April 22, 2026
X, the Everything App, is a horrible place, and I really shouldn't go on there. But, in the aftermath of the Virginia redistricting referendum, I just had to see what the right-wing psychos had to say. I encountered a surprising about of contrition, along with the expected histrionics. But one tweet in particular pissed me off:
There is no way to "end partisan gerrymandering." You might make it a little less extreme. But "nonpartisan" commissions, etc. don't work because the incentives still demand partisan gerrymandering and people are human.
Gerrymandering of the kind we're seeing is a symptom of the...
— Casey Mattox (@CaseyMattox_) April 22, 2026
For some reason, a bunch of people (both on the left and the right) seem to be under the impression that independent commissions are the only way that a federal gerrymandering ban could work. I don't hate independent commissions, and they factually have stopped gerrymandering in many jurisdictions (most other countries have some sort of independent body controlling redistricting, and that generally works fine). But, independent commissions are still pretty vulnerable to corruption. As much as it annoys me to hand it to them, the conservatives are correct that California's independent commission is a bit of a sham. And considering the politicization of all sorts of nominally independent bodies in this country, I have no confidence that my opponents would fail to undermine these commissions' independence as well.
However, independent commissions aren't the only way to draw districts. I hear some people suggest a public, deterministic algorithm that just makes them; while I'm not afraid of the code being racist or biased, I am afraid of it being buggy.
I think I have a superior method: let anyone submit a map, and then use the one that scores best in an objective metric.
To start, the ground-truth assumption that I'm relying on is that all anyone truly cares about is partisan proportionality; Every other concern is just a fig-leaf for proportionality concerns. Democrats are usually pretty normal about this, but Republicans like to hide this for some reason. However, if you actually look at their complaints about Democratic gerrymandering, it's all about partisan proportionality. Considering that, the core metric must be some measure of proportionality. The efficiency gap seems fine, if a little unsophisticated; maybe a modified version could be used. One might be tempted to also add some sort of competitiveness requirement here, but it is important to keep this as a single measure. We don't want to introduce the design decision of weighting various factors; the whole point is that there must be a single scalar value for each map, so that there can be an single, objective choice. You also need to make sure that you are using a standard set of previous elections in order to map out the partisanship of the state: a weighted average of presidential, congressional, gubernatorial, and state legislative races over the last ~5 years should suffice.
However, just using partisan proportionality will introduce some new problems (of varying importance). You'll have ugly, snaking districts, "communities of interest" will be split all over the place, incumbents will be unceremoniously ousted, a bunch of majority-minority districts will be eliminated, and there is no guarantee of competitiveness. We shouldn't care about incumbents and, in my opinion, communities of interest are mostly fake, so I don't care about that. And while Section 2 of the Voting Rights Act is very important in the status quo environment, in a system guaranteeing some level of proportionality, I think it's mostly unnecessary. It's not 1970 anymore: white Democrats are happy to vote for non-white Democratic representatives. I could be wrong, but I don't think the number of non-white representatives would go down in such a world, so I don't think this needs to be a concern. That leaves two "real" concerns: ugly maps and uncompetitive maps. Uncompetitive maps are obviously a real concern, and even though smart people don't care about the aesthetics of electoral districts, we must always factor in the intelligence of the average GOP voter, who will throw a tantrum the second they see a shape more complicated than a hexagon. For these we should add one more element to the puzzle: minimum standards. For a map to be considered (i.e., ranked according to the proportionality metric), it must meet some minimum thresholds of compactness and competitiveness. The metrics used for these should be simple, but they must be designed to never produce a situation where they are disjoint (i.e., where all sufficiently compact maps are insufficiently competitive).
For compactness, the simplest option seems to be the classic perimeter-area ratio, but I worry that it punishes districts on borders and I feel like the math behind it is a little unsophisticated. There's probably a better choice out there (perhaps some weighted distance from the centroid?). And Republicans love their "split-county" metrics, so I guess we can throw one of those in there too. For competitiveness, it's a little harder. Some states are going to be too small for true competition, and some are just going to have weird partisan distributions. But I'm sure some smart mathematical formula is out there. These criteria should, in general, be pretty light-handed. The main work here should be the optimization of the proportionality metric, not a set of hard-to-reach minimum requirements.
This system eliminates gerrymandering without needing to trust a nominally independent body. It allows anyone (the parties, academia, hobbyists, etc.) to generate maps using the fanciest methods available, and we just use the one that is objectively the best. No maintenance of public code, no need to update the law to account for new technology; just a uniform standard across the country about how maps are drawn (aside: they should also apply to state legislatures, but that might be a harder legal battle). FPTP sucks, but if we're going to use it, let's not be imbeciles about it.