Samuel Bagg As above,my response to these suggestions is race to "teach to the test"could hardly be expected to thoroughly pragmatic.While conceding the conceptual yield substantial benefits in vote quality-indeed,given possibility of unobjectionable epistemic qualifications, the massive expenditures required,the most significant I highlight the practical risks of giving political leaders consequence would be to intensify parties'dependence the power to design them.Though not quite as upon concentrated wealth. dangerous as the ability to appoint successors directly, Brennan concedes that his proposals could have the ability to implement epistemic qualifications would drawbacks,including the potential for manipulation also enable incumbents to entrench their rule-and (2016.222-30).However.he asserts that they also the hypothesized benefits here are even less likely promise vastly superior outcomes.Caricaturing any un- to materialize.Thus,Brennan's epistocracy ought to willingness to experiment with epistocracy as a species op//s remain off the table as well. of Burkean conservatism.he insists that we cannot Of course,democracies already enable incumbents know in advance whether such tradeoffs will be worth- to manipulate the pool of eligible voters in various while.I disagree.As I have already argued,we know ways.Some restrictions,like those concerning age and enough about politicians to predict that some would citizenship,are likely inevitable.Others,however- use these tools to entrench their power.Perhaps more including voter identification requirements,felon dis- importantly,we also know enough about voters to enfranchisement.and the gerrymandering of single- predict that disenfranchising those with low political member districts-are routinely used to entrench the knowledge would not substantially improve the epis- power of incumbents by tilting the electoral odds in temic quality of electoral results.Thus,even if in- their favor (Levitsky and Ziblatt 2018,183-6,208-11). cumbents(or independent bodies)imposed perfectly Such policies are typically justified on other grounds, benign epistemic qualifications,the benefits Brennan and some may be valuable on balance,yet in each projects would be very unlikely to materialize. case,this feature gives us reason to be suspicious (Is- Consider that voters in modern democracies do sacharoff and Pildes 1998:Shapiro 2016.86-8).By com not choose directly between different policies.Instead parison,then,Brennan's proposals would grant politi- they choose between parties or candidates.There is cians far greater leeway in choosing their own voters. thus substantial distance between a voter's overall level and should be regarded as far more dangerous. of political knowledge and the quality of her vote Consider the possibility of conditionalizing suffrage choice.Yet Brennan makes very little effort to demon- on educational attainment.Brennan himself speculates strate that higher political knowledge scores would be (2016,223)that in the US,Democrats would favor associated with greater likelihood of voting for the very low or very high qualifications,while Republi- "correct"party or candidate.12 Instead,he relies on an cans would prefer a mid-range qualification,such as intuitive sense that political knowledge must raise vote a high school diploma.Yet he fails to recognize what quality.13 a dramatic effect this could have:for either party,im- Perhaps,on the margin,it does.As Brennan himself plementing a properly targeted policy would instantly insists throughout his book,however,our intuitive un- translate a small temporary majority into a massive and derstanding of voting behavior cannot be trusted.In durable electoral advantage.Where traditional gerry- particular,it seems,rational assessment of the argu- mandering faces natural limits,moreover-imposed by ments for competing policies-which political knowl- 。101g the need to distribute a fixed voting population among edge tests purport to measure-plays a small role (at geographically contiguous districts-"epistemic"ger- best)in vote choice.As extensive research on"mo- rymandering could continuously compound incumbent tivated reasoning"has demonstrated,better informa- advantages through ever-more-targeted tests,recur- tion rarely changes our mind,it simply gives us better sively altering the voting population itself.Even if their tools for defending our intuitions(Lodge and Taber immediate effects were not so extreme,finally,such 2013).Meanwhile,education primarily alters our policy policies would still give incumbents strong incentives preferences indirectly,by changing our social context to manipulate educational attainment patterns for par- and identity.Where they exist,then,correlations be- tisan ends. tween political knowledge and vote choice are largely Implementing a competence test presents no solu- explained by unrelated,non-rational factors-centrally tion,as incumbents would have similarly strong incen including the set of identity groups to which we be- tives (and even greater leeway)to manipulate its con- long(Achen and Bartels 2016).By far the most signifi- tent.Once established,moreover,parties would likely cant consequence of disenfranchising those individuals scramble to prepare their supporters,much as they al- with low epistemic qualifications,therefore,would be ready conduct voter registration drives and get-out- the-vote campaigns in places without automatic reg- istration and mandatory voting.The resulting arms 2 We need not assume that political questions have objectively cor- rect answers to accept that elections present citizens with options that 11 The fact that literacy tests in the US South did not generate a similar arms race reflects their peculiarly pernicious,openly partisan a smattering of broadly libertarian preferences that he presumably nature.Since white voting rights were generally"grandfathered"in, considers correct(2016,33-4).While these studies control for certain only blacks were subject in practice to the tests,which were designed demographic variables,however,other potential confounds remain to be nearly impossible.Meanwhile,those who managed to pass were (e.g.,group identities elided by crude demographic controls).More often prevented from voting by other means.Under such conditions importantly,they do not establish that having such"superior"policy mass mobilization by Republicans would have been futile. preferences actually enhances vote choice. 898Samuel Bagg As above, my response to these suggestions is thoroughly pragmatic. While conceding the conceptual possibility of unobjectionable epistemic qualifications, I highlight the practical risks of giving political leaders the power to design them. Though not quite as dangerous as the ability to appoint successors directly, the ability to implement epistemic qualifications would also enable incumbents to entrench their rule—and the hypothesized benefits here are even less likely to materialize. Thus, Brennan’s epistocracy ought to remain off the table as well. Of course, democracies already enable incumbents to manipulate the pool of eligible voters in various ways. Some restrictions, like those concerning age and citizenship, are likely inevitable. Others, however— including voter identification requirements, felon dis￾enfranchisement, and the gerrymandering of single￾member districts—are routinely used to entrench the power of incumbents by tilting the electoral odds in their favor (Levitsky and Ziblatt 2018, 183–6, 208–11). Such policies are typically justified on other grounds, and some may be valuable on balance, yet in each case, this feature gives us reason to be suspicious (Is￾sacharoff and Pildes 1998; Shapiro 2016, 86–8). By com￾parison, then, Brennan’s proposals would grant politi￾cians far greater leeway in choosing their own voters, and should be regarded as far more dangerous. Consider the possibility of conditionalizing suffrage on educational attainment. Brennan himself speculates (2016, 223) that in the US, Democrats would favor very low or very high qualifications, while Republi￾cans would prefer a mid-range qualification, such as a high school diploma. Yet he fails to recognize what a dramatic effect this could have: for either party, im￾plementing a properly targeted policy would instantly translate a small temporary majority into a massive and durable electoral advantage. Where traditional gerry￾mandering faces natural limits, moreover—imposed by the need to distribute a fixed voting population among geographically contiguous districts—“epistemic” ger￾rymandering could continuously compound incumbent advantages through ever-more-targeted tests, recur￾sively altering the voting population itself. Even if their immediate effects were not so extreme, finally, such policies would still give incumbents strong incentives to manipulate educational attainment patterns for par￾tisan ends. Implementing a competence test presents no solu￾tion, as incumbents would have similarly strong incen￾tives (and even greater leeway) to manipulate its con￾tent. Once established, moreover, parties would likely scramble to prepare their supporters, much as they al￾ready conduct voter registration drives and get-out￾the-vote campaigns in places without automatic reg￾istration and mandatory voting.11 The resulting arms 11 The fact that literacy tests in the US South did not generate a similar arms race reflects their peculiarly pernicious, openly partisan nature. Since white voting rights were generally “grandfathered” in, only blacks were subject in practice to the tests, which were designed to be nearly impossible.Meanwhile, those who managed to pass were often prevented from voting by other means. Under such conditions, mass mobilization by Republicans would have been futile. race to “teach to the test” could hardly be expected to yield substantial benefits in vote quality—indeed, given the massive expenditures required, the most significant consequence would be to intensify parties’ dependence upon concentrated wealth. Brennan concedes that his proposals could have drawbacks, including the potential for manipulation (2016, 222–30). However, he asserts that they also promise vastly superior outcomes. Caricaturing any un￾willingness to experiment with epistocracy as a species of Burkean conservatism, he insists that we cannot know in advance whether such tradeoffs will be worth￾while. I disagree. As I have already argued, we know enough about politicians to predict that some would use these tools to entrench their power. Perhaps more importantly, we also know enough about voters to predict that disenfranchising those with low political knowledge would not substantially improve the epis￾temic quality of electoral results. Thus, even if in￾cumbents (or independent bodies) imposed perfectly benign epistemic qualifications, the benefits Brennan projects would be very unlikely to materialize. Consider that voters in modern democracies do not choose directly between different policies. Instead, they choose between parties or candidates. There is thus substantial distance between a voter’s overall level of political knowledge and the quality of her vote choice. Yet Brennan makes very little effort to demon￾strate that higher political knowledge scores would be associated with greater likelihood of voting for the “correct” party or candidate.12 Instead, he relies on an intuitive sense that political knowledge must raise vote quality.13 Perhaps, on the margin, it does. As Brennan himself insists throughout his book, however, our intuitive un￾derstanding of voting behavior cannot be trusted. In particular, it seems, rational assessment of the argu￾ments for competing policies—which political knowl￾edge tests purport to measure—plays a small role (at best) in vote choice. As extensive research on “mo￾tivated reasoning” has demonstrated, better informa￾tion rarely changes our mind, it simply gives us better tools for defending our intuitions (Lodge and Taber 2013).Meanwhile, education primarily alters our policy preferences indirectly, by changing our social context and identity. Where they exist, then, correlations be￾tween political knowledge and vote choice are largely explained by unrelated, non-rational factors—centrally including the set of identity groups to which we be￾long (Achen and Bartels 2016). By far the most signifi￾cant consequence of disenfranchising those individuals with low epistemic qualifications, therefore, would be 12 We need not assume that political questions have objectively cor￾rect answers to accept that elections present citizens with options that are, in some sense, better than others. 13 Brennan does cite two studies correlating high information with a smattering of broadly libertarian preferences that he presumably considers correct (2016, 33–4).While these studies control for certain demographic variables, however, other potential confounds remain (e.g., group identities elided by crude demographic controls). More importantly, they do not establish that having such “superior” policy preferences actually enhances vote choice. 898 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:04, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000527