26 March 2015

Gerrymandering

Yesterday, the U.S. Supreme Court in the case of Alabama Black Legislative Caucus v. Alabama, set forth detailed legal guidelines governing how special panels of three U.S. District Court judges that hear election law cases should analyze claims that a state legislative election districts fail to protect the rights of black voters in states that are subject to special scrutiny under the federal Voting Rights Act.  The same reasoning would apply, however, in the redistricting of any jurisdiction subject to the federal Voting Rights Act.

Background on Gerrymandering

Gerrymandering is an inherent risk of any system for electing multiple public officials, typically to Congress, the state legislature, or the governing board of a local government, from at set number of single member districts (typically three or more).

The requirement that these districts have equal population (a rule of thumb allows deviations in the population of each district based upon the most recent U.S. Census of not more than 1%), and that they be contiguous are well established and universally followed as basic rules of the redistricting game in essentially 100% of all cases.  They are followed in 100% of the cases largely because the ruled are perfectly clear and the remedies are obvious.

(Contiguous means geographically connected in a single unit, rather than two or more "islands" or territory, except (1) in cases of actual islands, in which case a geographical link over water is permitted and (2) in the case, for example, of county commissioner districts in counties that are not contiguous themselves.  For example, Arapahoe County in Colorado that has a non-contiguous lump called "Glendale" that is entirely surrounded by the City and County of Denver that could be included in an Arapahoe County Commissioners district together with additional non-contiguous territory that would have to be added to meet the equal population requirement.)

The tricky part is that there are a nearly infinite number of ways to draw more than two districts consistent with these two baseline rules of the game, and the outcome of the elections held using those districts will be very different from each other, even if every single voters vote is predetermined, depending upon which of the possible redistricting maps is adopted.

The mathematics involved in analyzing these situations is not very well developed, because they are quite complex.  You need to know more than just the proportion of the overall population that supports, for example, a particular political party, to know how much impact gerrymandering can have on an election.  You also need to know how "lumpy" the geographic distribution of supporters of each party is with considerable detail.

Suppose you have to draw a map with 100 contiguous, equal population districts in state with 3,000,000 people made up of 1,000,000 Democrats, 1,000,000 independents who are equally likely to vote for either party, and 1,000,000 Republicans.  The geographic distribution of supporters of each party greatly influences how much room there is to gerrymander the districts to favor one party over another.  If all the Democrats huddle together in one big blob, surrounded by all of the Republicans, and independents are distributed uniformly across the state, gerrymandering is limited to a few districts on the boundary between them.  If the Democrats are in half a dozen decent sized blobs, in contrast, there are far more potential gerrymandering strategies.

Gerrymandering also requires the person doing it to know not just the general partisan leanings of voters in a particular district, but how reliably loyal they are to a particular party or candidate.  If all voters have party affiliations and are perfectly loyal, a district with a 51%-49% split in favor of the paper you are trying to benefit is just as good as one with a 99%-1% split in favor of the party you are trying to benefit.

In practice, there is a trade off.  You can make more districts that lean in favor of the party you are trying to favor by getting closer to a 50%-50% split while giving the party you favor as many seats as possible, or you can reduce the number of districts that your party is likely to secure, but make the seats that it does win less competitive.

Generally speaking, the risk tolerance of people drawing a map to favor a particular party depends upon that party's overall strength.  The goal is usually to draw a map that leaves a majority of the seats plus some margin of error in the overall results from the map, with swing seats that are as safe as possible.

For example, if it was possible to draw a map that gave the party drawing it 55 of 100 districts in which that party had a 6 percentage point edge over the other party, or a map that gave that party 60 of 100 districts in which that party had a 2 percentage point edge over the other party, the first choice would be preferred to the second choice.

More generally, from a political party's perspective, the goal is usual to maximize the likelihood of securing a majority of the seats for your party, recognizing that the overall tide in favor of, or against, a political party, may shift during the course of the election campaign.

But, existing computer technology makes it possible to feed all available voter registration and census data into a model and optimize pretty much any objective that can be described with sheer brute force numerical methods (basically, drawing all plausible maps and comparing them, or starting with a plausible map and modifying it until it is optimized).

For example, one might want to draw a map that maximizes the number of seats held by Republicans in a manner that preserves seats for all incumbent Republicans with a realistic chance of being re-elected.  Someone else might want to draw a map that does the same thing for Democrats.

Usually, although not always, either the Democrats or the Republicans will control the process.  In situations where bipartisan support for a map is required, a failure to come up with a map at all, sending the matter to a court to draw a map is almost as common as a negotiated resolution.

The partisan outcomes aren't infinitely malleable.  For example, in a state like Colorado where the two political parties are roughly evenly matched in voter registration and each have regional strongholds, it is almost impossible to draw a map for its seven Congressional districts, 35 state senate districts, or 65 state house districts, that shuts out either party entirely.

But, it is frequently the case that it is possible for one map to leave one party with a strong likelihood of attaining a majority, and another map of the same districts at the same time to leave the other party with a strong likelihood of attaining a majority.

In general, the redistricting process tends to favor whichever party benefited most from the previous round of the redistricting process.

Intuitions about what maps are fair are mushy.

One set of intuitions is global.  If the map, taken as a whole, produces results with a partisan divide similar to those that would be achieved if a pure proportional representation system were used, the map is fair.  If the map deviates from that standard, it is unfair in the direction of whichever party benefits the most.

The trouble with a proportional representation standard is that it ignores a bias that is inherent in all single member district election systems.  Single member district election systems are inherently biased against evenly distributed populations that are a minority where they are found.  For example, if 15% of the population supports the Libertarian party, and that 15% is extremely homogeneous across the state, and the rest of the population is split in varying percentages between the Democratic Party and the Republican Party, then no Libertarian party members will be elected no matter how the map is drawn in any district where both a Democrat and a Republican run for office, even though they would get 15 seats out of 100 in a pure proportional representation system.

Another set of intuitions is aesthetic.  If the map generally follows existing county lines and municipal boundaries and tends to keep like-minded people in the same district, it is fair.  In contrast, if the map produces erratic and unnatural boundaries that are not obviously related to any pre-existing jurisdictional lines, have funny shapes that contort to achieve political objectives, and tends to split geographically contiguous groups of like-minded people into multiple districts, it is unfair, with the people who are split into multiple districts as the main victims of gerrymandering.

One way to conceptualize the issue is to imagine a state as made up of a lot of census blocks, each of which favor a candidate.  If you can corral enough blocks supporting a given candidate that are close to each other into boundaries equal to the predetermined per district population quota, then that candidate will be elected.  But, if those blocks are dispersed into multiple districts where other candidates have majority support, then that candidate will lose.

But, while districts that split communities of interest are unfair from one perspective, a map that creates lots of competitive districts is often viewed as more fair than a map that creates few competitive districts and preserves more communities of interest.

The Holding

The former Confederate states, including Alabama, have a history of racial discrimination against blacks that including Jim Crow laws and racially biased election laws through the 1960s.

As a result of that history, the Congress passed the Voting Rights Act with a provision that mandated in states that had this history of discrimination, not race neutral election laws and procedures, and instead required that changes in election laws and procedures not make black voters worse off.

This is conceptualized by prohibiting changes in electoral district that make it more likely that a geographically concentrated group of black voters will not be able to elect a candidate of their choice.

In practice, Southern whites now vote so reliably in favor of Republicans or conservative white Democrats, and Southern blacks now vote so reliably in favor of liberal black Democrats, that predicting the outcome of an election based upon how the districts are drawn is particularly easy.

The main holdings are that:

1.  A standing challenge to the organization bringing the suit probably lacked merit and was sprung on the organization inappropriately at the last minute without any party requesting it, or an opportunity to address to standing concerns.

2.  Gerrymandering in this context is to be addressed in light of individual districts that have been inappropriately gerrymandered and not the map as a whole.  Thus, even if race was a minor factor in drawing the map as a whole, it may have been a major factor in how a particular district was drawn.

3.  The baseline of equal population and contiguity are factors that don't count for or against a particular gerrymander's validity.  Permissible factors other than equal population and contiguity that were involved in drawing the district, however, may be considered.

4.  The question is whether a predominant purpose of the change as a whole in the district is to makes it harder for a group (blacks in this case) to elect a candidate of their choice, not, for example, whether the percentage of black voters has changed in the district.

These factors at least bring some measure of definition and applicability to a very slippery concept in a way that provides more clarity for voting gerrymandering cases.  It may be less than ideal and even conceptually problematic, but under the prior vague standards that were in place, it was almost impossible to challenge racial gerrymandering, whereas now that may be possible to do, and the threat of suit may also influence redistricting behavior more broadly.

The changes, in particular, may help blacks interested in having black Congressmen and state legislators represent them, at the expense of helping Democrats more broadly win majorities in Congress and the state legislature that would advance a legislative agenda favorable to blacks.  This is because the approach favors concentrating black votes in one district, rather than spreading them out where they might tip the balance in other contested races.  But, this concern is weaker in the South where the Voting Rights Act has special applicability than in other places, because race and partisan affiliation are so rigidly tied there.

No comments: