An impossibility theorem for gerrymandering
dustingmixon.wordpress.com
dustingmixon.wordpress.com
I think an eye opening example is to consider a hypothetical single state with 10 representatives. This state has a population, perfectly well distributed, of 35% democrat, 30% republican, 20% libertarian, and 15% green. And we draw our districts up with absolutely no bias whatsoever. What 'should' the election results be? I think any logical person would say about 3.5 democrats, 3 republicans, 2 libertarians, and about 1.5 greens. What we get in reality? 10 democrats as they win a plurality in each and every district. 0 gerrymandering, 35% of the support, 100% of the seats.
To get our above example to have the logical results we want, you end up having to create really absurd districts that would completely dwarf any sort of gerrymandering of today. So then the question is quite obvious. Why don't we do at large proportional elections instead of district based plurality elections? If a state has 10 representatives then any party is guaranteed at least 1 seat if they get 10% support, 2 for 20%, and so on.
In the past when politicians and the people were much closer, I think districts made a lot more sense. You were voting for somebody who you knew or were at worst split by the most minimal degrees of separation. But today this isn't true. The average congressional district is now up to 710,767 people [1] (as of 2010). In 1790, at the time of the first US census, the average district size was 37,082. Times have changed, but our electoral system has not.
[1] - https://en.wikipedia.org/wiki/List_of_United_States_congress...
[1] http://www.fairvote.org/cambridge-massachusetts-elections-a-...
The ideal is that candidates can rest on their own merits, with the party mechanic existing to provide support to candidates who nominally align with the party's interests. People vote for candidates, not parties.
Even the reality is that people are not determined by their party affiliation; New York City, for example, has repeatedly elected Republican mayors despite being overwhelmingly Democratic in state and national elections. This has even been used in a derogatory sense, as "Rockefeller Republicans" or "New York Republicans"; people affiliated with the national party but who skew to the preferences of voters in their region.
In your hypothetical state, once you created the districts, then the competition inside that district is between candidates who try to align themselves with that region; people who, for example, can be a "green-republican" to draw voters from both blocks, or just a person who speaks to the needs of voters in that district, regardless of party affiliation.
Using some sort of state-wide rank-preferential or approval system means that we remove a lot of the strategizing on the part of voters, which is a positive, but the strategizing on the part of political parties and individuals seeking office become much more complex, and not in a way that serves to present candidates that appeal to issues confronted by groups of people regardless of their party affiliation. Geography is just a heuristic for this, but not the worst one you could think of.
The reason for this is because of our electoral system. Imagine we have 4 parties each with about a quarter of the support. One of those parties is going to get 100% of the representation, and the rest are going to get 0. This incentivizes these parties to begin to merge. Two parties go 'Hey I know we have nothing in common, but if we work together we can guarantee our voices are at least heard in congress.' And the other two also see this going on and does the exact same thing. Next thing you know it you have things like libertarians and evangelical Christians both being supposedly represented by the same party, and you get a congressional approval rating in the teens.
Proportional representation changes this. Fractional support being more than sufficient to get seats in congress means people can break into parties that actually fundamentally represent their views - instead of being forced to clump up into super-parties. I think this effect would be particularly emphasized in our country. We have a phenomenally good system of checks and balances ensuring that even 'the little guy' in congress can have a meaningful effect. The problem is our electoral system all but precludes there being any little guys in congress.
Proportional representation, as I mention above, does remove most incentives to vote other than your actual preferences, but it does introduce very complex strategies on the part of candidates and parties. Every election becomes a state-wide election, so smaller candidates representing regional interests will find it harder to target their campaign. I'm not sure whether this is a net positive or negative.
In our current system the one and only power of smaller candidates is a spoiler effect. This enables them to have some influence on the super-party that's closest to their own ideology, even if implicitly. Rank preferential / Australian voting / etc not only fail to change the 10% support = 0% representation issue, but they even remove the spoiler effect which means that smaller parties can safely be completely ignored unless they look to gain a plurality themselves - which is rather contrary to the notion of smaller.
On this note, I'm not seeing why you think that proportional would make campaigning more difficult for 'smaller candidates.' It would mean they might actually stand a chance of finally getting a seat in congress. A basic example from Europe would be something like 'The Pirate Party.' They're never likely to gain substantial support, but they have been able to receive representation at the national level in a variety of nations, proportional to their support, exclusively due to proportional systems.
This doesn't really negate your point, I just figured you may be interested. They may not literally all be a D or an R, but it's trivial and I'm mostly being pedantic.
My bad, indeed.
(Mainah, ayuh.)
The total population of Maine is 1.3 million. The population of Vermont is 0.62 million. These states are smaller than many cities now a days. And you'll see a similar pragmatic politics in many of the other very small states. I think that's because you minimize the degrees of separation between representative and voter. When people know and meet other people it humanizes them and makes it easy to see that a letter beside somebody's name doesn't define them.
Our system was built for a different time, and we can see it's still a pretty reasonable system in the areas where that time is kind of emulated. But for many other parts of the country the system's age is showing. Even if we tried to keep up with how the founding fathers saw it we'd have literally thousands of individuals in the house of representatives which is probably similarly unworkable - and rather comical compared to the unchanging 100 of the senate.
We actually approved a change in our constitution because of the current situation. Our current govenor, he's an idiot, was elected with 34% of the votes. The previous election saw him win with something like 38% of the votes.
So, we had a referendum on the ballot and we will now have run-off voting.
I can, within reason, live anywhere on the planet. I've enough resources to get resident status and citizenship with pretty much every country. I retired and looked all over. I didn't 'settle' on Maine. I think that confers a negative view. So, I picked Maine because it best suits my needs.
[1] https://en.wikipedia.org/wiki/Proportional_representation#Hi...
Under MMP, you have two votes: 1) The delegate you want to represent your district (we call it an electorate) 2) The party you want to support.
In addition there are two changes: 1) Each of the parties publishes a ranked list of their candidates. This is imaginatively called their "list". 2) There are more seats in the house (we call it parliament) than there are districts (electorates). We have 71 districts but nominally 120 seats in parliament.
Once everyone has cast their two votes (quick brag: We had 92.4% of our population enrolled to vote, and of those enrolled, 79.8% voted -- which I think is pretty good going). The votes are counted like so:
1) The "electorate vote" determines the person who represents your district, in a simple most-votes-wins way. Anyone who wins a district is guaranteed a seat in the house (parliament).
2) The "party vote" is counted up nationwide, and each party gets that percentage of the total seats in parliament. If they won more seats with the party vote than the electorate (almost always), their seats are "topped up" from their list until they have the correct number.
(There are some subtleties: The exact way we assign seats is http://www.elections.org.nz/voting-system/mmp-voting-system/... )
This means that each electorate/district gets to choose its representative directly, so local representation is retained. At the same time, the house/parliament reflects the global percentage of votes, so smaller parties can still get proportional representation in government.
Because the seats going to the smaller parties would come at the expense of the parties currently in power, who are thus strongly disincentivized against changing the rules towards proportionality.
We could, in principle, get rid of the districts and just have some kind of statewide scheme for electing representatives proportional to the outcome of a popular vote. The idea behind retaining districts, as far as I understand it, is that people in similar locales will often have very correlated interests (e.g. with regard to decisions that affect that area) and that choosing representatives from particular locales is supposed to guarantee that people in that area are represented (and make the representatives to some degree beholden to them).
Ideally, we'd split the states into districts in a way that doesn't deviate from overall proportional representation very much, hence the idea of the efficiency gap. So gerrymandering can break that efficiency constraint.
But the notion of geographic correlation of interests makes it such that the population is unlikely to be distributed in the kind of uniformly random way you need to get weirdly shaped, efficient districts, and conversely having weirdly shaped districts at all effectively assumes that geographic correlation of interests isn't an important factor.
In other words: if we have to make weirdly shaped districts to maintain efficiency, then the assumptions that caused us to require districts in the first place probably fail and we should just ditch them.
That's why the efficiency gap is so exciting as a measure.
Every voting system has its flaws, but the US system seems to have particularly few advantages.
Many people prefer to spend time thinking within the context of what is likely to happen. That's not to say we don't need folks who dream big, but we need more than just those folks.
And by big-picture standards, because of the way the US is constituted, the questions that make their way to SCOTUS are usually pretty strange.
The data scientist in me wants centroids around circles with fuzzy edges. If we could all agree that fuzzy circles is the way forward we could solve gerrymandering over night.
The reality is though, that I create these amazing pieces of analysis then have to explain away why that district is odd, y'know, cos the shape is just weird.
Locally they are redoing the district boundaries which is giving me an opportunity to submit my own district shapes that work better with Census data :D
As a human who takes pride and joy in the Constitution and its tenets I find the crowbar-able gap space of redistricting an important valley of electoral attention that needs ample light of attention.
Or, we could go with proportional representation, make it a single district, and now we can reasonably expect to get 2-3 blue reps and 2-3 red ones, handling both the districting problem and efficiency/fairness problem in one go.
https://en.wikipedia.org/wiki/Single_transferable_vote#Findi...
The constituents also get much better representation, because they are much closer to the person representing them. The representative is intimately familiar with their constituents concerns because they live in and around them.
It also strips a lot of the money out of politics. Representatives don't have to spend all their time fundraising to raise war chests of millions of dollars. You can win an election in a 30,000 person district by knocking on doors and sending out inexpensive mailers.
Italics mine. Is there any reason to believe that any real life situation would fall under their impossibility? The key is to get a system which is good enough to prevent gerrymandering for the world that exists, not for any possible world.
Its a solid proposal, but I guess its up to the courts to decide if "Efficiency Gap" is a good metric. I guess I'll be interested in hearing the arguments as they come up.
Also, the concept of a 'blue/red' district can be very blurry - there are many examples of 'blue' districts (districts with democrat congress reps, and a history of voting blue) that voted red in the last presidential election, and vice-versa for previous elections.
Historically, Americans prefer to vote for people, not parties.
Should a new highway replace the housing development over there? Well, people who live close to that housing development might care one way or the other (Noise complaints, Ecological issues, Traffic Issues). People who live very far away from that center probably don't care at all. And the people who live in the housing development (who will be forced to move out), especially renters who probably won't be properly compensated, will care the most.
The placement of schools, the budgets of police, the design of zoning regulations (in particular "Enterprise Zones" of lower taxes to encourage business development in some areas)... these all are innately local issues.
But that create inefficiencies as that heavily favour some parties. So we have a chunk of undistricted delegates that are given out to even out the overall proportionality. This get you pretty good results as long as the amount of parties stay within reasonable boundaries[].
[] Which they of course haven't - we currently have 8 parties in the Riksdag.
I've always thought they should just reverse the order of operations. Instead of having people draw the districts, then machines evaluate them, they should have machines draw a bunch of potential "low efficiency gap" districts, then let people pick the best ones.
If a machine gave you a crazy-shaped district map because that's the only way to get "low efficiency gap", would you agree to them?
[0] https://www.census.gov/history/pdf/2010questionnaire.pdf Page 2 question #6.
100% agree. I think this can be further broken down into the questions:
(1) What are the problems caused by extreme gerrymandering?
(2) How will efficiency gap minimization alleviate these problems?
(3) What assumptions are we making in (2)? How likely are they to be true? And how robust is the solution to their falsification?
(4) How could this all go catastrophically wrong?
My intuition is that efficiency gap minimization is an incomplete solution and probably should be only one of several factors considered.
First of all, it seems highly unlikely such a distribution of red and blue votes would occur. But more importantly, blue is a majority of every 3x3 square - why shouldn't it win every district?
Because we are supposed to have a representative government. If there's 55% Blue voters and 45% Red Voters, you're supposed to ideally have 55% Blue representatives and 45% Red representatives.
The example is a counterexample of this concept. Instead of 55% Blue / 45% Red split, the map as a whole becomes 100% blue. This is almost the very definition of "Tyranny of the Majority".
Maryland is an example of a highly gerrymandered state (towards Democrats, although there are many Republican examples too). Maryland voted 60.5% Clinton / 35.3% Trump, which suggests that they have (roughly) 2 Democrats for every 1 Republican in the state.
However, due to the significant gerrymandering in Maryland, there is ONE Republican and SEVEN Democrats in the House. That's certainly not fair towards Republicans who live in Maryland.
EDIT: Apparently North Carolina is the Republican example to talk about, in case you want some "balance".
In Maryland's case it might be able to make 2 districts Republican, but that's assuming a very concentrated geographic minority which may or may not actually exist.
IMO, If we really want proportional representation then we should use a proportional system.
So once again, if you're talking about US Representatives and Voting Districts, it only makes sense to talk about well... Voting Districts. Name me one US State with 80/20 split and more than 1 US Representative. Ultimately, your hypothetical doesn't exist!
---------
That's the cool thing about politics: its real. There's no need to make up hypotheticals when we can just draw from the real world.
FYI: the discussion at hand is about US Representatives, and not about the Electoral College. Washington DC is therefore irrelevant, as it has no representatives. Washington DC does get 3 Electoral college votes, but that has nothing to do with Gerrymandering.
There's a variety of states who have "Representatives At Large", such as Wyoming, which are basically immune to Gerrymandering. The population is so low that Wyoming only gets 1-Representative, so the entire State is the whole district. There's really no "fair" way to cut up a single Representative (its all or nothing), but that's mostly due to the very low population of these states.
Basically, Gerrymandering can only be an issue in a state with more than 1 Representative.
https://www.dcboe.org/election_info/election_results/v3/2016...
So, no you can't dodge the question. And I ask you to draw a DC map with some Republican representation.
PS: AKA draw a map such that WARD EIGHT MEMBER OF THE COUNCIL is likely Republican.
But my general interest is how this whole event will play out in the greater scope of the country. I don't believe any US State with more than 1 Representative has an 80/20 split.
Perhaps Washington DC really is an edge case that needs to be thought out more. But I don't think its representative of the problems of Gerrymandering that exist in multiple states right now. (In particular, what this Gill v Whitford Supreme Court case is bringing up in Wisconsin)
Still, my concern is not about the results. My concern is how easy it is to game this proposal.
Approximately the same difficulty to game the US Court system.
At the end of the day, Judges will adapt the rules to future cases. Law is an innately human process, not the cold machine logic that most of us Computer Science folks deal with.
If a major mistake is made in the current rubric, future lawyers and judges will bring it up on a case-by-case basis. Unlike code, Law is pliable and changes on local conditions. The issue isn't the "edge cases" which are dealt with as they come up.
If the proposal isn't working for say, Washington DC, then you can expect a case to bubble up through the Washington DC Courts, to maybe the US Court of Appeals, and eventually back to the Supreme Court if its major enough. If its purely a Washington DC issue (and not applicable anywhere else), there might not be any need for the US Court of appeals or US Supreme Court to hear about the case!
Blue is indeed a slight majority of every "3x lattice"-aligned square. But if you take the top left square and shift one down and one right, you'll see that in fact red wins that 3x3 square decisively: 8 to 1.
> why shouldn't it win every district?
Maybe you think it should, but whoever drew the districts would run afoul of the newish gerrymandering measure the Supreme Court is considering.
Imagine that a few voters in each district change their minds, some from each party switch to the other. Even if the overall effect is a very slight swing towards red you could still end up with all the districts going to red. Or with the same swing but distributed differently you could still end up with blue winning (though not by as much), even though it had fewer total votes.
So it's not about what should happen in any one particular snapshot. That's just a single example. It's about how well the outcome of a series of elections should reflect the wishes of voters as a whole. I'm not a big fan of proportional representation, it has often lead to small minorities gaining kingmaking powers, but I do believe that the objective of electoral districting should be to produce results that are likely to fairly represent the wishes of the electorate.
A voting system should be explicit about exactly which demographic criteria should carry weighted bias, what value that weight should be, and what it's purpose is.
It should also be clear to each voter which demographic categories they fall under, what the final strength of their vote will be, and what the justification is for that outcome.
A fair metric minimizes average distance from a citizen to the centroid of their district. The lowest energy Voronoi diagram wins. Nobody gets screwed.
Trying to ensure proportionality in representation... is that the intended goal? Is this even a meaningful pursuit in a bicameral system? The Supreme Court makes rulings that span time infinitely forward, meaning that all those different parties that have existed since the inception of the Union have dissolved and new ones have been formed since, but the rulings of the Supreme Court have not this luxury of death and recycling. Nay, a ruling by the court is very potent and quite a footprint on the legal plain.
So there is no easy way to ensure that a district is Constitutionally Compliant, but we can definitely point out "unconstitutional districting" -- right? Do you agree?
Back in the day redistricting was even crazier. You could just draw an arbitrary number of districts with arbitrary population sizes. That seems bad on the surface. With this system, election outcomes could potentially be entirely determined by how districts were gerrymandered.
The efficiency standard seems like trick to get closer to proportional representation. All of this comes down to the 14th amendment being "one person one vote" rather than something like "all votes must have equal weight". A small language that has made things really complicated.
IMO just switching to proportional representation is a much cleaner solution. But the efficiency standard seems like a decent solution to keeping races competitive and from being severely gerrymandered.
Reading up on the 14th amendment, it does ensure that every male over 21 is able to vote (assuming they did not lose the right through crime, also a big issue today as a third of our populous is disenfranchised). It does not, however, say anything about how voting for reps is done at the state level, and I wonder if this becomes a States' matter as soon as we part from the 14th amendment. Pardon my gap-riddled knowledge of such important affairs
I believe the 14th amendment specifically applies to state elections under the "equal protection clause".
I know Maine allows it, and even encourages it. I actually started a campaign for State Senate at one point. As I got closer to election, I learned more about the system and decided I couldn't participate in that. So, I never turned in all the signatures and didn't run.
But, thanks for the correction. For some reason, I'd assumed the rest of the States allowed it. I understand that, in Maine, some candidates have even campaigned, in person, in the prisons. Notably, we don't actually have a lot of prisoners or very dangerous prisons.