Finding the Epistocrats

Abstract

Concerned about widespread incompetence among voters in democratic societies, epistocrats propose quasi-democratic electoral systems that amplify the voices of competent voters while silencing (or perhaps just subduing) the voices of those deemed incompetent. In order to amplify the voices of the competent we first need to know what counts as political competence, and then we need a way of identifying those who possess the relevant characteristics. After developing an account of what it means to be politically competent, I argue that there is no way for the epistocrat to identify such persons. Therefore, epistocracy cannot be implemented.

1. Introduction

It is a well-established fact that many citizens in democratic societies are radically uninformed about the issues the democratic process calls them to vote on. Concerned about this, epistocrats propose quasi-democratic electoral systems that amplify the voices of competent voters while silencing (or perhaps just subduing) the voices of those deemed incompetent (§2). To amplify the voices of the competent, though, we first need to know what counts as political competence (§3). Many epistocrats hold that political competence requires a capacity to reason well along with knowledge of current events and social science. This is true but incomplete. Political competence, I argue, also requires that persons possess moral knowledge about policies that may permissibly be implemented, as well as the motivation to act (i.e., vote) in accordance with this knowledge.

Even with a theory of political competence in hand, the epistocrat cannot yet implement her preferred electoral system. She must now identify who the politically competent are (§4). This is where the trouble begins. The main proposal for how to do this is to give voters a suffrage exam. While a suffrage exam can identify some characteristics of political competence, it cannot test whether one is motivated to vote morally which, I argued, is part of what it means to be politically competent.

How might the epistocrat respond to this challenge? There are three ways. First, the epistocrat might propose something besides a suffrage exam that can determine whether someone is sufficiently morally motivated (§5). I examine a proposal of this kind but conclude it fails. A different approach argues that we need not identify those who are morally motivated, because most persons are morally motivated when they enter the voting booth anyways (§6). I show that this is not so.

Finally, the epistocrat might concede that there is no way to identify those who are fully politically competent; however, epistocracy would still outperform democracy if those who are at least partially politically competent have their voices amplified (§7). This is the epistocrat's most promising route to circumnavigating my criticisms. I end by offering reasons to doubt that an imperfect epistocracy outperforms democracy. However, more empirical work is needed to fully adjudicate this issue.

2. Epistocracy and its varieties

Brennan (2011, 2016: Ch. 6) offers the most compelling argument in defense of epistocracy. He asks us to imagine five different criminal trials, each with a defendant who is being prosecuted for murder. In the first trial, the jury is ignorant. As the trial proceeds, these jurors “ignore the evidence presented to them. When asked to deliberate, they refuse to read the transcript” (Brennan 2016: 151). In the second trial, the jury is irrational. These jurors “evaluate the evidence in cognitively biased, nonscientific, or even antiscientific ways” (Brennan 2016: 151–2). They might make their decision based on wishful thinking or bizarre conspiracy theories.

In the third trial, the jury is impaired. The jurors genuinely want to pay attention to the evidence and evaluate it correctly, but “they are simply not competent to do so. Perhaps they are cognitively impaired or the case is too complicated for their mental capacities” (Brennan 2016: 152). In the fourth case, the jury is immoral. Here, the jurors decide contrary to the evidence because they are wicked. They might decide to convict because “the defendant is black, Jewish, Republican, or whatnot, and they dislike people like that” (Brennan 2016: 152). In the final trial, the jury is corrupt. In this case, the jury only finds the defendant guilty because they are bribed to reach that verdict.

In all five cases it is obvious that a misfire of justice occurs. If we discovered that a jury acted in one of the ways described above, then we would declare their decision void and demand a retrial. Moreover, there is a strong intuition that defendants who go on trial have a right to not be subject to a jury matching one of the above descriptions. That is, defendants have a right to a competent jury.

Few would disagree with Brennan's reasoning thus far. Yet the same sorts of cases that inspire the right to a competent jury also appear in democratic politics. Indeed, we often have an ignorant electorate: “the majority of voters pay no attention to the details of the election or the issues at stake” (Brennan 2016: 158). Beyond this, we also have an irrational electorate. Though many citizens do pay attention to the issues at stake, “they vote not on the basis of evidence but rather on the basis of wishful thinking and various disreputable social scientific theories” (Brennan 2016: 158). Third, we sometimes have an impaired electorate: “most of the discussion … is beyond [voters’] level of comprehension, requiring more intelligence than they in fact have” (Brennan 2016: 158).

Fourth, Brennan believes democracies often have an immoral electorate. He writes: “out of racism, the majority chooses a white candidate over a black one. Or, out of superficiality, they choose the better-looking candidate” (Brennan 2016: 158). And finally, we have a corrupt electorate: “the majority of voters choose a policy in their own self-interest, even though the policy severely harms or has a serious risk of imposing harm on the minority” (Brennan 2016: 158). Given the symmetry, consistency demands we proclaim persons have a right to a competent electorate, just as they do a competent jury.

If we grant that Brennan's argument succeeds, the question becomes: how do we ensure that citizens enjoy a competent electorate? There are many ways of institutionalizing epistocracy. Most methods involve amplifying the voices of the competent relative to the incompetent.¹ Perhaps the simplest way to implement epistocracy is by requiring “potential voters to pass a voter qualification exam … The exam would screen out citizens who are badly misinformed or ignorant about the election, or who lack basic social scientific knowledge” (Brennan 2016: 211–12).

John Stuart Mill's plural voting is another form of epistocracy (Harwood 2008; Mill 2015: 284–300; Brennan 2016: 213; Mulligan 2018). Everyone gets to vote according to this method, but more votes are given to individuals based on how competent they are. For instance, one's level of education might determine how many votes one gets.

Another version of epistocracy allows for universal suffrage, but gives an epistocratic council veto power over legislation. The council “has no power to make law. It cannot appoint anyone to any office, nor can it issue any decrees or regulations. It cannot instantiate any coercive regulations or rules on citizens. But it has the power to unmake law” (Brennan 2016: 216). According to Brennan, membership in the epistocratic council would be open to all citizens able to pass a competency test. Kindred spirit Bryan Caplan has proposed that the Council of Economic Advisers be given veto power over “uneconomical” legislation (Caplan 2006).

The final version of epistocracy is called simulated oracle, also sometimes called modelled democracy (Brennan 2016: 220–22; Ahlstrom-Vij 2022). Here, the electorate is divided up by demographic factors: age, race, gender, education background, religious beliefs, and so on. Then, a test is given to everyone. Those who do sufficiently well on the test get to cast a vote for themselves and also for everyone else in their demographic group unable to pass the test.

In sum, most versions of epistocracy secure the right to a competent electorate by implementing electoral systems that amplify competent voices over incompetent ones. This raises an important question: in the domain of politics, what makes one competent? What qualities must one possess to be one of the competent whose voice is amplified?² I answer this question in the next section.³

3. On political competence

What does it take to be politically competent? Brennan proposes an account consisting of the following components: competent voters “act on widely available, good information, if not always the best information available anywhere”; competent voters “avoid mass superstition and systematic error”; competent voters “evaluate information in a moderately rational, unbiased way”; and competent voters are “aware of their limits, and thus always look for more and better information on any high-stakes decision” (Brennan 2016: 165).

We can, I think, break Brennan's definition of political competence down into three distinct components. First, competent voters reason well. They are not, for instance, subject to confirmation and disconfirmation bias, availability bias, affective contagion and prior attitude effects, framing effects, peer pressure and authority, and so on (Brennan 2016: 43–8). Reasoning well also means that competent voters know their limits. If they are confronted with something they don't know or understand, then they do their best to correct this by apprising themselves.

The second and third components of Brennan's definition of political competence concern knowledge of a certain kind. There are two types of knowledge competent voters possess: knowledge of current events and social scientific knowledge. Brennan further describes these components in the following passage:

To be well informed, citizens also need to know the candidates’ policy platforms, how candidates are likely to vote in Congress, what policies the candidates are likely to support, whether these votes are likely to matter or not, and how much influence the candidates are likely to have if they win.

A well-informed voter needs to be able to assess whether the candidates’ preferred policies would tend to promote or impede the voter's favored outcomes. So, for example, suppose I know candidates Smith and Colbert both want to improve the economy, but Smith favors free trade, and Colbert favors protectionism. I can't make a reasonable choice between them unless I know whether free trade or protectionism is more likely to improve the economy; to know that, I need to know economics. (Brennan 2016: 28)

In sum, there are three components to Brennan's definition of political competence: the ability to reason well, knowledge of current events, and social scientific knowledge. If one possesses these three characteristics, then one's voice will be amplified in an epistocracy.

Brennan's definition of political competence is a good start, but it is not sufficient to address the problems that concern him. Recall the jury-electorate argument examined in the section above. According to Brennan, there are five ways juries and electorates can act incompetently and thereby treat persons unjustly. They can be ignorant, irrational, impaired, immoral, or corrupt. The problem with Brennan's definition of political competence is that it does not eliminate all possible ways in which electorates can behave incompetently.

To see this, suppose that an electorate is competent in Brennan's sense. This means that all voters in the electorate can reason well, they all possess sufficient knowledge of current events, and they all possess sufficient social scientific knowledge. Because all voters possess sufficient knowledge of current events and social science, the electorate is not ignorant. Because all voters can reason well (which means they are not subject to biases and faulty heuristics), the electorate is not irrational. And because reasoning well also entails knowing one's limits and seeking more information if one is confronted with an issue that one is unfamiliar with, the electorate is not impaired. Citizens are thus shielded from the injustice of having ignorant, irrational, and impaired electorates impose decisions on them.

Yet, the electorate as described can still vote in immoral and corrupt ways. An electorate satisfying Brennan's definition of political competence could vote to implement racial segregation, for instance. This is a clear example of an immoral electorate. Nothing about reasoning well and being well-informed about current events and social science prevents them from doing this.⁴ It is also possible that our well-informed, well-reasoning electorate might vote to redistribute the assets of the disenfranchised to themselves. This, I think, would qualify as a corrupt electorate. Again, nothing about reasoning well and being well-informed about current events and social science prevents voters from doing this. As we saw in the section above, these are kinds of incompetence that result in unjust political decisions.⁵ To ensure that the right to a competent electorate is fully protected, the definition of political competence must be expanded to exclude these possibilities.

Two further components must be added. First, competent voters must possess moral knowledge concerning the policies that may permissibly be implemented.⁶ A competent voter, for example, will know that it is impermissible to vote for racial segregation. A competent voter will also know that it is impermissible to redistribute to themselves the assets of those who cannot vote.

The above addition to Brennan's definition of political competence is still not sufficient to avoid immoral and corrupt electorates. This is because persons can know the ends that are permissible for them to pursue, but nonetheless fail to act on this information. For instance, an electorate might satisfy Brennan's definition of political competence, and also possess the requisite moral knowledge. This electorate knows that racial segregation is impermissible, but decides to implement it anyways. Perhaps they just cannot get over the prejudices inculcated in their youth. Despite their moral knowledge, this is still an example of an immoral electorate.

The definition of political competence must account for akrasia. Not only is moral knowledge necessary for political competence, but so too is the motivation to act on this knowledge. Competent voters not only possess moral knowledge concerning the policies that may permissibly be implemented, but they are also motivated to vote in accordance with this knowledge. By adding this fifth and final component to our definition of political competence, we finally take immoral and corrupt electorates off the table. Because a competent electorate knows the policies that are morally permissible to pursue, and because the electorate is also motivated to cast their votes based on this information, it will not act immorally or corruptly.

To sum up, political competence demands that voters reason well, are informed about current events, possess social scientific knowledge, possess moral knowledge, and are motivated to vote in accordance with their moral knowledge. With an account of political competence under our belts, we now turn to a deeply important question any account of epistocracy must answer: how do we find the politically competent? If this question cannot be answered, then epistocracy cannot be implemented.

4. Finding the politically competent

The politically competent, I have argued, consist of those who reason well, are informed about current events, possess social scientific knowledge, possess moral knowledge, and are motivated to vote in accordance with their moral knowledge. To implement her preferred system, the epistocrat must identify all persons who possess these characteristics, so their voices can be amplified. How can this be done?

Before jumping into this question, I want to first make clear that the question I am asking is distinct from a similar question in social epistemology. One important question in social epistemology is how non-experts can identify experts. Many think that non-experts can, in fact, do this (e.g., Goldman 2001; Anderson 2011). Others are skeptical (e.g., Friedman 2017). This literature tends to focus on how non-experts can identify scientific and social scientific experts. A related literature asks whether and how non-experts can identify moral experts (e.g., Cholbi 2007; Hoffman 2012; Riaz 2021).

This body of work is not relevant to my inquiry. The reason why is that in an epistocratic system there is no reason to think that the politically incompetent must be the ones who identify the politically competent. In fact, it is likely that those in an epistocracy who select the politically competent will themselves be experts of a certain kind. We can imagine identification of the politically competent being carried out by a bureaucratic agency that, like most bureaucratic agencies, is staffed by persons who are experts in the relevant fields. Of course, there are incentive-compatibility problems worth considering here (e.g., Klocksiem 2019; Manor 2022). I set these issues to the side, to see if epistocracy can be realized under ideal conditions.

So how will the expert bureaucrats find the competent voters? Brennan suggests a suffrage test. He writes:

The exam would screen out citizens who are badly misinformed or ignorant about the election, or who lack basic social scientific knowledge. The United States, for example, might use the questions on the ANES [the American National Election Studies]. Alternatively, the United States might require citizens to pass the citizenship exam, or score a three or higher on the Advanced Placement economic and political science exams. (Brennan 2016: 211–12)

The proposal is simple. Bureaucrats use their skills and expertise to craft an exam that tests for political competence. All citizens take the exam. Those who do sufficiently well on the exam are deemed politically competent; those who do not do sufficiently well are deemed politically incompetent.

Can a suffrage exam identify the politically competent? Testing for knowledge of current events and social science is straightforward. Political scientists use the American National Election Studies (ANES) to test for knowledge of basic civics and current events. And as Brennan notes in the passage cited above, we can use questions from Advanced Placement exams in economics and political science to test for social scientific knowledge. Testing one's capacity to reason well is also relatively straightforward. Social scientists, psychologists, and behavioral economists often use tests or surveys to examine the extent to which persons reason in a rational manner.⁷

Testing for moral knowledge is trickier. One cannot simply ask moral questions on an exam – “Is eating meat wrong?” – because moral experts disagree over how to answer these questions. Instead of testing for moral knowledge, one can test persons to see if they know how to reason well about ethical questions. This assumes that those who can reason well about ethical questions possess more moral knowledge than those who cannot, but this seems reasonable. There might be some questions about different moral theories. The exam could also ask questions about how to apply ethical theories to certain cases. Answering questions of this kind demonstrates that one possesses some capacity to reason about moral problems, which reasonably suggests some level of moral knowledge.

The big issue is how to test for moral motivation. One cannot simply ask on a suffrage exam “Do you desire to do what is right?” or “Do you always try to act justly?” because persons will answer “Yes” regardless of their true dispositions. Questions of this kind would be entirely uninformative, as those who lack moral motivation can (and would!) lie. It seems, then, that there is one component of political competence that cannot be identified through a suffrage exam. Suffrage tests may identify those who reason well and have sufficient knowledge of current events, social science, and morality. However, suffrage tests do not allow us to identify those who have the right kind of motivation.

5. Finding the morally motivated

In response to the challenge laid out in the prior section, the epistocrat might propose something other than a suffrage exam to find those who have the proper moral motivation. More specifically, there might be features of persons we can easily measure that are highly correlated with moral motivation. If characteristic c is highly correlated with moral motivation, then we can identify the morally motivated by finding those with c. The politically competent would then be anyone who passes the suffrage exam who also possesses c.

There are two issues with this proposal. The first problem is coming up with a list of characteristics that tightly correlate with moral motivation. Some have attempted this. A small literature in experimental philosophy asks whether ethics professors are more moral than professors who do not specialize in ethics (e.g., Schwitzgebel 2013; Schwitzgebel and Rust 2014). To answer this question, researchers come up with a list of activities highly correlated with being moral, and ask whether ethics professors engage in these activities more frequently than non-ethics professors. The list includes paying membership fees for professional societies, not eating meat, donating blood and organs, charitable giving, calling one's parents, etc.

Though the above list of characteristics supposedly correlates with being moral, we might reasonably wonder how tight the relation is. Charitable giving is tax-deductible in the United States. Giving to charity might reflect one's desire to help those in need, but it might also reflect one's desire to reduce one's tax burden. So, charitable giving is not sufficient for being morally motivated. Nor is it necessary, as there may be morally motivated persons who lack the financial means to give money away. Possibilities like this call into question just how tightly the proposed list of characteristics correlates with moral motivation.

But suppose, for the sake of argument, that we did come across a characteristic c that is highly correlated with being moral. In other words, persons who have characteristic c tend, with a high degree of frequency, to be morally motivated. Even if the epistocrat could find such a characteristic, she still runs into problems. In particular, the signal characteristic c conveys will, over time, become less and less reliable, until it is eventually no longer informative about persons’ moral motivation. This is because enfranchising persons (in part) based on c will incentivize persons to start acquiring characteristic c regardless of whether they are morally motivated or not. Hence, persons who are not morally motivated will begin acquiring c, which means that someone possessing c no longer tells us much.

That was a bit abstract, so let's consider an example. Suppose it's true at time t ₁ that volunteering is highly correlated with moral motivation. As such, the epistocrat proclaims that if someone volunteers enough hours every month, then they qualify as appropriately morally motivated. If this person also passes the suffrage exam, then they qualify as politically competent. At time t ₁, the epistocrat releases her criteria for determining whether persons are sufficiently morally motivated.⁸ Few volunteer a sufficient number of hours, and even fewer volunteer enough and can pass the suffrage exam, so it is a relatively small number of voices amplified. Many people, however, want to participate in politics. In order to achieve enfranchisement, persons now have a clear path forward: they must pass the competency test and must also volunteer a sufficient number of hours. So, those who want to participate begin studying and volunteering.

When the next electorate is chosen at time t ₂, many more people will be enfranchised, because many more people will have volunteered and passed the exam. The problem, though, is that not everyone in this new group of persons will be morally motivated. Many of them don't volunteer because it's the right thing to do; they volunteer merely so they can participate in politics. Volunteering was highly correlated with moral motivation at time t ₁ not because volunteering automatically makes one a moral person, but because moral persons tended to be the ones who made up the bulk of volunteers. But now, there are new reasons to volunteer that don't involve being morally motivated. Those who now volunteer for the wrong reasons dilute the composition of volunteers, such that volunteering is now no longer highly correlated with moral motivation at time t ₂.

The underlying dynamic generalizes. For any characteristic c that highly correlates with moral motivation, using c to select the competent will incentivize those who wish to exercise political power – whether they are morally motivated or not – to acquire c. The end result is that c is no longer highly correlated with moral motivation. Thus, relying on characteristics that act as a proxy for moral motivation is, in a sense, self-defeating. Once the characteristic is selected, over time it will become less and less correlated with moral motivation, because those who wish to exercise political power will acquire the characteristic.

In response to this argument, one might point out that spending time to acquire the characteristic is irrational, as acquisition requires a significant time investment for little payoff: one's vote, after all, is unlikely to make a difference.⁹ Therefore, we need not worry about people acquiring the characteristic just so they can vote. From a cost-benefit perspective, volunteering several hours a month is not worth gaining the franchise, for there is little benefit to voting in the first place. As such, it will still only be the morally motivated who acquire characteristic c even after possessing c becomes necessary for suffrage.

Whether it is rational to engage in a costly activity like volunteer work just so one can vote depends on why one votes in the first place. If the only reason one votes is to change the outcome of the election, then it will indeed be irrational to acquire the relevant characteristic, because one's vote has an infinitesimally small chance of being decisive. However, if this reasoning was descriptively accurate, then no one would vote in ordinary democratic elections in the first place, as the time required to register to vote and wait in line at the polls is outweighed by the infinitesimally small chance of changing the election's outcome. Those who vote must do so for reasons other than changing the outcome.

The most plausible theory of voter motivation is the expressive theory, which says that an individual votes “simply for the sake of the expression itself and without any necessary implication that the desired outcome will be brought about thereby. Revealing a preference is a direct consumption activity, yielding benefits to the individual in and of itself” (Brennan and Lomasky 1997: 33).¹⁰ An analogy is often drawn by proponents of the expressive theory between voters and sports fans. Sports fans spend significant time and money cheering for their team even though cheering probably won't make a difference in terms of the final score. For the fan, cheering itself is a consumption activity. Likewise, people vote (as well as attend rallies, put bumper stickers on their cars, etc.) so they can express support for their political team, even though this expression probably won't make a difference in terms of the outcome. For the voter, voting itself is a consumption activity.

Will expressive voters go through the trouble of acquiring the relevant characteristic just so they can vote? Surely not all of them. For many, the utility gained by expressing support for one's political team will not be worth the opportunity cost paid through volunteering enough hours each month. A significant number, however, will be happy to pay this price. This becomes clear if we once again consider voting as akin to rooting for one's favorite sports team. Countless sports fans spend large sums of money each year on tickets to games, jerseys, and other memorabilia. So too do they spend large amounts of time watching their team play along with consuming news about their team. If expressing support for a sports team leads many fans to pay significant costs, then expressing support for one's political team will likely also lead many voters to pay significant costs by acquiring the relevant characteristic.

To be clear, my claim is not that everyone who votes in an ordinary democracy will still vote in an epistocracy where one must acquire characteristic c to gain the franchise. My claim is that enough people will be driven by the desire to vote to acquire c such that the pre-existing pool of persons with c becomes significantly diluted by those who may or may not possess moral motivation. The result is that c is no longer informative.

6. Abandoning moral motivation

There is no plausible way for the epistocrat to find those who possess the proper moral motivation. One response here says that we need not worry about identifying the morally motivated because most people already have the proper motivation when they enter the voting booth. If we select for persons who are politically competent in all other respects, chances are these persons will possess the appropriate moral motivation anyways.

Why think that persons will, by default, be morally motivated when they enter the voting booth? According to Brennan, this is a well-established social scientific fact. He writes:

Political scientists have conducted numerous empirical studies of voter behavior, using a variety of methods. They overwhelmingly conclude that voters do not vote selfishly. Instead, voters tend to be nationalist and sociotropic. That is, they tend to vote for what they perceive to be in the national interest rather than in their self-interest. (Brennan 2016: 49–50)

Later, Brennan says that voters are “altruistic” and that they “want their elected officials to serve the common good of their country rather than their narrow self-interest” (Brennan 2016: 50). Note, this is not to say that persons are morally motivated all the time. The claim is that, based on the research we have, most people tend to behave morally when they enter the voting booth specifically.

I do not think this response to my argument succeeds. Before I detail why, I first want to examine a failed objection to this response, inspired by Piero Moraro's criticism of epistocracy. Moraro agrees with Brennan that persons do, by default, typically vote in a moral manner in existing democracies. But, he thinks it unlikely that the politically competent will continue to vote morally in an epistocracy. Hence, the epistocrat cannot forgo identifying the morally motivated. Let us look at the details of Moraro's argument.

Assuming Brennan is correct that our best social science says persons already tend to vote morally, what explains this fact? Key here is that an individual's choice in the voting booth is often not decisive, in that an individual's vote is incredibly unlikely to decide the outcome of an election. Since an individual's vote is unlikely to decide the outcome, she can indulge her moral values rather than pursue her self-interest. By doing so, the individual can feel good about how she voted, but doesn't sacrifice anything of importance to herself, because her vote was unlikely to make a difference anyways.

In other contexts, indulging one's moral values will come at a cost, thereby making moral behavior far less likely. For instance, Althea might have the choice of buying coffee each morning from two different shops, X or Y. Coffee shop X only buys fairtrade beans, and thus has moral business practices. As a result of this, X's prices are higher. Company Y does not buy fairtrade beans, so their practices aren't moral. But because they buy the cheapest beans possible, Y's prices are lower. In this case, indulging her moral values (by shopping at X) will cost Althea, so she has an incentive to be selfish and buy coffee from Y.

The key thing to note here, according to Moraro, is that the political context resembles the market context more and more as the electorate shrinks. As the electorate shrinks, one's vote becomes more and more likely to influence the outcome of the election, so who to vote for becomes more and more like Althea's choice of where to buy coffee. In Moraro's words: “As the electorate's size decreases, voters’ altruism tends to fade” (Moraro 2018: 210). But epistocratic electoral systems essentially shrink the size of the electorate. There is thus reason to think that competent voters are likely to vote selfishly rather than morally in an epistocracy. Because of this, the epistocrat cannot forgo identifying the morally motivated, as she must find those who will continue to vote morally in an epistocratic system.

Moraro's objection fails. The problem is that, in an epistocracy, it is unlikely the electorate will shrink enough such that voters are incentivized to begin voting in a self-interested manner. Experimental work on this question finds that a one-in-eleven chance of being decisive is low enough odds to induce some persons to vote morally who otherwise vote selfishly when they are more likely to be decisive (Feddersen et al. 2009: 176). That is, even when there is a roughly nine percent chance of being decisive, that is still low enough to induce some persons who otherwise vote selfishly to go ahead and vote morally. Though it will shrink the electorate a considerable amount, it is unlikely that epistocracy will make it such that each competent voter has a greater-than-nine percent chance of deciding an election.

How, then, should we respond to the epistocrat's claim that we need not worry about identifying the morally motivated, because people tend to vote in a moral manner anyways? The easiest thing to do is deny the empirical claim – that people tend to vote in a moral manner by default. Brennan, we saw above, says there is a large literature demonstrating that voters tend to vote morally. I do not think this is true. The literature Brennan cites to support this claim does not, in fact, show that persons typically vote in a moral manner.

The first problem is that many of the papers Brennan cites do not talk about voting at all. Many of them are about whether material self-interest can explain public opinion polling (e.g., Funk 2000), or whether material self-interest can explain persons’ perceptions, such as their perception of how the economy is performing (e.g., Conover et al. 1987; Funk and García-Monet 1997). The answer is no, but that's not the point. The point is that public opinion polling and perceptions are not always accurate indicators of how people vote. Suppose, though, that public opinion polls and perceptions are sufficient to track how persons vote. So interpreted, these studies show that material self-interest cannot explain how people vote. From this, though, it does not follow that persons ground their votes in moral considerations.

And in fact, the studies Brennan cites by no means suggest that people vote morally. One paper notes that partisanship plays a role in generating persons’ assessments of the economy (Conover et al. 1987). Voting on the basis of partisan allegiance, though, does not constitute moral voting. As another example, one paper Brennan cites shows that anti-busing attitudes among whites were not best explained by material self-interest, but by “racial intolerance and political conservatism” (Sears et al. 1979: 369). Voting on the basis of racial intolerance, though, is not moral voting. As a final example, a group of scholars found “various self-interest measures to have very little effect in determining either policy preferences or voting behavior. In contrast, symbolic attitudes (liberal or conservative ideology, party identification, and racial prejudice) had major effects” (Sears et al. 1980: 670). So yes, material self-interest is not doing the work, but morality sure isn't either.

The most compelling case Brennan can make is by appealing to papers that show voters tend to vote in the national economic interest, not their own material self-interest (e.g., Kinder and Kiewiet 1979). But even here, voting in the national economic interest is not the same as voting in a moral manner. Consider a case. Bertha is confronted with a candidate who says many racist things, and who wants to implement policies that discriminate against a racial minority. Bertha knows this is wrong, and that she shouldn't vote for the candidate. Yet, Bertha believes the candidate will best serve the national economic interest. Therefore, Bertha votes for the candidate anyways. It is true in this case that Bertha does not vote in her material self-interest, and it is also true that Bertha's behavior is consistent with (some of) the research we have on how persons actually vote. Bertha still fails to vote in a moral manner, though.

Finally, there is research that cuts against Brennan's claims about moral voting. One study shows that whether material self-interest informs policy preferences depends on how salient one's economic situation is made before one is asked about one's preferences (Sears and Lau 1983). But this means that voters may very well vote in their material self-interest, so long as they are primed before they enter the voting booth. Finally, there are cases where material self-interest clearly informs how persons vote. For instance, a look at California's tax revolt shows that those with a higher tax burden were more likely to support tax cuts while public employees were more likely to oppose tax cuts (Sears and Citrin 1982). In sum, the claim that persons tend to vote morally by default is not well supported by the empirical literature.

7. Abandoning moral motivation, again

The epistocrat has one final response to my challenge. Call the partially politically competent the group of voters who reason well, are informed about current events, possess social scientific knowledge, possess moral knowledge, but who may or may not be morally motivated. The best the epistocrat can do is identify the partially politically competent and amplify their voices. Given this, the epistocrat might say that though it is not ideal, an epistocracy with partially politically competent voters is still preferable to democracy. Essentially, this response grants that the central criticism of this paper is correct – there is a key component of political competence the epistocrat cannot identify – but says that a non-ideal epistocracy with partially competent voters nonetheless beats democracy.

It is easy to see why the epistocrat might think this. Recall that there are five ways electorates can behave incompetently: ignorance, irrationality, impairment, immorality, and corruption. In a democracy with universal suffrage, all five types of incompetence are permitted in the electorate. However, if we implement an epistocracy where only the partially politically competent are enfranchised, the ignorant, irrational, and impaired are disenfranchised. The partially competent electorate, of course, might still act in immoral or corrupt ways since this electorate may contain many voters who lack proper moral motivation. But it certainly seems better to only be confronted with two potential sources of injustice, rather than five.

In the above paragraph, the comparison between democracy and non-ideal epistocracy was made in terms of how many sources of injustice they permit. This is not the correct comparison. It does not matter how many potential sources of injustice there are. What matters is how frequently injustice occurs in both systems. Consider an analogy. There is Car A and Car B. Car A has five mechanical flaws that might potentially result in a fatal accident. Car B has two mechanical flaws that might potentially result in a fatal accident. From this fact alone we cannot conclude that Car B is superior to Car A, for it could be that Car B's two flaws occur with a far greater combined frequency than the combined frequency of Car A's five flaws. There are more ways Car A can fail, but Car B fails more often. Similarly, we should not care that democracy can produce injustice in more unique ways than non-ideal epistocracy can. Instead, we should care about how frequently democracy produces injustice when compared with non-ideal epistocracy.

Non-ideal epistocracy would produce injustice more frequently than democracy if the partially competent electorate makes more immoral and corrupt decisions than democracy makes ignorant, irrational, impaired, immoral, and corrupt decisions. If true, then democracy outperforms non-ideal epistocracy. How might this happen? There are a few possible ways.

First, it could be that voters who are partially competent tend to have especially low moral motivation, so they are especially prone to making immoral and corrupt decisions. In the voter behavior literature, we do find interesting correlations of a similar kind. For instance, more knowledgeable voters tend to be more partisan and ideological (for overviews of the literature demonstrating this, see Gibbons (2022) and Hannon (2022)). As Hannon summarizes the literature: “the most politically knowledgeable people also tend to be extremely partisan” (Hannon 2022: 31). Note, I am not saying that partisanship and polarization are akin to a lack of moral motivation. I am merely pointing out that we have found correlations between certain aspects of political competence and normatively undesirable political behavior, so it is within the realm of possibility that being partially politically competent is correlated with low levels of moral motivation. It could be that the partially politically competent are so prone to low moral motivation that the immoral and corrupt decisions non-ideal epistocracies make outpace the ignorant, irrational, impaired, immoral, and corrupt decisions democracies make.

Another possible mechanism need not assume that the partially politically competent are especially prone to low moral motivation. Suppose everyone has the same moral motivation and tends to (if enfranchised) vote in a self-interested manner. Now suppose that those who pass the suffrage exam and are thus deemed partially politically competent tend to come from only one demographic group, which Brennan says is likely to happen: “If the United States were to start using a voter qualification exam right now, such as an exam that I got to design, I'd expect that the people who pass the exam would be disproportionately white, upper-middle- to upper-class, educated, employed males” (Brennan 2016: 228). Since they tend to vote in a self-interested manner, the partially politically competent vote for candidates and policies that benefit their demographic at the expense of demographics who are disenfranchised. This, I think, would qualify as an immoral and/or corrupt electorate.¹¹ These sorts of collective decisions are less likely in a democracy because all demographic groups are enfranchised, which forces politicians to chase votes from multiple demographic groups, resulting in policy that is more (though by no means perfectly) focused on the interests of all.

In response to this argument, one might point out that democracy has its own version of this problem, because different demographic groups vote at different rates.¹² In the 2020 US election, for instance, 70.9 percent of eligible white voters cast a ballot, while only 58.4 percent of eligible non-white voters cast a ballot (Morris and Grange 2021). This disparity might result in a democracy where politicians favor the interests of white voters over the interests of non-white voters, simply because whites vote at a higher rate than non-whites. While democracy's version of this problem is troubling, it is less troubling than epistocracy's version of it. The reason why is that, in a democracy, even if a certain demographic group consistently chooses to vote at a low frequency, the group can still vote. In an epistocracy, however, a disenfranchised demographic does not have the option of voting. This difference is relevant because the mere threat that a low-voting demographic could turn out in large numbers can discipline politicians to not upset this demographic too much. Since the incompetent do not have the option of voting in an epistocracy, this threat is absent. Another way of putting it: a vote, even if not used, can act as a kind of deterrence. Low-voting groups in a democracy still wield their vote as a deterrent; the incompetent in an epistocracy have no deterrence.

In sum, it could be that the skewed demographics among those enfranchised in a non-ideal epistocracy will lead to more immoral and corrupt decisions than democracies make ignorant, irrational, impaired, immoral, and corrupt decisions.

The above mechanisms are not meant to be arguments for why non-ideal epistocracies will in fact produce unjust decisions at a greater frequency than democracies. They are possible explanations as to how this could happen. Ultimately, I do not know whether a non-ideal epistocracy would commit injustice more frequently than a democracy or vice versa. This is an empirical question, for which we currently lack sufficient evidence to come down on one side versus the other. However, it is clear that the epistocrat cannot simply claim that non-ideal epistocracy outperforms democracy because some sources of electoral incompetence are eliminated when only the partially politically competent are enfranchised. What matters is how frequently injustice occurs, not how many sources of injustice there are. We simply do not know enough about non-ideal epistocracy and partially competent electorates to render a firm judgment about how frequently such a system would commit injustice.

We do, however, know that an ideal epistocracy where only the fully politically competent are enfranchised would outperform democracy. This is because all sources of electoral incompetence are removed, meaning injustice would occur in such a system at a frequency of zero. But I have shown that such a system cannot be realized, because we cannot identify those who are fully politically competent. Epistocracy can only exist in non-ideal form, and who wins in the fight between democracy and non-ideal epistocracy is far from clear.