2 What is Political Polarization?
- Nolan McCarty
Commentators use few words to describe the American political scene as frequently as they use the word “polarized.” But unfortunately, the terms polarized and polarization have taken on such a wide variety of meanings among journalists, politicians, and scholars that they often confuse, rather than clarify, the problems that our political system faces. So one of my main tasks in this volume is to be more precise in the terminology in hopes of better explaining contemporary American politics. The formal definition of polarization is derived from that of polarity, which is the “state of having two opposite or contradictory tendencies, opinions, and aspects.”1 There are usages of polarization that span almost all possible political “tendencies, opinions, and aspects.” The public has variously been described as polarized over cultural norms and practices, religion, attitudes toward subgroups, policy preferences, and partisan attachments. In some cases, the definition is stretched to encompass social and political divisions involving more than two groups—such as when polarization is used to describe conflicts among social, ethnic, and racial identities.
This book, however, focuses on a much narrower set of definitions of polarization. I focus on those political phenomena where the public and its leaders have become increasingly divided. These areas include preferences over public policy, ideological orientations, and partisan attachments. The p. 9↵primary reason for this narrowing is that policy, ideological, and partisan polarization are those areas that have received far more attention from political and other social scientists and therefore have a set of arguments and findings that I believe “everyone needs to know.” Of course, there are links between cultural and social polarization that are important for understanding political polarization so I do not completely neglect these other forms of conflict.
Let me unpack the various forms of polarization: policy, ideological, and partisan. I start with policy polarization. A simple definition of policy polarization is a process where extreme views on some matter of public policy have become more common over time. As an example, consider attitudes toward government policies related to abortion. To simplify the discussion, let’s assume that voters are asked to evaluate three distinct policies related to abortion access. Under policy 1, abortion is legal under all circumstances and is not restricted in any way. Under policy 2, abortion is legal in most circumstances but restricted in some others. Policy 3 holds that abortion is illegal under all circumstances. We would say that policy preferences over abortion were polarizing if support for the two most extreme policies (policies 1 and 3) were growing over time relative to the centrist policy 2. Thus, polarization is distinct from uniform movements of attitudes in either a pro- or anti-abortion direction. We would not say opinion is polarizing if support for policy 3 was increasing while support for policy 1 was decreasing. Another implication of this example concerns how we measure polarization. When policy preferences are very polarized, the two extreme attitudes will have more support than the middle one. In the terminology of statistics, the distribution of polarized opinion is bimodal, as there are two distinctive, most common answers. Alternatively, we say opinion is unpolarized or centrist if it is unimodal, in that the centrist policy 2 is the single most common position. Polarization may also be related to how much variation there is in policy positions. In statistical terms, the variance of opinions p. 10↵represents the typical deviation of individual opinions from the average (or mean) opinion. In a situation of low polarization, most voters choose the same policy position and so the statistical variance is low. In the extreme case, where voters are equally divided between policies 1 and 3, the variance is quite large.
In addition to analyzing polarization on specific policies, political scientists often discuss it in terms of broader ideological differences among voters. For now, let us think of ideology as a general orientation to politics and governance. In the United States, we often imagine ideological orientations falling on a continuum from liberal positions to conservative ones and orient them so that they range from “left” to “right.”2 Conceptually, ideological polarization is similar to policy polarization. If most voters fall toward the ideological center, we’d say there is little ideological polarization. But to the extent to which liberal and conservative ideologies become more common relative to those of the center, we’d call that polarization. As before, we can identify polarization statistically by looking to see whether ideologies have become more bimodal or more variant in the population.
Figure 2.1 may be helpful in understanding what political scientists mean by polarization. The figure shows two curves representing different distributions of ideological orientations. The solid line represents what we might call a centrist distribution of preferences. In this case, the most typical position is one of moderation. Extreme liberal or conservative views are quite rare. The dashed line, however, represents a more polarized distribution. It is clearly bimodal in that the most common positions are distinctly conservative or liberal. Now moderate views are relatively less likely and extreme liberal or conservative views are no longer rare.
While these figures present polarization solely in terms of the distribution of ideological preferences, researchers often focus on how the positions of voters and politicians vary across political parties. Consequently, we can use partisan polarization p. 11↵to refer to situations where polarization is organized around parties. Most often, scholars use party polarization to describe situations where the policy and ideological differences between members of the Democratic and Republican parties have grown. However, as I soon discuss, this usage is controversial because it conflates two distinct trends about voters. Partisan ideological differences may grow either because there is ideological polarization between the liberals who tend to be Democrats and the conservatives who tend to be Republicans. Or partisan differences could increase without ideological polarization if there is a tendency over time for liberals to move into the Democratic party as conservatives move into the Republican party. These different trends and p. 12↵patterns have important implications for how we interpret the increased divergence of opinions across the parties and the likely consequences of those changes.
2.1. What is the difference between partisanship and polarization?
The terms polarization and partisanship are often used interchangeably, but such usage often obscures important differences. As discussed earlier, polarization generally refers to differences on policy issues, ideological orientations, or value systems, while partisan polarization may refer to these differences across members of different parties. Partisanship, however, can be more general in that it may refer to any partiality one feels toward one’s own party regardless of whether polarized preferences and attitudes are the source. In recent years, many scholars have argued that the rise in partisan conflict is best thought of as a rise in general partisanship that is unrelated to rising ideological or policy polarization. Many explanations have been offered as to why high levels of partisanship can persist even without underlying polarization. With respect to Congress and political elites, Frances Lee argues that the intense competition for majority control of the US House and Senate induces high levels of intra-party competition and inter-party conflict, which she dubs “teamsmanship.”3 Given the importance of majority control in setting policy and allocating patronage, this instrumental form of partisanship has been an important feature of American politics throughout its history. But in the current era of partisan parity, it has become much more salient.
Others have argued that partisanship at the mass level is less instrumental and is instead based on strong psychological attachments and social identification.4 From this perspective, the observed rise in political conflict in the United States is a reflection of the strengthening of “in-group” loyalties and “out-group” animosities. While partisan polarization might p. 13↵underpin these rising animosities, many scholars argue that differences on policy positions across the parties are caused by partisanship, as party loyalists adopt the positions favored by their own party.5 In chapter 4, I report on the research that has sought to explain the rising salience of partisanship and partisan identities.
2.2 What is the difference between mass and elite polarization?
Any discussion of polarization, its sources, and its consequences should distinguish between elite and mass polarization. Social scientists use elite polarization to refer to divisions among office holders, party officials, policy intellectuals, and activists. Alternatively, mass polarization refers to that associated with normal voters and citizens. While most people assume that elite and mass polarization are closely related, that is often not the case. As long as the political elites are not perfectly representative of the electorate or not responsive to ordinary voters, we could observe increasing political conflicts among elites that are not mirrored in the broader public. The politics of abortion are a good example of this pattern. Elected politicians tend to take polarized views on the subject. Most Republican leaders have adopted a pro-life position that provides for abortions only in exceptional circumstances, such as when the life of the mother is in jeopardy.6 Many Democratic officeholders take the near-opposite position that there should not only be few if any restrictions on the practice but that abortion services for the poor should be supported by tax dollars. A plurality of voters reject these positions, however, preferring instead that abortion be available in most circumstances but accepting restrictions based on term. Support for public funding is low.7 While voters’ views on abortion correlate with their partisan identification, large numbers of Democratic voters are pro-life while many Republican voters are pro-choice.8
Alternatively, society could become quite divided, but an elite consensus could persevere. A good example of this might p. 14↵be the Vietnam War. Attitudes about the continued conflict in Vietnam became polarized in the public well before the bipartisan elite consensus in favor of US involvement broke down. By May 1967, the American public was evenly divided over the question of whether it was a “mistake” to send troops to Vietnam, but the leadership of both parties remained committed to the war until after the Tet Offensive in 1968.9
As I discuss in chapters 3 and 4, polarization of political elites and the masses began at very different times and have followed distinct trajectories. Specifically, the current era of elite polarization appears to have begun in the mid-to-late 1970s, while similar changes in the mass public do not emerge clearly until the 1990s. Given these differences, distinguishing between elite and mass polarization is crucial for understanding the underlying causes and the likely consequences.
2.3 What is partisan sorting and is it different from polarization?
In discussions about polarization, it is often noted that Democratic and Republican voters have increasingly divergent opinions on many matters of public policy. For example, in a recent report, the Pew Research Center notes that the gap between Democrats and Republicans on the value of open immigration has grown markedly.10 Eighty-four percent of Democrats agreed that “immigrants strengthen the country with their hard work and talents,” whereas only 42% of Republicans shared this view. This 42-point gap grew from only a 2-point gap in 1994.
There are two logical ways in which such a partisan gap in views on immigration can emerge. The first is voter polarization. It might be the case that partisans have increasingly taken the extreme positions. Democrats may have increasingly adopted very pro-immigrant positions while Republican voters have become much more anti-immigrant. These changes of voter attitudes lead to the large partisan gap on the question about the contributions of immigrants.
p. 15↵But it is also possible that opinions about immigration have not polarized. Perhaps voters have just sorted into parties so that voters with pro-immigration attitudes now overwhelmingly identify as Democrats while immigration restrictionists have migrated into the Republican party. Such a pattern of party sorting can account for the increased differences across partisans even if the distribution of immigration attitudes in the population remains unchanged or moves uniformly in one direction or the other. In this case, it is clear that attitudes have shifted in a pro-immigration direction. Roughly 30% agreed that immigrants strengthened the country in 1994. In the 2017 survey, 65% did. So the most likely cause of the partisan gap is sorting.
Partisan sorting can arise in two different ways. First, voters can choose parties based on their agreement with the party’s position on salient issues. In the immigration example, an anti-immigrant Democrat might recognize that the Republican party has increasingly adopted positions closer to her own, and therefore she decides to switch her party allegiance. I call this ideology-driven sorting. Since party switching is relatively rare,11 ideology-based sorting is probably most pronounced for new voters entering the electorate. A new anti-immigrant voter in 1994 may not have recognized an important difference between the parties on immigration, but one entering the electorate in 2017 clearly would. Those who see immigration as a sufficiently important issue might use these differences in deciding which party to support.
The second mechanism is that partisans may decide to adopt the policy positions of their preferred party. So an anti-immigrant Democrat might alter her views about immigration to correspond to the dominant viewpoint of her party. The same might be true for pro-immigrant Republicans. This party-driven sorting mechanism is probably most pronounced in those cases where voters do not have strong views about immigration and are therefore susceptible to persuasion and social pressure from other partisans and party elites. To the extent p. 16↵party is an important social identity, many voters may simply decide that maintaining that identity requires supporting their party’s dominant view.
Throughout the remainder of the book, I try to distinguish between conclusions related to voter polarization and those related to sorting. But in many cases, it is not clear which of the mechanisms is responsible for the diverging views of partisans. I describe such findings as partisan divergence, which of course can be caused by either polarization or sorting.
2.4. What is belief constraint and ideological consistency?
Many scholars of public opinion are interested in another concept closely related to polarization. Ideological consistency is the propensity of a voter to have either all liberal, all moderate, or all conservative views. Since the seminal work of Philip Converse, this phenomenon is also called belief constraint, which Converse defines as “the success we would have in predicting, given initial knowledge that an individual holds a specified attitude, that he holds certain further ideas and attitudes.” For example, if we could predict a person’s position on tax cuts from her position on free trade or from that on gay rights, we’d say that those beliefs exhibit constraint and that the voter is ideologically consistent.
While the concepts are distinct, increases in ideological consistency and belief constraint have manifestations that are similar to polarization and sorting. A consistent liberal is not only likely to have liberal views across the board but is also likely to only support liberal politicians and is therefore likely to join the Democratic party. They disagree strongly with consistent conservatives. However, if beliefs were less constrained and consistent, the typical voter might support liberal positions sometimes and conservative ones at others. She might be likely to split her votes between Democratic and Republican politicians. Moreover, pairs of opposed partisans are more likely to agree on at least some issues.
p. 17↵ Chapter 4 reviews the evidence about the ideological consistency of voters and how it has changed over time.
2.5. Who is polarized—the public or the politicians?
As I stressed earlier, it is important to distinguish between mass and elite polarization. This is true not only because they are distinct phenomena, but because the evidence points to a much weaker relationship between polarization at the two levels than many people presume. The academic consensus that political elites have polarized over the past forty years is quite strong and is bolstered by both qualitative and quantitative evidence. Noteworthy are qualitative accounts, which often combine historical research and participant observation.12 There are also several excellent histories of the intra-party battles among partisan elites that culminated in our polarized party system.13
As I explain in some detail in chapter 3, the starting point for many quantitative studies of polarization is the robust observation of rising partisan differences in roll-call voting behavior in Congress. The bipartisan coalitions of the 1950s and 1960s have given way to the party-line voting of the twenty-first century. Also discussed in chapter 3, similar patterns of elite polarization have been documented for state legislatures, the judiciary, and large campaign donors. Both the quantitative and qualitative evidence suggest that the late 1970s were a turning point. To be sure portents of the intra- and inter-party conflicts that led to polarization and sorting were in play much earlier, but the predominance of the liberal wing of the Democratic party and the conservative wing of the Republican party was not cemented until the late 1970s.14
The extent to which the mass public is polarized is a topic of somewhat more vigorous academic debate that is taken up in detail in chapter 4. Longitudinal studies of voter opinion generally do not provide much evidence of polarization or significant sorting until the 1990s.15 Consequently, it is hard to p. 18↵sustain claims that mass polarization is the primary cause of elite polarization given that elite polarization precedes it by about fifteen years. Yet it does not appear that the centrist, unsorted electorate placed too many constraints on the efforts of the parties to reorganize themselves along ideological lines. The comparison of the 1964 and 1980 presidential elections is instructive, if imperfect. In both cases, a very conservative Republican candidate challenged a Democratic president from the moderate wing of the party. In the first instance, Barry Goldwater lost forty-four states plus the District of Columbia. In the second, Ronald Reagan won forty-four states. While there are many differences in the context of the two elections, it seems clear that the electorate was far more tolerant of a conservative message in 1980, despite the apparent lack of polarized public opinion.16
The debates about the magnitude and timing of mass polarization focus on how to interpret the increased difference between Republican and Democratic voters in terms of general ideological orientations and specific policy preferences. One school of thought, led by Morris Fiorina, argues that these differences can be explained almost entirely by the ideological sorting of voters into the parties.17 Fiorina and his coauthors often point to the fact that most voters remain fairly moderate in their expressed policy positions.18 Moreover in studies that produce estimates of voter issue positions that are comparable to legislator positions, representatives are generally found to take positions that are considerably more extreme than those of their constituents.19 Since voters do not seem to increasingly take on extreme positions, the partisan differences are likely caused by sorting, with liberal voters aligning with the Democratic party and conservative voters aligning with the Republican party.
This sorting interpretation has been challenged by Alan Abramowitz who observes that while many citizens are moderate, those most likely to participate in politics increasingly take extreme policy positions.20 The greater the level of p. 19↵engagement the more polarized are the preferences. Highly informed voters also appear to be polarized. While some moderate voters have chosen middle-of-the-road positions for substantive policy reasons, many others are uninformed, unengaged, or apathetic, checking off the middle position on surveys due to the lack of an opinion. Of course, at very high levels of voter engagement and sophistication, the lines between elite and mass begin to blur.
Despite the lack of evidence that voter polarization causes elite polarization, it is clear that both voter sorting and the polarization of the engaged electorate can reinforce if not exacerbate elite party divisions. Even if voters are merely sorted into parties, the incentives for parties to take positions that appeal to supporters of the other party will diminish—leading to greater partisan polarization and greater incentives for voters to sort.
2.6. Why is polarization bad?
Very few people use the word “polarization” to describe a healthy state of political affairs. It is almost always used as a near-synonym for dysfunctional conflict. But at the same time, we might imagine situations in which polarization were too low. If there is little polarization among the public, we might worry about the costs of conformity. Few citizens will challenge current practices and conventions, and there would be little impetus for social progress and reform. For example, the American electorate of the 1950s demonstrated a very high degree of consensus on the issues that were on the public agenda, but this consensus left issues related to the rights of African Americans, ethnic minorities, women, immigrants, and the LGBTQ community largely unaddressed.
Polarization among political elites and the parties is also not unambiguously bad. Indeed, the consensus among political scientists is that democracy works best when parties provide the voters with distinct menus of policy positions. Some degree of polarization is necessary for political representation and p. 20↵accountability. When the parties do not take distinctive positions, voters lack a clear choice with regard to policy. Moreover, heterodox parties reduce the usefulness of partisan cues as to which candidates to support. But when parties are distinct and coherent, voters can better register their views through their vote. Additionally, when parties push different policies, voters know who to hold accountable when a policy approach fails. These arguments, known as Responsible Party Theory, were summed up nicely in the American Political Science Association’s report from its Committee on Political Parties in 1950:
In a two-party system, when both parties are weakened or confused by internal divisions or ineffective organization it is the nation that suffers. When the parties are unable to reach and pursue responsible decisions, difficulties accumulate and cynicism about all democratic institutions grows. An effective party system requires, first, that the parties are able to bring forth programs to which they commit themselves and, second, that the parties possess sufficient internal cohesion to carry out these programs . . .
On the other hand, . . . a coalition that cuts across party lines, as a regular thing, tends to deprive the public of a meaningful alternative. When such coalitions are formed after the elections are over, the public usually finds it difficult to understand the new situation and to reconcile it with the purpose of the ballot. Moreover, on that basis it is next to impossible to hold either party responsible for its political record. This is a serious source of public discontent. 21
In sum, without some differentiation of the political parties, it would be almost impossible for the typical voter to have any influence over the direction of public policy. But as I discuss in chapter 7, there is considerable evidence that the level of polarization among the elites and the public is well to the warm side of the Goldilocks point.
Polarization has become a catch-all word used to describe almost any form of political conflict and disagreement. But understanding the causes of political dissensus requires distinguishing polarization from many other sources of partisan conflict. While partisanship, partisan sorting, and ideological consistency may be closely related to polarization, it is important to identify them as distinct phenomena. For example, the extent to which conflict reflects polarization or sorting has implications for the extent to which conflict is bottom-up from the voters or top-down from the elites.
It is equally important to consider who is polarized—elites and elected officials or regular ordinary voters. It is entirely possible that one group but not the other is polarized. The question of the extent to which elites are polarized is taken up in the next chapter, while voter polarization is considered in chapter 4.
Finally, it is important to remember that polarization is not always a bad thing. If the parties did not offer distinctive public policy positions, voters could hardly be in a position to influence public policy through their votes. We might also be wary of those calls to reduce polarization in the public that would involve repressing certain viewpoints. To riff on Madison in Federalist 10, there are two methods of removing the causes of polarization: the one, by destroying the liberty which is essential to its existence; the other, by giving to every citizen the same opinions, the same passions, and the same interests. Clearly, the first is worse than the disease, and the second is unlikely to happen given the diversity of American public life. But unfortunately, as I discuss in chapter 7, Madison’s constitution may not provide the needed relief in controlling polarization’s effects.