Diverse Representation: Districts vs. Block Plurality

If a jurisdiction wants more women to get elected, what voting method should it adopt? What if it wants an underrepresented minority to stop being underrepresented? What voting methods should be avoided? This is the first post in a four-part series on the effects of voting methods on female and minority representation:

1. Districts vs. Block Plurality
2. Comparing Single-Winner Voting Methods
3. Proportional Representation
4. Conclusions

Single-winner districts vs. Block Plurality

In the United States, the most common ways of electing a city council are through single-winner Plurality elections using a ward system (i.e. single-winner districts) and through a multi-winner at-large Block Plurality election (where each voter can vote for up to as many candidates as there are winners). The most notable feature about Block Plurality is that it is winner-take-all; a faction comprising 51% voters can ensure that they win every single seat. Of course, every single-winner contest is also winner-take-all — but when there are several seats up for grabs and different groups of voters electing the winners for different seats, it’s far easier for multiple factions to be represented. This is especially true when factions are determined according to race; due to some mixture of redlining and people wanting to have at least a few neighbors of their race, it’s very common for people of the same race to be clustered together, enabling the creation of majority-minority districts.

My favorite study comparing Block Plurality to single-winner wards (with Plurality) is The Context Matters: The Effects of Single-Member versus At-Large Districts on City Council Diversity by Jessica Trounstine and Melody E. Valdini, which looks at election results for city councils in over 7,000 US cities and compares the electoral success of Blacks, Latines, and women. They perform a regression analysis, controlling for obvious factors like demographics, election year, and Presidential election results. For Blacks and Latines, they limit their analysis to cities with an appreciable population of the group in question as compared to the size of the city council. The basic observations are that, compared to Plurality at-large, districts increase Black representation by 6% and Latine representation by 5%, but reduce female representation by 2%.

There are several complementary explanations for why at-large multi-winner elections are better for women. From the perspective of individual voters in a four-winner election, electing three men and one woman probably seems better than electing four men, and when you’re voting for four candidates it feels more like you’re choosing the gender composition of a city council than when you’re electing a single winner from your ward. Also, voting for four men and no women feels way more sexist than voting for one man and no women. From the perspective of a political party or another group that recruits and endorses candidates, it looks similarly bad to be seen as only endorsing men. Even if a party thinks the four best male candidates are each better than the best female candidate, it’s easiest to endorse her over a male candidate in an at-large race than with single-winner districts; in an at-large race they can just drop the fourth-best man, but with districts they might have to drop one of the stronger male candidates depending on where she lives.

Looking at the data more closely, the situation becomes far more interesting. First, the benefits of single-winner districts only apply to Black men; Black women perform no better with districts than with Block Plurality. (The authors had to use a more limited dataset for evaluating the intersection of race/ethnicity and gender, and no difference was visible for either Latinos or Latinas on the smaller dataset.) Second, Blacks and Latines only benefit from districts when they live clustered together. The benefits of districts are only apparent when Blacks/Latines are present in large enough numbers and live close enough together. Trounstine and Valdini measure this via the isolation index:

One measure is the isolation index, which ranges from 0 to 1 and represents the probability that group members will meet members of their own group in their census tract. A score of .6 for African Americans means that the average African American lives in a census tract that is 60% black. This measure has the benefit of being sensitive to a group’s size in addition to the distribution of the group throughout a community. It would be impossible to have a high isolation score unless a group composes a substantial portion of the total community.

They only find benefits in Latine representation where the isolation index is greater than .75. The findings for Black representation are quite nuanced:

For African Americans, the effect of districts goes from being negative at very low levels of concentration to significantly positive at high levels. Districts have the largest effect for cities in the third quartile, where moving from an at-large system to a district system increases the estimated probability of electing an African American council member by about 10 percentage points, from 14% to 24%. This is a powerful effect compared to the first quartile, where districts decreased both the probability of having any African American councilors (from 7% to 3%) and the expected proportion from .9% to .3%. When the isolation index is very high for African Americans the effect of districts becomes insignificant.

Ultimately, I think this makes a whole lot of sense. When minority voters aren’t clustered it’s impossible to construct majority-minority districts, so the mechanism by which we expect districts to work is ineffective. My best guess is that, with dispersed and relatively small minorities, electing minorities is similar to electing women: many White voters prefer to have at least slightly diverse city councils, and they’re more willing to give one of several votes to a person of color than to give their only vote to a person of color. The finding that districts don’t increase Black representation when Black voters are extremely clustered and numerous also makes sense; Blacks may constitute an outright majority in many such cities (meaning that they can also ensure their representation in Block Plurality contests), and there is a risk the single-winner districts would result in packing, i.e. having districts that Black candidates win by an enormous margin while White candidates defeat Black candidates by much smaller margins in other districts.

Where minorities aren’t clustered, districts don’t work. In places with a reasonably clustered Black population, using districts instead of Block Plurality is an effective means of electing more Black men at the cost of somewhat reduced representation for women (while doing basically nothing for Black women), and it’s similar with Latines. Fortunately, we may be able to avoid this trade-off with better single-winner voting methods and proportional representation.

Next post in this series: Comparing Single-Winner Voting Methods