WHY EPISTOCRACY?

That there is a certain intuitive appeal to epistocracy has been obvious since antiquity. Just as we would want to travel in a ship piloted by the most competent captain, so too it would seem—at least on first glance—that we should want to live in a political community run by individuals who rank highly with regard to the relevant skills. Nevertheless, the standard view in contemporary democratic theory is that epistocracy is unjustifiable. Typically, the reason for this is simply that insofar as we are equal members in a political community, it is unclear why some of us should have more power than others.

By itself, though, this line of thought cannot carry us very far. After all, in other domains we routinely distribute privileges according to assessments of competence. Nobody objects to administering exams to people who wish to fly commercial airliners and, based on the results, allowing only some citizens to do so. The reason for administering such exams is that when one pilots such a plane, one has power over the lives of other citizens. This explains why we put those who wish to fly commercial airliners, but not those who wish to go water skiing, through extensive competence testing. Although one might injure oneself water skiing, the dangers posed to others are marginal. But, if this is right, it would seem natural to suppose that we ought to distribute political power in a way that empowers the skilled—after all, political decisions importantly affect the lives of others.^{Footnote 4}

Furthermore, making such decisions intelligently is very demanding. In a modern representative government, legislators and political executives represent enormous numbers of citizens and exercise authority on a very diverse range of policy issues. In such settings, being an effective leader requires, inter alia, extensive knowledge across a wide range of policy areas. This is necessary to be in a position to propose plausible policy solutions to ongoing public problems, and to assess competently the arguments, advice, and proposals of experts, stakeholders, and fellow legislators. Effective officeholders also need the knowledge and ability to organize and bring off complex inter-temporal and inter-issue agreements with other legislators. This requires recognizing the needs and constraints created by the agency relationships in representation of one’s own constituency and of the constituencies of other representatives, the diversity of preferences and opinions across electoral constituencies, as well as the unavoidable uncertainty in assessing the merits of policy alternatives. We cannot assume that all citizens are readily equipped, nor interested in bearing the costs associated with becoming equipped, to make specific policy judgments that optimally advance the public good. Given the importance and complexity of public policy decisions, it is desirable, all else equal, to delegate them to particularly competent citizens.

One might think that this line of reasoning is mistaken—that given the epistemic demands of good governance, one should instead focus on limiting the size and scope of government [for arguments along these lines, see Caplan (Reference Caplan2011) and Somin (Reference Somin2016)]. However, even governments that are considerably more limited than those in the present-day United States or European liberal democracies would be epistemically demanding (albeit, of course, less so than more ambitious alternatives). Even relatively noncontroversial functions of government (such as enforcing criminal law or pursuing national defense) admit of a wide variety of approaches and entail different tradeoffs, which can be pursued in better or worse ways. Thus, the intuitive appeal of epistocracy is present even under quite limited forms of government, though of course its relative value grows with the size of government. This does not imply that prescription of small(er) government is wrong—merely that even with a smaller government, we should still be interested in whether there are defensible forms of epistocratic rule. In this article, we therefore set aside questions about the appropriate size of government and focus on whether well-designed representative institutions can generate epistocratic rule.^{Footnote 5}

The appeal of epistocracy notwithstanding, it faces—as we mentioned above—two important objections. According to the disagreement objection, the selection of any person or group to have more power than others would be controversial because there is reasonable disagreement about which citizens are distinctive in ways that make them well suited to exercise political power. In other words: although reasonable people can agree on which people are qualified to be pilots, there is widespread and reasonable disagreement about which citizens are well qualified to exercise political power. This objection is widely relied upon to rule out epistocratic arrangements. For instance, Thomas Christiano insists that because:

Advocates of the disagreement objection are typically willing to concede, then, that some citizens are more competent to responsibly exercise political power than others, yet any particular criteria used to identify the most competent will be the subject of reasonable disagreement.

Even if the disagreement objection were somehow overcome, however, there is yet another important concern about epistocracy. The demographic concern is that even if we could select a subset of particularly competent citizens (e.g., the educated) and trust them to make sincere judgments about the policies that would best promote the public good, there would nevertheless be a reasonable concern that they would be “no better able to rule wisely than others” because they “may disproportionately have epistemically damaging features that countervail the admitted epistemic benefits of education” (Estlund Reference Estlund2008, 215).^{Footnote 6} For instance, although we might think that the highly educated would be in a position to rule better than their less educated counterparts, in our society advanced education is

Thus, according to the demographic objection, even if we were able to select particularly competent members of the polity to entrust with political power, we may rightfully worry that they would share other demographic characteristics that would bias their political decisions.^{Footnote 7}

Thus, epistocratic arrangements are widely rejected because (1) there will be reasonable disagreement about which citizens count as epistemically superior and, even if this were not the case, (2) an epistemically superior subset of citizens may reasonably be objected to on the grounds that they may be biased (even if unintentionally) in ways that undermine their ability to generate superior political outcomes. The alleged upshot is that we ought to prefer democracy to epistocracy.

AN ASIDE ON METHOD

The two objections to epistocracy described above have particular presumptive power because they are meant to identify problems that such forms of governance would face even under quite idealized circumstances. Most importantly, both objections apply even if one assumes that representatives are motivated by the desire to maximize social welfare—that is, even setting aside the obvious litany of concerns associated with rent-seeking (Estlund Reference Estlund2008, 206). The point is that even given assumptions that are very generous to the epistocrat, such proposals face overwhelming objections. The disagreement and demographic objections are therefore meant to go to the heart of the notion of epistocracy.

But whereas the strongest argument against epistocracy hinges on showing that these objections apply even when political officials are ideally motivated, the most convincing response would show that such objections miss the mark even if we assume that political officials are not ideally motivated. With that in mind, we proceed with broadly realistic assumptions about the behavior of political actors: agents may not be perfectly motivated to pursue social welfare (indeed, they will sometimes use public power to gain personal benefits), they are not omniscient, they are responsive to incentives, and officials have a strong desire to continue to hold office.

In the next two sections, we show that at the core of representative democracy lie mechanisms that when properly institutionalized, can generate epistocratic rule in a way that circumvents the standard disagreement and demographic objections to epistocracy, even given these realistic assumptions. We proceed by way of describing two counterfactual models—stylizations of the world that abstract away from certain features to focus our attention on others that are of particular import to the argument. The first model, which we call the treatment model, captures the idea of effects that a given change in environment (treatment) may be expected to bring about relative to a counterfactual without changing the identity of relevant agents. This model identifies mechanisms that will lead officeholders in representative systems to make different choices and develop their capacities differently (capturing both responsiveness to incentives and the “formative” effect that that may bring about) than they would as citizens in a direct democracy. The second model, which we call the selection model, explains how a system of iterative elections can help bring forth a set of officeholders who are particularly well suited to live up to the demands of office-holding.

Before describing these models, and the way in which they evade the standard objections to epistocratic rule, it is worth clarifying their function. Perhaps most importantly, these models are not meant as positive descriptions of actually existing representative democracies—at least, not in any simple way. The epistocratic mechanisms that they highlight emerge out of the central features of representative democracy, but their bite is conditional on the broader institutional design of the representative system.

For this reason, these models are not meant to characterize any particular representative democracy, and it would be a mistake to conclude from the fact that existing representatives are not epistocrats, let alone defensible epistocrats, that the position laid out in the article must be incorrect. If a particular representative democracy fails (as many of our own representative regimes, to a considerable degree, have) to adopt institutions that allow it to capture the benefits identified in the treatment and selection models described here, its elected representatives will likely lack special competence. And, furthermore, because these models are not necessarily descriptive of actual representative regimes, they also cannot, in any straightforward sense, serve to justify such regimes.

Instead, the models we describe are aspirational, but in a very specific way. They explain, given realistic assumptions about the behavior and capacities of citizens and officials, how the central features of representative democracy can, if well designed, generate a distinctively defensible form of epistocratic rule. By identifying feasible mechanisms with that property, such models can play an important role in framing public discourse about the justification of such regimes, as well as about institutional reforms that could improve democratic governance.

THE TREATMENT MODEL

We begin, in this section, with an explanation of how treatment effects associated with representation may facilitate epistocratic rule in ways that circumvent the disagreement and demographic concerns. We will focus on two treatment-based reasons, the pivotality and accountability effects, which—when jointly operative—provide reason to think that elected representatives will exercise their political power in a more considered and socially beneficial way than would ordinary citizens in a majoritarian direct democracy.

Consider, first, the pivotality effect: because representative systems entrust political power to a relatively small group of elected officials, each one is far more likely to be pivotal in policy choice than an ordinary citizen in a direct democracy. Because the votes of representatives are far more likely to affect policy outcomes, they have significantly stronger instrumental reason to exercise their power with due care (including investing more time and effort into acquiring knowledge and expertise about relevant policy considerations) than citizens in a direct democracy.^{Footnote 8}

The second reason to think that a well-designed representative democracy will lead representatives to exercise political power more responsibly than would citizens in a direct democracy is the accountability effect. By linking continued office-holding to public approval, representative systems generate incentives for officeholders to exercise power with due consideration for the likely effect of policies on the welfare of ordinary citizens. Insofar as it is desirable to retain office, representative democracy thereby gives officeholders a reason to exercise their power with a level of care and consideration that does not exist for ordinary citizens in a direct democracy who are not stripped of valuable privileges for failing to use their power in a way that citizens think advances their interests.

These two effects, pivotality and accountability, are logically distinct, and both are necessary for the treatment model of epistocracy. In the absence of the accountability effect, there is nothing to prevent those citizens who lack sufficient public spiritedness from making choices that would benefit themselves or those close to them at the expense of the society as a whole. Indeed, their greater pivotality might, on its own, increase the temptation of those choices. The accountability effect acts as a(n imperfect) backstop to that temptation.^{Footnote 9}

To see the contribution of the pivotality effect given the accountability effect, note that the objects of accountability are, most plausibly, not votes or policies, but policy outcomes—complicated products of representatives’ choices and stochastic, unobservable, or both states of the world, including preferences and actions of other political actors, hidden institutional and transactional incentives, and sheer luck. The citizens face a classic team production problem vis-à-vis their elected officials: if other individuals can influence the same outcomes as you, your action becomes, all else equal, less consequential for determining those outcomes, and so harder to motivate externally (e.g., by citizens you may be representing). This problem is most acute when your pivotality is lowest (e.g., in a regime approaching direct democracy) and decreases as the pivotality goes up. In the language of comparative politics (Powell Reference Powell2000; Tavits Reference Tavits2007), with extremely low pivotality, the clarity of responsibility for outcomes is extremely low, and the incentives that the accountability effect would need to rely on to make a difference may be exceptionally, unreasonably, high. As the clarity of responsibility increases with the increase in the officeholders’ pivotality, the consequential accountability effect becomes increasingly feasible. The upshot is that the accountability effect is most promising in the presence of the pivotality effect.

Whereas representative democracy relies on ordinary citizens to hold representatives accountable, the epistemic demands of doing so are less significant than the demands associated with office-holding. Whereas citizens in a representative democracy only need to evaluate policy outcomes, the officials need to evaluate a range of possible policy options—usually, in the context of imperfect information, without the hindsight of knowing how things will work out—and the latter is a much harder task.^{Footnote 10} It is easier to know, for instance, that the economy is performing poorly or that there is growing economic inequality than it is to know what steps would rectify these problems (never mind which among those steps would bring about the best trade-offs). Similarly, it is easier to know that there have been no significant terrorist attacks than to know how to prevent them with minimal costs to civil liberties.

In addition to the epistemically different burdens associated with such choices, there is a further consideration that suggests that the demands on citizens may be fundamentally lower. Insofar as information about policy outcomes is readily available, officials could be motivated to pursue good outcomes even in circumstances in which citizens are paying little attention (see more on this in the article’s penultimate section). In effect, in such circumstances, to induce good performance by incumbents, the credible threat of learning more about incumbents’ choices is often sufficient to induce better choices even when that threat is largely not carried out.

To be sure, elections are a coarse instrument of influence, and the multidimensionality of government performance confounds perfect representation. Electoral choice bridges the expertise of officeholders and the experienced welfare of voters arguably well enough to ensure a certain floor level of welfare. But just how well it can harness that expertise, and so, ultimately, how high the welfare of voters in representative democracies is, depends on underlying political institutions and the social and political practices associated with them. Thus—for example—clarity of responsibility is clearly affected by a host of political institutions, including freedom of the press, details of electoral systems, and responsiveness of the bureaucracy to democratic outcomes. An important example of political practice that also critically affects accountability is public deliberation. In the face of multidimensional government performance, well-structured public discourse can create salience for some dimensions over others, thus allowing citizens a greater degree of effective control over policy choices that matter to them.

When the relevant officeholders are public welfare-motivated, the institutional details may have less bearing on the choices that they make (though, such details may still matter for how voters evaluate those choices). But when public welfare motivation is weak, effective institutional design becomes critical to ensuring that the bite of the accountability and pivotality effects is sufficiently strong. In a well-institutionalized system of representation, elected representatives may be expected to exercise their political power in a manner that generates epistocratic rule.^{Footnote 11}

Importantly, insofar as representative democracy generates epistocratic rule through these treatment effects, it is not—in contrast to commonly discussed epistocratic arrangements—vulnerable to the disagreement or demographic objections. It is not vulnerable to the disagreement objection because the mechanisms at stake do not require that citizens agree that there is a particular subset of citizens who are most capable. Instead, the mechanisms by which representative systems induce competence could work even if every citizen in the population were, ex ante, equally competent. Because the argument claims that any reasonably able member of the population would be expected to become more competent if exposed to the incentives associated with office-holding, it is irrelevant that there “is no public way of demonstrating the superiority of competence” that members of the epistocracy may claim to possess. The position does not depend on the claim that there is ex ante some subset of particularly qualified citizens whom we must agree upon.^{Footnote 12}

The epistocracy generated by representative democracy is also not vulnerable to the demographic objection, at least in the way that it is normally understood. Recall that this objection holds that the characteristics that epistocrats are selected for (in Estlund’s example, education) may be correlated with other characteristics (such as race) that may bias the decision-making of that group, despite their generally superior competence. The epistocracy that we described above is not vulnerable to this version of the objection because the relevant treatment effects work on any reasonably able citizen placed into office and do not require selection on special ex ante desirable characteristics. The idea here is that even if citizens believed that there were no important differences in competence between their compatriots, or agreed to an equality-motivated constraint on public reason implying that such differences ordinarily ought to be set aside, an epistocratic justification for representative democracy could still be given.^{Footnote 13}

However, a related, if somewhat second-order, concern might be thought to pose problems for the treatment model. It could be that—in virtue of being officeholders—representatives gain certain shared interests. For instance, insofar as challengers need to spend more than incumbents to run competitive races (to overcome the preexisting advantages of incumbents), incumbents may have a shared interest in limiting campaign spending and could bias legislation addressing the issue.

The iterative nature of the electoral process gives representative democracy a built-in mechanism to allay such concerns. If officeholders are systematically and appreciably biased in favor of policies that advance their shared interests at the expense of the citizens, this should undermine the very electoral advantage that incumbents were supposed to have, and so provide them with an incentive to deviate to less biased behavior. The iterative nature of elections, thus, allows citizens to filter officeholders in a way that can mitigate this bias.^{Footnote 14} This is very much in contrast to the types of epistocracy typically discussed in the literature, such as an epistocracy of the educated, which are ordinarily conceptualized in ways that lack such dynamic filtering mechanisms.

It is worth stressing that this filtering-out of bias need not be perfect, because ultimately the question is not whether epistocratic arrangements eliminate bias altogether, but whether the institutional regime has effective mechanisms for detecting and eliminating bias. The electoral incentives in well-designed representative democracy, which are instrumental in generating epistocratic governance, are precisely such mechanisms. Of course, one may ask whether those mechanisms are more or less promising than the corresponding mechanisms of alternative regimes—and in particular, of direct democracy. But we see no reason to believe either that direct-democratic governance would be less susceptible to what would, in effect, be an equivalent kind of “self-dealing bias” (after all, even direct democratic majorities may be self-dealing) or that it would be better equipped to mitigate it.

THE SELECTION MODEL

The second model of epistocratic governance under representative institutions focuses on the possibility of variation in candidates’ relevant competence or aptitude and on the ability of citizens to elect officeholders who possess high levels of those. A particularly well-known version of the underlying argument is in Federalist No. 10, in which James Madison argues that in a well-designed system, the effect of delegating the tasks of government “to a small number of citizens elected by the rest” is to:

The idea is that through elections, the citizenry can select particularly able individuals to take up important roles in government and pursue refined versions of the public’s commitments. When these individuals are put into office by means of elections, they can help the community pursue the basic political commitments of ordinary citizens more effectively than would a system that “convened” ordinary citizens for the purpose.^{Footnote 15}

Of course, this argument notwithstanding, in many actually existing representative regimes, it is doubtful whether elections effectively select highly qualified citizens for public office. There is ample reason to worry that in large political communities elections select the preternaturally ambitious, hubristic, shameless and power-hungry—at least, so long as they are also suitably famous, wealthy, and well-connected. Indeed, although it is less often quoted, Madison makes this point himself as the above passage continues:

Madison’s view is that this is the sign of a poorly designed or corrupt electoral process, and he prescribes large electoral districts as a way of combatting such outcomes.

Contemporary political science scholarship on elections has added a considerable list of further factors that influence the quality of electoral selection, including the campaign finance regime (Ashworth Reference Ashworth2006; Gordon et al. Reference Gordon, Huber and Landa2007), the presence or absence of term limits (Alt, Bueno de Mesquita, and Rose Reference Alt, Bueno de Mesquita and Rose2011), details of electoral systems (Cox Reference Cox1997; Myerson Reference Myerson1999), the strength of institutional determinants of incumbency advantage (Gordon and Landa Reference Gordon and Landa2009), and the extent to which policy-making authority is divided or unified in relation to the complexity of the underlying policy areas (Ashworth and Bueno de Mesquita Reference Ashworth and Bueno de Mesquita2017; Landa and Le Bihan Reference Landa and Le Bihan2018). Cumulatively, this literature can be understood as an attempt to specify the conditions under which the selection effect is likely to be effective—that is, the conditions under which desirable characteristics of incumbents (particular types of competence, integrity, judgment, temperament, etc.) become sufficiently transparent to the voters to facilitate the selection of better incumbents more reliably.

This, then, is the core underlying mechanism: voters take their cues from the best information available about the leaders’ performance and about the record of the potential challengers; both the leaders and those potential challengers decide whether the comparison of their respective records would implicate them as sufficiently competent to make it worthwhile to run for election; and the voters then make their decisions on the basis of those records and further competence-related information revealed in the course of campaigns. The key claim is that when properly institutionalized, the various mechanisms for improving the quality of selection create a quality-tracking property for electoral representation. It is not that selection of leaders in successive elections always improves the quality of the officeholders, with the quality of each successively selected leader at least as high as that of the previous. Rather, the idea is that the mistakes—low-quality incumbents—will tend to be less frequent and their magnitudes less significant than the correct (competence-improving) selections (Ashworth and Bueno de Mesquita Reference Ashworth and Bueno de Mesquita2008; Gordon et al. Reference Gordon, Huber and Landa2007; Zaller Reference Zaller1992). To summarize, then: the selection model should be understood not simply as implying that leaders are selected on the basis of considerations of merit, but also as entailing—in the institutions of electoral representation—mechanisms for iteratively improving the quality of selection.

To be sure, much here depends on proper institutional design. Elections cannot guarantee the selection of skilled individuals who will aim at the public good, but much as they can be designed to be factional tools—for instance, to promote class interests if property qualifications are put in place or if successful candidates need to raise money from a small group of ideologically homogenous donors—they can also be designed to minimize such effects. The selection model posits that a well-designed electoral system increases the chance of good policy outcomes (relative to other possible political regimes) by selecting individuals who are well suited to exercise political power.

We next turn to considering how vulnerable representative democracy, through the eyes of this model, is to the two objections to epistocracy that we laid out above. Although the case on behalf of the selection model is different and more demanding than that on behalf of the treatment model, there are good reasons to believe that the selection model also circumvents these objections.

Recall that the disagreement objection posits that there is no way to come to agreement on who is competent in a way that affords them the right to have special power over political decisions. Notice, though, that even if we live in a direct democracy, there will be disagreement about which policy decision is correct. Thus, regardless of the regime type, some will live under policies with which they disagree. The difference, related to disagreement, between epistocracy and democracy arises only if, in the former, those who enjoy positions of power have those positions permanently. Indeed, Estlund worries that “under unequal suffrage, some people are formally and permanently subjected to the rule of certain others. This is a ruling relationship that is not present under majority rule” (Estlund Reference Estlund2008, 37). Although we cannot avoid some living under policies with which they disagree, it is particularly objectionable to be permanently subject to the rule of others whose competence one does not accept (or, even if accepts, does not do so to the exclusion of other possible rulers). Think here of an epistocracy in which rulers must be chosen from a given pool of candidates, e.g., those with elite education, whom one does not take to be exclusively well-suited to the role. The disagreement objection takes its force not from the fact that one has to live under a policy one disagrees with, but from the connection between people’s disagreement about appropriate laws (or rulers) and epistocracy’s ostensible commitment to a permanently enabled class of rulers.

The key reason that the selection model of representative democracy need not be vulnerable to this objection is that the epistemic benefits associated with the selection mechanism critically depend on it being the case that citizens have regular opportunities to replace leaders from a diverse pool of potential candidates. Although there will, surely, be disagreement in representative systems about who ought to rule, that disagreement is of the same order as disagreement in systems of direct democracy about what the laws should be. The parity would, arguably, break down in relation to systems in which rulers were drawn from a limited pool. But the selection model does not require that representative democracy be such a system—whereas one can easily imagine representative systems with restricted pools of candidates, one can also readily conceive of a well-functioning representative system that eschews any such restrictions. In short, the disagreement objection should not be construed as posing a problem for selection mechanisms (within representative systems) as such.

Let us consider next the force of the demographic objection. If institutions of electoral representation are quality-tracking, then elections successfully generate epistocracy without relying on invidious comparisons. The worry underlying the demographic objection is that in selecting for competence, elections may also inadvertently introduce an adverse bias; those who have markers of excellence that are identified for selection, such as—for example—superlative education, may be reasonably expected to have a bias (e.g., higher levels of risk tolerance) that afflicts those with such markers.

This implies that there is an ordering of outcomes, such that the leaders afflicted with bias would produce lower-ranked outcomes than (the counterfactual) leaders without such bias. If so, though, the voters who observe such lower-ranked outcomes should remove the lower-performing leaders from office, replacing them with the candidates drawn from the same pool who are not known to have such bias. The process should be expected to continue for as long as the voters observe outcomes associated with bias.

The claim that such a process tracks quality implies here that we should expect it to move incrementally to the selection and re-election of candidates who are free of the bias. As with the treatment model, in a properly institutionalized electoral democracy, the iterative nature of elections filters out biased officeholders. So, once again, it turns out that criticisms of epistocracy implicitly depend on the assumption that such forms of government will make once and for all decisions about who is entitled to rule. The iterated nature of selection in a well-designed representative democracy generates a form of epistocracy that helps circumvent such problems.

We add three comments. First, the preceding argument abstracts away from some of the empirical complexities that characterize existing democratic societies, including, in particular, not only unequal rates of political participation among different socio-economic groups (e.g., Lijphart Reference Lijphart1997; Parvin Reference Parvin2018a) but also burdens of judgment that could lead to persistent disagreements over what counts as bias on the part of the incumbents. These complexities must surely lower our expectations of fully eliminating the concern at the core of the demographic objection—even well-institutionalized versions of representative democracy will not be free of demographic bias, though we may reasonably expect it to be mitigated by more conducive institutions, such as more egalitarian systems of campaign finance. However, the persistence of bias does not distinguish selection-based representation from direct democracy. Whereas the former may end up with the decisive coalition of the electorate ignoring the bias of the elected official against a minority, the latter may be expected to end up with a majority of voters choosing a policy that a minority recognizes as biased. Likewise, there is no reason to expect direct democracy to do better with respect to realizing equality of political participation than would a well-institutionalized representative democracy—indeed, the greater burdens of citizen participation in direct democracy may well exacerbate this problem.

Second, it is important to recognize that term limits, which are sometimes popularly seen as ways of decreasing the bias of officeholders are, in fact, detrimental to the ability of the electoral process to track quality in the selected rulers. Similarly, the incumbency advantage, which is also associated with the presence of a systematic bias in favor of, but also among, the incumbents, would, in a properly institutionalized setting, be a mark of better selection—selection of leaders with smaller or no bias. In other words, in a well-designed representative system, incumbents would normally defeat challenges, but this would owe more to the iteratively-reinforced superiority of incumbents than to structural features of the electoral systems or of office-holding that arbitrarily favored those already in office.

Third, it is instructive to compare the treatment and the selection models with respect to their consistency with egalitarian commitments. The selection model is clearly consistent with effective equality of opportunity to run for office. In fact, the model provides its own reason to insist on such equality. After all, were a departure from equal opportunity to produce worse performance, quality-seeking voters would want to enlarge the pool of available candidates. This means that the institutional provisions that create meaningful equal opportunity to run for office (such as access to quality education) must be a part of the institutional framework underlying the selection model. In contrast, under the treatment model, there is no reason to prefer some candidates over others, making that model consistent with a more radical notion of egalitarianism (according to which all citizens are regarded as equally capable of effectively discharging the responsibilities of public office).

THREE ALTERNATIVES/OBJECTIONS

In this section, we briefly consider three common types of attempts to dull the challenge presented to democrats by advocates of epistocracy or, what often amounts to the same thing, to improve the epistemic performance of democratic systems in ways that are compatible with egalitarian commitments. In the former sense, these attempts are, implicitly, objections to the underlying premise of the critical importance and role of expertise; in the latter sense, they are alternatives to the approach we laid out above. Our aim here is not to offer a comprehensive response to these attempts, nor to argue that they are always deficient—each of them is predicated on an important idea that sometimes yields a genuine improvement in democratic governance. Rather, we aim to highlight the appeal of our approach by identifying worries about these alternatives that make them unsatisfying as a general, encompassing, response to the challenge of epistocrats.

STILL MORE EPISTOCRACY?

We have explained how an epistocratic defense of representative democracy might proceed by showing how the treatment and selection mechanisms that are central to that form of government should, in a well-designed system, lead elected officials to acquire greater knowledge and take greater care to meet the expectations of citizens than would voters in a direct democracy. However, once one has adopted this epistocratic perspective, it might seem that s/he cannot resist following it even further—why not pursue further epistemic gains by, for instance, limiting voting rights to those who can pass a competence exam?

The standard reasons to be wary of such arrangements are evident in the preceding discussion: in particular, the disagreement and demographic objections raise important concerns about such proposals (that, again, do not extend to the mechanisms lying at the heart of traditional forms of representative democracy). In fact, even proponents of such schemes often accept that the disagreement and demographic objections raise important worries about their proposals. For instance, Jason Brennan allows that “restricted suffrage is unjust and morally objectionable” for the reasons identified by the disagreement and demographic objections, but argues that a system built around universal suffrage is also unjust because it allows incompetent individuals to impose political decisions on others (Brennan Reference Brennan2011, 716). His view, then, is that although both types of systems are unjust, we should accept restricted suffrage as the less unjust alternative:

The important point, for our purposes, is just that—on Brennan’s account—we have to weigh the respective injustices of the two types of systems and select the lesser evil (though, see Brennan Reference Brennan2018). The central difference here between Brennan and Estlund is not that Estlund denies that democratic arrangements have significant epistemic limitations, but rather that he views egalitarian considerations as imposing a side constraint on institutional arrangements, leading him—unlike Brennan—to favor universal suffrage.

However, the preceding discussion suggests that the options facing us may be considerably less stark than either advocates of epistocracy or advocates of universal suffrage imagine. The treatment and selection mechanisms that lie at the heart of representative democracy provide a way of maintaining the egalitarian appeal of universal suffrage even while addressing the epistemic concerns associated with allowing ordinary citizens to make policy decisions. In other words, once one appreciates the treatment and selection mechanisms that lie at the heart of representative democracy, the ex ante appeal of proposals for restricted suffrage is considerably dulled. After all, the treatment and selection models reveal representative democracies already to have powerful, in-built mechanisms for epistemic gains that have the advantage of not being susceptible to the important objections faced by schemes hinging on restricted suffrage.^{Footnote 18} Thus, insofar as one finds the disagreement and demographic objections to be concerning, the obvious response to the worries raised by epistocrats is to work to strengthen the effectiveness of the treatment and selection mechanisms identified above.

In a review of Brennan’s book, Christiano observes that there seems to be an inconsistency between, on the one hand, well-known empirical studies about the ignorance of ordinary citizens and, on the other, evidence that consolidated democracies tend to perform fairly well, at least in a number of important areas (such as the protection of human rights). This leads Christiano to conclude by noting that:

The treatment and selection models provide an important part of an answer to Christiano’s question. Representative democracies may work reasonably well despite the informational limitations of ordinary citizens because representative democracies are not pale imitations of direct democracies; instead, they—at least when well-designed—capitalize on treatment and selection mechanisms that allow them to perform substantially better than would systems that simply turned over policy decisions to citizens en mass. This epistemic appeal of representative democracy is so often missed in the contemporary democratic theory literature because participants on both sides of the debate tend to ignore the important insight of historical proponents, such as Madison, that representative democracy is not an approximation of direct democracy, but an importantly distinct form of government, and one, as we sought to argue, with considerable and unique epistemic virtues to recommend it.

A further, and complementary, part of the answer is that the disjunction between relatively uninformed and disengaged citizens and reasonably well-functioning democracies is paradoxical only in appearance. In regimes with robust freedom of the press and high government transparency—that is, regimes in which the cost to voters of obtaining information about public officials is low and the quality of potentially accessible information is high—public officials can readily anticipate that signals of malfeasance or poor performance could be substantiated with relative ease. Given common knowledge of this anticipation, voters in those regimes have, in equilibrium, less reason to pay the costs associated with accessing information, because incumbent officials who place a high value on their positions are more deterred from egregious misbehavior. In effect, the easy potential accessibility of high quality information is the watchdog that allows its owner to nap: it makes the impediments to oversight from poor knowledge and low participation less binding and so motivates elected officials to act in ways that advance their constituents’ interests.^{Footnote 19} The bottom line is that instead of being at odds with one another, a relatively ignorant and disengaged citizenry and the strong macro-performance of well-designed representative systems can be (and in liberal societies, arguably, are) aspects of the same internally consistent (equilibrium) pattern resulting from a system with effectively functioning treatment and selection mechanisms. This suggests that common hand-wringing about the ignorance of voters may be overwrought, and how concerning it ought to be must turn on more nuanced details of institutional contexts than the current state of the discourse suggests.^{Footnote 20}

As a final point, note that in addition to helping explain the gap between the knowledge level of ordinary citizens and the performance of well-constructed representative democracies, this insight also allows us to think more constructively about how to make such regimes work better by highlighting important mechanisms that must be strengthened to accomplish that goal. Juxtaposing the treatment and selection models to the Condorcet jury theorem may help clarify this point. Most obviously, if we understand democratic politics through the lens of the latter, it makes sense to support reforms, such as referenda on major policy issues and limitations on the discretion of elected officials, that may allow us to better capitalize on the epistemic gains associated with the law of large numbers. By contrast, if we understand democratic politics through the lens of the treatment and selection models, it makes sense to be quite anxious about such reforms because they can undermine the very mechanisms that those models suggest can help generate better policy results.