INTRODUCTION
Should turnout in national and state elections be compulsory? There are currently 27 countries that have adopted a system of compulsory voting and at least 12 others that have experimented with the system in recent history (IIDEA 2023). Legislators have recently introduced bills to make turnout mandatory in Canada, the United Kingdom, and France (none of which were successful).Footnote 1 Former US President Barack Obama publicly endorsed the introduction of compulsory voting as a short-term strategy for increasing turnout among young, poor, and minority voters (Yan Reference Yan2015), and there are editorials regularly calling for its adoption in the United States (Moyo Reference Moyo2019; Stephanopoulos Reference Stephanopoulos2015) and the United Kingdom (Klemperer Reference Klemperer2023; Padmanabhan Reference Padmanabhan2015).
Normative political theorists are sharply divided on this question. Proponents often support compulsory voting (henceforth, CV) because it incentivizes more citizens to fulfill their civic duty to vote while promoting democratic values such as equality, participation, representativeness, or legitimacy (Brennan and Hill Reference Brennan and Hill2014; Chapman Reference Chapman2019; Elliott Reference Elliott2017; Engelen Reference Engelen2007; Hill Reference Hill2006; Reference Hill2010; Reference Hill2017; Lijphart Reference Lijphart1997; Umbers Reference Umbers2020). Skeptics often resist proposals to make turnout mandatory on the grounds that such proposals are morally unnecessary, ineffective at realizing the relevant democratic values, or morally unacceptable due to the element of coercion involved (Brennan and Hill Reference Brennan and Hill2014; Lever Reference Lever2010; Saunders Reference Saunders2010; Reference Saunders2012; Reference Saunders2018; Reference Saunders2020; Volacu Reference Volacu2019).
This article contributes to this debate by introducing an original argument for the importance of increasing turnout: that introducing compulsory voting in highly polarized majoritarian democracies such as the United States can reduce political polarization, defined as the gap between two parties’ platforms in the dominant policy space. Polarization is an increasing focus of both academics and policymakers, in part because of concerns that it can undermine support for democracy and lead to democratic backsliding (Binder Reference Binder2003; Graham and Svolik Reference Graham and Svolik2020; Levitsky and Ziblatt Reference Levitsky and Ziblatt2018; McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2006; Svolik Reference Svolik2019; Reference Svolik2020). We draw on two existing formal models of political competition to make two claims. First, that divergence between parties arises in part because, in voluntary voting systems, citizens can threaten to abstain from voting if neither party is sufficiently close to their own preferences. That is, the threat of abstention due to alienation increases polarization. Second, we show that adaptations of both formal models predict a decrease in polarization if turnout increases. This directly suggests that compulsory voting could decrease polarization and lead both parties to locate closer to the preference of the median voter.
Our argument in this article is restricted to majoritarian electoral systems, with the US serving as our primary case study. Although one might expect to find similar effects of CV in systems with proportional representation, the details of electoral competition under PR are sufficiently distinct to be outside the scope of our current article and of the formal models discussed here. Given that 40.4% of democracies have a version of plurality or majority voting (88 countries total), our argument for the adoption of CV in majoritarian electoral systems applies to a large number of democracies, many of which are currently experiencing high levels of political polarization (IIDEA 2023). While our discussion of polarization draws heavily on the US case, as this is where much of the empirical work has been done, we view this as a general argument and support it with evidence from multiple countries and regions.
By integrating the formal and empirical literatures on elections and compulsory voting, our approach has three other advantages. First, we are able to use these literatures to develop scope conditions for when CV is most likely to reduce polarization, namely in high-capacity majoritarian systems with mid-level turnout. Second, both the formal models we use, and the existing empirical literature, allow us to examine a number of potential objections to our argument, including the possibility of spoiled ballots; whether voters mobilized by CV will vote rationally; the effect of CV on affective polarization; and the possibility that CV generates opposition to democracy. We argue that these may reduce, but not eliminate, the benefits of CV in reducing polarization.
Finally, our approach provides a framework for how to incorporate formal theory into normative work, and how to use this to think more carefully about scope conditions as part of normative debates. Arguments rarely apply universally, and normative theorists should acknowledge that certain arguments apply to some systems more than others. Formal theory is one of the tools that can be used to elucidate the assumptions and mechanisms underlying many normative claims, allowing us to make more precise normative assessments and prescriptive recommendations in core areas of democratic theory such as the decision whether to have voluntary or compulsory voting.
Our argument proceeds as follows. First, we review the normative debate about compulsory voting and discuss existing theory and evidence that political polarization can negatively impact democratic systems. We then introduce our main argument; we first provide an intuition for why CV could reduce polarization, then adapt two existing formal models to provide support for our argument and clarify mechanisms and scope conditions. Finally, we examine potential objections to our arguments and conclude.
The Normative Case for Compulsory Voting
We use the term “compulsory voting” to refer to policies mandating turnout at the polls for a range of national and/or subnational elections. There is significant variation in the ways in which CV has been implemented across the 27 countries that currently use it, particularly the extent to which it is enforced both in the law and in practice. Empirical studies of CV suggest that introducing compulsory voting will increase turnout anywhere from 7 to 10 percentage points with no enforcement to 14.5 to 18.5 percentage points with enforcement (Kostelka, Shane, and Blais Reference Kostelka, Shane and Blais2022).Footnote 2 We argue in favor of a system of “strong CV,” in which there are meaningful, enforced penalties for abstention in order to substantially increase turnout. This means that our argument will depend on a sufficiently well-organized state with the capacity to enforce the law. The term “compulsory voting” should perhaps more accurately be called compelled turnout, as voters can refuse to cast a valid ballot once compelled to turn out. We discuss this further in the section on objections and limitations.
This section summarizes the existing normative case for CV, grouping arguments into three lines of argument: (i) the moral duty to vote, (ii) the threat of low turnout to democratic legitimacy, and (iii) the problem of unequal turnout. We discuss each in turn, considering some prominent objections. We show that, if successful, our argument about polarization would significantly strengthen the case for CV.
The first argument, pursued by many advocates of CV, is what we call the moral duty to vote. According to Brennan and Hill (Reference Brennan and Hill2014), “[v]oting is not, as is commonly thought, a waivable privilege, but an inalienable duty-right” (155); we simultaneously have a right to vote that governments should respect, and we have a duty to vote that governments can enforce. Theorists have argued that the moral duty to vote is necessary for minimally just governments (Maskivker Reference Maskivker2018; Reference Maskivker2019); for the preservation of democracy (Brennan and Hill Reference Brennan and Hill2014); or to avoid free-riding on others’ political participation (Lijphart Reference Lijphart1997, 9). In all of these cases, advocates argue that CV allows a larger number of citizens to fulfill this moral obligation to vote by penalizing unjustified free riding (Umbers Reference Umbers2020), creating the right habits (Chapman Reference Chapman2019), serving as a pre-commitment device (Elliott Reference Elliott2017), creating a social norm of voting (Engelen Reference Engelen2007, 41–2), and/or forcing the state to make voting more accessible (Brennan and Hill Reference Brennan and Hill2014, 121–4).
There are three primary objections to these arguments. First, the majority of political philosophers reject the claim that there is an unconditional moral duty to vote (Brennan Reference Brennan2009; Lomasky and Brennan Reference Lomasky and Brennan2000; Maskivker Reference Maskivker2018; Reference Maskivker2019). Second, even if one accepts the claim that there is a moral duty to vote well and that this requires turning out to vote, Saunders (Reference Saunders2020) and Volacu (Reference Volacu2019) argue persuasively that one cannot derive a duty to vote from the duty to vote well, as voting badly is morally inferior to not voting. Third, even if voting is a moral duty, it is not clear that CV itself can be justified as a way to help citizens meet their moral obligations; CV may be paternalistic or an overstep of government intervention (Brennan and Hill Reference Brennan and Hill2014; Lever Reference Lever2010).
The second approach is what we call the democratic legitimacy argument. Advocates of CV often argue that high turnout is necessary for the realization of key democratic values such as political equality and democratic legitimacy.Footnote 3 According to Chapman (Reference Chapman2019), near-universal turnout forms “extraordinary spectacles” that reinforce the value of democracy (103), while elections with low turnout fail to realize the relevant democratic values and raise questions about the legitimacy of the political system.
Critics of CV deny that high turnout is necessarily beneficial for democratic values. Saunders (Reference Saunders2012) argues that low turnout only signals a problem for democracy when it is due to unequal barriers to voting. Voters may choose to abstain because democracy is going well, and the stakes of the election are therefore low. If low turnout is due to unequal access to voting, the solution is likely to directly solve those issues, rather than institute CV. Brennan and Hill (Reference Brennan and Hill2014) deny that democracies with low turnout are illegitimate, and note that such a claim would paradoxically imply that such democracies lack the political legitimacy to impose CV.
The third and final approach relies on identifying what we call the problem of unequal turnout. In many modern democracies, turnout is disproportionately low among younger, poorer, and less educated voters. According to Lijphart (Reference Lijphart1997, 2), “low voter turnout means unequal and socioeconomically biased turnout.” Proponents of CV argue unequal turnout is democratically illegitimate on procedural grounds. Engelen (Reference Engelen2007), Brennan and Hill (Reference Brennan and Hill2014), and Lijphart (Reference Lijphart1997) further argue that these procedural biases can, and often do, lead to electoral outcomes that favor groups who vote at higher rates, especially wealthy voters. This view is supported by evidence that the introduction of CV leads to policies that favor the interests of the least well-off citizens, resulting in lower inequality (Chong and Olivera Reference Chong and Olivera2008), and that higher turnout within developing countries is associated with more government redistribution (Mahler Reference Mahler2008, 178).
Critics of CV deny that the higher turnout generated through CV would in fact lead to a government that is more responsive to the poor, at least within developed countries such as the United States. According to Brennan and Hill (Reference Brennan and Hill2014), it is campaign donations and non-electoral features of the United States that drive pro-rich policy, not voting patterns, and so CV would not address the fundamental problem of a government that is biased in favor of the wealthy.
We have now introduced the three main existing normative arguments for the introduction of compulsory voting: that there is a moral duty to vote; that CV will increase democratic legitimacy; and that it will increase democratic equality and representation by eliminating unequal turnout. The first two arguments require agreeing with controversial normative arguments about the value of voting. The successful case for the moral duty to vote requires a compelling account of the distinctive value of voting that (i) does not depend on whether one votes well and (ii) does not accept other forms of political participation and contributing to the collective good as moral substitutes for voting. The successful case for the democratic legitimacy argument similarly implies a commitment to a particular normative theory of democracy and/or political legitimacy in which electoral participation has a uniquely valuable role. These arguments are unlikely to appeal to skeptics who do not ascribe the same high moral value to the act of voting or to near-universal turnout. The unequal turnout argument is less normatively controversial, but it is not clear that CV can solve the underlying problems of unequal political influence.
The normative argument in this article aims to strengthen the case for CV within majoritarian electoral systems by adding a reason that does not depend on the controversial normative premises above.
Why Polarization is Bad for Democracy
As a first step in our argument in support of CV, this section establishes that polarization is normatively problematic and can seriously undermine democracy. The next section then shows that there are reasons to expect that, under a broad range of conditions, increasing turnout through CV will lead to a decrease in polarization and, therefore, a lower risk of democratic backsliding.
Political polarization refers to a family of concepts describing a growing divide between different political actors. In this article, we focus on elite polarization in majoritarian systems, defined as “high levels of ideological distance between parties and high levels of homogeneity within parties” (Druckman and Peterson Reference Druckman, Peterson and Slothuus2013, 57). Assume there are two main parties. If the equilibrium policies offered by the two parties converge, then we would consider the levels of political polarization to be low. If in contrast the equilibrium policies diverge, we would consider the distance between these policies to measure the degree of political polarization, with larger distance indicating more polarization.
Empirical political scientists agree that elite polarization has increased significantly in the United States since the 1950s (Hetherington Reference Hetherington2009; Layman, Carsey, and Horowitz Reference Layman, Carsey and Horowitz2006; McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2006). Using a well-known measure of ideology compiled from Congressional voting records (DW-NOMINATE), Figure 1 illustrates the significant increase in ideological divergence between the two main US parties during the last few decades.Footnote 4
Figure 1. Polarization in the US Congress House of Representatives 1979–2019
Early studies of political polarization tended to emphasize potential benefits more than harms. For example, APSA (1950) report on the United States argued in favor of more clearly differentiated “responsible parties,” understood as well-organized national parties running on coherent party platforms.Footnote 5 Similarly, Layman, Carsey, and Horowitz (Reference Layman, Carsey and Horowitz2006) note that potential benefits of polarization include voters who are more ideologically sophisticated, who more strongly identify with a political party, and who generally care more about politics. In theory, the clear and distinct choices facilitated by party divergence should make it easier for voters to identify the ideologically closest party and to hold politicians accountable.
However, Hetherington (Reference Hetherington2009) notes that polarization had consequences that the 1950 APSA report did not anticipate, particularly the potentially negative effects on cross-party compromise and intra-party accountability. In recent decades, a consensus has started to emerge that polarization is dangerous for the stability and sustainability of democratic institutions in ways that should concern all political theorists who consider the survival and well-functioning of democracy to be normatively preferable to the alternative of democratic decline or dysfunction. Here we consider three of these: that polarization hurts voters’ ability to credibly reject anti-democratic politicians in their own party; that it leads to an erosion of fundamental democratic norms; and that it leads to gridlock and democratic dysfunction.
The first concern is that polarization forces voters to choose between their preferred policies and their commitment to democracy in ways that make it more costly to vote against undemocratic candidates (Graham and Svolik Reference Graham and Svolik2020; Svolik Reference Svolik2019; Reference Svolik2020). Following the formal model developed by Graham and Svolik (Reference Graham and Svolik2020), one can think of voters as having preferences across two dimensions: they prefer candidates (1) whose policy platforms align with their own preferences (e.g., over taxation or education), and (2) who support democracy and do not undermine democratic institutions (like press freedom and election integrity).
When the difference between parties’ policy positions—and therefore polarization—is low, voters face a simple choice in protecting democracy. Even if a voter prefers the policy platform of an undemocratic candidate, the policy cost of voting for the more democratic candidate is relatively low, and voters will tend to favor the democratic candidate. When political polarization is high, candidates will offer very different policy platforms. If one of the candidates is anti-democratic, that candidate’s partisans must choose between voting for a democratic candidate with a very different policy platform or voting for their preferred policy at the cost of democratic decline. Polarization therefore increases the cost of protecting democracy for voters, potentially increasing voters’ willingness to support an anti-democratic candidate. Graham and Svolik (Reference Graham and Svolik2020) find strong experimental confirmation for their suspicion that voters are less likely to penalize undemocratic moves in highly polarized contexts.
The second negative effect of polarization is the erosion of democratic norms among elites. Levitsky and Ziblatt (Reference Levitsky and Ziblatt2018) argue that political polarization can lead political elites to behave in ways that undermine democracy through the erosion of informal democratic norms like mutual toleration and institutional forbearance, norms that constitute the so-called “guardrails” of democracy. Mutual toleration refers to the willingness to tolerate members of the opposition as adversaries in the political game rather than “as treasonous, subversive, or otherwise beyond the pale” (102). In the absence of mutual tolerance, political elites “may decide to employ any means necessary to defeat them,” and the peaceful transition of power may break down (104). Institutional forbearance involves abiding by informal norms like term limits, self-limiting use of executive prerogatives like the executive order, avoiding partisan gerrymandering, and limited filibustering. According to Levitsky and Ziblatt (Reference Levitsky and Ziblatt2018), when political elites are highly polarized, it is very easy for political competition to degenerate into no-holds-barred conflict with existential dimensions—the exact type of conflict that kills democracies.Footnote 6 Although it is possible for political elites to counteract polarization through cooperation and compromise, the authors note that such cooperation is highly unlikely. A decrease in polarization is needed to incentivize a return to cooperation across party lines.
A final potential downside to polarization is legislative gridlock, which has negative short- and long-term effects (Binder Reference Binder2003; Jones Reference Jones2001; Mann and Ornstein Reference Mann and Ornstein2012; McCarty, Poole, and Rosenthal Reference McCarty, Poole and Rosenthal2006, 175–83). Jones (Reference Jones2001) argues that in cases where control is split between the legislative and executive branches, polarization increases party cohesion and makes it less likely that centrist politicians vote with the opposing party to pass legislation, leading to gridlock.Footnote 7 McCarty, Poole, and Rosenthal (Reference McCarty, Poole and Rosenthal2006) argue that party polarization increases gridlock by (a) shrinking the feasible set of policies that could garner a majority, and (b) making even some feasible policies difficult to achieve because of incentives to squeeze as much as possible from opponents. Such brinkmanship may be especially likely if voters are also polarized, and reward politicians who refuse to compromise (or punish those who do); Binder (Reference Binder2003) makes a similar argument.
The general consensus in the literature is that polarization makes gridlock more likely, and that gridlock is a costly process that prevents governments from addressing pressing problems. Over time, gridlock becomes a form of democratic dysfunction and can lead to high levels of dissatisfaction with democratic institutions. These high levels of dissatisfaction further exacerbate the risks of democratic backsliding as outsider candidates with little commitment to existing democratic institutions become more appealing to a public dissatisfied with the political status quo.
How CV Can Reduce Political Polarization
The previous sections argued that existing arguments in support of CV are incomplete and introduced the problem of political polarization. This section develops a theory of how CV could reduce polarization. We first present the intuition and assumptions behind our argument. We then adapt two existing formal models of voter turnout and elections to show that both predict that an increase in turnout will lead to at least partial convergence between party platforms, decreasing polarization. Critically, neither model’s authors address compulsory voting, nor do they discuss the normative implications of CV, voter abstention, or polarization. The next section discusses a number of potential objections to, and scope conditions on, our theory.
To understand the effect of compulsory voting on party polarization, we first need a basic theoretical model of political competition. For the rest of this article, we make the following assumptions, common to spatial models of politics. First, we assume a unidimensional policy space. This means that voters’ ideal policy preferences can be represented on a single left–right spectrum, with policies on the furthest left (right) appealing to the most left-leaning (right-leaning) voters. This policy space could represent a single dimension, like taxation, or bundles of policies.
Second, we assume that each voter has a single ideal policy that they prefer, and that we can model voters’ utility as a function of the distance between their ideal point and a party’s proposed policy: the larger the gap, the lower the utility of the voter. This directly implies that, if a voter casts a ballot, they will prefer to vote for the candidate/party whose position is closer to their ideal point. Third, we assume that parties choose their campaign platform policy to maximize vote share. Finally, we assume that the political system is majoritarian: this follows other literatures (both empirical and normative) on polarization and party competition.
Consider a political system with two dominant parties. Leading up to an election, each party must choose a party platform from the unidimensional policy space. Voters then choose whether to vote: if they do vote, we assume that they vote for the candidate whose proposed policy is closer to their own preferred policy. A pair of policy proposals by the two parties constitutes an equilibrium if neither party can improve their expected vote share by moving to the left or right, holding the other party’s platform fixed. The models below take this general model and adapt it to explore the effects of abstention.
In this kind of spatial model of politics, party polarization can be defined as the absolute distance between the policies proposed by the two dominant parties in an equilibrium. That is, when the equilibrium is for both parties to propose the median voter’s ideal point, polarization is zero. If Parties A and B propose policies x and y, then the level of polarization is the distance between the policies,
$ \mid x-y\mid . $
Drawing on this definition, we argue that CV decreases polarization if it reduces the absolute distance between the parties’ platforms, even if party platforms do not completely converge.
If compulsory voting affects party platforms, it must do so by affecting voter turnout. We next consider individual citizens’ turnout decisions. In order to vote, citizens must pay a cost in terms of their time and (potentially) lost wages. Voters will therefore abstain if they do not think voting is sufficiently “worth it.”Footnote 8 We follow previous literature in considering two key reasons for abstention. First, voters may abstain due to alienation; that is, a voter abstains when even the closest party is perceived as too far from their preferred policy. Second, voters may abstain due to indifference. In this case, the voter believes that the parties’ platforms are sufficiently close to one another that it is not worth voting. Note that this can occur even if the voter is not strictly indifferent between the two candidates; all that is required is that a voter’s utility from each platform is similar.
We argue that, while each type of abstention may have different effects on party platforms, the average effect of abstention in general is to increase the degree of party polarization. To understand the intuition behind this argument, consider abstention due to alienation. Say there is an election in which both parties’ platforms are relatively close to the median voter’s preference. Voters toward the left and right end of the spectrum may abstain due to alienation. If a party moves away from the median voter (i.e., toward the left or right end of the policy space) they may lose some voters from the center, who now vote for the other party. However, this move also induces some previously alienated voters to turn out. If the latter group is larger than the former, the party will be better off moving away from the median.
Now imagine this same polity introducing compulsory voting. Abstention of either type is no longer possible. This implies that many previously alienated voters will now turn out and cast valid ballots for the closest party, even if there is some ballot spoilage or protest voting (see discussion in the “Objections and Limitations” section). This means that parties can no longer gain votes from moving toward the extreme, and instead will have incentives to pick policy platforms that are closer to the ideal point of the median voter. This policy convergence will reduce the degree of party polarization in the polity.
Formalizing the Intuition
The argument introduced above, while intuitive, relies on a number of assumptions. To better understand those assumptions, and the conditions under which CV might decrease polarization, we consider two different existing formal models of party competition. These models are part of an extensive formal literature analyzing majoritarian party politics. In the most traditional, Downsian models of spatial politics, the unique equilibrium outcome is that both parties propose the policy preferred by the median voter—this is the well-known median voter theorem (MVT) (Downs Reference Downs1957). An equilibrium, in this context, is a set of policy proposals where neither party wishes to deviate; it is a stable point. Subsequent work has established that additional institutional features are necessary in order to support equilibria in which two dominant parties propose different policies. For example, divergence can be supported by introducing partisanship along with uncertainty about the median voter’s location; by introducing party primaries; and a number of other conditions (see discussion in Grofman Reference Grofman2004).
We focus on two existing models that incorporate voter turnout: Callander and Wilson (Reference Callander and Wilson2007) and Adams and Merrill (Reference Adams and Merrill2003). While these models share some elements, they also use different mechanisms to explain why party platforms diverge. Critically, in both models, abstention, particularly due to alienation, is a key force driving party polarization. By looking at two models, we show that our theory holds across multiple views of how elections work and why polarization arises in the first place.
In both models, abstention can generate party divergence, and thus polarization, only when combined with another aspect of the political system. In the Callander and Wilson (hereafter C&W) model, both abstention and the threat of entry by a third party are necessary. In the Adams and Merrill (hereafter A&M) model, divergence is driven by abstention combined with citizens who have partisan preferences separate from their preferences over policy. While neither model explicitly addresses compulsory voting, both can be used to help us understand the likely effects of an increase in turnout from CV, both when CV increases turnout to 100%, and when there is still some abstention or spoiling of ballots. This section provides an overview of each model and how it can be applied to understand compulsory voting; Supplementary material S1 contains a more detailed, and formalized, discussion of each model.
Alienation and Third-Party Entry: The C&W Model
C&W model a polity with two existing parties and a potential third-party entrant. We first discuss the model’s assumptions, and how abstention works in the model, and then discuss the results. We then show how the model can be applied to better understand CV. In the C&W model, each of the two parties proposes a policy platform. Next, a third party decides whether to enter the election (and propose a policy of its own) or stay out.Footnote 9 Citizens then observe the available candidates and choose whether to cast a ballot or abstain. Following the spatial model assumptions outlined above, C&W assume that voters have an ideal policy in the unidimensional policy space and, if a citizen turns out to vote, she votes for the candidate whose policy is closest to her own ideal policy. After votes are cast, the winner of the election takes office.
A critical piece of the model is citizens’ turnout decisions. C&W assume that citizens get an expressive, non-economic benefit from voting; this gives citizens an incentive to vote even if their vote is unlikely to be pivotal.Footnote 10 However, citizens also suffer disutility proportional to the distance between their own ideal point and that of the candidate for whom they vote. This can generate abstention due to alienation: if both candidates’ proposed policies are sufficiently far from the citizen’s own ideal point, the citizen prefers to abstain rather than vote for a candidate whose policies are not in line with their own. For example, in the 2016 US election, voters on the far left may have abstained rather than vote for Clinton.
The first result of C&W is therefore that abstention due to alienation, combined with the threat of a third party in a two-party system, can lead to party divergence. If both parties locate at the median voter’s preferred policy, a third-party candidate may be able to enter toward the left or right of the distribution and pick up a substantial number of otherwise disaffected voters, much like Ralph Nader did in the 2000 US election. Parties may be willing to locate further from the median voter’s preferred policy in order to avoid this. C&W show that, under a wide range of different possible distributions of voter preferences, the threat of abstention due to alienation, plus the threat of entry by a third-party candidate, leads to divergent equilibria in which one party locates to the left, and the other to the right, of the median voter’s preferred policy. This by definition creates some degree of polarization.
Recall that we define political polarization as the distance between the platforms offered by the main parties. C&W’s model’s second prediction is that the level of turnout will have a direct impact on the degree of party divergence. As this divergence is our preferred measure of polarization, we can use it to examine how an increase in turnout due to compulsory voting would affect party platforms. Recall that, when the gap between a voter’s own preferred policy and that proposed by the closest party is too large, a voter prefers to abstain from voting altogether. This is because the disutility of voting for such a far-away candidate exceeds any perceived benefit from voting. C&W consider the largest acceptable policy gap for a particular voter their alienation threshold. As this threshold increases, voters are willing to tolerate a greater level of alienation before they refuse to vote, and therefore overall voter turnout increases.
C&W show that, when voter turnout increases, it has predictable effects on party platforms. Critically, past a certain point, increases in voter turnout will actually lead both parties to move away from the edges of the distribution and toward the policy preferred by the median voter. That is, polarization (the gap between party platforms) decreases. At very high alienation thresholds, the parties fully converge to the median voter’s preferred policy. This finding of the model has direct implications for the effects of compulsory voting.
Because compulsory voting increases the costs of not voting (by imposing fines or other penalties) it will drive some citizens who previously abstained due to alienation to vote. This will lead to significantly higher voter turnout, even if CV is not completely successful at mobilizing all potential voters. In terms of the C&W model, CV increases the alienation threshold (that is the distance citizens are willing to accept between their own ideal point and their preferred party’s platform) and we can apply their finding that such an increase leads to partial or total party convergence in equilibrium. This is, by definition, a decrease in party polarization. This result holds even if not all voters turn out, or if some voters spoil ballots (see discussion in the next section).
Alienation and Non-Policy Preferences: The A&M Model
The C&W model relies on abstention due to alienation plus the threat of third-party entry to generate polarization. While this kind of entry (or its threat) is in fact quite common in majoritarian political systems, we may worry that CV will not have the same effects in the absence of a potential third party. We therefore draw on a second model of politics, Adams and Merrill (Reference Adams and Merrill2003) (hereafter A&M) to show that the predicted effect of compulsory voting is robust to a different modeling approach.
Like C&W, the A&M model assumes a polity with two dominant parties who each select a policy from a unidimensional policy space. They also assume that voters have a preferred policy (ideal point), and all else equal prefer a candidate whose platform is closer to that ideal point. However, in contrast to C&W, A&M do not allow for third-party entry. Instead, in order to generate divergence in party platforms, they make three assumptions, each of which we discuss more below. First, they assume that voters can abstain either due to alienation or due to indifference. Second, they assume that voters have a partisan bias toward one of the two parties that is separate from the policies those parties propose, and so all else equal will prefer their own party to the other. Finally, they assume that these non-policy-related (partisan) leanings will in fact be correlated with policy preferences. That is, voters with left-of-center policy preferences are more likely (but not certain) to also prefer the left-wing party for non-policy reasons.
The non-policy preferences in A&M are central to the model’s predictions, and so it is important to understand them properly. A&M argue that citizens typically have one or more characteristics that make them predisposed toward one of the two major parties. This could include demographic characteristics like race, gender, income, or age. For example, in the US Black voters are more strongly associated with the Democratic party and are also more liberal on average; the same is true for members of the LGBTQ community. This implies that, regardless of the precise platforms chosen by each candidate in an election, most voters are predisposed to vote for their preferred party, in part for reasons independent of those platforms. A&M summarize this for most of the article into a partisan identity; in the model, voters get utility from voting from whichever party is closer to their preferred policy, but also from voting for the party that shares their partisan identity.
A&M use these assumptions to model both abstention due to alienation (similar to C&W) but also abstention due to indifference, which C&W do not consider. A&M argue that a citizen will abstain due to alienation when even voting for her preferred party generates too little utility to make voting worthwhile; this is more likely when one’s preferred party proposes a policy that is too far from one’s own preferences. Abstention due to indifference arises when a voter’s utility if they vote for their own party is too similar to their payoff if they vote for the opposing party; that is, the two candidates’ policies are too close together. As A&M put it, “According to our model, a voter votes for her preferred candidate if her utility for that candidate exceeds her alienation threshold and the differential between her utilities for the two candidates exceeds her indifference threshold” (Adams and Merrill Reference Adams and Merrill2003, 166). Otherwise, she abstains.
A&M show that, if abstention is possible and citizens have partisan identities, then party platforms will diverge. When voters can credibly threaten to abstain due to alienation, and citizens have partisan preferences, parties have little incentive to move to the center in order to sway the other party’s voters; given partisan identities, such a move is unlikely to succeed. Instead, parties focus on ensuring that their own partisans are sufficiently motivated to vote, particularly those at the extremes of the policy space. This gives parties a strong incentive to move away from the median voter of the electorate writ large, and toward the policy preferred by the median voter among their own partisans. This generates policy divergence. Interestingly, A&M describe this as a desirable political feature, one that gives citizens more choice between parties. We argue, instead, that excessive divergence is by definition polarization, and can lead to the negative consequences discussed earlier.
Abstention from indifference does not drive policy divergence in the A&M model. Rather, when abstention is driven solely by indifference, then parties will tend to move toward the center. This is because, while the party may lose some voters on the edge of the policy space to alienation, it will also make some of the opposing party’s partisans stay home by making them indifferent between the two parties. When A&M allow both types of abstention (alienation and indifference) in combination with sufficiently strong partisanship, they find that the equilibrium outcome is for each party to locate away from median, and close to the median policy preferred by their co-partisans.
How, then, can we understand compulsory voting in this framework? A&M show that, absent the threat of abstention due to alienation, partisanship by itself cannot sustain policy divergence: under complete turnout, both parties will propose the policy preferred by the median voter. This directly suggests that compulsory voting should lead to greater policy convergence and therefore decreased levels of polarization. Even if CV does not increase turnout to 100%, we should still expect polarization to decrease in the model, as there will be fewer votes to be gained from deviating away from the median voter’s preference, and thus incentives for polarization are weaker. Thus, we should still expect a decrease in polarization even if some voters still abstain under CV.
Empirical Support for the Model’s Predictions
There is limited empirical evidence on the linkages between party positions, voter turnout, compulsory voting, and political polarization. This is in part because the argument presented above is novel, and thus has not been directly tested. While we can draw on related literature on turnout and party positions, doing so is likewise problematic. This is because the distribution of voter types (including tolerance for alienation and indifference, as well as voter ideal points) in a particular election determines both the equilibrium position of parties, and the level of turnout. This endogeneity makes it difficult to disentangle the effect of party positions on turnout, for example. However, we can use existing empirical evidence to examine two claims: that voters will abstain due to alienation, and that compulsory voting will lower party polarization.
First, consider abstention due to alienation, which implies lower voter turnout among those whose preferred policy is far from even the nearest party’s platform. This is a key assumption that is central to both models’ results. There is support for this hypothesis from studies of US presidential elections and US senate races that assess both abstention from alienation and abstention from indifference. In US presidential elections from 1968 to 1980, Zipp (Reference Zipp1985) finds that both abstention from alienation and abstention from indifference make a significant difference in voter’s probability of turning out to vote; a one standard deviation change in alienation lowers the probability of turning out between 1.5 and 4.5 p.p. Using ANES data on 1980s US presidential elections, Adams, Dow, and Merrill (Reference Adams, Dow and Merrill2006) find that 18–21% of respondents abstained from alienation, 13–14% abstained from indifference, and 11–18% abstained from both, leading to total abstention of 46–50%. Finally, Plane and Gershtenson (Reference Plane and Gershtenson2004) examine US Senate races from 1988 to 1992 and find that a one standard deviation increase in a voter’s alienation level leads to a 6 p.p. decrease in the probability of voting, with a change from minimal alienation to maximal alienation leading to a 29.5 p.p. decline in the probability of voting.
Note that abstention from alienation does not require that voter turnout is lower among more extreme voters: in fact, when party positions are highly polarized, we argue that it is specifically in order to ensure that these extreme voters turn out. Muñoz and Meguid (Reference Muñoz and Meguid2021) find some support for this claim. Exploiting the multiround nature of the 2012 and 2017 French national elections, they find that polarization increases turnout among voters who now find themselves ideologically closer to one of the more polarized candidates.
Second, our theory’s main claim is that introducing compulsory voting will reduce polarization. This claim has not been the explicit focus of much empirical work; in a recent book-length treatment of the consequences of CV beyond turnout, Singh (Reference Singh2021) notes that “there is a limited amount of theoretical and empirical research on the relationship between compulsory voting and party behavior” (135). Singh’s excellent work goes some way toward addressing this limitation and provides some evidence in support of our theory. Using cross-country comparisons, Singh (Reference Singh2021, 142–66) finds evidence that mainstream parties in countries with CV moderate their messaging on several issue areas, and that the effect is stronger when compulsory voting is enforced more strongly. He also finds some support for his theory that nonmainstream parties may become more extreme as they attempt to mobilize the most alienated voters who will refuse to vote even under CV. Comparing the theory presented by Singh (Reference Singh2021) with the formal models presented in this article suggests that there may be significant differences between majoritarian systems in which one requires a plurality or majority of votes in order to win elections and PR systems in which multiple parties can position themselves across the ideological spectrum and gain seats even with very low levels of support. We therefore expect that our results on polarization primarily hold in majoritarian systems rather than in systems with PR. This discussion also suggests that further empirical testing of the hypothesis regarding CV and polarization would benefit from treating countries with CV and majoritarian institutions differently from countries with CV and PR.
Beyond Singh (Reference Singh2021), there are a few country-level studies of CV that suggest its introduction could serve to reduce political polarization. In the case of Australia, multiple authors attribute the limited degree of political polarization to CV, and Malkopoulou (Reference Malkopoulou2020) argues that CV serves to reduce the appeal of rightwing populism. One dissenting view is de Leon and Rizzi (Reference de Leon and Rizzi2016), who claim that forced voting in Brazil results in political polarization. However, Brazil uses a PR system. While important, none of these existing studies investigate political polarization understood as the ideological distance between parties. Singh (Reference Singh2021) examines whether party manifestos show evidence of moderation (that is by deemphasizing issue positions central to their ideologies such as social justice and equality for parties on the left and law and order for parties on the right). Malkopoulou (Reference Malkopoulou2020) focuses on support for rightwing populism. de Leon and Rizzi (Reference de Leon and Rizzi2016) focus on support for extreme left parties. Although each of these measures could be strongly correlated with the type of political polarization that we argue leads to democratic backsliding and dysfunction, we do not have any direct evidence of this as of yet.
Scope Conditions
One advantage to drawing on the C&W and A&M models is that they help to clarify the scope conditions on our theory. First, both models assume majoritarian political systems with two dominant parties. This matches other work on polarization, as there is (as of yet) no consensus on how to best measure or conceptualize polarization in the multi-party systems characteristic of PR systems. Second, both models show that the threat of abstention due to alienation is critical for driving polarization. If there is a political system where this is not the case, then the results from the models may not hold. However, note that under both models, if parties are polarized, we should see citizens with more extreme policy preferences voting at higher rates than they did under less polarized conditions. This is because, if the threat of abstention by extreme voters is effective, parties will locate away from the center and thus those voters will in fact turn out to vote.
Finally, the C&W model suggests that turnout must increase beyond some threshold level—about 35% in their simulations—before it leads to party convergence. While this is encouraging in that it suggests full turnout is not necessary to achieve lower polarization, it does suggest that implementing CV in a country with extraordinarily low turnout (such that even CV leaves turnout under 35%) is unlikely to have the desired effects. However, C&W do show that their main results hold under a wide range of potential voter distributions, including those that best represent large majoritarian systems like the US and Great Britain. The A&M article likewise shows that their main result—that abstention due to alienation drives party divergence—holds under a set of very general conditions on voter preferences and distributions (see Supplementary material S1 for technical discussion).
Objections and Limitations
There is widespread concern that polarization is bad for democracy; the previous section presented an argument for why compulsory voting, by increasing turnout, could decrease elite polarization and ameliorate at least some forms of democratic backsliding. Two formal models of party competition support this argument, showing that decreasing or eliminating the possibility for citizens to abstain, particularly due to alienation, can decrease pressures for parties to locate toward the preferences of their more extreme members. While existing empirical evidence is limited, it supports the models’ assumptions and our theory. This section considers four potential limitations and objections to our theory: the potential for spoiled ballots to undermine CV; the potential lower quality of compelled votes compared to voluntary votes; whether CV might decrease ideological polarization at the cost of increasing affective polarization; and the potential for CV itself to undermine democratic survival by increasing dissatisfaction with democracy among compelled voters.
Objection 1: What If Alienated Voters Spoil Ballots?
One concern with CV is that, while voters may be compelled to show up to the polls, they are not required to cast a valid ballot. It is possible that voters who would have abstained absent CV will simply spoil their ballots, negating CV’s impact on the number of valid votes. This would be of particular concern if voters who would have abstained due to alienation are those most likely to spoil ballots.
In every election, some fraction of ballots cast is invalid. There are typically two reasons for this. First, citizens can cast protest votes when they object to the election overall, or to all candidates—for example, a voter could submit a blank or X-ed out ballot or write in “Mickey Mouse” or “none of the above.” Second, a voter could intend to vote, but fill in the ballot incorrectly or incompletely, resulting in the ballot being invalidated. This is often due to a ballot being confusing or difficult to fill out for some voters.
If CV increases ballot spoilage by alienated voters, or if opposition to the implementation of CV increases overall ballot spoilage, then the effects of CV could be attenuated or even eliminated, depending on the size of the effect. This raises two empirical questions: whether CV actually increases the fraction of spoiled ballots in an election, and which types of voters are most likely to spoil ballots.
There is a consistent finding in cross-country studies that CV increases the number of invalid ballots when compared to voluntary voting systems (Hirczy Reference Hirczy1994; Kouba and Lysek Reference Kouba and Lysek2016; Martinez i Coma and Werner Reference Martinez i Coma and Werner2019; Power and Garand Reference Power and Garand2007; Singh Reference Singh2019; Reference Singh2021; Uggla Reference Uggla2008). The effectiveness of CV in reducing polarization will therefore depend on the expected magnitude of this increase in invalid ballots relative to the increase in turnout. The majority of cross-country studies of the effect of CV on invalid voting estimate that the move from a voluntary voting system to a strictly enforced CV system will lead to an increase in invalid ballots of 5 to 7 p.p. (Kouba and Lysek Reference Kouba and Lysek2016; Martinez i Coma and Werner Reference Martinez i Coma and Werner2019; Uggla Reference Uggla2008). Studies exclusively focused on the Americas tend to find a higher increase in invalid ballots, with Power and Garand (Reference Power and Garand2007) estimating an 8.52 p.p. increase in their sample of 80 legislative elections across 18 democracies in Latin America and Singh (Reference Singh2019) estimating an increase of 12 p.p in a sample of Latin American and Central American democracies. In contrast, looking at national variation at provinces in Austria with and without CV, Hirczy (Reference Hirczy1994, 68) finds an increase in invalid ballots of 1.9 p.p. in federal elections and 2.4 p.p. in provincial elections.
Table 1 uses a set of equilibrium simulations, based on those in C&W, to consider a set of possible scenarios for the increase in invalid ballots. First, we consider three different baseline levels of turnout: 40%, 50%, and 60%. These cover the range of turnouts seen in recent US congressional and presidential elections over the past two decades. For each, we assume that CV increases total voter turnout by 16.5 percentage points; this is in the middle of the empirical estimates of 14.5–18.5 percentage points (see the “Normative Case for Compulsory Voting” section). We then consider the net effect of turnout on polarization under three different levels of spoiled ballots: (1) the most common estimate of 6.5 p.p. (Kouba and Lysek Reference Kouba and Lysek2016; Martinez i Coma and Werner Reference Martinez i Coma and Werner2019; Uggla Reference Uggla2008), (2) a medium-level estimate of 8.5 p.p. (Power and Garand Reference Power and Garand2007), and (3) the highest possible estimate based on Singh (Reference Singh2019) of 12 p.p. This gives us net turnout increases of 10, 8, and (rounded down) 4 p.p.
Table 1. Simulations of the Effect of CV on Polarization
Abbreviation: CV means compulsory voting.
For each scenario, we estimate the percent change in polarization. Supplementary material S1, Supplementary Table 1A| reports more detailed calculations, while Supplementary Figure 1A (based on Figure 2 from Callander and Wilson [Reference Callander and Wilson2007, 1054]) provides a visual illustration of the predicted effects. Across all nine scenarios, we see significant decreases in polarization due to higher turnout. As expected, the estimates are higher when baseline turnout is lower, and when the net increase in turnout is higher. Even in the “worst-case” scenario when spoiled ballots mean that turnout only increases from 40% to 44%, we still see a meaningful 22% decline in polarization.
These estimates also assume that all spoiled ballots caused by CV are caused by alienated voters. Ballots spoiled for reasons independent from party platforms do not affect the incentives for convergence. The empirical literature reveals that there are multiple motives for compelled voters to cast an invalid ballot, including ballot complexity combined with low literacy or inexperience (Hill and Young Reference Hill and Young2007; McAllister and Makkai Reference McAllister and Makkai1993; Power and Roberts Reference Power and Roberts1995), low interest in politics, distrust in government, and dissatisfaction with democracy (Singh Reference Singh2019).Footnote 11 We therefore expect the predictions above to be the lower bound on the effect of CV on polarization.
Objection 2: What If Newly Mobilized Voters Cast Lower Quality Votes?
A second potential concern is whether voters who are newly mobilized by CV will, as our theory assumes, be proximity voters: that is, will they vote for the candidates whose preferences are closest to their own? For example, if those who abstain under voluntary voting are poorly informed about politics or do not have well-defined policy preferences, they may vote almost at random, or use non-policy-based decision rules, either of which would make it hard to predict how parties will respond and what equilibrium will result.
Fortunately, existing empirical evidence suggests that, while some voters mobilized by CV may vote in these ways, most do not. Dassonneville et al. (Reference Dassonneville, Feitosa, Hooghe, Lau and Stiers2019) find that the most reluctant voters (i.e., those who say they would not vote if voting was voluntary) are on average 10% less likely to vote for the most ideologically proximate party, but the effect is only statistically significant in 50% of the elections they studied and the effect is especially likely in countries with PR and large numbers of parties such as Belgium. Singh (Reference Singh2022) finds no evidence that compelled and voluntary voters differ in the way that they select candidates or in the weight they attribute to policy considerations. While his study was primarily looking to see if CV increases the level of political sophistication and therefore improves the quality of the vote, his null result shows that, even if CV does not lead to better votes, it also does not lead to worse votes. We therefore expect that, while these non-proximity voters may slightly reduce the impact of CV, they will not eliminate its effects.
Objection 3: What If CV Increases Other Forms of Political Polarization?
Our theory focuses on elite polarization, examining the potential for CV to reduce the distance between the platforms of the dominant parties in a majoritarian system. One potential worry is that while the introduction of CV will reduce the distance between the main parties, it will simultaneously increase the levels of other types of harmful polarization, particularly affective polarization among citizens. Affective polarization is generally defined as the tendency of partisans “to view opposing partisans negatively and co-partisans positively” (Iyengar and Westwood Reference Iyengar and Westwood2015, 691). In recent years, affective polarization has increased significantly in the United States (Iyengar et al. Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019; Iyengar and Westwood Reference Iyengar and Westwood2015) and across a range of other democracies, many with majoritarian political systems (Gidron, Adams, and Horne Reference Gidron, Adams and Horne2020; Reiljan Reference Reiljan2020).
One way in which the introduction of CV might increase affective polarization is by increasing the number of people who identify as partisan. The empirical evidence does generally suggest that CV is associated with higher rates of partisan identification among voters and slightly higher strength of partisan attachment (Dalton and Weldon Reference Dalton and Weldon2007; de Leon and Rizzi Reference de Leon and Rizzi2016; Singh and Thornton Reference Singh and Thornton2013), although the finding is not universal (Rau Reference Rau2022). While there is no consensus regarding the causal effect of partisanship on affective polarization, cross-country data reveal a positive correlation between partisanship and affective polarization (Wagner Reference Wagner2021). This suggests that, if CV increases partisanship, then CV could increase affective polarization, and existing empirical evidence cannot rule this out.
At the same time, it is also possible that CV could reduce affective polarization if a reduction in the distance between the main parties decreases animus between the two sides. As opposing parties offer increasingly similar policies, voters might see opposing partisans as more similar and therefore experience less “fear and loathing” for them. The relationship between party distance and affective polarization receives some support in the literature. Using survey experiments in the United States, Rogowski and Sutherland (Reference Rogowski and Sutherland2016) and Webster (Reference Webster and Abramowitz2017) find that increased perceptions of ideological distance between parties generate significant increases in affective polarization, particularly reported dislike for opposing partisans. This suggests that elite polarization can influence polarization among ordinary citizens. Moral and Best (Reference Moral and Best2023) show that the causal arrow generally runs from elite polarization to mass polarization, with the largest and most immediate effects among the most politically sophisticated citizens. However, his study focused on ideological rather than affective polarization, so further research is required in this matter.
Establishing which effect predominates requires a clearer account of whether partisanship or policy distance has the stronger causal impact. There is currently no consensus on this matter in the empirical literature on affective polarization, although there are a few papers that have tried to assess the effects of both partisanship and policy distance on affective polarization. Both survey experiments (Lelkes Reference Lelkes2021) and cross-national observational data (Hernandez, Anduiza, and Rico Reference Hernandez, Anduiza and Rico2021) show that both mechanisms matter but that ideology has a stronger effect. However, Dias and Lelkes (Reference Dias and Lelkes2022) find that the two effects are linked and a large portion of the effect of ideological distance is explained by partisan identity. There are, to the best of our knowledge, no studies of the relationship between CV and affective polarization. Although existing evidence suggests that on balance CV should reduce affective polarization due to lower elite polarization, we cannot rule out the possibility that CV might increase affective polarization. One of the contributions of our article is to clearly spell out the possible mechanisms through which CV will influence political polarization in ways that can advance the empirical conversation.
Objection 4: What If CV Undermines Support for Democracy?
A final objection is that introducing compulsory voting could upset citizens who believe voting should be voluntary, potentially lowering support for democracy (and increasing democratic backsliding) even if CV successfully reduces polarization. This could lead to either higher support for autocrats and the overthrow of democracy, or to higher votes for extremist candidates.
The empirical evidence on both possibilities is mixed. Singh (Reference Singh2018) finds that CV can exacerbate dissatisfaction with democracy among those who already have anti-democratic attitudes. Yet, Singh (Reference Singh2021) finds no evidence that CV amplifies the negative relationship between dissatisfaction with democracy and lower support for authorities. Empirical studies of the relationship between CV and votes for extremist parties likewise are mixed. Singh (Reference Singh2021) finds that CV can exacerbate the extent to which those with negative attitudes toward democracy intend to vote for extremist parties. Miller and Dassonneville (Reference Miller and Dassonneville2016) find that CV in the Netherlands and Belgium accounts for a 3 to 5 p.p. increase in the vote share of rightwing parties. However, Birch (Reference Birch2009) does not find any significant increase in the vote share of extremist rightwing parties in countries with CV, except in Belgium (which she considers a significant outlier).
We expect that the propensity of CV to increase the vote share of extremist parties will depend on the electoral system and will be most likely in systems with PR—particularly PR systems with low thresholds for gaining seats. This represents a further reason why our argument about CV and polarization may be restricted to majoritarian systems. More generally, our normative conclusion about CV is contingent on its ability to increase rather than decrease democratic resilience and more empirical evidence is necessary to ensure that CV’s positive effects outweigh potential negatives.
CONCLUSION
Normative political theorists often worry about persistently low turnout in national elections as a sign of democratic dysfunction, but there is significant disagreement about when and why low turnout is a cause for concern. Our article introduced a new argument about the normative significance of electoral turnout: in majoritarian systems where voters can threaten to abstain due to alienation, low turnout can lead to political polarization and therefore threaten the survival of democratic institutions. In these cases, the introduction of compulsory voting can reduce polarization and protect democratic institutions from antidemocratic threats. Our argument strengthens the normative case for compulsory voting, which otherwise tends to require a commitment to contested theories about the moral duty to vote and/or about democratic legitimacy. One can support the introduction of CV as a method of reducing polarization regardless of whether one is a participatory democrat, deliberative democrat, epistemic democrat, or realist democrat, and regardless of whether one considers voting to be a duty or a right, provided one is committed to democracy as superior to authoritarian alternatives.
Our article also contributes to a growing literature in politics, philosophy, and economics that investigates political institutions using interdisciplinary tools from political science, normative political philosophy, and economics (Brennan and Hamlin Reference Brennan and Hamlin1998; Reference Brennan and Hamlin2000; Waldron Reference Waldron2016). In the article, we repurpose two existing formal models of the relationship between voter turnout decisions and political party behavior to investigate the effect of introducing compulsory voting. Both models show how voter abstention, particularly due to alienation, can generate divergence in party platforms. However, neither model takes polarization as a central concern nor examines directly how their models could affect polarization. Likewise, neither model considers compulsory voting. The use of the formal models allows us to develop scope conditions under which compulsory voting would be effective in reducing political polarization and to estimate the magnitude of these effects under a range of empirically realistic scenarios.
More theoretical and empirical work will be needed to fully understand the impacts of CV and how it is best implemented. In the case of the United States, for example, one would have to decide whether voters should also be compelled to participate in primary elections. Although a full assessment of the consequences of compulsory voting in primary elections is beyond the scope of this article, we believe that CV can still substantially reduce political polarization even if participation in primaries remains voluntary. This is because the national election with CV will still exert significant pressure on candidates to move toward the center or otherwise lose the election to more centrist opponents. At the same time, we would expect compulsory voting in primaries to further reduce polarization as the winner of each primary would be closer to the preferences of the median voter under CV than under voluntary voting.
Although our article focuses on majoritarian systems, there may be circumstances under which CV would reduce polarization in systems with proportional representation or large numbers of parties. For example, under voluntary voting, one might expect abstention from alienation to motivate parties on the fringes to move closer to the most extreme voters, increasing the distance between the leftmost and rightmost parties. The few empirical studies that have investigated polarization in multi-party systems confirm our argument that systems become more polarized when turnout is low. Most recently, Dreyer and Bauer (Reference Dreyer and Bauer2019) looked at 11 Western European democracies and found that “parties adopt more extreme positions in response to higher voter polarization and that the size of the effect declines as turnout increases.” This offers some evidence regarding the effect of CV on polarization in multi-party systems. However, the effects of CV on polarization under PR require a separate formal and empirical treatment that is beyond the scope of this article. To the extent that a similar effect exists, we would support the introduction of CV in these political systems as well.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055423001399.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/QIA26P.
ACKNOWLEDGEMENTS
Tragically, Prof. Brennan passed away on July 29, 2022, prior to the second round of revisions to the manuscript. The project would have never been possible without his inspiration, his generosity of spirit, and his persistence in thinking through these ideas even while battling cancer. His co-authors attribute to him all the best elements in the final version, but we reserve to ourselves all remaining errors and infelicitous expressions. The authors would also like to thank Susan Bickford, Eric Cheng, Keith Dowding, Jeff Spinner-Halev, Lars Moen, Daniel Muñoz, and Daniel J. Stephens, for helpful comments on earlier versions of this article. We would also like to thank Steven Callander for sharing his replication code. We are also very grateful to the APSR editors and to the six anonymous reviewers. All of your comments have significantly improved the quality of the article.
CONFLICT OF INTEREST
The authors declare no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The authors affirm this research did not involve human subjects.
Open access
