Gerrymandering with geographically compact districts
dustingmixon.wordpress.com
dustingmixon.wordpress.com
All to say that, while different methods for representation should absolutely be considered and discussed, it's important to remember that none of these are a panacea, and we should be upfront about the nuances and disadvantages of every option.
Well, other than that more proportional results, by whatever means, seem consistently across modern democracies to provide government with which the people are more satisfied, and also provide a higher dimensionality to the range of political ideas and debate. (See Lijphart’s Patterns of Democracy.)
Every options has positive and negative traits, but single-member FPTP districts are about the worst, on balance.
Well, worst would be that plus having districts that don't even try to have similar population, so that the results are even less proportional than you'd expect with FPTP in single-member districts.
The solution to that, however politically difficult, is to break the larger states into smaller ones. New York City, for example, shares very little politically or economically with the rest of New York. Similarly with the various regions of California -- Silicon Valley and Los Angeles are as different from each other as both are from the areas north of San Francisco.
If a liberal state does it, it costs liberals votes. If a conservative state does it, it costs conservatives votes. It seems pretty clear that such a system is fairer, and yet no-one wants to actually do it.
> As shown above, in 1804 each Representative represented approximately 40 thousand people. Today, the average population of congressional districts is nearly 700 thousand and growing.
> In order to restore the House to the citizenry, our total number of federal Representatives will have to be increased substantially. Achieving this goal is essential to extending the ground of public confidence in our government and ensuring the beneficent ends of its institution.
It's doubtful whether that still exists as anything more than a foil/justification though, given Congress's expansion was stopped a century back, and each Representatives now stands in for at least half a million people (RH has the lowest number of citizens per rep' at 530k) and up to a million (Montana, population 1042520, currently has a single rep'), with the average closing in on 740k (up from ~710 during the last census).
Can a rep' really vouch for the specific needs of 700000 people any more than a proportional delegation would?
It's unclear to me that the House of Commons manages to do that, and they've got 650 MPs for 65 million people (or ~100000 people per MP on average, though the average electorate is lower, and it has very extreme low ends, Na h-Eileanan an Iar has an electorate of under 22k).
[]: I don't mean to say FPTP is undemocratic - Canada is plainly democratic - just that America isn't democratic in much the same way one says Russia isn't democratic. If the government is intent on not hearing the will of the people, it will find a way.
Not if they establish their own parties.
TL;DR: With FPTP, each belligerent tries to form the smallest winning coalition, resulting in a near even split two-party system.
Proportional Representation is the remedy for assemblies, councils, houses.
Approval voting is the remedy for executive races (eg sheriff, mayor, governor).
Additional benefits of both PR and Approval Voting:
- no more primaries
- fewer elections means reduced voter fatigue
- increased turnout, because real competition boosts participation
- greatly reduces mudslinging, because everyone wants to be your second choice
- no more spoilers, making an incubator for third parties
You may think for various reasons that it's better to form coalitions at one level of the voting process than the other, but let's not pretend PR avoids it.
The reason why we have so many bills falling through is because of the razor thin margins in the Senate, so that a few dissenters can affect things. But if you look at how votes in legislature went down historically, we're pretty much at peak partisanship - it used to be a lot more common for congressmen to vote across party lines.
There are many other accounts, giving different perspectives, of course. The Politics of Attention and Unorthodox Lawmaking come to mind.
> The difference is that legislative coalitions don't have to be permanent, the way parties are.
Just like the parties that make up a legislative coalitions can change, the voters making up a political party can and do change.
> Various factions in the legislature can come together for a particular bill, or series of bills on some issue that's relevant to them all. But on something else, they might realign differently
This happens all the time on individual bills in a two-party systems as well. ("Crossing the aisle".) But the coalitions forming control of the legislative body rarely change except immediately following elections (in both PR and two-party systems).
> Also, even when coalitions are more persistent, the terms are effectively renegotiated every election - party platforms are much more stable.
The analog of a particular legislative coalitions in a two-party system is not a particular party but the party in power. And the platform of the party in power changes dramatically when power shifts from one party to another.
Some PR countries have strong two party systems (see most obviously Malta).
There is something additional in America that causes a two party system which may or may not continue past a switch to PR because we don't know what it is.
Also, it is good to specify your goals. You might want to make it easier for unconnected people to enter the assembly, or to make it easier to throw out the choice of the primary voters, or literally just a more fragmented assembly. It is easier to come up with useful compromises if you specify your specific objectives because not every system of PR will have any of those goals. Closed list PR with primaries is probably the antithesis of what you want, but it's an extremely common kind and what a lot of people hear when you say "PR".
Over simplistically:
#1 - Our FPTP elections leads to two major parties.
#2 - Our gerrymandered districts (local, state, congressional) has led to increased partisanship. As noted elsewhere, because contests are won during the low-turnout primaries where the motivated base participates.
#3 - Ever increasing campaign costs has led to concentrated influence (power law distribution of attention). Who ever controls the contributions controls the agenda.
#4 - We have city, district, county, state, federal elections. The party system starts at the bottom. Memories are long and defections are never forgiven (eg Nader & Green Party).
Having now done it... Everyone should run for office, at least once. Because most commentary, complaining isn't even wrong.
My goal, always, is to empower, enfranchise people. The various reforms I support increase voter choice and participation. At every level by various (sometimes non-intuitive) means.
I wrote something about this. https://asitoughttobe.com/2010/07/18/score-voting/
I point out this problem because it is one of the few times you will find Supreme Court cases and writings by the Founders on what algorithm to use. A bill changing the apportionment algorithm was subject to the very first presidential veto. The first proposed amendment to the Constitution (https://en.wikipedia.org/wiki/Congressional_Apportionment_Am...) was on the subject of apportionment. When else will you find the highest levels of government debating the merits of algorithms and mathematics?
What's the ideal size for a constituency? To maximize responsiveness, accountability, effectiveness? 100k? 200k? 400k?
I don't know, but I'd like to find out.
But yeah, I think at a few hundred thousand, though that would likely mean over one thousand representatives, could probably work.
This seems to be a pretty ideal number to me - it's about the size of the township I grew up in. Many people in their district would have personal ties to their representative, as a further detriment to attempts at lobbying/corruption.
The US constitution used to have a similar provision, but this was abandoned by the ammendment process in the late 1920's.
I have no idea if that's the ideal rule or how you might determine that.
There can be an amusing side-effect to aggressive gerrymandering. Let's say a state has a small majority of party A, but districts that are strongly partisan (let's say 80%) for A or B. You only need 50% to win your district -- that extra 30% is "wasted" and would be more valuable if those voters were pushed into one of B's districts.
Which is one of the things partisan redistricters do. Now they have more districts with a comfortable 55% margin and an assured lock on more districts, right? Umm, well, now they've made it easier for the other side to capture districts too...and it happens.
There's no magic way to avoid bizarre results (check out the wikipedia pages on the districting, and on voting methods, for some eye-opening problems), but something close to algorithmic probably ends up the least controversial and most comprehensible in the long run.
PS: IMO it's really a scale problem. If districts give a ~3% advantage that's huge in aggregate, but races still feel competitive. However, we are currently in a situation where nobody bothers to run in many races and that does not promote democracy.
One suggestion in this direction is the shortest split line algorithm: http://www.rangevoting.org/GerryExamples.html
(The OP also uses split lines, but they are optimized for maximum partisan effect rather than minimum length.)
"Assessing significance in a Markov chain without mixing"
We present a new statistical test to detect that a presented state of a reversible Markov chain was not chosen from a stationary distribution... We illustrate the use of our test with a potential application to the rigorous detection of gerrymandering in Congressional districtings.
When the first stage elections are contested among isolated ideological subsets of the electorate their results tend to extreme positions.
What I'm saying is this: it is inevitable that the REDMAP project that yielded the kinds of district-level results shown in this article will continue to generate more and more extremist election results.
Ryan and McConnell, and the whole US federal house of representatives, have bought and paid for their dysfunction with this systematic gerrymandering. The situation isn't going to improve as long as this holds.
Pouring money into contesting the general elections in these districts is entirely a waste. It enriches nobody but robocall and television companies, and doesn't affect the outcome.
It's still necessary to contest every race. Because of down ballot races. Moving the needle 5 points in a statewide race can mean winning a lot of local elections. Ditto putting initiatives on the ballot, which motivate your base to turn out.
On the other hand, primaries attract more motivated voters than the general. And those motivated voters tend to be motivated by their ideological positions - in other words, they're more extremist. From the perspective of candidates in the primary, they have to pander to those more extreme voters to get to the general - but they have no incentive to pander to the center, because even the more centrist members of their own party will vote for them over the opposition candidate anyway.
This also sets up a positive feedback loop, where even less ideologically radical voters, forced to vote for more and more extreme candidates of their party in the general, adopt at least some of the positions of those extreme candidates over time (making it easier to cast such a vote). Which in turn makes primaries more extreme, and shifts the overall party platform in that same direction.
We've already seen Republicans walk down that road, and where it got them. Now Democrats are going through the same process (Indivisible etc). It looks like we're undergoing the last political realignment possible in the current system, and at the end of it, we'll have two "ideologically pure" parties, with most voters going for straight party ticket, and with most legislative votes going down party lines. At which point the system is going to deadlock, because it pretty much requires some compromise to function properly.
It reminds me somewhat of a program on The History Channel called "How the States Got Their Shapes." And one of the really interesting parts was how states in the eastern part of the country had their lines drawn before the railroads and most in the west after the railroads. And so you see many more rivers, mountains and other natural features affecting state borders back east and a lot more straight border lines in the west. There's a very good reason (the Potomac) why Maryland's border looks the way it does, but if you just showed people the shape and told them it was a congressional district, it wouldn't be hard to convince them it was the result of gerrymandering.
I wonder whether there's a better way to visualize voting districts, perhaps using Google Maps transit data to visualize public transport/drive times to polling stations. There would still be the issue of deciding where to place polling stations, but it seems like once that's been decided, a district's shape should be derived from including all addresses that are most easily able to travel to those polling places. This would naturally create odd shapes, especially along freeways, bus/train lines and such. But making it as easy as possible for people to vote seems, to me, more important than having a simple shape.
And because I've never seen a social media post about gerrymandering favoring democrats, I'd like to add the efficiency gap chart from [2]. 8/26 states listed on the efficiency gap chart here are listed as favoring democrats. I fully recognize right now gerrymandering favors republicans significantly, however I'd like to note in some areas democrats are favored.
[1] https://fivethirtyeight.com/tag/the-gerrymandering-project/ [2] https://www.azavea.com/blog/2017/07/19/gerrymandered-states-...
CGP Grey did a great video on STV: https://youtu.be/l8XOZJkozfI
What could potentially help is to have a legislative body not determined by geography, possibly some kind of proportional party system (the downside being you need to make parties a formal part of the system, but they'll always be present anyways), that must ratify redistricting proposals and can challenge currently drawn districts.
PR is no panacea. To avoid gerrymandering, you need an adequately high district magnitude. After all, majority-preferential is just quota-preferential with a magnitude of 1. And FPTP is just List PR with a magnitude of 1.
Anyway, I was responding to this:
>>> I don't think STV presents any function that would really prevent gerrymandering.
I still believe that STV-PR resists gerrymandering as much as List PR does.
> PR is no panacea. To avoid gerrymandering, you need an adequately high district magnitude.
I agree, but I expect any form of PR would deal a substantial blow to gerrymandering, even if it isn't a panacea. Perhaps the only way to eliminate gerrymandering is to elect all representatives from a single multi-member constituency. (I expect any other system with enough compensatory List PR seats would be practically as resistant to gerrymandering.) Of course, the sticking point for STV-PR is that a ballot paper for STV435 wouldn't be very user friendly :) .
What do you think of STV+ as a compromise? Under STV+ most representatives are elected under STV-PR and a minority are elected through compensatory List PR to increase the precision of party proportionality.
I think Germany has a good solution where there's single-winner districts along with a proportional list. I also think giving authority of redistricting to an authority that can't directly benefit from it could also work.
That is what I mean. (In theory you could have an election with one enormous constituency under STV, but I accept that it wouldn't be very practical at government level, even if it works well for electing the boards of non-profit organisations.)
> Anything involving districting is, you can still split up populations so they have weaker representation, even under proportional systems, by making their representation much weaker in many districts and overly strong in a few.
True, but if I understand correctly that's about unequal district population, rather than gerrymandering. Proportional systems with a single constituency or compensatory List PR largely avoid concerns about districting.
> I think Germany has a good solution where there's single-winner districts along with a proportional list.
Germany does indeed have a good system, but I'd prefer it without the single-winner districts and with most representatives elected by STV (ie, the STV+ system described in the last paragraph of my other post: https://news.ycombinator.com/item?id=15970289 ).
No, it's not. With any proportional system (including STV), the more members per district, the less distortion is theoretically possible through differences in how district lines are drawn.
STV with multimember districts (not the degenerate single-member district form also known as IRV) limits the degree of distortion that you can get through how district lines are drawn (whether intentional gerrymandering or otherwise.)
Score Voting and Instant Runoff Voting are a bit more fair than Approval Voting, but much more difficult to implement, execute, audit, describe / educate, etc.
Both scores and approvals are ambiguous in meaning and different people with the exact same preferences would map ballot papers differently. (In situations other than standard public elections, this may not be the case: approval is great for voting on group activities where either “approve” is a binding commitment to participate if the choice is chosen, or “disapprove” is a binding opt-out if the option is chosen.)
Approval Voting seems like the best compromise between fairness and simplicity (to better ensure election integrity).
The OP also uses split lines, but they are optimized for maximum partisan effect.
Thanks for the summary. I’m at work and won’t be able to review he actual paper until later.
You eliminate gerrymandering by eliminating single-member FPTP legislative districts in favor of multimeter districts with a system producing proportional results (Single Transferrable Vote or some similar candidate-centered method, or a party-list method, though I prefer the former.)
STV works best in limited magnitude districts (three or five seems to be preferred) or else the number of options becomes too difficult for humans to express a coherent useful and accurate set of preferences. Once you limit magnitude, you can draw your districts to benefit you. For instance, perhaps you'll discover a preference for two or four seaters in reasonably strong Democrat areas and three seaters in reasonably strong Republican areas. You need fifty percent plus one to get a majority of seats in an odd numbered district, but fifty percent plus one only gets you fifty percent of the seats in an even numbered district. I can imagine Americans spending huge amounts of political capital for years getting STV in, only to discover that one party still has a manufactured majority because they wasted votes efficiently.
Its true that gerrymanders per se become harder and harder the higher the magnitude, but if your problem is you have a political system that is opposed to hearing the will of the people they're going to find a way to stuff their ears no matter what voting system you pick.
(Aside: Open list PR is both candidate centred and party-list. You vote for Fred Nurk of the Purple Party. That helps the Purple Party win more seats, and it moves Fred Nurk up in the list. It seems to be the only tick-a-box election system which doesn't devolve into FPTP in a district magnitude of 1 so it might be appealing in the US, where several states only elect a single representative. e.g. You vote for Gregory Hsien of the Maroon Party, who stands in a group with Josephine Yusuf of the Red Party. The Red-Maroon group win one more vote than the Blue-Cyan group even though the Blue party got more votes than the Red party. Reds win because you voted for their ally, the Maroons.)
The world underneath is full of other facts on the ground, like municipal boundaries, county lines, and the like. Districts that deviate too far from these other entities -- entities that are already being used for governance and administration -- then clearly some trickery is going on to deliberately cherry-pick people for some reason.
Tying districts to more closely resemble existing administration boundaries won't magically solve this issue -- packing, cracking, nonsensical groupings are still a danger -- but it's a sensible baseline expectation that can be used to build on. When applied together with nonpartisan redistricting, perceived compactness, and efficiency gap metric, as the article notes, it can produce reasonable results.
That's true, but the point of the demonstration is simply that the snaky, salamander-looking shapes are not necessary.
There's every reason to believe that you can achieve stunning examples of partisan gerrymandering under various "nice shape" criteria. You are correct that the "n-1 split lines" is simply very easy to state mathematically.
Yah think? Read David Daley's book Ratf*cked. http://www.worldcat.org/title/ratfucked-the-true-story-behin...
[0]http://michiganradio.org/post/redistricting-drive-step-close...
I'm not sure exactly how it would work but I imagine some kind of flood fill where you start with a population center and grow outward from it. Those should fill up quickly and by the end you should have large rural districts to finish off the state.
Which is ultimately the problem with algorithmic approaches; "fair" isn't an objective measure.
And the various form of elections are absolutely more or less fair, eg FPTP vs Approval Voting.
And it depends on selecting binary sides. How do you objectively deal with a popular moderate?
I'm very political, partisan. I'm seeing first hand the rise of interest group organizing as the political parties continue to decline. This sea change is making me very optimistic.
https://en.wikipedia.org/wiki/The_Diamond_Age
https://en.wikipedia.org/wiki/Phyle
The Blueprint: How the Democrats Won Colorado http://a.co/gQ7VIS7
1. No congressional district can span more than 1 partial city and/or county.
This will force districts to be drawn to cover whole cities/counties and would stop ridiculous districts such as the Texas 34th which covers parts of at least 8 counties from the gulf coast up to San Antonio
State legislatures control the definition of administrative subdivisions of the state just as much as they do Congressional districts, so that rule does nothing. A state that is really committed to gerrymandering will just move city and county lines when it redistricts.
A smarter but still corrupt state legislature that wants to gerrymander without disrupting functional local government will make “city” and “county” names of ceremonial subdivisions used for limited purposes, so that redrawing their lines doesn't have much practical effect, and adopt a different set of subdivisions with overlapping lines for most functional local government purposes (“urbanizations” and “parishes”, perhaps.)
https://medium.com/@jameson.quinn/place-voting-explained-129...
I agree with others on this thread that the only solution to gerrymandering is some form of PR. PLACE voting is interesting because it doesn't require a change in the ballot and retains the notion of geographic districts.
When you are changing how ballots work, this should be unavoidably obvious, even to people who spent the time the change was being debated in a coma or Antarctic research expedition.
Also, STV in small (~5 member) districts gets much improved proportionality from the status quo, keeps geographic districts, and increases the proportion of people with someone representing both their ideology and district simultaneously.
I don't think that's something you can really do with voting systems, though. Lots of voting systems are based around the same input assumptions, i.e., people are submitting a rank-order of preferences. And nothing's saying you'd have to switch to PLACE from STV rather than from FPTP. Regardless I think it doesn't make a lot of sense to associate ordinal ballots with STV specifically rather than just, well, what they are, which is a rank-order of preferences, that will get used somehow.
You can sometimes, you can't others.
> Lots of voting systems are based around the same input assumptions, i.e., people are submitting a rank-order of preferences
But FPTP is not based on that input, so changing from FPTP to a more proportional system (many of which rely on that kind of input) provided a clear opportunity to have a clean break without false familiarity.
> And nothing's saying you'd have to switch to PLACE from STV rather than from FPTP.
The false familiarity is FPTP -> PLACE.
There are no easy answers. I believe the game makers advocate for independent non-partisan commissions who are restricted in how they're allowed to create maps.
1. decide how many congresspeople there should be
2. allow anyone to run
3. allow anyone to vote for anyone
4. once someone has #voters/#congresspeople, they become a congressperson
5. no one else can vote for that person
So, you're not actually competing against another person. You're competing get enough people to say "this person represents me".
First, in order to make the lockout in #5 work it would seem that the votes need to be tallied in real time as they are cast.
This could be addressed in a couple of ways.
(A) Allow votes of the form:
Vote for candidate C1
If C1 is locked out, vote for C2
If C2 is locked out, vote for C3
...
(B) Instead of a lockout, allow a candidate to receive more votes than the #voters/#congresspeople threshold. Each voter starts with a weighted vote with weight 1, and provides a list of alternative candidates similar to from (A).
If C1 gets a total weighted vote above the election threshold, choose t such that if each voters weight is multiplied by t the total for C1 will exactly equal the threshold.
For each C1 voter, split their vote giving t of it to C1, and 1-t of it to their C2.
Iterate through this until it converges. (I think it must converge, but offhand do not have a proof).
Second, you need to know the number of voters in advance to set the threshold. Note that you need the actual number of people who will vote, not just the number of eligible voters, because if you use the later you will be setting the threshold too high.
Third, if the number of people running is large, it is quite possible nobody reaches the election threshold. You need a procedure to whittle down the number of candidates.
Could a person have their vote assigned to more than one person if none of them have won yet? (Probably with some sort of ranking to resolve any ambiguities as to who their vote goes to)
I'm not sure if that would be the optimal threshold for number of votes needed.
Here is an idea for a small modification:
First, have all the voters vote on the number of congresspeople to have, and then pick the median (if there is an even number of votes, and the middle two are different, pick between the two at random).
Then, after the result of the first vote has been decided, have everyone vote on the vote threshold (making sure that they are informed of what the voters/seats number), and again, take the median of the votes (and again, if a tie between 2 values, pick between the two at random. Don't average them.).
Then do what you said, with the chosen number of seats and the chosen vote threshold.
There are no doubt some downsides to this method. I don't yet have enough understanding of them to fully recommend it, but it seems worth looking into!
In an STV election people make a list, at step 4 above the highest vote on your list that's still viable accrues to someone, if they get more votes that required the portion of your vote that's needed to elect the person stays with them and the rest of your vote moves down your list ... whenever the process gets stuck the person with the least number of votes is removed from contention and their voters get to try the next person on their list
Not ideal but much better than what has been going on since the 2000s.
In the current system the question is who can be the most conservative/liberal. There is little downside to pushing further because there are more votes there but few people you’ll drive away.
Your change to gerrymandering doesn't make pushing for the extremes a worse strategy. Even in a perfectly split district, roughly 25% vote Democrat, 25% Republican, and 50% don't vote. That 50% is a better target than any voters on the fence.
Maybe in the Reagan years, but it's become more difficult since then. Orwell said "The real division is not between conservatives and revolutionaries but between authoritarians and libertarians." Both Republicans and Democrats have grown increasingly more authoritarian over the past 20 years.
Your most important vote is in the primaries (especially in areas dominated by a single party), and I think the country as a whole is beginning to realize it.
The reason that the parties have influence over the ballot is because they like it that way.
I used to agree, but the parties have no ideology attached to them. Their names are just generic terms regarding democracy and you can see how much their platforms have changed and moved over the last century. At this point they are just the Urban and Rural divisions of our election playoff.
They are just the ideology of the people elected. If you change the candidates of the party, you have changed the ideology of the party. Trump figured this out and instead of sticking with the Reform party, he just took over the Republican nomination.
That statement is simply incorrect. Both parties have a clear ideology that ebbs and flows for the sake of remaining in power
> If you change the candidates of the party, you have changed the ideology of the party.
That is also incorrect. The 2 parties may slightly adjust their ideology for the sake of power but the core of their ideology remains the same.
There's a significant difference in the modern era in that gerrymandering can be much more precise. The 2010 districts the Republicans drew in several states are very sophisticated works of statistics, GIS, and political science. I fear what machine learning techniques are going to do in 2020.
Also, I think it's partly because at the country's founding, most hot political issues had a clear large-scale geographic distribution, and so geographic districts were considered an adequate way to capture the overall mood.