▲ 129 points
back
203 comments
There's a far better solution to gerrymandering than these creative ways to draw lines on a map. It's called proportional representation, and it's used in many countries in Europe (usually combined with the d'Hondt formula). FPTP leads to far more silliness than PR and therefore countries using PR have healthier political discourse and systems. Sad that the UK rejected the proposal to abolish FPTP in the early 10s.
Edit: changed this tautological sounding sentence - "None of the countries using PR suffer from the sillinesses arising from FPTP."
Yes, but we have to consider what is politically feasible or plausible.
The US is not going to switch to proportional representation at any point in the near to medium future. It's just not going to happen. There's zero public awareness or desire -- right now we're having a hard enough time just making the much smaller incremental improvement of ranked voting.
On the other hand, there's no strong tradition of any kind behind drawing districts. If someone really can come up with a fair, objective way of doing it, it's far easier to see it getting adopted. Gradually at first (state by state), and then possibly nationally. Still difficult, but within the realm of possibility (unlike PR).
> None of the countries using PR suffer from the sillinesses arising from FPTP
Frankly, I'd be surprised if they did. It seems much more likely that they'd suffer from the problems of PR instead.
I strongly disagree. Its not just the ratio that matters, but whether your representative is geographically close to you.
The American ideal is NOT just that you have a Representative, but that Representative knows YOUR community.
Proportional representation could result in all my States Representatives coming from LA or SF, but not the rural areas.
Ideally, I'd like Representatives that I'd have realistically met before in normal day to day living in the community. We need it representative for every 25,000 people, in my opinion. and that representative coming from my local community of 25,000 or so people.
It's also an incredibly effective way of getting fascists elected - who can then use their platform to gain legitimacy and support. I look at Europe and see fascist parties gaining seats in every country - except the UK, because in the UK FPTP will ensure they never get one. Even UKIP, never won a seat, despite being hugely influential populists.
Regarding the "alternative vote" referendum in the UK: the voting system (also called "ranked-choice voting" or "instant-runoff voting") is an improvement to first past the post, but it's worth noting that it is not proportional representation and would do little to dissuade gerrymandering.
I don't think PR would work for the US, at least in its most common(?) form where people vote for parties and each party has a list of candidates.
A big problem with Gerrymandering is that it favors extremists. When a district is competitive, a candidate must appeal broadly to people of various political positions. When a district is artificially created to cater to party X, a candidate only has to be selected by party X, because then they're practically guaranteed to win. Thus candidates who are more loyal to party X is more likely to be elected, regardless of how they're viewed by the general population.
PR has the same problem - someone who's at #1 position of Party X's list is as good as elected, and the position is determined by internal party politics. Thus, the top positions will be filled by those who always toe the party line and never compromise.
I don't hear all that much about gerrymandering from Britain. And the referendum to abolish FPTP was IIRC not to abolish single-winner districts.
Although I agree PR is better. The German version also even keeps the districts if you really like them.
While PR seems good in principle if we look at the actual outcomes they don’t end up so great.
Unless you have an electorate wide multi-member electorate (not as common as proportional systems rare) you will require geographic electoral boundaries to be drawn.
The UK’s AV referendum was not for proportional representation but instant runoff voting for single member electorates. Which is a better system but in the UK in 2011 clearly would benefit the Lib Dems much more than actual multi member electorates would.
You cut, I choose becomes I choose, you un-cut.
Note that this paper contains no optimality proofs.
There's a lot of mathy-looking stuff, but then instead of a proof we get "yeah we took this one particular Markov-chain algorithm and had it play against itself and it failed to generate a gerrymander".
If deployed in real life, an immense amount of power -- and therefore money -- will be at stake: certainly more than enough money to hire persons with intelligence equal to or greater than the creator of "the ShotBurst Algorithm" as consultants.
If an optimality proof is not possible, the only alternative is to test using real humans and (very substantial) real rewards.
Part of the problem is the so-called "two-party system" that we effectively had. George Washington had the right idea - political parties will divide and ultimately destroy the nation. We see this happening right now, today, as the Democrats and Republicans argue over high-profile "moral" issues while colluding in the background to benefit the large corporations and disadvantage the individual citizen.
Abolish political parties now, before it's too late. It may already be too late.
I only read the abstract and very briefly skimmed the article, but...
So it claims that it doesn't require bipartisan agreement to reach 'optimality'. Fine, I won't argue that point as I haven't looked that deeply into it.
But it still requires the parties to agree to adopt this Define-Combine procedure in the first place! And why would they do that? They're not, when the rubber hits the road, interested in "optimality" (however you define it), but power. Power at the expense of the other party, to be precise.
Gerrymandering is a political process; it can't be "solved" with math. If there were a technical solution to the 'gerrymandering problem' we'd've used any of the standard tools for drawing such maps.
The major sociopolitical issue is that gerrymandering can be used for both good (minority-majority districts) and "bad" (supermajority plural houses when there's only a bare majority; or, worse).
The root of the issue is that map-drawing is unfair and we'd like a fair system. I think for most of the interested parties, you can find middle ground on fair electoral mechanisms. For parties just trying to keep power you can make this argument: gerrymandering moves the election to the primary; the turnout for primaries is abysmal, and the result is that our government is weaker for having gerrymandering.
In my state, Texas, there are some districts where the outcome of the election was decided by as little as 2% of the voters, i.e., only 3.5% of the voters showed up to the primary.
Can't you just draw the N districts that you want, then break out N new districts from those by extracting a "ring" from each. Then you have N districts each fully surrounded by another district. The only way to merge pairs of adjacent districts is to restore the original layout.
Seems like a pretty trivial way to break the algorithm. But I probably misunderstood something.
On the one hand -- I love this spirit, of trying to adapt the "I cut you choose" technique to something that might work for gerrymandering. We need as many ideas as we can get!
On the other -- unfortunately I don't see how it would actually help, because the first party which defines the half-districts is generally going to "win".
If party 1 is the Republican party which wants to maximize rural votes, they just split up every city like a pie, with each slice extending far into the countryside, halfway to the next city. Then it doesn't matter which slices party 2 (Democrats) combine, because they're always stuck with rural votes outweighing city votes.
While if party 1 is the Democratic party which wants to concentrate city votes, they do the opposite and surround the city with two halves of a ring that is half-urban half-rural, so the city can't be diluted. Party 2 (Republicans) can combine rural areas only with the ring, while the inside remains totally protected.
So one strategy is skinny pointed daggers to invade the cities, the other is about skinny ring shields to protect the cities. By controlling the initial shapes, you control nearly everything.
(Of course with real geography it becomes more complex, but the general principle of districts shaped as skinny daggers or shields is still going to be the primary strategy here -- you're not going to win all the districts, but it'll get you the extra 10% or 15% you need.)
I'm one of the authors. Thanks for reading our paper. Happy to answer any questions.
If you're interested, here is a (still in-progress) simulator I wrote where you can try out Define-Combine on a simple grid. https://mpalmer.shinyapps.io/DefineCombine/
To everyone in this thread who is stuck on the refrain "technical solutions to social problems never work": broadly shared agreement among people of flesh and blood to use a certain technical solution is ultimately a social solution. The value of a fair technical solution is as a rallying point around which a social solution can be formed. Besides, by the same token any democratic voting system would be deemed a "technical solution to a social problem", yet clearly some of those were adopted and now work.
If the analysis holds up in practice (and the methodology looks good at first glance) I really like this. It will allow fair districts to be drawn even by completely divided jurisdictions, without needing any kind of neutral arbitrator to step in.
(The UK has reasonably neutral boundary commissions, but it seems for whatever reason there's no hope of that approach working in the US -- at best there's just an almighty legal battle in state supreme courts.)
I've always felt that a simpler solution would be to just increase the number of representatives.
In 1912 (the last time we added more), the US population was ~90 million, and it's currently at around ~340 mil and we still have only 435 representatives.
1912: 1 rep per 20,000 people
2024: 1 rep per 780,000 people
This doesn't "solve" gerrymandering but it does blunt the effectiveness of it (among hopefully some other positive effects.)
The good thing about this is that it removes the problem of finding selfless competent people to carry out the process impartially.
It works when both sides are cynical crooks.
Dirt don't vote, so why are we drawing circles on a map? Just section off a list a of eligible voter' names or IDs. After all you're not representing area, you're representing unity.
The solution to gerrymandering is multimember district with proportional representation inside the districts, whether small (e.g., 5 member districts with STV) or large (whole-legislature “district" with party-list proportional) or hybrid (any form of Mixed Member Proportional).
Better proportionality in representarion also has a whole host of other benefits.
Everything with messing with how you draw lines in single-member winner-take-all districts is just shifting representation problems around and not actually solving them.
In addition to these big-move structural proposals that will take a long time to actually be adopted, there's something everyone can do now that can have an immediate effect: vote in primaries. Gerrymandering has the effect of making the general election meaningless, so the dominant party's primary becomes the whole game. If you're a Republican in Berkeley, for example, you don't need a lot of Republican and moderate voters to tilt the local Democratic primary, because not a lot of people vote in them.
https://www.theatlantic.com/ideas/archive/2022/05/primary-el...
This solution is still based on a concept of defining districts by presumed voting patterns of the inhabitants. Is it really that hard to define districts sensibly purely based on population density/distribution?
I wish they could come up with a solution to this that made sense. The things I find particularly troubling about gerrymandering in the US isn’t only that one party gets a few more representatives either way. The reason it’s absolutely terrible in the United States is that it creates uncompetitive districts on both sides and we get extreme candidates that don’t appeal to people in the middle at all and seem to have no ability to compromise and get basic things done. I’m not sure that proportional representation addresses this issue.
How about communities decide themselves which district they want to be part of?
A community is a village, town, administrative parts of a city, etc.
Every 10-15 years, there is a vote across the country in which every community rates adjacent and nearby communities according to who they want to be in the same district as.
Then you run some pre-decided fixed algorithm that uses these ratings to join adjacent communities in a province/state in roughly equal populations. The design of this algorithm is left as an exercise to the reader.
What are the advantages over "I cut, you freeze"?
https://www.cmu.edu/news/stories/archives/2017/november/i-cu...
Couldn't we just use proportional representation?
Both parties heavily invest in gerrymandering
and both parties apepar to (think) they get an
advantage out of doing it.
Neither wants the practice to stop. Politicians will pretend outrage when they personally are negatively impacted by a change of course.
I think there ought to be precise rules and regulations for how and what you can do, but of course for the reason above, this wont happen.
The real problem is the “gerrymandering” of state boundaries and the Senate. CA has around 10 million non-Democrat voters who have had zero representation in the Senate since 1992. In fact the entire West Coast hasn’t had any non-Democrat senatorial representation since 2009.
This is why states like the Stare of Jefferson need to be allowed to exist.
I always thought a simple geometric solution would make a huge difference (e.g. a district can have no more than 360 degrees of internal arc (with some modification for districts that border non linear state lines).
It would be far from perfect, but it would take the precision out of the gerrymandering process which I think by would be a net positive.
What happens when there's collusion, e.g. protecting incumbency at all costs? This tends to be the M.O. here in Florida, where the Democrats in office are satisfied without statewide power as long as their asses continue to be the ones in the allotted seats. Algorithms prevent this situation but your way doesn't.
Even if it's stable in a game theory sense, it depends on the political class being partitioned into exactly two parties. Otherwise the two biggest parties would have an incentive to smother all smaller competitors.
I think the answer has to lie in an algorithmic solution dependent on a high-resolution population density map.
You do still need the controlling party to agree to enact this solution, which is a form of compromise. If you're willing to do that you may as well be willing to cede control to a nonpartisan third party. In either case the outcome will be a reduction in your party's power.
This problem has been solved by mathematicians:
A partisan districting protocol with provably nonpartisan outcomes https://arxiv.org/abs/1710.08781 by Wesley Pegden, Ariel D. Procaccia, Dingli Yu
There's also at least one method to create N districts canonically -- no human intervention:
> 2 contiguous, equal-population districts
What this seems to ignore is the legitimate desire to keep towns or counties in the same district. Most people would consider it silly to have the dividing line be a side street, or even in between two houses.
Is this a generalized version of sharing a cake?
- I get to cut the cake in half
- you get to choose which slice to eat
This assumes the people doing gerrymandering want a solution
They don’t
The practice of gerrymandering is self acknowledged adversarial social engineering intended to create an unfair split to consolidate power. Is intentional and domination based
The singular and only possible way to execute equitable democracy is enthusiastic engagement by the population in the running of the system.
Until you can create a society that gives individuals the time, space, and resources they need to make fully informed decisions on the complexities and externalities of human action, societies will continue to be run by a small handful of high energy narcissists.
I am unaware of any large scale (>1000 people) human communities that have demonstrated long term peaceful self organized society, spanning the lifetime of a citizen (~100 years) without a major destabilizing event in their community.
Haha, like my granny used to say when there was cake to share. "You cut, I choose".
A worthy variation of "I cut you choose" --- the only fair way to divide anything.
Technology for direct democracy makes the need for elected representatives obsolete.
This is yet another technical solution to what is fundamentally an incentives problem. Whoever benefits from the status quo (gerrymandering) will naturally be incentivized to resist any change which is objectively more 'fair'.
That's the real problem. It's nothing to do with how we draw lines on a map, it's how do you get whoever is currently in power to agree to a more fair system, given that it's (by definition) not in their interest.
This methods only works for a two-party electoral system...
Is this actual political science?
If so, where I can I find more?
The problem isn't that there isn't good solutions to this problem. The problem is that the parties in charge don't want good solutions to the problem.
" One party defines a map of 2 equal-population contiguous districts. Then the second party combines pairs of contiguous districts to create the final map of N districts."
What a load of crap and what dream world do they think that could be accomplished? We can't even get ranked voting to work, and most places this past year where they did use it, it's gonna go away because they didn't get the nutjub they wanted, and this is the age of the nutjub.
Most Americans want to gerrymander, so do their representatives. It's a total myth this will get fixed or people have the will to do it or that you can somehow have these 2 parties work together, at all. Even if that's some silly "contiguous and combine" setup. Shit, one party would get angry with you for even using a word like "contiguous", that's the bar you have to get passed. Good luck with that.