Redistricting: Does Shape Matter?
The Census Bureau finally released its much-awaited 2020 redistricting data across the various states, thereby officially launching the redistricting process for the decade. State legislatures and commissions will use this information to draw boundaries representing constituents at the congressional and state legislative levels, with each district equally populated within-state. The process will be far from technocratic, however — legislators will have considerable ability to choose their voters and thus all but ensure their own reelections.
Where single parties dominate, the majority party will almost certainly attempt to hold onto power via gerrymanders, even when they fail to secure a majority of the statewide vote, as seen in states such as Michigan, North Carolina, and Pennsylvania. While various ways exist to measure and quantify the impact of gerrymanders, political commentators and judges overwhelmingly rely on the oddity of shapes to single out gerrymanders for suspicion. The phrase “I know it when I see it” when judging obscenity can often be applied here — districts shaped like elongated reptilian creatures are probably not made with the purest intentions.
Does shape matter, though? Do oddly shaped and tortuous districts directly impair representation? The answer: Well, sort of. It is not so much that the shapes themselves impair representation, but rather the action of ignoring geographies of communication central to the constituent-representative link. Oddly shaped districts often suggest that legislatures have split up and violated these boundaries, ranging from counties to media market areas to ZIP codes.
Redistricting discourse centers on images in both the popular and legal sphere. Most often, these images adorn the front pages of news articles on the topic or of evidence presented in litigation. The Washington Post in 2014, for example, presented an analysis of the “most gerrymandered districts” based on how they deviated from compact shapes (i.e., circular or square), highlighting the oddities and labeling them “crimes against geography.” Critiques of odd district shapes have a history going back to the American republic’s early days, with the original odd collection of townships that formed a salamander-esque district in order to thwart Federalists from securing power in the Massachusetts legislature. Presently, 32 states require that districts be compact for state legislatures; 17 make the same requirement for congressional districts. The Supreme Court in Shaw v. Reno (1993) ruled that in regards to questions of race and redistricting “is one area in which appearances do matter,” starting off a wave of litigation where plaintiffs used a district’s odd shape as their primary evidence of a racial gerrymander
For example, consider Maryland’s sixth congressional district. The district stretches from the suburbs of Maryland within Washington, D.C. to the northwestern reaches of the state. It is certainly oddly shaped and has no single main highway connecting the entirety of the district, making it hard to traverse. The Brennan Center of Justice found the boundaries significantly changed from the previous cycle and just on the edge of violating equal population requirements. In litigation alleging an illegal partisan Democratic gerrymander, a federal judge described the district’s design as imitating that of “a broken-winged pterodactyl, lying prostrate across the center of the state.
As for why oddly shaped boundaries matter, political scientists and legal scholars traditionally give two reasons. First, mandating regularly shaped districts can act as a constraint on redistricting. While regular shapes are not good in and of themselves, they can limit the extent to which legislators choose individuals to be part of districts in an attempt to gerrymander a state. Compactness constraints can prohibit the creation of districts that zig-zag across highways and split municipalities, thus forcing legislators to settle for presumably less biased maps.
Second, regularly shaped districts can be worth striving for by improving representation. As I have noted before, the reason that the Founding Fathers created the House of Representatives in the first place was to enshrine the constituent-representative link. James Madison in Federalist 56 called for cohesive districts on the principle that a “representative ought to be acquainted with the interests and circumstances of his constituents.” Political scientist Jay K. Dow cites the rationale of Representative Thomas Arnold’s (W-TN) rationale for mandating single-member congressional districts in the Apportionment Act of 1842 – that the “majority govern, but the minority be heard.” Should a district be overly large and tortuous in design, constituents could be confused about who represents them, and representatives about whom they represent. In these districts, only the majority party would be heard, and the minority would suffer geographic obstacles to political organizing or contacting their representative. Compact districts can reduce confusion, thereby strengthening the link between constituents and representatives, even when they are members of opposing parties. On this reasoning, then, if non-compact districts actively impair representation, then they would amount to direct proof of representational harm incurred in the attempt for some type of political gain.
So far, however, the evidence for using district shape as a metric for gerrymandering is fairly weak. As a constraint, political scientist Anthony McGann and his coauthors find compactness to primarily constrain Democratic and not Republican gerrymanders. Using redistricting simulations, political scientists Jowei Chen and Jonathan Rodden find that Democrats tend to be overly clustered geographically, which necessitates more oddly shaped districts just to ensure that they have a chance to win the same proportion of seats as Republicans for the same proportion of the statewide vote. That said, just because compactness as a constraint does not perform as well in constraining Republican gerrymanders does not mean that it should be discarded. Identifying and limiting Democratic gerrymanders is a pressing public concern, too, especially if non-compact shapes directly lead to poorer representation – though evidence for this contention remains thin.
Political scientist Richard Engstrom, for example, fails to find evidence that people participate less electorally in more oddly shaped congressional districts. Political scientist Daniel Bowen finds evidence supporting a weak correlation between more compact districts and more communication and knowledge between constituents and their representatives, though the analysis is sensitive to how one measures compactness. Additionally, the effects of compactness pale in comparison to the increased attachment between constituents and representatives arising from competitive districts and sharing the same party. Using shape can be a poor metric in the effort to ensure that a district can be easily traversed, given that oftentimes roads themselves can be oddly shaped. Insofar as one wants to reduce travel times across a district, metrics of functional compactness as opposed to shape-based measures would be the way to go. These results suggest that compactness is correlated with a weaker constituent-representative link, but perhaps not the cause per se.
A possible explanation for the weak effects: Americans tend to be infamously geographically illiterate, often unable to identify cities within their home state. Constituents commonly identify who their representative is by searching by ZIP code in the House representative-search engine or analogous state legislative search engine. When representatives conduct outreach to their constituents (regardless of party), they have one of three options. First, they can send mailings only to those who previously voted and are therefore already in the state’s voter file. However, from my own archival research of congressional communications, up to 25 percent of entries within these databases can be outdated within a year’s time. Additionally, reliance on such voter files misses those who recently moved into the district or are new voters. When representatives engage in mass communication for official outreach, they are by law required to target only their own constituents – when that is not possible, they must minimize the extent to which they communicate to people outside their own district.
The second means of outreach is electronic, via television, radio, social media, and so on, by some combination of designated media market area and county. The drawback of mass communication is the failure to specifically target messages, thus often wasting money on non-constituents. The third means, for the most specific outreach, representatives can engage in mass mailings via the Postal Service’s simplified addresses – that is, every four-digit extension inside a ZIP code with which a district overlaps.
These geographies of communication break down, however, when multiple districts split up these geographic units. Political scientists Jonathan Winburn and Michael Wagner find that when congressional districts split counties and media markets, those living outside the district covering a majority of the county/media market are less likely to recognize the name and party of their representative. Tyler Steelman and I find a similar relationship for when districts split ZIP codes, where constituents are less likely to recognize the name and party of their representatives and less likely to contact them. Additionally, constituents in these split-up ZIP codes perceive greater ideological distance, even when controlling for a wide variety of potential confounding factors.
To illustrate the impact, consider ZIP code 27713in North Carolina in the corridor connecting two portions of North Carolina’s fourth congressional district, one of the least compact districts in 2013. Within the ZIP code, districts one, two, four, and six all encompass some portion of the population. Searching one’s member by ZIP code would therefore reveal one of four potential members. We estimate that people residing in ZIP codes as split up as 27713 to see a reduction in recognition of one’s member by approximately 20 percentage points.
These results carry three substantial implications. First, name recognition is essentially a necessary condition for voters to cross party lines and vote for an opposing party candidate, as confirmed by rigorous survey analyses by political scientists Cindy Kam and Elizabeth Zechmeister. Second, constituents who do not even know the name of their representatives are unable to contact them following elections for nonpartisan concerns, such as dealing with bureaucracy or voicing concerns over public policy. Third, people confused about who their representative is might contact the wrong representative and potentially overwhelm the system. Tyler Steelman and I confirmed this third issue when we found 28 representatives posting messages noting severe miscommunication directly due to split ZIP codes. One illustrative message can be found on Democratic Rep. Sheila Jackson Lee of Texas’s page, which reads:
“There are multiple Representatives who share the same 5-digit zip code that was entered. Due to the large volume of US mail, email and faxes I receive, I am only able to accept messages from residents of the 18th congressional District of TEXAS. A 9-digit zip code including the zip+4 extension is required to identify your correct Representative.”
We see, then, that ignoring and splitting these geographies of communication directly impair every stage of representation.
Ultimately, the answer to the question of whether district shapes matter is inconclusive. While oddly shaped districts might not directly impair representation, they are a potential indicator that redistricters almost certainly ignored and split geographies of communication, such as media market areas, counties, and ZIP codes in an attempt to advance incumbent or partisan interests. While some splits almost certainly need to occur to ensure population equality, the most oddly shaped districts – such as the Democratically drawn “broken-winged pterodactyl” Maryland sixth district, or the Republican-drawn Texas-21 – unnecessarily impair the constituent-representative link in a partisan attempt to bias maps.
Redistricting reformers should use oddly shaped districts as a means to narrow the scope to identify where geographies of communication are most ignored. From there, it is possible to identify and quantify the specific harms to representation that arise independent of election outcomes. Oftentimes these oddly shaped districts act as keystones, where removing them causes the entire map to implode. Such an approach would avoid complaints that it is not the Supreme Court’s role to choose election outcomes, as Chief Justice Roberts declared in his majority opinion in Gill v. Whitford (2019). While district shapes should not be the be-all and end-all in judging gerrymanders, in the proper context, district shapes do matter.
John Curiel, Ph.D., is a research scientist at the MIT Elections Data and Science Lab. The information and opinions expressed in this column represent his own views and research, and do not necessarily represent the opinions of the MIT Election Lab or MIT.
