What, If Anything, Is Wrong with Gerrymandering

What, If Anything, Is Wrong with Gerrymandering? In contrast to other pathologies of American democracy, gerrymandering has few public advocates. Few openly contend that it is good thing, or even that measures to curb it would be bad. By contrast, many contend that unfettered, private spending on campaigns is a good thing. It informs voters, and measures to curb it would violate freedom of political speech. And many contend, even if disingenuously, that voter ID requirements are a good thing. They prevent fraud. The closest anyone comes to openly arguing in favor of gerrymandering is that there’s no workable legal test of whether districts have been gerrymandered, or that gerrymandering does not violate any constitutional guaranteed right.1 This defensive posture reflects a “visceral reaction” (Alexander and Prakash 2008) that gerrymandering is objectionable: unfair, undemocratic, or at any rate fishy. This paper asks what this objection, of political morality, rather than law, might be. Let me begin by defining some terms. Gerrymandering is a problem only in district systems, where, as I’ll call it, a superdistrict’s voters are sorted into districts. Typically, the districts are associated with territories, with a district’s voters being just those who primarily reside in the territory. Each district independently elects by majority (to fix ideas) a delegation of district representatives to the superdistrict assembly. The superdistrict assembly then makes decisions by majority (again to fix ideas) for the superdistrict as a whole. District systems tend to divide representatives between two major parties. 1 See, in particular, Alexander and Prakash 2008. Many of their rebuttals of alleged constitutional objections, however, go deeper than that, and are likewise rebuttals of alleged moral objections. Indeed, many of the points in this paper simply repeat points made in theirs. 1 Some district systems do not give each voter equal voting power (more later on how this is to be understood) across the superdistrict, either because some voters have greater voting power within a district, or because the share of the voting power of a district’s delegation to the assembly is not proportional to the district’s share of population in the superdistrict. In the United States, the Senate and, albeit to a lesser degree, the House and the Electoral College, give some districts—that is, states—greater voting power. However, since the reapportionment cases of the 1960s, it has been harder to bring this objection against congressional districts within a given state. Those cases found, roughly, a constitutional requirement that state’s congressional districts have equal populations, against a background assumption that each congressional representative has the same voting power as any other. A district system that gives every citizen equal voting power, however, can still suffer from gerrymandering. Let us provisionally define “gerrymandering” as drawing districts with the intent, based on a belief about how people are likely to vote, to bring about certain electoral outcomes. One way to gerrymander is to “pack” the other party’s voters in a smaller number of districts, thereby giving one’s party smaller, but still reliable, majorities in a greater number of other districts. Another way to gerrymander is to “crack” the other party’s voters evenly across all of the districts, thereby ensuring that they constitute a majority in none. This definition of gerrymandering may not cut, as it were, at the moral joints. In particular, intent may not matter morally. Even without any intent to bring about electoral outcomes, compactness and contiguity criteria, along with the fact that Democratic voters are more likely to reside in more densely populated areas, say, may naturally “pack” Democrats into fewer districts with the same result as an intentional gerrymander. Some scholars suggest that such “clustering” or “natural gerrymanders” are no less objectionable (Wang 2016, Beitz 2018). 2 Gerrymanders can aim for different kinds of “electoral outcomes.” (1) District representation: Who represents the district in the assembly. (2) Group representation: How many representatives of a given party or group are in the assembly. (3) Party control: Which party controls the assembly. (4) Policy: What laws, policies, etc. the assembly enacts. Gerrymanders come in three main varieties, distinguished in part by the electoral outcomes they seek. (1) Racial gerrymanders obstruct outcomes favored by a racial minority. Remedies typically aim to increase group representation by candidates from or favored by the racial minority. (2) Incumbent-protection gerrymanders aim for district representation by the incumbent. (3) Partisan gerrymanders aim for party control (or at least greater group representation by a particular party) as a means to policy. 1. Results So what is wrong with gerrymandering? One answer is that gerrymandering has bad results. Results are “better,” I’ll say, when there is a fairer distribution of the satisfaction of “substantive interests,” defined negatively as interests neither in the correspondence of political outcomes with preference as such, nor in the influence of choices over political outcomes as such. To be sure, I think that partisan gerrymandering, in particular, has led to worse results for most of the current decade. But to a great extent I think this for “partisan” reasons. It has led to Republican control of Congress. And Republican control of the levers of power, at least since the departure of big-state progressives like Nixon, has led to significantly worse outcomes than Democratic control would have had. Compounding this, the creation of safe Republican seats has led to worse representatives. With safe seats, the only real electoral pressure is the Republican primary, and Republican primary voters support worse policies than Republicans, let alone the electorate, as a whole. 3 If I left it at this, my objection to gerrymandering would be, perhaps uncomfortably, close to the recent defense of gerrymandering by North Carolina Representative David Lewis: “I think electing Republicans is better than electing Democrats. So I drew this map to help foster what I think is better for the country.”2 Moreover, in both popular and academic discourse, the objections to gerrymandering tend to be at least superficially nonpartisan. What I wrote in the previous paragraph is not the sort of thing that one is supposed to say in a law or philosophy journal. True, one can take a broader view, and argue that gerrymandering leads to “systemically” worse results, which don’t depend on the policies of the gerrymandering party. For instance, it might be said that gerrymandering has the effect of spacing elections further apart, which has worse results over the long run than elections spaced closer together. Still, as common as arguments of this kind are—that some putatively “undemocratic” feature of institutions leads to systemically worse results in the long run—I find them exceedingly hard to evaluate. It’s not just that the data are so scanty. It’s hard to know how even to frame the hypotheses. How generally is the feature specified? Which alternatives count as relevant? Which other conditions are held fixed? How long is the “long run”? At any rate, I am going to bracket whether gerrymandering has bad systemic results, and ask whether any further objection to gerrymandering remains. Many people seem to think that a further objection does remain. First, when people say why they object to gerrymandering, they often don’t say anything about results, but instead contend, for instance, that gerrymandering wastes or dilutes votes, which is wrong, or anyway a bad thing, in itself. 2 As quoted in the District Court Memorandum Opinion in Rucho v. Common Cause. 4 Second, even when people don’t say why they object to gerrymandering, the immediacy, certainty, and heat of their objections suggests that they do not rest on evaluation of qualified, empirical predictions about systemic effects. It seems hard to explain the outcry against gerrymandering as simply reflecting a social-scientific hypothesis that fewer years should pass between Congressional elections for optimal long-term results. Imagine the Constitution gave congressional representatives, like senators, six-year terms, and political scientists were advertising the benefits of shortening them to two years. Would the debate have the same temperature? Finally, when people object to racial or partisan gerrymandering, their objection often seems to be partial: on behalf of only part of the electorate, such as black or Democratic voters.3 But systemic defects, such as that elections are sub-optimally spaced, don’t seem partial in this way. If anyone has an objection to them, everyone does. 2. Preferences What, then, is the objection to gerrymandering? Perhaps that it makes policy insufficiently responsive to the superdistrict. Policy is more responsive to the superdistrict at a time, let us say, simply to the extent that, on each given policy question, the policy answer preferred by a majority of the superdistrict at that time obtains. Gerrymandering seems most to threaten responsiveness when it violates Majority Proportionality: that is, when a party with a minority of support in the superdistrict nonetheless 3 For example, it has been suggested that partisan gerrymandering violates freedom of speech, since officials impose burdens on members of the disfavored party for their prior support of it (Kennedy in Vieth). This argument presupposes that gerrymandering imposes special burdens on the disfavored party. This is also why I don’t consider this free speech objection independently in the text: namely, that it presupposes that gerrymandering imposes some special burden. As we will see, it’s hard enough to identify what this special burden is. And once we have identified what it is, we have already identified an objection to gerrymandering. 5 enjoys party control. This already suggests one limitation of this appeal to responsive policy. While partisan gerrymandering may lead to violations of Majority Proportionality, there is less reason to expect that racial and incumbent gerrymanders do. Why does responsive policy matter, setting aside faith that it leads to better results? The most straightforward answer, it seems, rests on two premises. First, each of us has an interest in correspondence: in the satisfaction of our policy preferences. Second, insofar as policy is responsive, the satisfaction of this interest is fairly distributed. I don’t find either premise convincing, for reasons I try to explain in Kolodny 2014a.4 I just gesture at a few reasons here. Regarding the first premise, we have an interest, to be sure, that our substantive interests be satisfied. But the policies that we prefer are not always reliable guides to the policies that are in our substantive interests. Regarding the second premise, why should we suppose that insofar as policy is responsive with respect to each policy question, the satisfaction of these interests in the satisfaction of policy preferences will be fairly distributed among people? If one has prioritarian leanings, then a fair distribution would be something like maximizing the satisfaction of the policy preferences of those with the least satisfaction over time—something like Rawls’s “difference principle,” as applied to the satisfaction of policy preferences. To simplify, suppose that one’s policy preferences are satisfied to the extent that one’s own party, in a two-party district system, holds a greater share of legislative seats. Then, assuming that there is at least one voter in each party, a 50-50 split in the legislature would maximize minimum preference satisfaction. A shift to 51-49, by contrast, would reduce the minimum from 50 to 49. Note that this is true even if 99% of voters support one party and 1% of 4 However, I doubt that this argument has anything to fear from social choice theory. It doesn’t need a method for aggregating preferences, only a principle of distributive justice telling us how to fairly trade off the satisfaction of one interest against another. 6 voters support the other. Rather than indicate some problem with partisan gerrymandering, therefore, the appeal to interests in preference satisfaction seems to argue for continual partisan gerrymandering in favor of the minority party. A further difficulty arises if we say that not only the satisfaction of preferences for policy, but also the satisfaction of preferences for district representation matter. Then Democratic voters “packed” into a district are more likely to see their preference that their district have a Democratic representative satisfied, whereas Democratic voters “cracked” into districts are less likely to see their preference satisfied. It seems odd, however, to think that “packed” voters lack a complaint that “cracked” voters have. 3. Equal influence Setting aside results and preferences, one might argue that gerrymandering distributes influence over political decisions unequally. Once again too briefly, there seem two main reasons to care about equal influence: or, more fully, equal opportunity for informed, autonomous influence. There’s the activity that such opportunity makes possible: the activity of bringing one’s convictions to bear on public policy. Call this the “Activity Argument” (Kolodny 2014a). And then there’s the objectionable relation that one stands in to others when they have greater opportunity for political influence than one has. Call this the “Status Argument” (Kolodny 2014b). I won’t try to make either argument here.5 What I want to emphasize is simply how hard it is to identify an objection to gerrymandering, even if we grant these arguments. When is one’s opportunity for influence as a voter, as opposed to an official or persuader of other voters, equal (provided that one has equal access to relevant information)? Bloodless though it sounds, I think that one’s opportunity for influence as a voter over an outcome should 5 For a sense of the challenges such arguments face, see Arneson 2003 and Brennan 2016. 7 be measured as the a priori chance that one’s vote is decisive over the outcome, where one’s vote is decisive just when had one’s vote been different, the outcome would have been different. By one’s “a priori chances,” I mean one’s chances assuming that every pattern of other votes is just as likely as any other pattern. You and I enjoy “a priori equality” with respect to an outcome just when my vote and your vote would have equal chances of being decisive over the outcome, assuming that no pattern of other votes is more or less likely than any other pattern. A priori equality is violated by mal-apportionment, plural voting, and less-than-universal suffrage. But a priori equality isn’t violated by gerrymandering. So if a priori equality is the right measure of equal influence, gerrymandering does not upset equal influence. Why not say instead that the right measure of equal influence requires, instead, equal actual decisiveness? First, it’s not clear why it should matter on either the Activity or Status arguments. It is not as though we participate in politics only insofar as we are decisive, or as though we are no longer equal citizens just when elections are decided by a single vote. Second, it is hard to accept that there is reason to equalize, for each of us, something, such as decisiveness, that depends on how the rest of us actually vote. Suppose the yeas are 2, the nays 1, abstracting from my vote. If I vote yea, decisiveness will be equal. If I abstain, decisiveness will be unequal. It seems implausible that that is a reason for me to vote yea. What we have reason to provide equally, it seems, is opportunity for influence, not what actually results from the exercise of that opportunity, given how others exercise their like opportunities. (Compare equal opportunity for employment. Whether or not you and I enjoy equal opportunity for a job should not depend on whether either of us actually applies for it, even though your applying for it may lower my chances of getting it.) 8 Finally, it’s hard to see how the objection to gerrymandering could be that it distributes actual decisiveness unequally. For actual decisiveness is almost always equal, except in rare cases in which the outcome turns on a single vote. One way to minimize inequalities in actual decisiveness, then, would be to minimize outcomes decided by a single vote. It might be said that this is done by making districts less “competitive”: that is, reducing the difference in vote share between winner and second-place finisher. But this is precisely what “packing” generally does. Indeed, a frequent complaint about gerrymandering is precisely that it makes districts less competitive. One could say that the objection is only to “cracking,” which makes districts more competitive. But again it seems odd that “cracked” voters should have an objection that “packed” voters lack. Why not say that what should be equalized is not actual decisiveness, but instead ex ante decisiveness: the epistemic probability of being decisive, given what the relevant evidence suggests about how others will vote? Setting aside the new difficulty of saying whose evidence (Nate Cohn’s? Karl Rove’s?) when (a week before the election? the eve of the election?) determines whether you have equal influence, it’s not clear how the shift to the ex ante perspective helps. So long as the epistemic probability converges to the long run frequency, precisely the same institutions satisfy the imperatives: “Equalize ex ante decisiveness!” and “Equalize actual decisiveness over time!” As an alternative measure of equal influence, one might propose equal actual “success,” where one is successful when one’s preferences not only are realized but also contribute to their own realization.6 And, indeed, people generally care more about being successful than they care about being decisive: more about whether what they preferred prevailed than about whether it 6 I assume that it is the success of one’s preference, not one’s vote, that matters. People don’t have some standing reason to change their votes just to be on the winning side. 9 prevailed by a single vote. But even so, an interest in success adds little at this point. Assuming that people vote, the imperative “Equalize success!” recommends the same as the already considered imperative “Equalize preference satisfaction!” Might there be something else, other than decisiveness or success, equality of which gerrymandering objectionably disturbs? The “efficiency gap” measure of partisan gerrymandering suggests the answer: the equality of “non-wasted” votes (McGhee and Stephanopoulos 2015). The number of non-wasted votes in a set of elections is the sum of the minimum number of votes, in each election, required to elect the successful candidate. Note, however, that this is not equalization among individuals, but instead among parties. Indeed, if my vote is successful but not decisive, then there’s no fact of the matter whether it was or was not “wasted”: whether or not my vote belongs to “the” set of votes that was just enough to elect the candidate. The reason to care about the efficiency gap, if there is one, has to do with neutrality among parties, which we will consider later. 4. Proportional representation The objection to gerrymandering, it is often said, is that it upsets “proportional representation.” “Proportional representation,” however, can mean quite different things. (1) “Multi-party proportional representation,” as we might call it, requires that a party’s representation in the assembly be proportional to votes cast for that party’s representatives across the superdistrict. This conception is perhaps best realized by party list systems. (2) “Nonpartisan proportional representation” requires that each representative be associated with a distinct, equinumerous set of constituents who voted for that representative. This conception is perhaps best achieved by single transferrable vote. (3) “Two-party proportional representation” is more closely 10 approximated as the division of legislators between the two major parties more closely approximates the division of support for those two major parties in the electorate. Proportional representation is often invoked in defense of gerrymandering, in the following way: (1) Gerrymandering is objectionable only insofar as it departs from proportional representation. (2) So gerrymandering is objectionable only if proportional representation is required. (3) If a district system is permissible, then proportional representation is not required (since any district system is bound to depart from proportional representation). (4) So, if a district system is permissible, gerrymandering is not objectionable (McGann et al. 2016 42–43). However, one might accept (1) but reject (2). One might grant that proportional systems are not required, since the best district system is no worse than the best proportional system. Still, one might argue, district systems that are more two-party proportional are better than district systems that are less. The objection to gerrymandering is precisely that it makes district systems less two-party proportional than necessary to achieve the benefits of a district system. I assume that it is two-party proportionality that is at issue here, since that is the kind of proportionality that limiting gerrymandering might be expected to increase.7 As Beitz 2018 argues: “the effect of partisan gerrymandering is to impose an unjustifiably large share of the costs of [a district system] on those whom it disadvantages.” 7 The benefits might be that, since district systems tend toward only two major parties, they compel the formation of coalitions before, rather than after, elections, which makes control over the legislature after elections smoother. Or the benefits might be that district systems make representatives more closely bound to specific, geographical constituencies. 11 The question is then what is good about two-party proportional representation, apart from better results or responsiveness.8 What are the “costs” to which Beitz refers? He suggests that “voters are less likely… to have a representative in the legislature whose political commitments track their own reasonably closely” (344). This brings us back to preferences, however what matters now is not the satisfaction of one’s preferences for policy, but instead minimizing the “distance” between one’s policy preferences and the policy preferences of the closest representative in the legislature. Suppose that the A party, with 51% of the vote, controls 60 seats, whereas the B party, with 49% of the vote, controls 40 seats. Beitz’s suggestion seems to be that if instead the split was more two-party proportional, say 51 seats to 49, the maximal distance any voter must suffer would be reduced. Setting aside why minimizing the maximum distance should matter, why think that making the split more two-party proportional would minimize the maximum distance? Perhaps we are to suppose (i) that no A-representative is closer to any B voter than some B representative (and vice-versa), (ii) that the greater the number of B representatives, the greater the range of Bvoter preferences represented by someone in the legislature (and vice-versa), and (iii) the spread of B-voter preferences is just as wide as the spread of A-voter preferences. For example, there might be 200 evenly spaced points in A-space each occupied by the same number of A voters (and vice-versa). Each successive A representative elected to the legislature then occupies the midpoint of the widest gap between any two adjacent A voters. Thus, the first A representative is at point 100, the second at 50, the third at 150, the fourth at 25, and so on. 8 It is sometimes suggested, as a kind of reductio, that if representation by party must be proportional, then, absurdly, representation by every affiliation must also be proportional. But it doesn’t seem arbitrary to me to single out party affiliations, since parties are expressly organized to advance a policy platform by electing representatives. 12 Assuming all this, it is then true that moving from a 60-40 to 51-49 split would reduce the maximum distance between any voter and the closest representative. But this has nothing to do with two-party proportionality. It simply reflects the fact that, under the assumptions, whatever the distribution of the votes, the maximum distance is minimized by approaching a 50-50 split. If the A party had 60% of the vote, moving from a 60-40 to a far less two-party proportional 5149 split would still reduce the maximum distance. Perhaps, then, the idea is that even if the spread of possible B-voter preferences is as wide as the spread of possible A-voter preferences, the ratio of the spread of actual A-party preferences held by some voter to the spread of actual B-party preferences is proportional to ratio of votes for the A party to votes for the B party? But why think that this is generally true? Suppose that there are 10,000 distinct points in B-voter space. One might expect that with, say, 40,000 B voters, at least one voter actually holds each possible B-voter preference. That’s enough to argue for a 50-50 split, even if there are 60,000 A voters.9 5. Majority rule We noted earlier that partisan gerrymandering, if not racial or incumbent gerrymandering, threatens Majority Proportionality. Granted, district systems cannot guarantee Majority Proportionality. But given a district system, one might complain that gerrymandering makes Majority Proportionality less likely than necessary. But why accept Majority Proportionality? One argument is just by analogy. 9 It is sometimes suggested that voting rules should also respect “quasi-anonymity”: that two profiles of votes should deliver the same outcome if we swap the party support of any two voters. Quasi-anonymity rules out district systems. For example, if we swap the opposing votes of a voter, Ty, in a tied district and a voter, Lopseid, in a far from tied district, then we change the outcome (Christiano 1996: 234; Still 1981: 382). However, quasi-anonymity does not support, in any clear way, proportionality, rather than say an at-large but winner-take-all superdistrict. Moreover, it’s unclear why quasi-anonymity matters. It’s not clear what Lopseid’s objection is if not to not being actually decisive, when Ty is. 13 (1) Members of the majority have an objection about departures from majority rule by district voters over district representation. (2) Members of the majority have an objection about departures from majority rule by representatives over policy. (3) Departures from majority rule by superdistrict voters over party control are sufficiently like either of these other departures. (4) Therefore, members of the majority have an objection about departures from majority rule by superdistrict voters over party control. As a psychological explanation of why many protest against partisan gerrymandering, this has a certain verisimilitude. People find it objectionable just because it’s similar to violations of majority rule that people find objectionable. But how does it fare as a normative explanation of why, if at all, partisan gerrymandering is, on reflection, objectionable? Is there a more principled case for majority rule? Some might argue for majority rule on the grounds that it leads to better results or to the satisfaction of preferences. But we have already considered those possibilities. Is there some more “intrinsic” democratic reason in favor of majority rule? The standard answer is that majority rule is entailed by “neutrality” between outcomes. Indeed, the nub of complaints about “partisan asymmetry,” the “efficiency gap,” and “mean-median difference” (Wang 2016) seems to be neutrality between parties, which seems a special case of neutrality between outcomes, where the outcomes are party representation. But why accept neutrality between outcomes? The Activity and Status Arguments may require equal influence among people. But that doesn’t require neutrality between outcomes. For simplicity, consider supermajority requirements on referenda. Granted, these favor the status 14 quo: keeping policy over changing it. But they don’t distribute opportunity for influence unequally among people. Each person has the same opportunity as has any other person to influence change—even if the (equally enjoyed) opportunity to influence change is less than the (equally enjoyed) opportunity to influence stasis. Presumably, what matters is that people have equal opportunity to influence decisions, not that decisions have equal opportunity of being made. Now, one might reply: “Supermajority requirements do not give people equal opportunity for influence. For, taking as given which outcome which people actually prefer, supermajority requirements give those who prefer stasis greater opportunity to influence the outcome that they prefer than they give others to influence the outcome that they prefer.” In the Status Argument, however, the reply finds no foothold. Consider activities that have value for someone only insofar as they flow from her choices or attitudes: such as career or marriage. Where such activities are at stake, there are at least the makings of a complaint, on the part of person A, that it is harder for A to satisfy that interest given A’s actual attitudes than it is for B to satisfy that interest given B’s actual attitudes. According to the Status Argument, however, what ultimately matters is not some activity that one pursues by exercising an opportunity—such as some activity of civic participation that one engages in when one casts a vote—but instead a status one has simply by having an (equal) opportunity. In the Activity Argument, by contrast, there are the makings of such a complaint: that supermajority requirements give B less opportunity to satisfy B’s interest in political activity, given B’s preference for change, than they give A to satisfy A’s interest in political activity, given A’s preference for stasis. It’s not clear to me that supermajority requirements do this. But even if they do, this complaint strikes me as misplaced. When evaluating the distribution of 15 opportunity, it’s not clear why we shouldn’t treat B as “free,” as Rawls 1975 puts it, and not gauge the extent and quality of B’s opportunities relative to the attitudes that B actually has. But, one might protest, partisan gerrymandering that violates Majority Proportionality does treat people unequally; it favors (say) Republicans over Democrats. For example, Democrats may need 60% of the vote to gain control over the assembly whereas Republicans need only 40%. But this isn’t so. Partisan gerrymandering doesn’t exploit any rule that says that Democrats need 60% of the vote to gain partisan control over the assembly whereas Republicans need only 40%. Instead, it exploits control over the districting process gained by winning earlier elections. Had Democrats won those earlier elections, they would now need only 40% of the vote. The bias is a kind of lagged advantage to the winners of earlier elections. And that just looks like favoring the status quo. 6. Incumbent-Protection Gerrymanders: Improper Representation We have asked so far whether the objection to gerrymandering is to how it distributes influence among voters. Might the objection be instead to how it distributes influence between the official in charge of districting, Gerry, and ordinary citizens? This would resonate with the complaint that “voters should choose their representatives, not the other way around” (Berman 2005). But two doubts arise straightaway. First, there would presumably be no objection to Gerry’s greater influence if he had used it to district without gerrymandering. Second, I suspect that many gerrymanders that seem objectionable when performed by officials would still seem objectionable if (cutting out the middle-man) they were effected by direct plebiscite. Still, if one is attracted to the idea of equal influence, whether from the Activity or Status arguments, there might seem to be a general objection to representative government in general: to officials having special powers of any kind, whether to district or otherwise. After all, this 16 gives them greater opportunity for influence than ordinary citizens have. One response, admittedly vague, is that such greater opportunity is acceptable, when and because the official stands in relations of “proper representation” to their constituents: when and because the official stands as an “agent” to the citizenry as collective “principal” (Kolodny 2014b). This reply suggests, in turn, another possible objection to gerrymandering: namely, that it disrupts relations of proper representation between official and constituents. As a first way of pressing the objection, it might be said: (1) Gerrymandering leads to less competitive districts (2) In less competitive districts, the incumbent’s responsiveness to her district—her acting so as to satisfy a majority of the district’s preferences (contrast: policy being responsive to the superdistrict)—does less to improve her chances of reelection. (3) So in less competitive districts, the incumbent will be less responsive to her district. (4) Proper representation requires that representatives be responsive to their district. (5) So, gerrymandering leads to violations of standards of proper representation. Among other doubts, this suggests, counterintuitively, that strict term limits would be as objectionable as gerrymandering. After all, term limits likewise make the incumbent’s responsiveness powerless to raise her chances of reelection (which are stuck at zero no matter what she does). Moreover, if the objection is to being represented by a less responsive representative, then the objection would be one that, oddly, all voters in a less competitive, “packed” district would have to a greater degree than all voters in a more competitive, “cracked” 17 district.10 In any event, I don’t think that proper representation does require that representatives be responsive in this sense. A second way of pressing the objection is: (1) Gerrymandering makes districts’ elections of their representatives less frequent or less direct. (2) Proper representation requires that districts’ elections of their representatives be sufficiently frequent and direct. (3) So, gerrymandering leads to violations of standards of proper representation. It’s not clear that partisan or racial gerrymandering does this. True, some claim that partisan gerrymandering of congressional districts replaces the direct election of representatives every two years with an indirect election, via the election of the state government that controls districting, every ten years. But what partisan gerrymandering seems more clearly to do is instead to make the party composition of the state’s congressional delegation indirectly determined every ten years. Each district’s election of its representative still occurs directly, every two years. Indeed, partisan gerrymandering might be superimposed on term limits. There is more plausibility, however, to the claim that incumbent-protection gerrymandering makes the district’s election of its representative insufficiently frequent for the standards of proper representation to be met. Consider an extreme analogy: once elected by a majority, a representative can unilaterally change the vote threshold for reelection to be safely below a poll of likely support. This seems close to electing representatives for terms that extend as long as the representative chooses. This does seem like an “intrinsic,” democratic objection, Alternatively, if the objection is that having fewer competitive districts has bad systemic effects, then that is an objection that everyone has (Beitz 2018 351). But we have already discussed systemic effects. 10 18 at least to incumbent-protection gerrymanders. Whatever else proper representation requires, it requires that representatives should not have the power to give themselves life terms. [USD incorporation stopped here.] 7. Incumbent-Protection Gerrymanders: Corruption And there is another, perhaps more basic, objection to incumbent-protection gerrymanders. The objection is not that official Gerry has greater influence than ordinary voters, but instead that Gerry uses that greater influence corruptly. Elsewhere (Kolodny ms), I suggest that corruption consists in an official’s deciding how to use the office, and so its superior power and authority over others, for a reason that (roughly) does not serve the values that justify the office: reasons such as bribes, nepotism, a perceived debt of gratitude, a pet project, or a weakness for flattery. Those subject to that superior power and authority have a complaint about that power and authority becoming the extension of the will of a particular person in this way. If an office would serve the justifying values no worse if officials were to exclude a certain class of reasons, I suggest, then they should exclude them. Whether Gerry’s gerrymandering counts as corrupt, therefore, depends on whether the reasons for which Gerry gerrymandered serve the values that justify the office. If Gerry is a legislator, then offhand one might expect that reasons of policy—that this will bring about good or preferred policies—do serve the values that justify his office. If legislatures serve any justifying values, they presumably serve them by legislators being sensitive to reasons of policy. This makes it hard to pin the charge of corruption on partisan gerrymanders, which may well be made for reasons of policy. To that extent, the North Carolina State Rep. quoted earlier has a point. 19 By contrast, when incumbent Gerry indulges in an incumbent-protection gerrymander, then his reason may well be to get reelected even at the expense of policy. To protect their seats, incumbents often strike deals with incumbents of the other party, which opposes their policies. Drawing the maps merely to get oneself reelected, damn the policy consequences, seems far less likely to serve the values that justify the office of legislator. It seems more like feathering one’s nest.11 Insofar as the objection to gerrymandering is an objection to corruption, it depends on intent. A decision is corrupt is not because of its content, but instead because of the reason for which it was made. Hiring a qualified contractor need not be corrupt. But hiring the same contractor for a bribe is. Similarly, suppose that Gerry corruptly gerrymandered a district map. If Gerry had drawn the same map, but not for reasons that do not serve the justifying values—if Gerry had simply applied some randomizing procedure—then it would not be corrupt. 8. Racial Gerrymandering: Discrimination Perhaps, then, we have identified objections to incumbent-protection gerrymanders: namely, that they violate the standards of proper representation and that they are corrupt. But what about racial and partisan gerrymanders? In the case of racial gerrymanders, the answer might seem obvious: wrongful discrimination. But, on closer inspection, several challenges come into focus. First, we need a conception of wrongfully discriminating against a group that does not require giving members 11 One might distinguish between two offices that Gerry holds: legislator-on-matters-ofdistricting and legislator-on-other-matters. And one might argue that even if the office of legislator-on-other-matters serves the justifying values when the legislator acts for reasons of policy, the office of legislator-on-matters-of-districting does not. So partisan gerrymandering, even when influenced by policy, can be corrupt. But if “serving the justifying values” here means “having good results,” then the argument that partisan gerrymandering is corrupt would seem to presuppose a prior argument that partisan gerrymandering leads to bad results. 20 that group less of some good than members of another group. For we have yet to identify a good that gerrymandering gives people less of: an “impact,” as it were, that might be “disparate.” Perhaps wrongful discrimination can take other forms, however, besides the lesser provision of some good. Perhaps an effort to thwart electoral preferences is one such form. Perhaps that is, in itself, hostile or adversarial treatment. Second, we would need to explain what is wrong with such treatment.12 Since what is wrong, we are supposing, is not the lesser provision of some good, perhaps it has to do with the attitudes expressed. If it does have to do with the attitudes expressed, then wrongful discrimination, and so this objection to racial gerrymandering, may in turn depend on intention. It may imply that natural racial gerrymanders would not be objectionable in the same way.13 Finally, we need to identify which groups, when treated in those ways, are wrongfully discriminated against. If there are any such groups, presumably racial groups are among them. Racial groups are more or less defined by a history of group subordination, of a kind that involved far more than simply thwarting electoral preferences by gerrymandering schemes. But if we grant this, why not also say that partisan gerrymandering is wrongful discrimination against political party, just as racial gerrymandering is wrongful discrimination against race? To be sure, it might be wrong, say, to refuse to hire someone because of their party ID. But, first, it’s not clear whether what makes this wrong, when it is wrong, is simply refusing to hire someone without good reason, or whether refusing to hire someone because of their party ID, like refusing to hire someone because of their race, is wrong in a distinctive, further way. And, second, one doubts that it can be wrongful discrimination, in general, to try to thwart the 12 For a recent survey of the challenges, see Alexander 2016. It would thus be hard to explain the 1982 amendment to Section 2 of the VRA, requiring only discriminatory outcomes, not discriminatory intent, other than as an “evidentiary dragnet.” 13 21 electoral preferences of members the opposing political party—let alone simply to try to improve the electoral fortunes of one’s own political party. To be sure, racial gerrymanders might also be objectionable because they have independently objectionable effects. The effect might be that policy unfairly disserves the substantive interests of minorities. Or it might be that there are no, or disproportionately few, representatives from a racial group. Or it might be that a racial group is consistently outvoted. Note that even if the racial group has representatives, it might still be outvoted on questions of policy in the assembly.14 Offhand, there seems no more reason to expect these effects from intentional than from natural gerrymanders. Having one’s interests unfairly underserved is objectionable on its face. But is being underrepresented or being consistently outvoted objectionable, in itself? It’s true that when the group is outvoted, the preferences of its members are not satisfied. But we have already set aside preference satisfaction. It can’t be said that, because they are reliably outvoted, members of the minority have less opportunity for influence than members of the majority have. As individuals, everyone’s a priori chances of decisiveness are equal and everyone’s actual decisiveness is equally zero. It’s common to dramatize the complaint of members of persistent minorities by saying that their votes “made no difference.” But members of persistent majorities can almost always make the same complaint.15 14 The creation of majority-minority districts can increase the representation of the minority. But this may not improve, and indeed may worsen, the prospects of control by the party, or of the pursuit of policies, favored by the minority. 15 It is true that a persistent majority as a group enjoys greater influence—indeed decisiveness—than a persistent minority as a group. Perhaps—although this is admittedly speculative—in addition to having reason to care that one not be subordinated as an individual to another individual, one also has reason to care that one’s group not be subordinated to another group, with which one’s own group has a claim of equality. Compare how Iraqis might have objected, had Iraq been annexed as the 51st state. And perhaps such group subordination could 22 9. Partisan Gerrymandering: Asymmetric information What are we to say about the remaining case: partisan gerrymandering? Consider an example. The letter of the law says that an election can be held on the 1st or 2nd of the month, at the discretion of the election commissioner. However, elections have always been held on the 2nd, and few even realize that the 1st is a legal possibility. The election commissioner, a partisan of the Fête Party, holds the election on the 1st, informing only members of his own party. When the members of the opposing Fiesta Party are confronted on the 2nd with the fait accompli, they would seem to have a complaint. Their complaint isn’t that elections on the 2nd somehow have better long-term results or are intrinsically more democratic than elections on the 1st. The complaint is that they, the members of the Fiesta Party, weren’t informed, whereas members of the Fête Party were informed, about when the election would be. But why is that grounds for complaint? An answer might lie in the idea, supported by either the Activity or the Status argument, that citizens have a complaint about being deprived of equal opportunity for informed, autonomous influence over political decisions. The adjective, “informed,” is key. Citizens must have equal opportunity to know what the decision-making procedure is. This is not to say that you were deprived of equal opportunity simply because you do not in fact know what the procedure is. For others may have led you to water: done all that could be expected of them to put you in a position to know. But if you form reasonable but false expectations about what the procedure is—or even if you formed false and unreasonable expectations, but it would have been cheap and easy for others to correct the mistake—then you may well have been deprived of equal influence. be constituted by the fact that the other group is reliably decisive. But, again, this is all speculative. 23 On an admittedly extremely stylized depiction, Republican gerrymandering around 2010 redistricting bears some comparison to machinations of the Fête commissioner. Democrats assumed that everyone was complying with the rules of the more restrained “Traditional Contest,” according to which partisan make-up of Congress is largely determined by elections of representatives every two years. They understood that control of state governments might result in some gerrymanders, but these would be kept within traditional bounds. Republicans, by contrast, had no plan to comply with the rules of the Traditional Contest. Instead, they were complying with the rules of a fewer-holds-barred “Lagged Contest,” where partisan make-up of Congress is determined mainly by elections of state governments every decade. That is, they systematically targeted elections of state governments in 2010, in order to gerrymander, to an unprecedented extent. Democrats falsely believed that both sides were competing in the Traditional Contest, and this was a reasonable belief—or at any rate cheap and easy for Republicans to correct. Democrats might complain that they were deprived of equal opportunity for influence. It was as though Republican knew, but withheld from Democrats, when the “real” election—that is, the one that would determine partisan composition of State’s congressional delegation—would take place. Of course, it’s a further question what relation this stylized depiction bears to the reality. A REDMAP-er might well reply: “Don’t blame us! You should have known better!” As the architect of REDMAP, Chris Jankowski puts it: “there’s people who play on the other team who should do their job” (Daley 2016 75). He has a point. Karl Rove, for instance, made no secret of what was afoot (xvii). On the other hand, students in high-school civics courses and applicants for citizenship are certainly encouraged in the impression that everyone plays the Traditional Contest: that members of congress are elected directly every two years, not that the party 24 composition of the state’s congressional delegation is determined is by whatever state government happens to be in power at the time of redistricting. (Compare McGann et al. 2016 192 on “transparency.”) And even politically savvy actors, while aware that full-tilt gerrymandering was legally possible, seem to have assumed that the other side would not push things to the edge of legal possibility. Daley 2016 describes REDMAP as a “secret plan” and a “surprise,” and reports shock from old hands, such as NC congressman David Price: “This is not same-old, same-old. This has been taken to an extreme” (41). McGann et al. 2016 describe Vieth v. Jubelirer in 2004 as an “unnoticed revolution” in partisan gerrymandering, explaining in some detail why it went “unnoticed” even by otherwise well informed observers (14). The objection to partisan gerrymandering, I am suggesting, might be an objection to depriving some of equal opportunity for informed, autonomous influence over decisions by depriving them of opportunity to be informed about the decision-making procedure that others in the system will follow. It’s hard to see how natural gerrymanders—pure demographic shifts, against randomly drawn maps meeting contiguity and compactness constraints—would give rise to such an objection, because it is hard to see how this would involve anyone’s exploiting asymmetric information. This objection to partisan gerrymandering depends, as we have seen, on which decisionmaking procedures it is reasonable for participants to expect other participants to follow. And what is reasonable to expect depends on contingencies of history and context. If it was common knowledge that everyone played the Lagged Contest, this objection simply would not apply. If we all get used to gerrymandering, in other words, it will cease to be objectionable, at least on these grounds. 25 If this seems counterintuitive, notice that gerrymandering is a special case of a more general category: selecting a particular a priori equal procedure from among other a priori equal procedures that, given the predicted distribution of electoral sentiment, will favor a particular outcome. Such selection needn’t involve drawing districts. For example, an incumbent might replace majority rule with plurality rule for fear that the presently divided opposition would otherwise unite in a run-off. Or a party may call for elections now in the summer, when their poll numbers are soaring, rather than later in the winter, when they may have regressed. We might call this “temporal gerrymandering.” If this is objectionable, surely it isn’t because some deep principle of democracy, independent of some complicated, empirical argument about long-run systemic effects, favors majority rule with run-off over plurality rule—much less summer elections over winter elections. It seems more plausible that the objection has to do with deviating from the “established” or “accepted” equal chances procedure: the procedure that it was reasonable to expect others to follow. By contrast, when it is “established” or “accepted” that parties will select among equal chances procedures to favor themselves, no one seems to call foul. “Temporal gerrymandering” in some parliamentary systems, for example, is a matter of course. Everyone knows that everyone else is playing a lagged game, in which the governing party has an advantage in when it calls elections. Similarly, if civics classes in the US broadcast that congressional delegations are determined every ten years by state legislatures, that that is when the real congressional elections occur, one wonders whether gerrymandering would continue to strike people as unfair. 10. Summary To sum up: While some of the concern about gerrymandering is concern about the results that it leads to, some of the concern is that it is itself somehow objectionable: that it is undemocratic or 26 unfair or foul play or what-have-you. If there is such an objection to incumbent-protection gerrymandering, my best guess is that it violates conditions of proper representation or that it is corrupt. If there is such an objection to racial gerrymandering, my best guess is that it is wrongfully discriminatory. And if there is such an objection to partisan gerrymandering, my best guess is that it exploits asymmetric information about how participants in the electoral system will behave, thereby depriving some of equal opportunity for informed influence. In each case, especially the last, the objections depend on a number of contingent conditions. 27