Must Realists Be Pessimists About Democracy? Responding to Epistemic and Oligarchic Challenges

In this paper we show how a realistic normative democratic theory can work within the constraints set by the most pessimistic empirical results about voting behavior and elite capture of the policy process. After setting out the empirical evidence and discussing some extant responses by political theorists, we argue that the evidence produces a two-pronged challenge for democracy: an epistemic challenge concerning the quality and focus of decision-making and an oligarchic challenge concerning power concentration. To address the challenges we then put forward three main normative claims, each of which is compatible with the evidence. We start with (i) a critique of the epistocratic position commonly thought to be supported by the evidence. We then introduce (ii) a qualified critique of referenda and other forms of plebiscite, and (iii) an outline of a tribunebased system of popular control over oligarchic influence on the policy process. Our discussion points towards a renewal of democracy in a plebeian but not plebiscitarian direction: attention to the relative power of social classes matters more than formal dispersal of power through voting. We close with some methodological reflections about the compatibility between our normative claims and the realist program in political


Introduction
Political scientists often accuse normative democratic theorists of theorizing in ways unrelated to actually existing democracies. The accusation is that the best evidence we have about the actual functioning of contemporary democracies uncovers constraints that challenge our normative models-and even the desirability of democracy itself. Whatever the merits of the accusation, we propose to take the most pessimistic-yet solid-empirical results about actually existing democracy, and see how normative political theory can respond.
We will discuss two main sets of results, not because they are the most accurate, but because they are often thought to be the greatest reason for pessimism in our understanding of advanced democracies-the United States, mainly, but also Western Europe.1 Broadly, they are both results about democracy's unresponsiveness to voters' preferences. For now let us anticipate them in crude terms. The first is the finding, by Achen and Bartels (2016) and others, that neither policy selection nor leadership selection motivate voting behaviour (not even in a negative sense). Rather, voters vote based on their group identity, which does not amount to a coherent stance on either leaders' performance or policy priorities. The second one is the finding by a number of scholars (for the United States see Bartels 2008;Winters and Page 2009;Hacker and Pierson 2010;Gilens 2012;Gilens and Page 2014;Schlozman, Brady, and Verba 2018. For Western Europe see Hopkin and Lynch 2016;Schakel 2019) that portray the democratic process as generating outcomes that reflect the preferences not of the median voter, but of economic elites and organised (mostly pro-business) interest groups. This leads to an oligarchic bias within public policy-it is mostly a class-level phenomenon concerning the top 10% of the economic pyramid and, at the limit, there is ample space for individual super-wealthy people to exert outsized influence. Crudely, the empirical claims we engage with show that voters don't choose policies nor leaders, yet somehow policies are reliably aligned with elite preferences. This problem intersects with wider concerns about the role of wealth concentration in contemporary capitalism (Piketty 2014;O'Neil 2017).
We argue that these findings, when viewed together, help set the terrain for a realist democratic theory that properly acknowledges how failures of citizen competence intersect with the problem of disproportionate elite influence: that is, how the "epistemic challenge" to democracy intersects with the "oligarchy challenge." These phenomena each have independent causes, yet they coexist in a perverse symbiosis. If citizens struggle to make well-informed decisions on policy issues and are unable to deploy electoral mechanisms to secure meaningful accountability over elites, then these epistemic deficiencies only magnify the oligarchic threat to democracy. They provide more space for the wealthiest citizens to bypass popular scrutiny and exert political influence through more shadowy, non-electoral mechanisms: lobbying, influence-peddling, and various other forms of oligarchic capture of the public policy process (Winters 2011;Lindsey and Teles 2017). Achen and Bartels' epistemic critique carries its greatest force when juxtaposed against these other empirical dilemmas. But importantly, we argue that epistocratic elitism, of the sort advocated by Jason Brennan (2016), is not a path out of this dilemma. If the challenge for democracy is that citizens are bad at selecting leaders and wealthy elites capture the political process for their own ends, it is hardly clear that more elitism would solve these problems.
The upshot, then, is that successful efforts to reduce oligarchic influence in contemporary democracies will require democratic innovations that try to overcome some of the epistemic challenges discussed by authors like Achen and Bartels without falling into the camp of epistocratic elitism. And this will require both closer attention to the empirics and more aggressive institutional reforms than many political theorists have hitherto endorsed.
In this spirit, we make three main contributions. First we offer an argument against epistocratic elitism as a corrective to democracy, though we also show that, second, the epistemic challenge does have damning consequences for certain forms of plebiscitarian democracy. Third, we begin our positive contribution by exploring one compelling roadmap for institutional design that remains compatible with core realist commitments, but is also radical, in the sense of departing from mainstream democratic theory: the neo-Roman tribunate model endorsed most recently by "plebeian" democratic theorists like John McCormick (2011).2 The focus here is amplifying the socioeconomic identity of non-wealthy citizens by formalizing their power in class-specific institutions that can act on behalf of "plebeian"interests in a representative but not plebiscitarian mode. This model, while not fully curative, can help address some of democracy's current challenges, by offering a counterpoint to elite-dominated institutions and by increasing citizen consciousness of important policy issues, and we move beyond McCormick by developing the model in some fresh directions.
In addition to those three normative points, we envision this article as an innovative contribution to the realist program in democratic theory. While realism has now moved on from a purely methodological debate (Rossi 2015), thus far most contributions have primarily focused on carving out space for non-status quo-biased forms of realism (e.g. Raekstad 2018, Cross 2019, Miller 2019, Rossi 2019, or on offering non-moralised versions of extant normative positions (e.g. Jubb 2015). We hope, though, that by putting some of the relevant empirics front and center and constructing our normative conclusions on their basis, we can begin to chart a new way of doing realist political theory.

The Evidence
In this section we present the main relevant empirical findings, and discuss some of the main responses offered by political theorists so far.

The group theory of voting
In their widely discussed book Democracy for Realists, Christopher Achen and Larry Bartels (2016) take aim at the "folk theory of democracy," which, they claim, casts a large shadow over much work in political science. This is the idea that well-informed citizens successfully translate their preferences into public policy by electing representatives who are responsive to their will. Such folk theories are manifestly false, the book argues. Citizens neither determine policies (as Dahl 1998 would have it), nor do they select leaders (Schumpeter 1942), nor even enforce effective retrospective accountability (Key 1966). Rather, Achen and Bartels advance a "group theory" of voting, whereby specific ethnic, religious, racial, occupational, and socioeconomic ties are decisive in determining party affiliation, which then conditions the policy positions voters adopt as their own: "group ties and social identities are the most important bases of political commitments and behavior" such that "election outcomes have little real policy content" (Achen and Bartels,p. 319). Crucially, the authors suggest that these group ties are decisive even for the higher-educated voters whom one would expect to be driven by pure policy considerations (Ibid,p. 310).
In one sense, the book simply repackages core concerns raised by social choice theorists like Kenneth Arrow and William Riker, who demonstrate the difficulty of aggregating individual preferences into coherent collective choices. We won't rehash their argument here, though some scholars have raised important criticisms from a more pro-democratic perspective (Mackie 2003). In any case, Achen and Bartels go well beyond Arrow and Riker, with a range of comprehensive empirical claims. They find, for example, that Republican and Democratic lawmakers behave much differently in office, even when their constituents have similar preferences (Achen and Bartels 2016, p. 48): elections do not force these representatives to respond to anything resembling the median voter preference, regardless of whether the system is majoritarian or proportional (Ibid, p. 49). The authors also strongly criticize the retrospective theory of accountability, which posits that even when voters are ignorant about public policy specifics, they can still assess leaders' performance in office, and so retain some control over outcomes (Ibid,p. 91). Retrospective accountability requires voters to accurately assess blame, to discern how the decisions of incumbents impact their well-being, and voters often struggle to do this (Ibid, p. 115), engaging in"blind retrospection" by punishing incumbents for events out of their control, as when shark attacks in 1916 reduced Woodrow Wilson's vote share in New Jersey, in the book's most memorable example (Ibid,. The authors concede that voters do enforce some forms of issue-based accountability, often judging presidential incumbents on economic performance. But these voting patterns tend to be myopic, as voters base their assessments of economic performance on conditions in the months before elections rather than holistically across a term (Ibid, pp. 158).
In earlier work, Bartels (2008) argued that this myopic time-scale has benefitted Republican presidencies, which have experienced more concentrated income growth in election years (Ibid,. Moreover, voters often misjudge how particular economic policies impact their own bottom line. Many Americans express concerns about inequality in the abstract, while supporting tax cut measures whose benefits disproportionately flow to the more affluent (Ibid, pp. 163). Many Americans who supported repealing the estate tax mistakenly believed that their own taxes would be lowered as a result, for example (Ibid,. Achen and Bartels cite social-psychology experiments, in which white Americans, who are informed about the possibility of the country becoming "majority minority" through demographic change, then express more conservative views not just on immigration but also on issues like defence and healthcare (Achen and Bartels 2016, p. 265). The implication, which seems apropos post-Trump, is that Republicans can latch onto a heightened sense of white identity to push a broader policy agenda favoured by socioeconomic elites within the party, but not necessarily responsive to the economic preferences of constituents. But group voting is not just limited to white Republicans, as African-American Democrats have also been shown to strongly support politicians who match their racial identity (Achen and Bartels 2016, p. 313).
In effect, "voters choose a party validating their social and political identities, then rationalize their decision with appropriate party supplied reasons," (Ibid, p. 311) such that "election outcomes are, in an important sense, random" (Ibid, p. 176). This means that voters' actual knowledge of and commitment to specific policy goals is remarkably thin, except insofar as it reflects group identity; the result being that "conventional democratic ideals amount to fairy tales" (Ibid, p. 7), having "collapsed in the face of modern, social scientific research" (Ibid,12). We believe such sweeping statements are too harsh, and must be qualified, for example, against recent epistemic defences of democracy, which we won't discuss here, but which certainly command attention (Ober 2008;Landemore 2012). Nonetheless, democratic theorists of all stripes still have an obligation to take the more pessimistic findings seriously.

Criticisms
Of course, a controversial book like Democracy for Realists is bound to attract considerable pushback, some of it warranted. For example, while Achen and Bartels lodge empirical criticisms of participatory citizen initiatives, the evidence is far from complete, relying mainly on a small series of qualitative case studies, as Frega (2018, p. 8) stresses. We agree that well-designed participatory institutions of the sort that existed, for example, in democratic Athens (Ober 1989) have played a "decisive role in empowering citizens" (Frega 2018, p. 9), and Achen and Bartels would do better to acknowledge this point. Even episodic acts of voting can give citizens incentives to acquire additional information that they would not acquire otherwise; as Frega argues, Achen and Bartels' own data demonstrates that electoral primaries have a learning effect on citizens (Frega 2018, p. 14; see Achen and Bartels 2016, pp. 60-89).
Simone Chambers criticizes the book from a deliberative democratic perspective, arguing that its focus on pathologies inherent to individual rationality occludes the rationality gains that can come from group deliberation (Chambers 2018). This is a valid point, though research on the political significance of cognitive bias (Lodge and Taber 2013) is still emerging, and its pessimistic implications for theories of deliberative democracy cannot be dismissed prematurely. Andrew Sabl complains that "real democratic theorists, i.e., political theorists who write on democracy….have rarely imagined that democracy normally translates public preferences into policy, or judged that democracy's health depends on whether it does" (Sabl 2017, p. 157). This point is echoed by Niko Kolodny, who argues that many of Achen and Bartels' arguments for the primacy of group identification (i.e. Southern whites voting Republican) can still be explained in standard, issue-based terms, "broadly consistent with the 'folk' idea that people affiliate with the parties they do because they independently believe that those parties will satisfy their political preferences" (Kolodny 2017).
We agree that the relationship between group identity and issue-based preferences is complex. For example, in her pathbreaking study of rural voters in Wisconsin, Katherine Walsh (2012) traces the formation of a group identity she calls "rural consciousness," (Ibid, pp. 517-518) demonstrating how this identity is intricately bound up with informed positions on economic issues-for example, the idea that specific economic policies are rigged in favour of urban areas. The result, she argues, is a sophisticated conception of distributive justice that is, nonetheless, still highly responsive to "rural" values. Group identities and policy preferences are often so closely intertwined that it can be difficult to claim one as derivative of the other. As Kolodny emphasizes, "it is because creditors are members of the group 'creditors', for example, that they want low inflation," but "even if group identification drives these policy preferences, still those policy preferences drive party affiliation and voting, just as the folk theory would have it" (Kolodny 2017).
Democracy, as an ideal, clearly entails more than simply responsiveness to the preferences of the median voter. So when such responsiveness is not achieved, this fact isn't inherently a cause for alarm. Indeed, if democracy was only about preference fulfilment, then the source of these preferences would be, in some sense, inconsequential (Kolodny 2017). By implication, Achen and Bartels' critique works best when we acknowledge that democracy is also a form of anti-oligarchic counter-power, whose function is constraining elites. From this perspective, Achen and Bartels (and other similar critics) are most instructive when highlighting the limitations of existing (electoral) accountability mechanisms. If voters have difficulty assigning blame, making judgments about economic performance, voting on economic issues, and so on, then the central activity of electoral democracy-enforcing accountability against elites-is at least partially defective, a point stressed by other political scientists (Przeworski, Stokes, and Manin 1999).
The "realist" implications of their analysis are thus quite clear, and they concern the basic problem of elite capture of public policy, as we argue in the next section of the paper (Lindsey and Teles 2017;Bagg 2018). This means that a realist democratic theory must do more than simply "identify procedurally democratic nudges or barriers to counteract such forces of clannishness and emotion," as Cohen (2017, 153) suggests. It must aggressively work to minimize oligarchic power. And it must do so cognizant of some of the democratic defects noted by Achen and Bartels.
We now explore a second set of empirical findings which are heavily related to the epistemic issues explored above, but which must be treated as an independent set of concerns. We can start with Bartels' 2008 work Unequal Democracy, which launched a scathing critique of the so-called "new Gilded Age" political economy, demonstrating that contemporary democracies are systematically more responsive to the preferences of the more affluent, a point seconded by scholars like Martin Gilens and Benjamin Page, who show that when the preferences of economic elites and organized interest groups are controlled for, ordinary Americans have a "miniscule, near-zero, statistically nonsignificant impact upon public policy" (Gilens and Page 2014, 575; also Gilens 2012) leading to the "nearly total failure of 'median voter' and other Majoritarian Electoral Democracy Theories" (Gilens and Page 2014, p. 575). How should the disproportionate influence of the affluent be measured, empirically? The most solid studies focus on the mass affluent, i.e. the top decile. Public policy is indeed more responsive to this group, in part because its members are better organized, more likely to hold office, volunteer, donate, and do other things associated with active citizenship (Gilens 2012). That is worrying enough. But what those studies don't even rule out, and perhaps what ought to command our scrutiny even more, is the narrower cohort of super-rich elites, those with personal access to tens or hundreds of millions, or even billions. We'll refer to such individuals as oligarchs. Not only do oligarchs access massive wealth, but they can deploy that wealth for considerable discretionary influence in the public domain (Winters 2011;Arlen 2019). They can fund Super-PACs to lobby for their favoured political agenda, with the Koch Brothers' 501c3 organization, Americans for Prosperity (AFP), the paradigmatic example of an oligarchic driven special interest machine; a kind of shadow political party, as Skocpol and Hertel-Fernandez (2016) argue.
Oligarchs are empowered, then, to pursue their policy preferences through mechanisms that ordinary citizens can hardly fathom, mechanisms that have very little to do with electoral processes. It's notoriously difficult to get good survey data on the super-rich, to do sufficiently powered large-N studies, which can make these subjects appear inaccessible to social scientists, who focus instead on studying the mass-affluent. And yet, while the preferences of the mass-affluent and super-rich citizens are certainly aligned to some degree, the two groups should not be conflated. In one survey of wealthy Americans, with average net worth of around $25 million, the interviewees were shown to hold views diverging from most non-wealthy Americans; more preoccupied with national debt, more opposed to regulatory programs favoured by the general public; more sceptical of welfare programs (Page, Bartels, and Seawright 2013). But crucially, those with $40 million or more in personal wealth displayed more extreme views on these issues than did those with $5 million or less (Ibid,. While the precise mechanism of oligarchic influence can be difficult to measure empirically, Winters and Page have developed a so-called material power index, designed to quantify the disproportionate political power of the super-rich. According to their model, the average member of the Forbes 400 retains about 22,000 times more materially based power than does the average American (Winters and Page 2009, pp. 733-738). Of course, such estimates are highly speculative, but they do dramatize an important point: democratic modes of formal equality are perfectly compatible with massive amounts of substantive political inequality.
An obvious question is whether these findings can be generalized to other Western democracies. In a recent analysis of Dutch politics, Wouter Schakel (2019) successfully replicates some of Gilens' (2012) core findings about the link between affluence and influence. The Netherlands features some of the world's lowest levels of income inequality, and a highly proportional electoral system that limits the role of financial donations. Yet even in this "least-likely case," responsiveness is "strongly skewed towards wealthier citizens;" findings Schakel deems "strikingly similar" to those of the US (Schakel 2019, p. 11 and p. 17). He hypothesizes that corporate lobbying plays an especially strong role in Dutch politics, among other mechanisms. Other comparative studies document how both domestic inequality within Eurozone countries and between them have exacerbated American-style "winner-take-all" political dynamics, albeit in different forms (Hopkin and Lynch 2016;Matthijs 2016).
In short, whether we fixate on the mass affluent as a cohort that enjoys superior organization and access, or on the smaller group of oligarchs who exert personalist influence over the political process, the damage to democratic equality is real, if hard to measure precisely. Several implications follow. First, the pessimistic empirical results advanced by Achen and Bartels must be interpreted in terms of these trends towards "elective oligarchy." For these trends amplify the consequences of group voting insofar as not all groups are equally positioned to advance their preferences. If electoral mechanisms are mainly about harnessing the power of numbers, then the inadequacy of electoral accountability (be it retrospective or prospective) means that non-electoral mechanisms associated with oligarchic power will have more space to operate.
Pluralists inspired by Dahl (1961) might argue that the problem is less serious than it appears, because socioeconomic elites are dispersed through cross-cutting cleavages persisting along cultural and religious lines. Certainly, many different "groups" are represented among both the mass affluent and among the oligarchic elite (Christian evangelical businessmen vs. Silicon Valley tech titans), even if both groups do skew more white and male. But in fact, as Jeffrey Winters argues, there is a remarkable degree of cohesion within the oligarchic elite around core socioeconomic issues involving wealth and income preservation (Winters 2011, pp. 20, 217-254). All oligarchs can draw upon the "income defence industry," the high-priced lawyers and accounts adept at constructing complex tax sheltering strategies (Ibid).
If electoral democracy is primarily about group identity, then care must be taken to ensure that certain groups are not systematically advantaged. As Achen and Bartels stress: "The theory of group politics provides a clearer explanation for why a more egalitarian society would result in a more egalitarian political process," that is, a democracy in which political power better tracks under-represented groups (Achen and Bartels 2016, p. 326). Far from warranting complacency, then, their analysis actually offers up an empirical manifesto for the sort of aggressive reforms needed to curb elite influence: "Real politics is much more complex-and much more strongly shaped by unequal clout-than the fastidious vision of the folk theory suggests. Serious political reform must face that fact squarely" (Ibid, p. 327).
In short, while we have seen that electoral results bear no relation to policy outcomes, they are not entirely random. They reliably track the preferences of a small subset of the population. In terms of current theories of democracy, the evidence points towards what has been called elite domination, or at best "biased pluralism" (Gilens and Page 2014, pp. 567-568).

Two Realist Challenges
Let's take stock of the argument thus far. Our purpose in exploring this empirical evidence has been to help set the terrain for a realist democratic theory by establishing a "two-pronged" challenge to contemporary democracies. The first prong is the "epistemic challenge" associated with deficiencies in citizen competence and the failure of electoral mechanisms to enforce effective accountability. The second prong is the "oligarchy challenge" associated with rising wealth concentration and discretionary forms of elite influence that allow super-rich actors to elude popular scrutiny while pursuing their own political objectives.
Crucially, these challenges can be treated in tandem since the one amplifies the other. Of course, both have independent causes. The oligarchic threat to democracy reflects exogenous trends in political economy, and it would be reductionist to attribute these economic trends to epistemic errors in the voting population. Likewise, epistemic errors are not simply attributable to oligarchic manipulation of public opinion, though certainly, oligarchic funded mass-media can shape public opinion in deleterious ways. Nonetheless, our central claim is that scholars should focus on the interrelation between these phenomena to achieve proper conceptual clarity about the threats facing democracy today. This leads to some important normative conclusions, which we explore now.

Against Epistocracy
Achen and Bartels express concerns about citizen competence from within a framework oriented towards strengthening and defending the democratic project. There is a clear risk, however, that the empirical apparatus developed in their work might be deployed to buttress more radical forms of epistocracy that are hostile to democracy.
Indeed, one critic suggests that Achen and Bartels should have endorsed a more explicitly epistocratic model on the basis of their own premises; "I struggle to see how one can be a realist of the Achen and Bartels type and still maintain a commitment to anything resembling democracy" (Ahlstrom-Vij 2018, p. 11). But to suggest that realism is incompatible with democracy (if only Achen and Bartels had acknowledged this!) is to overlook the actual thrust of their critique, which actually recommends against epistocracy, or at least against some common arguments for epistocracy, as we will try to show. For this reason, it's important to examine the more radical version of the epistemic challenge, as represented by authors like Jason Brennan (2016; see also Caplan 2007).
Though Brennan's book is entitled Against Democracy, he realizes that "ideal epistocracy isn't a live option" (Brennan 2016, p. 207), and so his argument takes the form of an epistocratic corrective to democracy, rather than a full-on defense of minority rule. Specifically, Brennan defends a version of epistocracy based on weighted voting, so-called "government by stimulated oracle" (Ibid, p. 220).
Nonetheless, Brennan's argument remains quite radical, from a contemporary democratic sensibility. Brennan seems content to support forms of "rule by the rich" so long as they secure epistemic goods, arguing that the connection between affluence and influence is actually "reason to celebrate": "democracy works better than it otherwise would, because it doesn't exactly work" (Ibid, p. 198). We think this is a rather misguided response to the evidence presented above (see also Arlen and Rossi 2018). For those studies provide little compelling evidence that the wealthy have disproportionate influence because of epistemic virtue per se. That's not to deny that some among them may possess such virtue, but simply to insist that the main mechanism of their influence is not primarily epistemic.
One way to grasp this point is by considering the classical Aristotelian distinction between "aristocracy" as rule by the few who are virtuous and "oligarchy" as the rule of the rich who lack virtue and exert power solely on account of their wealth. Putting aside his numerous departures from Aristotle, Brennan's argument for "government by stimulated oracle" might be interpreted as a modern-day reformulation of the ideal of rule by competent aristocratic guardians. Any argument for aristocratic guardianship pivots on clearly distinguishing aristocrats from oligarchs so to ensure that those who benefit from epistocratic correctives (like plural voting) actually command authority on account of their virtue, not simply on account of superior wealth (Arlen 2019). But in practice, epistocratic correctives will likely favour socioeconomic elites (including the super-rich), regardless of how these correctives are implemented. Indeed, it is precisely this concern about pernicious"demographic" biases in epistocratic mechanisms that causes prominent democratic theorists, like David Estlund, to recommend against them (Estlund 2008).
Brennan worries that elections allow ignorant citizens to deploy their votes as unwieldy "bosses," imposing bad decisions on others. In a sense, then, Brennan is idealizing the same theory of democracy he criticizes; one based on people directing policy outcomes by imposing preferences electorally. If this is not, in fact, what voters do, as Achen and Bartels suggest, then Brennan's concerns have a lot less steam. If even informed voters tend to vote along group lines, on the basis of social identity, then even a perfectly informed electorate would not really meet Brennan's epistocratic ideal of competent rule. Thus, Brennan's epistocratic corrective to democracy only reproduces some of the dilemmas of group voting discussed above.
In short: Brennan is on firm ground in acknowledging the existence of an epistemic challenge to democratic ideals. But we worry that his proposed solutions won't have the epistemic payoff he anticipates. Moreover, even if fully realized, Brennan's model only addresses the epistemic challenge while neglecting or even amplifying the oligarchy challenge. There's an important methodological point here: democratic theorists cannot succumb to solutions that respond to one dimension of the two-pronged challenge to democracy while neglecting the other. Achen and Bartels and studies like theirs should not be used as the empirical pretext for a normative turn towards epistemic elitism. Indeed, their criticisms of the folk theory of democracy are motivated by a sense that it actively "props up elite rule;" "it is unrepresentative elites that most profit from the convenient justifications it provides for their activities" (Achen and Bartels 2016, p. 327). Any response to their findings must come on more solidly democratic grounds.
Put differently: the salient question is less whether citizens are good or bad rulers according to some epistemological criteria, but rather whether citizens can rule themselves in such a way that they can check elite influence. And their inability to do so indicates less an intrinsic failure of cognitive ability, than a weakness in the prevailing institutions to which normative democratic theory can propose remedies. If one problem with epistemic failures is that they risk enabling a corrupted rule by the few, the answer cannot be endorsing solutions that only cement rule by the few.

Against plebiscitarianism
The epistemic challenge, however, cannot be fully dispatched simply by rejecting epistocracy. We do acknowledge that certain models of radical democracy come out poorly. Specifically, crude plebiscitarian models that rely on aggregating popular voice through issue-based referenda (often bypassing intermediary institutions) should be considered suspect in light of recent empirical evidence. Whereas plebiscitarian theories were developed by earlier figures like Weber, Schmitt, and Schumpeter, few mainstream democratic theorists today endorse plebiscitarianism as a normative ideal. But as Nadia Urbinati argues, many contemporary democracies are experiencing a revival of plebisciatarian impulses in the face of mass-media, the declining prestige of traditional parties, and the rise of charismatic executive authority which has led to a "revolt" against intermediary institutions (Urbinati 2015;. This revival is linked to the rise of populist movements of which Urbinati is also quite critical, but she makes a point to distinguish the two: "Unlike populism, which embodies the ideal of mobilization, plebiscitary democracy narrows the role of active citizenship to stress instead people's reactive answer to the promises, deeds, decisions, and appearances of the leader" (Urbinati 2014). At the core, Urbinati views plebiscitarianism as both a challenge to democracy's "procedural form," (Ibid, p. 182) and as a "radical rejection of individual judgment in politics" (Ibid,p. 191). Under the imperfect epistemic conditions described by Achen and Bartels, the dangers of plebiscitarianism are aggravated further. One way to grasp this point is by considering referenda movements, like Brexit, as an extension of the plebiscitarian impulse described by Urbinati. Most contemporary referenda are issue-based, and Achen and Bartels suggest that people don't generally vote on issues: "…issue congruence between parties and their voters, insofar as it exists, is largely a byproduct of these other [social identity-based] connections, most of them lacking policy content" (Achen and Bartels 2016, p. 301). Of course, some voters are still passionate about specific issues, and eager to express those preferences. A referendum in favour of stricter gun control measures will likely attract many voters who do care passionately about the issue, on both sides. But again, if we take Achen and Bartels seriously, then this policy preference remains, in important ways, conditioned by a larger group identity.
In these conditions, issue-based referenda, on their own terms, are a poor decision-making tool in actually existing democracies. But that is too crude. By issue-based referenda we mean referenda that address a policy question that is not supposed to have to do with group identity, since that is what ultimately moves voters. The 2016 Brexit referendum provides an informative illustration of our point. Analyses of the vote have shown quite convincingly that policy issues specifically to do with EU membership were probably a factor, but hardly the determinant one (Clarke, Goodwin and Whiteley 2017). So, insofar as the referendum was ostensibly about those questions, it was a poor way of addressing them. Now, one may hold that an identity-based temperature check on supranational integration was welcome, in which case our position is that the referendum should have been explicitly framed in those terms. In fact, and without taking a view on the Brexit referendum specifically, we are tempted to hypothesize that an explicit identitarian framing might have altered the outcomes of some polls whose framing and subsequent campaigns have tended to divert attention away from a public conversation centred on what actually motivates voters.
Epistemic critics have good reason, then, to be concerned about the plebiscitarian impulses driving some contemporary populist movements. But these concerns are not reason to fall into the epistocratic camp criticized earlier. We refuse to disavow all populist energies, many of which have considerable antioligarchic potential (see Vergara 2019; Mulvad and Stahl 2019). And as we argue in the next section, issue-based referenda can have an important function when they align more explicitly with the socioeconomic identity of citizens. The challenge for a realist democratic theory is to respond to oligarchy not with crude plebiscitarianism, but with a form of popular politics suitable to the empirical challenges that have been the focus of this paper.
We now explore what such a response might look like.

Plebeian tribunes vs oligarchs
In light of the oligarchic challenge described earlier, democratic institutional design has a twofold task: it must (1) try to compensate for some of the epistemic shortcomings plaguing contemporary publics while also (2) offering up resources to help counteract the threats associated with oligarchic power. It must recognize that these phenomena are intricately related, and that epistemic correctives, if poorly constructed, can exacerbate oligarchic power. In what follows, we endorse one strategy that should command more attention from mainstream democratic theory: the neo-Roman tribunate model which has been revived, most recently, by democratic theorists like John McCormick.
The Roman tribunes of the plebs composed a classspecific office, occupied by plebs, whose main function was mediating between the plebeian and patrician orders for the purpose of protecting the former's interests. Contemporary representative democracies are "class-neutral," that is, political offices are assigned irrespective of socioeconomic distinctions. While class-neutrality undoubtedly offers advantages over the formal property-based exclusions which characterized previous aristocratic and monarchical regimes, it holds notable disadvantages that have become more glaring with rising inequality. Class-neutrality deprives ordinary citizens of formal spaces for enacting their class-based disadvantage, and for seeking institutional redress within a constitutional democracy.
Within the framework outlined here, class-neutrality exacerbates both the "epistemic" and the "oligarchic" challenge to democracy. It exacerbates the former by fostering a democratic public less capable of articulating class-based grievances and filtering those grievances through the policy process; it exacerbates the latter by rendering socioeconomic elites more insulated from institutional scrutiny and thus more able to exert their influence unchecked. The oligarchic challenge is our main focus in this section. Our aim is to develop the tribunate model in a direction that puts it directly in service to the task of reducing oligarchic influence on the policy process.
Specifically, the tribunate model offers one roadmap for moving beyond class-neutrality, without jeopardizing hard-fought modes of formal equality. How might this work? Inspired by Machiavelli's analysis of the Roman tribunes, McCormick proposes a People's Tribunate, an assembly composed of 51 nonwealthy adult citizens selected by lottery for one-year terms, with a variety of oversight powers, such as the ability to veto one national legislative proposal or initiate impeachment proceedings against one lawmaker (McCormick 2011). On McCormick's rendering, the tribunate has carefully circumscribed powers, compatible with existing American constitutional checks and balances. But the tribunate remains class-specific, in the sense of excluding citizens above a certain wealth threshold from participation in the body.
Our idea, then, is to extend McCormick's proposal so as to even more directly tackle the problem of unequal influence highlighted by the empirical evidence discussed earlier. The idea is to task the tribunate specifically with reviewing a range of political activities that, while strictly speaking legal, can threaten democratic equality.
Specifically, we might identify five different categories of oligarchic influence over public policy, five domains in which personal access to massive concentrated wealth can be deployed for discretionary public influence3, and our claim is that each domain merits individuated scrutiny from plebeian institutions: (1) occupying formal elected office or serving as an appointed officeholder; (2) lobbying candidates and officeholders directly through interest-group activity; (3) lobbying candidates indirectly through (in the American context) groups like Super-PACs; (4) influencing public policy through ownership of mass media; (5) influencing public policy through elite philanthropic activity. These categories apply best to the American case, but also have salience in other advanced democracies, given the empirical trends discussed earlier.
Each of these five domains, while interconnected, manifests oligarchic power in slightly different ways. Each involves different institutional processes. Each requires different responses. Thus, whereas McCormick advocates a singular, overarching People's Tribunate, we think it's more fruitful to think about a wider tribunate system; an interconnected set of plebeian assemblies, each operating within a designated policy area.
To illustrate: McCormick's proposed People's Tribunate has three core formal powers: (1) it can "veto one piece of congressional legislation, one executive order, and one Supreme Court decision" (McCormick 2011, p. 184); (2) it can initiate impeachment proceedings against one federal official from each of the three branches of government; (3) it can call one national referendum (Ibid). These powers are applicable within one-year non-renewable terms. But we worry that by limiting these powers to a single People's Tribunate, the assembly will prove less capable of simultaneously regulating different political pathologies at once. Thus, the tribunate system described below proves more expansive. One plebeian assembly would be assigned to scrutinize formal officeholders, and thus would retain the power to initiate impeachment proceedings, as described by McCormick. However, in contrast to McCormick, the veto and referenda proposing power would be dispersed among four additional plebeian assemblies each nestled within a specific policy domain, and limited to that domain. Each of these assemblies are composed in the manner McCormick describes (51 non-wealthy citizens selected by lot), but we think two-year terms are preferable to McCormick's proposed one-year terms because they allow for a longer time horizon to develop expertise within these domains.5 The system enforces a division of labor around specific accountability mechanisms, since only one assembly has the power to initiate impeachment proceedings, while the other four are empowered to initiate referenda and exercise a limited veto power. To illustrate, consider the case of elite philanthropy. American political scientists have shown that the charitable tax deduction is both regressive (in the sense of benefitting wealthier taxpayers more) and contributes to wider forms of plutocratic philanthropy in which the wealthy use their giving to support public policy agendas (Saunders-Hastings 2017;Reich 2018). However, philanthropy can support important pluralist goods such as innovation and discovery (Reich 2018), and represents an important ethical dimension of wealth. Either way, the issue is ripe for greater public scrutiny by the ordinary citizens impacted by large donations. Suppose that the U.S. Congress proposed a new law dramatically expanding the scope of the charitable tax deduction to cover activities that we might commonly think of as outright political activism (for example, using philanthropic spending to fund climate change scepticism). The plebeian assembly (A) devoted to studying private philanthropy could spend its term studying and debating this proposed legislation (including calling witnesses and holding public hearings). On McCormick's model, it could then, on majority vote, veto the legislation, and Congress would have to wait an additional year before proposing it again (McCormick 2011, p. 184). Simultaneous to this activity, a second plebeian assembly (B) could call a national referendum for the purpose of overturning the Citizens United court decision which has led to the dramatic proliferation of soft-money in the American context. Finally, a third plebeian assembly (C) could simultaneously initiate impeachment proceedings against a lawmaker suspected of taking bribes from a wealthy donor. Each assembly operates on separate tracks, but the activities of one assembly may influence the activities of the others in dynamic ways.
We recognize that giving ordinary citizens the ability to initiate impeachment proceedings or exercise a veto power over national legislation is a radical step. We follow McCormick's proposal in emphasizing that such power should only be exercised sparingly, and should be subject to various institutional checks and balances such as scrutiny by other branches of government. Our goal here is not to address all potential objections to this model, nor work out all the particulars. Rather, we want to think within the terms of McCormick's proposal, to underscore some of its advantages, and show how it might be developed as part of a broader tribunate system.
Epistocrats like Brennan would certainly reject this proposal, and realists like Achen and Bartels would likely consider it naive. But we believe such resistance overlooks the substantial promote issue expertise. But we still advocate the longer two-year term to maximize expertise. epistemic potential of plebeian assemblies. Well-constructed plebeian assemblies can amplify the socioeconomic identity of a specific group-the non-wealthy-by dramatizing socioeconomic faultiness that cut across the polity. In addition to its concrete powers, by excluding wealthy citizens from participation, plebeian assemblies would reinforce the point that other electoral institutions are oligarchic in composition, despite their putatively egalitarian formal properties. If Achen and Bartels are correct that voters often identify along group-lines, then this act of amplifying group identity, and directing it towards specific policy aims, might offer important epistemic returns in addition to its direct antioligarchic function.
Again, rather than working out all the kinks in the model, our goal here has simply been to highlight some of its main advantages. We do not suggest that democratic tribunes are sufficient to curb oligarchic influence or correct for all the challenges confronting contemporary democracies. As McCormick notes, ordinary citizens have a more "elongated learning curve," often unable to "foresee what is beneficial or deleterious for common utility as quickly as grandi foresee what is in their own interest" (McCormick 2011, p. 90). However, the experience of other premodern governments, like democratic Athens, does leave room for optimism. As classicists have shown, Athenian institutions like the Assembly and courts were successful precisely because they allowed ordinary non-wealthy jurors and assemblymen to feed off their socioeconomic group identity in discursive interactions with elites (Ober 1989).
Our argument thus contributes to an ongoing discussion, within democratic and republican theory, about the "mixedregime" metaphor and its continued relevance (Manin 1997;Pettit 1997;Rosanvallon 2008). The classical mixed-regime was predicated on formal distinctions between different constitutional forms (i.e. democracy vs. aristocracy) or different classes and orders (i.e. patricians vs. plebs). We believe the "mixed-regime" metaphor still holds force in contexts of formal legal equality; for the plebeian system outlined here can function as a counterweight to more "patrician" institutions like the US Senate, securing greater constitutional balance. Reformers might also, however, focus on mixing together different democratic institutions, some plebeian and some non-plebeian, approaching what Arlen (2019) calls a "new Mixed Regime." Here the focus is balancing within different conceptions of democratic authority. For example, class-based plebeian assemblies can operate in coordination with other deliberative mechanisms that are fully inclusive and non-class based. Likewise, plebeian institutions can operate alongside electoral institutions, and in conjunction with core liberal legal norms, such as formal equality and protection of property rights.
Crucially, unlike Marxist approaches which attempt to overcome class distinctions, the plebeian approach is focused on managing these distinctions. But unlike some liberals, plebeians do think class is a central feature of politics. The plebeian approach is thus compatible with core liberal commitments to private wealth accumulation, with the understanding, however, that the unequal political influence associated with private wealth must be contained more aggressively than many liberals allow. Reformers must anticipate and address the normative and institutional trade-offs that occur between plebeian and liberal conceptions of politics: it is possible to use plebeian institutions to temper liberal democracy rather than supplant it (Green 2016). Of course, more radical, non-liberal plebeian models are possible too. But they are not a prerequisite for getting plebeian reforms off the ground.
There's strong justification, then, for institutional innovation that moves beyond the relatively tame measures advocated by Achen and Bartels. The exact contours of such innovation are still to be worked out. But it's vital to emphasize that realism about the empirical threats to democracy should be met, not with acquiescence, but with a hard-edged attempt to innovate on the status quo.

A methodological conclusion
In the preceding sections we have seen that even the most pessimistic empirical results about actually existing democracies need not spell doom for normative democratic theory. In fact, we hope to have at least gestured towards a way in which political philosophy can offer evaluation and prescriptions compatible with the evidence, while avoiding some of the most brazen forms of idealism and moralism that some of its critics-including the empirical ones-accuse it of. By way of conclusion, we would like to briefly point out how, in addition to our three main normative contributions-the rejection of epistocracy, the critique of plebiscitarianism, and our version of the plebeian tribunatecontending with the grim picture presented by Achen, Bartels, Gilens, Page, Winters, and other scholars yields some methodological lessons for political philosophers. More specifically, as we noted at the outset, we take the preceding discussion to be an innovative contribution to the realist program in normative political theory (Rossi & Sleat 2014)-innovative insofar as the empirics drive the search for normative arguments and yet are not employed as feasibility constraints.
Indeed, the first thing to note is how the centrality of the empirics allows us to make sure that none of the normative conclusions we have drawn depend on pre-political moral commitments-for this is the relevant sense of "realism" here and in the wider realist program, and not the misleading one that equates realism with the pursuit of feasible prescriptions.6 While we have sought to find strategies to realize the egalitarian potential of democracy, that commitment can be seen as practicedependent, i.e. a normativity that "emerges within specific, institutionally mediated political and social contexts." (Sangiovanni 2008, p. 164). 7 We do not rely upon a prior commitment to equal respect or any such abstract moral notion. Rather, we interpret the practice of democracy as a distinctive political form characterised by an aspiration to an egalitarian distribution of power. This is not a pre-political moral commitment to some abstract ideal of equality, but an interpretation of what makes democracy a distinctive political practice. In fact we use no claims about moral intuitions nor do we rely on any account of moral powers or rights to ground our proposals. As in most other realist contributions, our normative stances are driven by the idea that there is a normativity internal to politics itself.8 However, our approach differs from other forms of broadly realist egalitarianism (e.g. Jubb 2015) because we did not develop a non-moralized defence of forms of egalitarianism typically defended in moralistic terms, but rather developed our egalitarian position as a direct response to the challenges we identified through the lens of the empirical results.
Relatedly, we hope to have also shown how realist theorising differs from mainstream, moralistic non-ideal theory (see Rossi 2019). Non-ideal theory is, broadly speaking, driven by feasibility constraints, whereas our arguments aren't. That is to say, we did not use the empirics to set limits on our evaluative and prescriptive goals, but rather used them to individuate those goals-which does not condemn the realist outlook to acquiescence towards the status quo but, as our proposal for a plebeian tribunate system shows, opens quite radical vistas.