Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

The North Carolina Senate remedial map shows reduced bias but is weak for minority representation

September 15th, 2019, 11:45am by Sam Wang

This post has been modified to remove reference to the Voting Rights Act.

(click the map above for greater detail)

The Senate remedial map, as passed by the entire chamber (“Consensus v3”), is reminiscent of the House map in its character. It has less partisan skew – but some still remains.

PlanScore [2018 map] [Remedial] [Consensus v3] is currently using a 2016/2018 model in which the statewide vote is 52% Republican, 48% Democratic. As I did for the House calculations, I have adjusted PlanScore’s estimated Total D/R seats to reflect a split vote of 50% R/50% D.

If the statewide vote were perfectly divided, it would elect 23 Democrats and 27 Republicans. So there’s still some partisan advantage to Republicans in the plan.

I can think of two reasons why the Senate was relatively successful in removing bias. One is bipartisan cooperation, since this map had input from members of both parties. Another reason is technical: with fewer boundaries, there are fewer opportunities to commit gerrymandering offenses.

In several places the remedial map has rearranged minority voters to give them more opportunities to elect members of their choice. However, there is one notable exception.

The remedial map unpacks minorities from Districts 15, 32, and 37. In these regions, substantial numbers of whites vote in alignment with blacks, and placing blacks there in high percentages is not necessary for them to elect candidates of their choice. The remedial map spreads black voters from these districts into neighboring districts, giving them broader influence.

However, District 11 poses an opposite problem. It’s in Johnston County, which is part of a six-county cluster flagged by the Court:
Johnston County has about 200,000 people, similar to the average population of a Senate district – but because of its location it has to be split. The proposed new Senate District 11 (shown in red) is 30.3% black (partisan data is in the PlanScore model).

Johnston County is changing: its northwest border includes the growing exurbs surrounding Raleigh, such as Clayton. Meantime, more eastern communities such as Hares Crossroads, Shoeheel, and Moores Chapel would probably feel better represented in the 10th District (the one with the two horns, shown at left in pink). Appropriate adjustments could increase the black population in the 11th District, a change that would make the district more competitive, and also help bring the statewide map into better overall balance.

How much higher would the percentage of black voters need to be? An expert report by Moon Duchin in the Bethune-Hill racial gerrymandering case showed that in opportunity-to-elect districts nationwide, 37% was enough to make the minority group’s vote determinative. However, the exact target is not known. Calculating it would require information about partisan voting habits in the region – which legislators are not allowed to use.

(Feel free to comment on Johnston County neighborhoods in comments. All I really know about that area is that it has great barbecue and ham.)


Finally, a word about the partisan metrics. The three metrics describe different features of the plans.

The sharp reduction in efficiency gap indicates a decrease in wasted votes, defined as either massive wins (packed districts) or close losses (cracked districts). This is reflected in the reduced difference between average Democratic wins and average Republican wins. By these measures, the parties would be treated more equitably than under the gerrymander.

Despite its generic name, partisan bias is a technical measure that describes deviations from a 50-50 split in power if the vote is split 50-50. It is useful in a closely divided state like North Carolina. Again, the decline is sharp.

The same is true for the mean-median difference, which compares the median district with the statewide average vote. Some difference is expected from the way voters arrange themselves. (It is possible to estimate the built-in geographic advantage by calculating the mean-median difference for counties and precincts. I will save that for another day.)

Tags: 2020 Election · Politics · Redistricting

3 Comments so far ↓

  • ALS

    I’m actually defending a Voting Rights Act lawsuit at the moment, so this is interesting. A couple things:

    I found this information, but it would be helpful to readers if you said how many Senate districts there are and how big NC’s black population is. Absent Section 5, retrogression in and of itself is no longer a problem, so whether 7 or 13 opportunity districts are too few becomes, broadly speaking, a question of proportionality.

    Why do you view 40/55 as the critical threshold? Bracketing the necessity of setting Democratic vote share at 55% (by vote share do you mean Democrats as a percentage of VAP, or actual vote share — if the latter, I don’t know why you need a 5% cushion to afford opportunities as opposed to safe seats), it seems to me that 40% is unnecessarily high, as it affords black voters, supposing equal turnout in Democratic primaries, 72%, supermajority control in those primaries. A district where black voters are 33% of the population in a 55% Democratic district is a district where, assuming black voters’ cohesion — which of course must be present in order for the Voting Rights Act to require drawing an opportunity district in the first place — black voters’ preferred candidate can and will win Democratic primaries and can and will win the general.

    • Sam Wang

      40% was just a threshold to allow easy counting of potential opportunity districts. I agree that in a region where the other conditions are met for an opportunity district, the threshold should be lower – maybe 37%, which is a threshold that Moon Duchin highlighted in her expert report in the Bethune-Hill case.

      The other problem is that I had applied this concept to regions that don’t have racially polarized voting, Wake and Mecklenburg Counties. That was an error. I’ve edited the post to de-emphasize this aspect.

      Thank you for your input.

  • Bill

    hard to use 2018 as a base year… it was a wave year for democrats… that would be to advantage Team GOP

Leave a Reply to ALS (Cancel)