the contentious elections of 2016 and 2020. Focusing on these three elections will help to shine a light on this high- profile and relevant era, as well as update previous research. The parties were selected over their respective ideologies (liberal and conservative, respectively) as they are quite readily presented as the face of those ideologies in popular culture. Public opinions of Democrats and Republicans provide clearer data than public opinions of liberals and conservatives in general, as these categories are harder to define. Notably, the criteria by which the electorate was identified was “Party ID,” not “Party of Registration,” as the latter, while more precise and less subjective, cuts out the significant portion of Americans who are not registered to vote. For all three years, respondents were asked “generally speaking, do you usually think of yourself as [a Democrat, a Republican / a Republican, a Democrat], an independent, or what?” (ANES, 2020, p. 52). As for the measurement of opinion, the “Feeling Thermometer” was put to use. Feeling thermometers measure a respondent’s “warm” or “cold” feelings towards a particular subject on a scale of 0 to 100 (0 being the coldest and 100 being the warmest). As described by the 2020 questionnaire, “Ratings between 50 degrees and 100 degrees mean that you feel favorable and warm toward the person…ratings between 0 degrees and 50 degrees mean that you don't feel favorable toward the person 11 and that you don't care too much for that person…[and] you would rate the person at the 50 degree mark if you don't feel particularly warm or cold toward the person” (ANES, 2020, p. 37). Of course, different subjects will have different understandings of what “warm” or “cold” means, but this measure has been used consistently in ANES surveys over many years and provides a relatively stable measure. Two feeling thermometers in particular were analyzed: one for respondents’ feelings toward the Democratic Party and one for respondents’ feelings towards the Republican Party. With the previous variable in mind, the data provides an idea of how members of both parties view the opposing party and their own party. Furthermore, feeling thermometers were also analyzed from the perspective of independents, members of other parties, and those who have no party preference, such that a view from outside both parties could be noted. While this is similar to the approach taken by Groenendyk (2018), it differs in that it examines both one’s rating of their own party and one’s rating of the opposing party, whereas Groenendyk focuses solely on the former. Additionally, Groenendyk also factors in “strong partisans, weak partisans, and leaners,” whereas this analysis simply uses the basic categories established by the “Party ID” question (Groenendyk, 2018, p. 163). Iyengar et al. also measure similar data in their 2012 analysis, but this will serve to update their work for recent election cycles (it also focuses specifically on ANES, while their analysis covers multiple sources). Methods of Data Analysis Anonymized ANES data (with no way of identifying respondents) is publicly available through the Inter-university Consortium for Political and Social Research (ICPSR), an organization that stores decades of social science data. Through this 12 resource, the data packages from 2012, 2016, and 2020 were downloaded, containing responses from thousands of participants (5,915 from 2012, 4,271 from 2016, and 8,280 from 2020). The variables of Party ID, Feeling Thermometer: Democratic Party, and Feeling Thermometer: Republican Party were isolated and cross-referenced in several Microsoft Excel pivot tables. This produced a grand total of six tables: with Party ID cross-referenced with two feeling thermometers over three data sets. Due to the varying numbers of respondents between years, these tables were organized by percentage of Party ID instead of number of respondents. For instance, Table 1 demonstrates that in 2012, 29.65% of respondents who identified as Democrats gave the Republican Party a score of 0 on the feeling thermometer. To simplify the data (as there are, in theory, over 100 measures included on the feeling thermometer axis), responses were sorted into twelve “bins” that included feeling thermometer responses of roughly ten units. Any response between or including 10 and 19, for example, was included in the bin labeled “15.” The other bins were organized similarly, with 20-29 being included in “25,” 30-39 being included in “35,” and so on. The only exceptions were the bins at the extreme ends of the spectrum. The responses of 0 and 100, being the most common responses, were given their own categories. As a result of this, “5” was limited to 1-9 instead of 0-9. It is worth noting that both the “5” and the “95” bins will appear rather empty due to the proximity and non-inclusion of the 0 and 100 ratings, as respondents tended to round to the nearest ten, particularly when it came to the maximum and minimum rating. The data tables have been combined into several graphs to visualize both the cross-references for each year and the change in these variables across the years. 13 Additionally, the average (mean) of each Party ID category’s responses to the feeling thermometer questions has been included for all six tables (these averages were based on the original values, not the bins). For the purposes of this presentation, the Party ID categories of “Don’t Know,” “Refused,” and “Technical Error” (when applicable) have been omitted from the graphs, as these categories tended to be much smaller and not significant to the larger investigation (although they are still included in the “All” category). 14 Data and Analysis Table 1: Party ID and Feeling Thermometer: Republican Party, 2012 ANES Time Series 0 5 15 25 35 45 55 65 75 85 95 100 R/DK Democrat 29.65% 1.95% 18.47% 1.95% 14.02% 11.77% 13.93% 5.42% 1.82% 0.76% 0.08% 0.04% 0.13% Don't know 6.82% 0.00% 4.55% 0.00% 2.27% 13.64% 27.27% 18.18% 2.27% 4.55% 0.00% 0.00% 20.45% Independent 9.86% 0.76% 8.62% 1.30% 10.57% 13.33% 25.85% 13.55% 9.81% 3.47% 0.22% 1.14% 1.52% No Preference 3.03% 0.00% 6.06% 1.52% 3.03% 10.61% 39.39% 16.67% 6.06% 6.06% 0.00% 0.00% 7.58% Other 16.15% 1.86% 10.56% 2.48% 7.45% 8.07% 22.98% 16.77% 7.45% 3.73% 0.00% 0.62% 1.86% Refused 10.42% 0.00% 4.17% 0.00% 0.00% 2.08% 43.75% 8.33% 6.25% 0.00% 0.00% 0.00% 25.00% Republican 0.65% 0.22% 1.01% 0.36% 1.80% 2.95% 8.64% 21.38% 27.65% 22.17% 4.18% 8.78% 0.22% Grand Total 15.67% 1.12% 10.72% 1.35% 9.57% 10.01% 17.28% 12.26% 10.62% 6.80% 1.08% 2.45% 1.07% Figure 1: Party ID and Feeling Thermometer: Republican Party, 2012 ANES Time Series, Stacked Column Graph 15 Figure 2: Party ID and Feeling Thermometer: Republican Party, 2012 ANES Time Series, Clustered Column Graph The data for the Republican Party in 2012 shows some unsurprising patterns when viewed in isolation. 29.65% of Democrats rated the Republican Party a 0 on the feeling thermometer. This serves as a good indicator of polarization, as it is the lowest possible rating, and would therefore imply that about 30% of Democrats find absolutely nothing redeemable about the Republican Party, equating it with the things the respondent dislikes the most. As such, the 0 ratings will be an important aspect to note going forward. Democrats tended to rate the Republican Party poorly in 2012, but the ratings are reasonably spread out across the lower half and middle of the feeling thermometer. The Republicans, as is to be expected, had a higher opinion of their party than the Democrats. The most popular ratings were in the 70-79 range (denoted as “75”), 16 with 27.65% of Republicans selecting this category. The “65” and “85” bins were also popular, but it is worth noting that a rating of 100 is somewhat rare, with only 8.78% of Republicans selecting it. The more neutral categories (independent, other, no preference) are seemingly in agreement. The most popular category for each of these groups is the “55” bin, which contains the perfectly neutral value of 50. The measure of all respondents demonstrates the same trend. Along with the relatively even spread across the continuum, it seems that most respondents in these categories had mixed or neutral feelings about the Republican Party. The “All” group also indicates that the extremes of the Republicans and Democrats mostly seem to cancel each other out. However, it is worth noting a slight spike at 0 for all of these categories, suggesting a distaste for the party amongst a solid contingent of these seemingly middle-of-the-road groups.1 Table 2: Party ID and Feeling Thermometer: Democratic Party, 2012 ANES Time Series 0 5 15 25 35 45 55 65 75 85 95 100 R/DK Democrat 0.64% 0.17% 0.68% 0.17% 0.68% 1.69% 5.89% 13.51% 22.07% 28.21% 3.43% 22.74% 0.13% Don't know 0.00% 2.27% 0.00% 0.00% 0.00% 6.82% 36.36% 18.18% 6.82% 4.55% 2.27% 6.82% 15.91% Independent 7.48% 0.60% 6.23% 1.36% 8.73% 12.03% 24.34% 16.69% 12.09% 6.72% 0.43% 1.84% 1.46% No Preference 0.00% 0.00% 0.00% 0.00% 7.58% 10.61% 36.36% 18.18% 16.67% 6.06% 0.00% 0.00% 4.55% Other 21.74% 0.00% 11.18% 1.24% 5.59% 11.18% 22.36% 9.94% 7.45% 5.59% 0.00% 1.86% 1.86% Refused 10.42% 0.00% 0.00% 2.08% 0.00% 4.17% 35.42% 12.50% 4.17% 6.25% 0.00% 0.00% 25.00% Republican 21.17% 2.23% 20.16% 2.95% 16.34% 14.76% 12.89% 6.05% 2.30% 0.50% 0.00% 0.43% 0.22% Grand Total 8.23% 0.79% 7.25% 1.23% 7.07% 8.40% 14.54% 12.73% 13.59% 13.78% 1.52% 9.86% 0.98% 1 Note: In 2012, ANES polls significantly more Democrats than Republicans (2361 Democrats vs. 1389 Republicans) 17 Figure 3: Party ID and Feeling Thermometer: Democratic Party, 2012 ANES Time Series, Stacked Column Graph Figure 4: Party ID and Feeling Thermometer: Democratic Party, 2012 ANES Time Series, Clustered Column Graph 18 The 2012 data for the Democratic Party shows similar trends. 0 is the most popular rating for the Republicans at 21.17%, while “85” is the most popular bin for the Democrats at 28.21%. The three unaffiliated groups, as well as the “All” group, have their most popular rating at the middle ground of “55.” However, there are slight differences. The Republicans present a flatter curve towards the Democrats than the Democrats present towards the Republicans (Figure 4 vs. Figure 2). The “15” bin is almost as popular as 0 for the Republicans, but the “15” bin is over 10% lower than the 0 rating for the Democrats. However, these differences in the curve seem to be mainly driven by the 0 rating, which the Democrats were much more likely to utilize in 2012. The Democrats also seem to have a slightly higher opinion of their own party than the Republicans, as their rating of the Democratic Party peaks at one bin higher than the Republicans, and they were much more likely to rate their own party a 100 (22.74%). The spread of the unaffiliated groups and the “All” group seems to be more positive towards the Democratic Party, with a decreased propensity to rate it a 0 and a higher proportion of ratings on the positive side of the feeling thermometer. Table 3: Party ID and Feeling Thermometer: Republican Party, 2016 ANES Time Series 0 5 15 25 35 45 55 65 75 85 95 100 R/DK Democrat 21.92% 5.31% 20.26% 2.21% 16.54% 11.65% 11.85% 4.76% 2.55% 0.90% 0.07% 0.41% 1.59% Don't know 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 60.00% 10.00% 0.00% 10.00% 0.00% 0.00% 20.00% Independent 8.34% 2.19% 9.58% 1.68% 12.73% 13.02% 21.65% 15.51% 8.05% 3.66% 0.44% 0.51% 2.63% No Preference 8.16% 0.00% 10.20% 0.00% 8.16% 6.12% 40.82% 10.20% 8.16% 2.04% 0.00% 0.00% 6.12% Other 22.30% 2.03% 9.46% 1.35% 10.81% 10.14% 17.57% 14.86% 7.43% 1.35% 0.00% 0.00% 2.70% Refused 0.00% 0.00% 0.00% 0.00% 0.00% 6.67% 33.33% 0.00% 13.33% 0.00% 0.00% 0.00% 46.67% Republican 0.81% 0.49% 2.52% 0.41% 3.74% 5.93% 7.96% 25.02% 22.34% 18.60% 3.41% 7.88% 0.89% Grand Total 11.22% 2.72% 11.12% 1.45% 11.24% 10.28% 14.59% 14.45% 10.28% 6.93% 1.15% 2.58% 2.01% 19 Figure 5: Party ID and Feeling Thermometer: Republican Party, 2016 ANES Time Series, Stacked Column Graph Figure 6: Party ID and Feeling Thermometer: Republican Party, 2016 ANES Time Series, Clustered Column Graph 20 The 2016 data for the Republican Party does not demonstrate a dramatic change from the 2012 data. The unaffiliated and “All” groups remain relatively neutral. The main distinction is that the Democrats were much less likely to give a rating of 0 (dropping to 21.92%) and that the Republicans’ most popular rating of themselves dropped from the “75” bin to the “65” bin (the latter receiving 25.02% of Republican responses). One could interpret this as a slight move away from the extremes, at least in terms of the public view of the Republican Party. Table 4: Party ID and Feeling Thermometer: Democratic Party, 2016 ANES Time Series 0 5 15 25 35 45 55 65 75 85 95 100 R/DK Democrat 0.76% 0.48% 1.10% 0.28% 1.72% 2.76% 6.75% 15.02% 24.53% 26.26% 5.44% 14.20% 0.69% Don't know 0.00% 0.00% 0.00% 0.00% 0.00% 20.00% 30.00% 20.00% 10.00% 0.00% 0.00% 0.00% 20.00% Independent 7.97% 2.27% 9.66% 0.95% 10.53% 12.58% 20.92% 15.65% 9.95% 5.56% 0.66% 0.88% 2.41% No Preference 6.12% 0.00% 10.20% 0.00% 8.16% 12.24% 36.73% 10.20% 4.08% 6.12% 0.00% 0.00% 6.12% Other 22.97% 2.70% 10.14% 1.35% 5.41% 12.16% 16.22% 12.16% 6.08% 5.41% 0.00% 2.03% 3.38% Refused 0.00% 0.00% 0.00% 0.00% 0.00% 6.67% 26.67% 0.00% 13.33% 6.67% 0.00% 0.00% 46.67% Republican 25.51% 6.58% 21.12% 2.36% 15.19% 11.37% 9.59% 3.98% 2.68% 0.41% 0.00% 0.41% 0.81% Grand Total 11.03% 2.88% 10.02% 1.12% 8.62% 8.87% 12.90% 11.85% 12.62% 11.10% 2.06% 5.29% 1.64% 21 Figure 7: Party ID and Feeling Thermometer: Democratic Party, 2016 ANES Time Series, Stacked Column Graph Figure 8: Party ID and Feeling Thermometer: Democratic Party, 2016 ANES Time Series, Clustered Column Graph 22 A potential softening of extremes in 2016 could be supported by the Democrats’ decreased likelihood to rate their own party a 100 (dropping to 14.20%). However, this is contested by the Republicans’ increased likelihood to rate the Democratic Party a 0 (climbing to 25.51%). The other groups remain roughly as neutral as they did before, but the Republicans seem to take a distinctly more negative stance towards the Democrats, with a notable increase in the “15” bin and even the “5” bin. Table 5: Party ID and Feeling Thermometer: Republican Party, 2020 ANES Time Series 0 5 15 25 35 45 55 65 75 85 95 100 R/DK Democrat 38.06% 2.16% 21.93% 1.85% 13.34% 7.40% 8.66% 2.97% 1.57% 0.63% 0.03% 0.31% 1.08% Don't know 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 50.00% 0.00% 25.00% 0.00% 0.00% 0.00% 25.00% Independent 14.96% 1.27% 11.91% 2.45% 10.05% 9.66% 17.57% 12.54% 9.93% 5.38% 0.55% 1.70% 2.02% No Preference 0.00% 0.00% 14.29% 0.00% 0.00% 0.00% 28.57% 28.57% 0.00% 14.29% 0.00% 0.00% 14.29% Other Party 26.39% 1.86% 5.58% 1.12% 5.20% 9.67% 16.73% 13.01% 9.29% 4.83% 0.37% 1.86% 4.09% Refused 9.09% 0.00% 4.55% 0.00% 4.55% 4.55% 20.45% 4.55% 9.09% 4.55% 0.00% 0.00% 38.64% Republican 0.74% 0.16% 1.21% 0.55% 1.68% 3.24% 4.76% 12.68% 22.19% 29.02% 3.04% 19.70% 1.05% Technical Error 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 100.00% 0.00% 0.00% 0.00% 0.00% Grand Total 18.86% 1.24% 11.81% 1.59% 8.39% 6.85% 10.53% 9.25% 10.82% 11.04% 1.14% 6.79% 1.68% Figure 9: Party ID and Feeling Thermometer: Republican Party, 2020 ANES Time Series, Stacked Column Graph 23 Figure 10: Party ID and Feeling Thermometer: Republican Party, 2020 ANES Time Series, Clustered Column Graph The 2020 data for the Republican Party demonstrates a much more dramatic shift than the one that appeared in 2016. The Democrats shift directions and become significantly more likely to rate the Republicans a 0 (38.06%). Furthermore, the Republicans also change course as their most popular bin for rating their own party becomes “85” (29.02%). Perhaps even more interesting is the shift of the “Other” and “All” categories, which had previously held stable at a most popular rating of “55.” Although they still seem to peak around the middle for the most part, both of these groups are now most likely to rate the Republican Party a 0 (26.39% for “Other” and 18.86% for “All”). Even “Independent,” whose most popular category remains at “55,” sees a marked shift into the lower ratings (with 0 at 14.96%). The change in the “All” response is perhaps the most noteworthy, as it seems to demonstrate a general “cooling” 24 of feelings towards the Republican Party in 2020, one that overcomes the often- counteracting forces of Democratic and Republican ratings.2 Table 6: Party ID and Feeling Thermometer: Democratic Party, 2020 ANES Time Series 0 5 15 25 35 45 55 65 75 85 95 100 R/DK Democrat 0.66% 0.07% 0.91% 0.10% 1.40% 2.30% 5.24% 15.15% 27.58% 29.19% 2.23% 14.04% 1.12% Don't know 0.00% 0.00% 0.00% 25.00% 0.00% 0.00% 25.00% 0.00% 25.00% 0.00% 0.00% 0.00% 25.00% Independent 13.65% 0.91% 9.10% 1.66% 8.07% 10.29% 18.08% 16.62% 12.43% 5.82% 0.32% 1.35% 1.70% No Preference 14.29% 0.00% 0.00% 0.00% 14.29% 14.29% 28.57% 14.29% 0.00% 0.00% 0.00% 0.00% 14.29% Other Party 24.54% 2.23% 12.64% 0.74% 12.64% 7.06% 14.87% 8.18% 8.18% 2.97% 0.00% 1.49% 4.46% Refused 11.36% 0.00% 2.27% 0.00% 6.82% 4.55% 22.73% 2.27% 4.55% 4.55% 0.00% 4.55% 36.36% Republican 47.85% 1.05% 19.81% 2.30% 9.01% 7.10% 6.24% 3.55% 1.09% 0.70% 0.12% 0.27% 0.90% Technical Error 0.00% 0.00% 0.00% 0.00% 0.00% 100.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Grand Total 20.08% 0.70% 9.65% 1.29% 6.20% 6.41% 9.90% 11.70% 13.97% 12.21% 0.91% 5.42% 1.55% 2 “No Preference” only polls 7 respondents in 2020, making it somewhat insignificant 25 Figure 11: Party ID and Feeling Thermometer: Democratic Party, 2020 ANES Time Series, Stacked Column Graph Figure 12: Party ID and Feeling Thermometer: Democratic Party, 2020 ANES Time Series, Clustered Column Graph 26 Here we see arguably the most dramatic shift of all. Almost half of the Republicans polled rated the Democratic Party a 0 in 2020 (47.85%). No rating in any category comes close for any of the years. Democrats’ ratings of themselves remain relatively stable compared to 2016 (29.19% in the 85 bin), but “Other,” “Independent,” and “All” echo their own movements in the Democratic feeling thermometer (with their 0 percentages at 24.54%, 13.65%, and 20.08%, respectively). Here, the shift in “All” is somewhat more suspect due to the relative stability in Democratic ratings of their own party and the massive negative shift of the Republicans, but the shift towards 0 is still noteworthy. Table 7: Party ID and Feeling Thermometer: Republican Party, Average Feeling Thermometer Rating Across All Three Years Table 8: Party ID and Feeling Thermometer: Democratic Party, Average Feeling Thermometer Rating Across All Three Years 27 Figure 13: Party ID and Feeling Thermometer: Republican Party, Average Feeling Thermometer Rating Across All Three Years, Line Graph Figure 14: Party ID and Feeling Thermometer: Democratic Party, Average Feeling Thermometer Rating Across All Three Years, Line Graph 28 By looking at the average feeling thermometer ratings from each group over the years, we can put the data from above into a new, more easily digestible light. We can see that the average Democratic feeling thermometer rating of the Republican Party remains roughly stable (even increasing slightly) before dropping in 2020 (24.50 to 25.74 to 18.96). The seeming convergence in 2016 and subsequent divergence in 2020 can also be seen in the Republicans’ average rating of themselves (73.30 to 66.56 to 75.65). Interestingly, in looking at the averages, we can see that the unaffiliated groups and the “All” group remain relatively stable at around the 40s. The shift to 0 (at least in terms of the ratings of the Democratic Party) was not as pronounced as it may have seemed at first. The Republican feeling thermometer rating of the Democrats sees a slightly steadier decline than the latter’s rating of the former (27.18 to 23.17 to 16.11), while the Democrats’ ratings of their own party remain relatively stable (77.62 to 74.05 to 74.47). We can also note what appears to be a slight decline in the unaffiliated groups and the “All” group. Once again, the “All” group may be inordinately affected by the large drop in the Republicans’ ratings. That being said, it seems that the large shift to almost half of the Republicans rating the Democratic Party a 0 has not brought down the average an obscene amount. 29 Figure 15: Party ID and Feeling Thermometer: Republican Party, All Years, Stacked Column Graph Figure 16: Party ID and Feeling Thermometer: Democratic Party, All Years, Stacked Column Graph 30 Reflecting on the aggregate graphs along with the rest of the data, one can make a few general notes. There does seem to be an increase in negativity towards the opposing party in recent years, specifically in 2020. Focusing on the propensity of groups to assign a rating of 0, we do seem to see an increase as the decade comes to a close, the most dramatic being the Republicans’ shift in 2020. This applies to the more neutral groups as well as the partisan ones. This indicates an increasing amount of complete disapproval of the major political parties, especially from members of the opposing party. However, the story is not quite so simple. For instance, the 2016 hiccup in the ratings of the Republican Party showed multiple groups, including the Democrats, seeming to warm towards the Republicans slightly (and the Republicans even cooling towards themselves somewhat). It is difficult to say what caused this, but it disrupts any sort of firm statement on a trend regarding people’s views towards the Republicans over the past decade. For the Democratic Party, we can see a slightly clearer trend of negativity over the three data sets, but even this is not particularly extreme when we consider the averages. Expanding this research across multiple decades could present us with some clearer trends over time, but in terms of change, this past decade does not seem to demonstrate a major shift in the electorate as a whole. There are still some intriguing shifts, particularly when examining the change in 0 ratings. But this would seem to indicate a change in individuals, not necessarily the electorate as a whole. The more concerning data point would be the fact that the ratings from opposing parties were so low in the first place, indicating a preexisting amount of polarization entering the decade, but it does not seem certain that this decade has brought about a major 31 change in the degree of this polarization. The 2020 data is the main factor that hints at a major change, but it is not certain that this will be long-lasting, and it does not seem to imply a consistent trend of rapidly spiraling across the past ten years. We might also consider that each party has experienced its own unique change. When considering feeling thermometer ratings for the Republican Party, the average of the “All” category hovers around equilibrium. The Republicans and Democrats seem to diverge, but this seems to roughly cancel out. Compare this to the “All” rating of the Democratic Party, which on average seem to decrease slightly as the Democrats’ rating of their own party does not increase enough to counteract the Republicans’ decrease. We can only speculate whether this is the result of increased disillusionment within the Democratic Party, unusually high animosity of Republicans towards the Democratic Party, or some other factor. Regardless, the fact that the ratings of each party do not match hints that perhaps there is no uniform trend across politics and the electorate as a whole. It is also worth noting the trends within the unaffiliated groups. While they demonstrated an increased likelihood to assign a 0 rating to one of the major parties, their average ratings remained much more stable over the years. A sharply decreasing average for these groups would indicate an increasing disdain for politics in general, but this does not seem to be demonstrated either. While the political situation indicated by the data does not necessarily seem to be stable, it is not completely falling apart, as some might argue. 32 Conclusion In the analysis of recent political polarization through public opinion of political parties, it is difficult to make a definitive statement that describes the situation completely and accurately. While there are trends that indicate a high amount of polarization in general, such as the already low opinions of party members towards the opposing party, there is less to definitively back up the claim that this polarization has been increasing in recent years. The data from 2020 indicates something of a sharp turn, and the increasing tendency to rate the opposing party a “0” indicates some increase in animosity, but the unstable trend in terms of average Democratic opinion of the Republican Party (which improved slightly in 2016) fails to suggest an increase in polarization across the board. Furthermore, while the more neutral groups of independents and members of other parties saw a small uptick in “0” ratings towards the Republican and Democratic parties, their averages remained reasonably stable. One could read this as an indication that the major parties are not moving away from the center, but one could also interpret this as a group of people unaffiliated with major parties who simply do not pay much attention to politics, while the partisans grow further apart. These groups are more difficult to analyze without knowing their motivations or the nature of their membership. Further research could combine the data above with changes in party registration over the past decade, analyzing the “party sorting” angle and seeing whether members of the “Independent” or “Other” groups are shifting or staying in place. 33 While the “All” category might not be shifting too dramatically, it is difficult to judge whether this is indicative of polarization or indicative that polarization is not occurring. Republicans and Democrats moving apart to the same degree would have no effect on the “All” category. In the shift from 2016 to 2020 in Figure 13, we can see what appears to be a widening gap between Republicans and Democrats cancelling each other out. On the other hand, the American public on the whole not changing their minds about the parties could indicate that the parties are not becoming increasingly alienated, and that things actually have not changed that much. Further analysis is needed to get a clear picture of the country as a whole. In the end, this analysis was only ever going to be a piece of the puzzle. It was an investigation from a particular angle looking at specific variables, and the result of this investigation would be up for interpretation. Even the ANES data itself is something of a flawed measure, since the Party ID variable was one of the aspects it was investigating and not a variable that was held constant. This led to a significant imbalance between Democrats and Republicans in 2012. There are also many shifting numbers over the years (the raw number of people polled in 2020 was almost double the number polled in 2016). While the sample size is still large enough to provide a significant picture and the data is well-regarded enough for many researchers to use it for their investigations, it may not be possible to get a perfectly accurate sample that represents the U.S. as a whole, just as it may not be possible to completely answer the question of polarization. Besides the avenues for potential future investigation mentioned above, further research could attempt to apply this data collection strategy to past decades, adding 34 more data points and establishing a clearer trend line across many years. As long as the questions remain relatively consistent across the years, overarching changes could be easier to see with more data. Furthermore, the shortcomings of ANES data could be counteracted in the future by actually performing a survey over successive years that would control for these particular variables (keeping the number of respondents in each category constant and focusing on these questions). In an ideal world, the same participants could be surveyed repeatedly to mark shifts in individual polarization. Qualitative evidence could also support this endeavor, providing opportunities for nuance and elaboration in measuring the participants’ shifting attitudes. With some indications that the parties are polarized and some indications that they are not being driven apart at a constant or extreme rate, perhaps the only conclusion we can come to is that polarization is not certain. The various shifts in public opinion seem to show that none of this is set in stone. The U.S. is not necessarily barreling towards a complete fracture with no hope of reconciliation. There is cause for concern, but there is also cause for hope. The data above demonstrates that four years can change many things, as can a single year. Nothing is guaranteed, and perhaps refusing to accept that the U.S. is doomed is the very action that is necessary to avoid dooming it. 35 Bibliography Abramowitz, A. I., & Saunders, K. L. (2008). Is Polarization a Myth? The Journal of Politics, 70(2), 542-555. https://doi.org/10.1017/S0022381608080493 The American National Election Studies (ANES). ANES 2012 Time Series Study. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2016-05-17. https://doi.org/10.3886/ICPSR35157.v1 American National Election Studies, University of Michigan, and Stanford University. ANES 2016 Time Series Study. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2017-09-19. https://doi.org/10.3886/ICPSR36824.v2 American National Election Studies, University of Michigan, and Stanford University. ANES 2020 Time Series Study. Inter-university Consortium for Political and Social Research [distributor], 2021-07-13. https://doi.org/10.3886/ICPSR38034.v1 Ansolabehere, S., Rodden, J., & Snyder, J. M., Jr. (2006). Purple America. The Journal of Economic Perspectives: A Journal of the American Economic Association, 20(2), 97-118. https://doi.org/10.1257/jep.20.2.97 Baldassarri, D., & Gelman, A. (2008). Partisans without Constraint: Political Polarization and Trends in American Public Opinion. The American Journal of Sociology, 114(2), 408-446. https://doi.org/10.1086/590649 Fiorina, M. P., & Abrams, S. J. (2008). Political Polarization in the American Public. Annual Review of Political Science, 11(1), 563-588. https://doi.org/10.1146/annurev.polisci.11.053106.153836 Fisher, D. R., Waggle, J., & Leifeld, P. (2013). Where Does Political Polarization Come From? Locating Polarization Within the U.S. Climate Change Debate. The American Behavioral Scientist (Beverly Hills), 57(1), 70-92. https://doi.org/10.1177/0002764212463360 Groenendyk, E. (2018). Competing Motives in a Polarized Electorate: Political Responsiveness, Identity Defensiveness, and the Rise of Partisan Antipathy. Political Psychology, 39(S1), 159–171. https://doi.org/10.1111/pops.12481 Iyengar, S., Sood, G., & Lelkes, Y. (2012). Affect, Not Ideology: A Social Identity Perspective on Polarization. Public Opinion Quarterly, 76(3), 405–431. https://doi.org/10.1093/poq/nfs038 36 Kozlowski, Austin C, & Murphy, James P. (2021). Issue alignment and partisanship in the American public: Revisiting the ‘partisans without constraint’ thesis. Social Science Research, 94, 102498–102498. https://doi.org/10.1016/j.ssresearch.2020.102498 Layman, G. C., Carsey, T. M., & Horowitz, J. M. (2006). Party Polarization in American Politics: Characteristics, Causes, and Consequences. Annual Review of Political Science., 9(1), 83-110. https://doi.org/10.1146/annurev.polisci.9.070204.105138 Westfall, J., Van Boven, L., Chambers, John R., & Judd, Charles M.. (2015). Perceiving Political Polarization in the United States: Party Identity Strength and Attitude Extremity Exacerbate the Perceived Partisan Divide. Perspectives on Psychological Science, 10(2), 145–158. https://doi.org/10.1177/1745691615569849 37