Trade-offs in Redistricting

This is slightly renamed second major iteration of “Trade-offs in Redistricting”. It’s still a work in progress.

The purpose of this revised site is to illustrate some quantitative policy trade-offs in congressional redistricting.

The Notable Maps in Dave's Redistricting (DRA) are the maps that individually maximize five quantifiable policy dimensions: proportionality, competitiveness, opportunity for minority representation, compactness, and county-district splitting. For the 2020 congressional redistricting cycle, these maps demonstrated that there were tradeoffs between these objectives. For example, compact districts aren’t always fair, and vice versa. In this site, we explore these trade-offs further in several marquee states.

Our preliminary results show several things:

1. Pairs of Dave’s Redistricting (DRA) ratings dimensions in unbiased ensembles of redistricting plans for a state are not correlated as you might expect them to be.
2. Nonetheless, there is a clear trade-off frontier between the pairs of dimensions. Since all DRA ratings are “bigger is better,” these Pareto frontiers are to the upper right (north-east) in the scatter plots on state pages.
3. These unbiased frontiers can be automatically “pushed” farther to the north-east, to approximate the trade-offs that people face when drawing maps.
4. The ratings for most human-drawn maps fall near the edge or outside the range of the ratings for unbiased ensembles. Whether for bad or for good, people are intentionally optimizing for something.
5. The most compact human-drawn maps are noticeably more compact than the most compact maps in the “pushed” frontier, because they smooth out district boundaries.
6. The trade-offs vary by state.
7. The trade-offs are probably different for the state legislative maps for the same states, because of the different granularity of the districts and the relative sizes of counties and precincts.

These findings are meant to be directionally illustrative rather than specifically descriptive.

We’ve only enabled a few state & plan types combinations so far:

• NC – Congress, upper state house, lower state house.
• PA – Congress, upper state house, lower state house.
• MD – Congress, upper state house, N/A - uses multi-member districts
• IL – Congress, upper state house, lower state house.
• SC – Congress, upper state house, lower state house.
• TX – Congress, upper state house, lower state house.

More will follow.

What we do here is different from the first iteration of the site, due to limitations that we discovered in that approach. The methodology we used this go around is summarized on the Method page.