Districting for a low-information electorate.

• Author: Elmendorf, Christopher S.

FEATURE CONTENTS INTRODUCTION I. VOTER IGNORANCE, BRIEFLY II. ON BIPARTISAN GERRYMANDERS AND THE SUBSTITUTABILITY THESIS III. MEDIA-MARKET DISTRICTS A. The Evidence B. Lessons for District Design 1. Which Media Market? Newspapers vs. Television 2. Integrating the Constitutional Requirement of "One Person, One Vote" 3. Do Newspapers (or Broadcast Television) Still Matter? 4. Tradeoffs IV. DISTRICTING TO IMPROVE THE QUALITY OF PARTY BRANDS A. What Makes Party Labels More Useful for Voters? A Nonexhaustive List B. Districting To Create More Useful Party Brands C. Implications CONCLUSION INTRODUCTION

People who talk about election law generally do not talk much about voter ignorance. The preferences of voters, and the wisdom of those preferences, are taken as a given. The main questions we ask are whether election laws hinder citizens from registering their preferences or from aggregating votes to elect their "candidates of choice." But expressed voter preferences are not necessarily wise--that is, they are not necessarily the same as informed preferences, or what voters would think if they knew basic facts about politics or policy. And this turns out to matter very much for how we design our electoral institutions.

Electoral districting at first glance may seem unrelated to voter knowledge. Drawing lines in one place rather than another does not, as such, provide information to or withhold it from voters. This Essay contends, however, that political science research on what voters know, and how they make decisions when they do not know very much, sheds light on longstanding debates about gerrymandering. District design can either ameliorate or exacerbate problems associated with voter ignorance.

We begin with a critique. A number of prominent scholars defend bipartisan gerrymanders--schemes that lump Democratic and Republican voters into safe districts for each party--on the ground that moving from a competitive map to a bipartisan gerrymander merely relocates the type of election in which voters hold politicians accountable, from general elections to primary elections. But this proposition, which we call the "substitutability thesis," rests on a fundamentally mistaken premise about the equivalence of voter performance in these two types of elections. In general elections, voters benefit from political party labels that summarize candidates' positions on the issues and enable voting based on the citizen's "running tally" of observations about a party's past performance when exercising power. If you, like most voters, don't know much about the individual candidates in a race, it frequently will not matter that much. Having a feel for the ideology and past performance of Democrats and Republicans as a whole is usually enough to determine which candidate to support even if you know nothing about the candidates beyond their party affiliation. Political scientists continue to debate how well voters perform with party labels on the ballot, but no one doubts that party brands have great potential to help voters leverage the little information they have. Or, as E.E. Schattschneider once said: "[M]odern democracy is unthinkable save in terms of the parties." (1)

In primary elections, however, ordinary voters receive no similar assistance. There are no ballot labels to help voters identify and understand candidates' affiliations with internal party factions. As a result, voters are simply less informed in primaries and less capable of using these elections to achieve accountable and responsive government. Because primary elections do not provide ordinary voters with the tools (ballot labels) that allow them to overcome their lack of knowledge about individual candidates, narrow and well-organized groups tend to control the outcomes. Primary election competition, which is more likely to occur in safe districts, simply is not an adequate substitute for the general election competition we see in districts whose electorates are split between Democrats and Republicans. Bipartisan gerrymanders shift the locus of accountability to an information-poor environment.

After developing that critique, we turn to the affirmative question of how districts should be designed in light of the fact that ordinary voters pay little attention to politics. We make two claims. More confidently, we assert that districts should be designed for congruence with media markets, so that district lines match the distribution area of a newspaper or television network. Numerous studies have shown that voters in such "media-market districts" are more cognizant of their representatives and that their representatives behave as if on a tighter leash. This finding is consistent with evidence that voters are passive, haphazard consumers of political information, learning political facts as they go about their daily business rather than through focused searches. Newspapers are more likely to cover the exploits of politicians who represent their whole market, rather than just a part of it, and this helps to educate voters who do no more than skim the news headlines en route to the cartoons or sports. Media-market congruence at the very least belongs on the list of good-government districting criteria, alongside such traditional considerations as respect for communities of interest and political subdivision boundaries. (2)

More speculatively, we offer some guidelines for drawing districts so as to induce the development of party brands that are more instrumentally useful to low-information voters. The meaning of a party's brand is partially determined by the positions that party-affiliated lawmakers take and by what the legislature does under each party's rule. Although other factors may matter more--most notably, the performance of the President--districting can have consequences for the content of party brands because of how it influences the makeup and incentives of the parties' legislative caucuses. Because voters' ability to use elections to produce competent, responsive governance depends on the informational value of party brands, policymakers should account for the effects of districting on these brands.

To foster the development of useful party brands, we offer three districting guidelines. First, districts should be drawn such that the median voter in the polity as a whole is also the median voter in the median district. This will give both parties an incentive to develop platforms that appeal to a majority of voters. Second, there should be a substantial number of median-voter districts, i.e., districts whose median voter is also the median voter in the polity. This will result in a large "winner's bonus"--a disproportionate number of seats for the party that wins the median voter--which strengthens the incentives of party-affiliated lawmakers to build a competitive, coherent party brand (against any other interests they may have). Because of the large winner's bonus, the majority party will generally have a supermajority of legislative seats, enabling it to govern and helping voters to see which party deserves credit (or blame) for the legislature's output.

Finally, there should not be too much interdistrict heterogeneity in the ideological position of the median voter across districts; that is, districts should not be too different from one another in their ideological makeup. As interdistrict heterogeneity increases, the major parties tend to become extreme or diffuse in order to forestall third-party challenges in some districts. Either outcome is problematic from the perspective of a low-information electorate seeking to achieve policies that are accountable and responsive to majoritarian preferences. Greater heterogeneity in district medians will lead either to party brands that are less meaningful or to parties that take positions further away from the preferences of the median voter.

We discuss how existing districting practices and criteria likely fare in terms of our guidelines (generally, not so well), and we suggest some alternatives. The reforms we outline are far from comprehensive, and they do not take into account all of the factors that might reasonably matter when districting. But, whatever else is taken into account, policymakers drawing district lines should consider what voters know--and more importantly, what they do not.

1. VOTER IGNORANCE, BRIEFLY

   If there is any well-accepted fact in political science, it is that most voters pay little attention to politics and know little about the basic institutions of government. (3) Fifty years of survey research bear out the hypothesis of "rational ignorance": because the probability of tipping an election's outcome with one ballot is vanishingly small, individual voters have no material incentive to become informed about politics and policy. (4) And so, for the most part, they do not. (5) What political information they do have, they frequently acquired adventitiously, as a byproduct of, for example, paying the tax collector, noticing a political headline when scanning the tabloids for celebrity gossip, sending kids to school, or losing a job, rather than as the result of a vote-motivated search. (6)

   Yet political scientists who study voting are not altogether despondent. An electorate comprised of fairly disinterested and uninformed voters may nonetheless perform reasonably well, thanks to the statistical properties of aggregation and the role of political parties.

   Aggregation can neutralize uninformed votes. Ballots cast for one candidate by citizens whose decision is essentially a coin flip will offset those cast for her opponent, leaving the election's outcome to be determined by voters possessed of relevant information. (7) Moreover, the Condorcet Jury Theorem establishes that if votes are just a little bit better than random, the electorate as a whole will converge on the "right answer" with high probability even though each voter individually is almost as likely to be...