Mathematicians looking to make elections in the US more representative • Hacker News Board

Re: Mathematicians looking to make elections in the US more representative

#31

Post by letitgo12345 » Thu, Jun 08, 2017, 4:03 AM UTC

MichaelGG wrote: post #26

letitgo12345 wrote: post #10

Earlier quoted context omitted.

Well speaking just for myself, I honestly would prefer an even more party insiders driven process for the primaries. Too much democracy seems to lead to insane levels of populism leading to characters like Trump. At any rate, US democracy guarantees one a vote in the general -- parties are private orgs and it's not "corrupt" or "wrong" for them to select candidates however they want.

Except since there's no chance for a 3rd party, these two parties hold too much power and it suddenly matters much more. If it was like other countries with tons of parties that rise and fall, you'd be correct. Also what do you mean by insane levels of populism? Isn't that just a derogatory way to say "what the people want"?

There would be more of a chance for a 3rd party if people bothered to care about the house rather than just the presidency...

But yeah, today there's little chance. But so what? Plenty of progress, far more than today was made in the past before the primary system. Several great presidents (including Lincoln, FDR, TDR) came from the old process. Meanwhile the new process gave us Trump...Not to mention, tons of members of Congress afraid to do anything sensible because they are too afraid of primary challenges by hyper partisan people on their side.

UK doesn't have a primary system (mostly -- they did for Labour who are now saddled with Corbyn) and they don't have a stupidly gridlocked govt. Neither does Canada. Both of those countries have politics dominated by mostly 2-3 parties. In fact almost no country on Earth has it to the level that the US does.

Re:"what people want" -- for one thing what people want is often contradictory. For another, there's a reason why the US doesn't have mob rule by majority. What people want is not always coherent or wise. IMO our goal as a society should be a functioning govt and a liberal polity for which democracy is a means. But by no means is democracy a good in an of itself for me at least.

Re: Mathematicians looking to make elections in the US more representative

#32

Post by kirrent » Thu, Jun 08, 2017, 4:06 AM UTC

aaron-lebo wrote: post #17

ValG wrote: post #2

There are a ton of important issues, but, I believe that gerrymandering is a core issue in the United States because : Partisan gerrymandering allows for politicians to secure their seats - this leads to less voters being represented (allowing the politicians to become more polarized) e.g. when races are tight, politicians tend to move towards the center. [2] Partisan gerrymandering is strongly disliked by both parti…

One of my professors at UT Dallas has a book that argues that districts in fact should be partisan. It's been a few years but the logic is something like if you've got a 50-50 district, you're gonna end up with 50% of the people in the district unhappy. Where as if you intentionally designed districts to be like like 90-10 (or whatever possible), more people are happy. Not sure if I believe him, but it's interesting.…

I think single transferable vote is a much better way of addressing that criticism. We have it where I live and it has tended to produce fairly boring yet representative elections.

Re: Mathematicians looking to make elections in the US more representative

#33

Post by yarg » Thu, Jun 08, 2017, 4:06 AM UTC

ValG wrote: post #2

There are a ton of important issues, but, I believe that gerrymandering is a core issue in the United States because : Partisan gerrymandering allows for politicians to secure their seats - this leads to less voters being represented (allowing the politicians to become more polarized) e.g. when races are tight, politicians tend to move towards the center. [2] Partisan gerrymandering is strongly disliked by both parti…

The fact that gerrymandering is even possible is a symptom (albeit an severely exacerbating one) - not the cause - of the problems in the US electoral structure.

It could be countered quite effectively by a system using larger jurisdictions each with multiple representitives.

Re: Mathematicians looking to make elections in the US more representative

#34

Post by kirrent » Thu, Jun 08, 2017, 4:11 AM UTC

ouid wrote: post #12

skierscott wrote: post #9

One of the most reasonable rules I've heard IMO for deterring voting districts: > In each section (at first a state, etc), find a dividing line that splits the population in two. This seems somewhat reasonable, and there's no way for humans to bias it. Here's some pictures of what this looks like [1], and the write up of this idea [2] [1]: http://rangevoting.org/alRS.png [2]: http://rangevoting.org/SplitLR.html

First of all, this only allows for a power of two of representatives. Second, any degree of freedom creates variance, and you can optimize it. There are many straight lines that divide a population in two.

For number 2, there could be any number of solutions to fix those degrees of freedom. Intuitively, a maximum margin separator for determining the division wouldn't seem to be a terrible idea.

Re: Mathematicians looking to make elections in the US more representative

#35

Post by hibikir » Thu, Jun 08, 2017, 4:18 AM UTC

skierscott wrote: post #9

One of the most reasonable rules I've heard IMO for deterring voting districts: > In each section (at first a state, etc), find a dividing line that splits the population in two. This seems somewhat reasonable, and there's no way for humans to bias it. Here's some pictures of what this looks like [1], and the write up of this idea [2] [1]: http://rangevoting.org/alRS.png [2]: http://rangevoting.org/SplitLR.html

The problem is not bias, but uneven representation. We can get uneven representation without bias quite easily: I've not ran your system in practical terms, but I'd be very surprised if, in the US, this gave you republican house majorities even when they lose the popular vote. The reason is simple: Republican voters are more dispersed geographically, so it's very hard for them to get clumped into a district where they have an overwhelming advantage.

This is why plenty of people prefer to do things like maximize the chance that a voter feels well represented, or minimize the safety of incumbents, instead of going for something that leads to simple lines.

Another important problem of straight lines is that it ignores the physical limitations of actual voting: Each voting location is expensive (go ask in North Carolina, where magical budget cuts make it harder to vote in areas that tend to vote democratic). inconvenient, mathematical lines will give you higher average costs to get to your own polling place, and will require more polling places, or places that have to corral different people to different booths with different options. Good luck doing that without doing everything fully digitally, which is, IMO, not necessarily a good idea.

Re: Mathematicians looking to make elections in the US more representative

#36

Post by mozumder » Thu, Jun 08, 2017, 4:22 AM UTC

aaron-lebo wrote: post #17

ValG wrote: post #2

There are a ton of important issues, but, I believe that gerrymandering is a core issue in the United States because : Partisan gerrymandering allows for politicians to secure their seats - this leads to less voters being represented (allowing the politicians to become more polarized) e.g. when races are tight, politicians tend to move towards the center. [2] Partisan gerrymandering is strongly disliked by both parti…

One of my professors at UT Dallas has a book that argues that districts in fact should be partisan. It's been a few years but the logic is something like if you've got a 50-50 district, you're gonna end up with 50% of the people in the district unhappy. Where as if you intentionally designed districts to be like like 90-10 (or whatever possible), more people are happy. Not sure if I believe him, but it's interesting.…

The problem is gerrymandering results in only one district that's 90-10 while 9 other districts are 55-45.

Without gerrymandering, districts would have a normal distribution from 60-40 to 40-60.

Re: Mathematicians looking to make elections in the US more representative

#37

Post by prodikl » Thu, Jun 08, 2017, 4:25 AM UTC

ianai wrote: post #5

I think we need approval voting.

Is this like.. where you choose 1, 2, 3?

so if I choose

1. Dem 2. Green 3. Repub

and you chose

1. Repub 2. Green 3. Dem

They'd all be equal? (four points each)

Re: Mathematicians looking to make elections in the US more representative

#38

Post by LaughingGoat » Thu, Jun 08, 2017, 4:25 AM UTC

woopwoop wrote: post #22

It's easy to gerrymander a surface. It's hard to gerrymander a line, or a circle, where the connectedness requirement has teeth. One solution to gerrymandering is to determine congressional districts by birthday, or last name, rather than by geographic location.

Interesting concept! I wonder if this is hackable based on demographic popularity of last names, by race. According to Wolfram Alpha, the most common last names are: Smith, Johnson, Williams, Brown, Jones, which don't share a common 1st letter. But maybe some ethnic names might: Rodriguez, Ramirez, Chan, Chen, Chang...

But ultimately, what political needs would a community of last-name-starts-with-C need?

Re: Mathematicians looking to make elections in the US more representative

#39

Post by ajmurmann » Thu, Jun 08, 2017, 4:25 AM UTC

nathancahill wrote: post #20

aaron-lebo wrote: post #17

Earlier quoted context omitted.

One of my professors at UT Dallas has a book that argues that districts in fact should be partisan. It's been a few years but the logic is something like if you've got a 50-50 district, you're gonna end up with 50% of the people in the district unhappy. Where as if you intentionally designed districts to be like like 90-10 (or whatever possible), more people are happy. Not sure if I believe him, but it's interesting.…

That's assuming a polarizing representative (one that 50% like and 50% are unhappy with. Could more balanced districts lead to more centrist representatives?

Maybe even having a system that makes more than two choices viable?

Re: Mathematicians looking to make elections in the US more representative

#40

Post by ajmurmann » Thu, Jun 08, 2017, 4:29 AM UTC

labster wrote: post #21

acjohnson55 wrote: post #6

To me, the problem is winner-take-all districts to begin with. It all but ensures some constituents are going to be sorely disappointed. Why not consider something like multi-member districts? We could tune it even further by awarding variable voting power. Then the way the borders are drawn becomes far less relevant. I'm sure there are tradeoffs, but it would certainly be more representative.

Most of the arguments against this center on the fact that ranked systems would be too complex for voters to understand. Which I used to think was pure hogwash, but Americans have since proved that they can't handle even two candidates for a major office. Hint: the worst one was the one with zero government experience.

The parent didn't mention ranked choice voting unless I read it wrong. I read them to argue in favor of multiple representatives per district. You can have that even with the current voting system. The top n candidates get the job with n > 1. There only difference is right now n = 1. That being said there is also the much simpler approval voting.