Should dumb people be allowed to vote? That’s the provocative question posed by Jason Brennan, an associate professor of strategy, economics, ethics, and public policy at Georgetown University and the author of the recently released book Against Democracy, in an essay published last week on Aeon and on Quartz.
Brennan thinks there’s a solid case to be made that no, they shouldn’t, and that we should claw back the universal franchise:
Consider an alternative political system called epistocracy. Epistocracies retain the same institutions as representative democracies, including imposing liberal constitutional limits on power, bills of rights, checks and balances, elected representatives and judicial review. But while democracies give every citizen an equal right to vote, epistocracies apportion political power, by law, according to knowledge or competence.
The idea here is not that knowledgeable people deserve to rule – of course they don’t – but that the rest of us deserve not to be subjected to incompetently made political decisions. Political decisions are high stakes, and democracies entrust some of these high-stakes decisions to the ignorant and incompetent. Democracies tend to pass laws and policies that appeal to the median voter, yet the median voter would fail Econ, History, Sociology, and Poli Sci 101. Empirical work generally shows that voters would support different policies if they were better informed.
Brennan, who is also the author of Libertarianism: What Everyone Needs to Know, isn’t quite calling for an IQ test — more a knowledge-based test in which only those who can answer a high-enough proportion of American citizenship test–type questions are granted voting rights. Overall, he seems very open to the idea that we could do better than democracy: At the very least, he writes, “which system, warts and all, would work best” is an “interesting question.”
The potentially dire problems with such a proposal are fairly straightforward, and they go beyond gut-level objections to discarding a bedrock principle of the United States’ identity as a country. Perhaps most obviously, there’s a very good chance that an epistocratic system would vastly increase the power wealthy people have in society, even relative to the already-quite-unequal system we currently have. Whatever cutoff point you set for You Must Have This Much Knowledge to Ride the Epistocracy, it would in all likelihood be strikingly easy for rich people to meet that threshold, simply because of how money and privilege and education work, and the vast majority of the people who couldn’t get past the sign would probably be poor — and therefore disproportionately nonwhite as well. In the past, of course, knowledge tests and their ilk have been used specifically to exclude nonwhite voters and, in some cases, white ones who were effectively deemed too poor to be entrusted with the responsibility to vote. It’s hard not to hear echoes of these (supposedly!) regretted-by-everyone past practices in Brennan’s proposal, especially given that his proposal would likely disproportionately disenfranchise the same groups.
But setting aside these important issues, overall, Brennan is engaging in a very human form of wishful thinking, which is to bemoan the fact that If only people were more knowledgeable, they’d have better politics. Everyone has a different view of “better politics,” but this is a fairly common sentiment. To support his particular version of this argument, though, Brennan distorts some political science. Let’s go back to his sentence “Empirical work generally shows that voters would support different policies if they were better informed.” (It’s pretty clear, from context, that Brennan is being coy here — he means better policies, not merely different ones, or else how is that sentence supporting the idea that we should flirt with epistocracy?) I emailed Brennan to ask him which research he was referring to, and he sent me a bunch of citations, as well as this paragraph from Against Democracy:
Political scientist Scott Althaus also finds, using the American National Election Studies data, that well informed and badly informed citizens have systematically different policy preferences. Althaus finds that poorly informed people have systematically different preferences from well-informed people, even after we correct for the influence of demographic factors, such as race, income, and gender. As people (regardless of their race, income, gender, or other demographic factors) become more informed, they favor overall less government intervention and control of the economy. (That’s not [to] say they become libertarians.) They are more in favor of free trade and less in favor of protectionism. They are more pro-choice. They favor using tax increases to offset the deficit and debt. They favor less punitive and harsh measures on crime. They are less hawkish on military policy, though they favor other forms of intervention. They are more accepting of affirmative action. They are less supportive of prayer in public schools. They are more supportive of market solutions to health care problems. They are less moralistic in law; they don’t want government to impose morality on the population. And so on. In contrast, as people become less informed, they become more in favor of protectionism, abortion restrictions, harsh penalties on crime, doing nothing to fix the debt, more hawkish intervention, and so on.
So Althaus, a professor of political science and communication at the University of Illinois at Urbana-Champaign, looms large in Brennan’s research. One of the sources Brennan pointed me to was page 129 of Althaus’s 2003 book Collective Preferences in Democratic Politics: Opinion Surveys and the Will of the People. If you go there, you’ll see that the page, which is in a chapter called “The Impact of Information Effects,” is given over to the following table:
The table is based on statistical simulations of what a more informed populace would prefer, and at first glance it certainly seems to support the paragraph from Brennan’s book. But according to Althaus himself, Brennan is seriously missing the point of his research, and ignoring several pages of caveats at the end of that very chapter cautioning against exactly the sort of reasoning Brennan is using to prop up his argument. “My argument in the book about the value (as well as the limits) of the simulation method runs entirely counter to Brennan’s proposal,” Althaus said in an email. “There is no statistical magic that can tell us what the American public would really want if it only knew better. These simulations do not and cannot tell us what the American people really want.”
Understanding why Brennan’s argument misfires involves understanding exactly what these simulations are. Althaus sent me a PDF of the chapter in question, which goes into detail. The simulations rely on the fact that certain surveys which gauge people’s political beliefs and preferences also gauge their current level of political knowledge, allowing researchers to draw correlations between the two. The basic conceit is that if you do the right type of number-crunching, you can take a sample of voters, all with different beliefs, demographic characteristics, levels of knowledge, and so on, and simulate a situation in which you wave a hypothetical magical wand which endows them all with the maximum amount of political knowledge it is possible to possess in the situation in question — while holding all their personal and demographic characteristics constant — and see whether and in what ways this new knowledge would theoretically affect their beliefs.
Althaus writes about a series of such simulations which employs data from American National Election Studies surveys. In each, he takes the actual preferences of voters, does that statistical trick to give them all what he calls “fully informed opinion” — again, that’s imagining that these voters are exactly the same in terms of gender, race, class, and so on, jacking up their level of knowledge to the ANES equivalent of 100 percent, and seeing how the model predicts their beliefs would change as a result — and reporting the results.
Some of those results are interesting: For instance, if you take a bunch of ANES-respondent men and gift them with full political information, it increases their support for full, unrestricted abortion rights just a little. For women, on the other hand, the so-called “information effect” is fairly huge — the percentage of women who support full abortion rights jumps from 46.4 percent in the real-life ANES sample to 61.9 percent when they are gifted, via simulation, with full knowledge. Elsewhere in the chapter, Althaus reports results showing that, relative to real voters, simulated fully informed voters are more likely to prefer free-market solutions to economic problems as compared to strong-government ones, for instance, and that they are more willing to raise taxes to pay down the deficit. As is true with the abortion results, many of these simulations contain important subtleties and nuances: Sometimes there are really big differentials in the size of the information effects along class or partisan lines, for example, and sometimes ceiling effects constrain the size of the information effect — that is, if 75 percent of a sample already believes something, in other words, information can only have so much of an effect.
Setting aside those nuances, some patterns do emerge from the simulation, and from them Brennan believes we can infer two: (1) these are the policies a more informed populace would prefer; and (2) these policies would be better. He states (1) outright in his Aeon/Quartz article, and he strongly implies (2) by offering these findings as support for an argument that universal voting rights should be reconsidered. (Again, “A higher degree of knowledge is correlated with different policies” couldn’t possibly be an argument against universal voting rights, but “A higher degree of knowledge is correlated with better policies” could be, if you’re into that.)
But if you read just nine pages past the above table, Althaus anticipates both these claims, and unequivocally cautions against making either. He lays out all sorts of different reasons why the simulations should not be used in such a straightforwardly predictive real-world manner. For one thing, models like the ones being used in the simulations have some really serious limitations. Take the hypothetical example of a low-income out-of-work steelworker who has one set of beliefs that he reports during an ANES survey, and that the model report will have those beliefs changed in X or Y direction once the worker is granted “full” political knowledge. In the model, everything about him is the same except for the knowledge that was magically shot into his skull. In real life, we know that the process of obtaining and integrating knowledge is really complicated, shot through with all sorts of sociological and other influences, and hard to fully understand. The model is an approximation that simply might not apply to real life. For a real-life out-of-work steelworker who suddenly bones up on politics, that new knowledge might filter through him ways the model can’t possibly anticipate.
But we don’t even need to speculate, because according to Althaus there’s already evidence these simulations don’t map onto real life. In another excerpt from his book that he sent me, Althaus notes that “Simulated measures of fully informed opinion accurately predict when collective policy preferences will remain stable but are decidedly less precise when forecasting that surveyed opinion should change in significant ways, at least in the short term.” So Brennan is simply assuming that the results of the simulation mean it’s easy to predict in which direction a more knowledgeable voter base would go, policywise, but there’s no solid reason to believe this is the case.
Finally, and perhaps most obviously, why should we grant such profound weight to what the most enlightened simulated voters want? In real life, enlightened voters can be wrong about stuff — are wrong about stuff — all the time! Tippy-top well-educated elites are afflicted by groupthink and dogma just like everyone else, and are perhaps more susceptible in some cases precisely because they are elites: How could someone as smart and well-educated as they are be wrong about, say, the housing bubble? As Althaus puts it in his book, “While statistically modeling a seemingly more informed citizenry can provide important insights into the relationship between opinions and political knowledge, the results of such an exercise should be accorded no greater normative status than the surveyed opinions from which they are derived.” In other words: Don’t take the preferences of these simulated enlightened voters and assume that, if they were enacted, the world would be better. Which is certainly what Brennan seems to be doing.