Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
Dead Reckoning
Sep 13, 2011

Hieronymous Alloy posted:

Here's Kennedy's concurring opinion that we're all discussing:
https://www.law.cornell.edu/supct/html/02-1580.ZC.html
I'm aware of the opinion, I just don't think he'll actually follow through because of the precedent it would set.

Rigel posted:

Last time around Kennedy said this is probably unconstitutional, but threw up his hands and complained about how it was currently impossible for a court to easily, quickly, and objectively discern that. Well, now thats no longer the case.
That's the big step though. The Supreme Court would have to say, "this equation is objective, mathematical proof of unconstitutional unfairness." It's an academic's wet dream: the most vexing problems of government reduced to a simple formula.

It raises a gently caress ton of questions though. What makes this particular formula the "objective" measure of fairness? Who gets to decide what that is? With a subject as fraught and charged as political representation, where we have intentionally constructed systems that distribute power in ways other than "one person one vote for everything", having the Court rule "beep boop, this professor's computer says your elected officials are being unfair, so we've voided their decision" poses a huge normative problem.

What other violations of rights can be proven by mathematical modeling?

Adbot
ADBOT LOVES YOU

twodot
Aug 7, 2005

You are objectively correct that this person is dumb and has said dumb things

Dead Reckoning posted:

What other violations of rights can be proven by mathematical modeling?
Literally any violation of rights that affects large amounts of people? If a police station has a database of arrests and it turns out that 100% of their arrests are left handed people, we don't need a political progress to figure out they are being discriminatory.
edit:
I'd be willing to entertain an argument that gerrymandering doesn't actually violate any rights before I'd entertain the idea we can't use math to investigate whether things are happening or not.

twodot fucked around with this message at 20:01 on Sep 6, 2017

Rigel
Nov 11, 2016

BirdOfPlay posted:

Would you say that, if Kennedy is convinced, it might move this into a 6-3 decision with Roberts coming over because of legacy concerns/getting a voice on the governing opinion? I know little about Roberts' jurisprudence save for his concerns of legacy and valuing state's rights.

Because 5 of the Justices were not on the court for Vieth v. Jubelirer, are they expected to fall in along "party" lines?

Its definitely possible to get Roberts, especially if Kennedy comes over. However Roberts is a very "defer to politics/states" guy so he might come down on the side of "gosh gerrymandering sure sounds bad, but courts are poorly equipped to deal with it, its a political issue, etc". Thomas is a firm no, and I don't think there's any chance in hell that Alito can be persuaded, he's more deferential on this sort of thing than Roberts. I have no idea on Gorsuch, we could cynically call him a no, but I'll wait and see while also not counting on him.

Rigel
Nov 11, 2016

Dead Reckoning posted:

It raises a gently caress ton of questions though. What makes this particular formula the "objective" measure of fairness? Who gets to decide what that is? With a subject as fraught and charged as political representation, where we have intentionally constructed systems that distribute power in ways other than "one person one vote for everything", having the Court rule "beep boop, this professor's computer says your elected officials are being unfair, so we've voided their decision" poses a huge normative problem.

Its not completely arbitrary. They started out with a subjective principle that everyone agrees on, namely a map where one party will always outperform their vote share under all reasonable scenarios. Once you start from there, you crunch the numbers and get to "well an efficiency gap of 7% would mean that you'll outperform the other party in all likely voting scenarios for all 5 elections in the decade 90% of the time". If you want to be conservative, then say 10% efficiency gap which starts entrenching your party in even rather unlikely voting scenarios.

You describe what unfair means as clearly as possible, and once you can do that, you can calculate what produces the result you describe based on literally thousands of maps across the country over 4 decades.

Its also important to note that the court isn't being asked to completely ban partisan gerrymandering, because thats too hard. Most maps will favor one party or the other a smidge, what we want to avoid is a case like in Wisconsin where 47% of the vote gets you over 60% of the seats, while the other party would need 55% of the vote to gain just a bare majority.

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

Rigel posted:

Its also important to note that the court isn't being asked to completely ban partisan gerrymandering, because thats too hard. Most maps will favor one party or the other a smidge, what we want to avoid is a case like in Wisconsin where 47% of the vote gets you over 60% of the seats, while the other party would need 55% of the vote to gain just a bare majority.

How much does this model bake in the assumption of only two parties? In Canada it's pretty common to form a majority government with 40% of the popular vote.

Rygar201
Jan 26, 2011
I AM A TERRIBLE PIECE OF SHIT.

Please Condescend to me like this again.

Oh yeah condescend to me ALL DAY condescend daddy.


Subjunctive posted:

How much does this model bake in the assumption of only two parties? In Canada it's pretty common to form a majority government with 40% of the popular vote.

A lot? We have a two party system. It's baked right in.

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

There are other parties that run, no? What's it baked into?

Rygar201
Jan 26, 2011
I AM A TERRIBLE PIECE OF SHIT.

Please Condescend to me like this again.

Oh yeah condescend to me ALL DAY condescend daddy.


Subjunctive posted:

There are other parties that run, no? What's it baked into?

Almost every election here is first past the post in a single member district. Splitting a sympathetic electorate just loses you a seat someone ideologically inclined to you could win. Over the centuries, this has lead to two main parties becoming a facet of American politics. Our parties are more like governing coalitions of parties in parliamentary democracies. Not every Democrat or republican believes the same things, or has the same priorities. They work together because that is how you get into power though.

joepinetree
Apr 5, 2012
The mathematical model being presented does not require or imply 2 parties. The idea of an efficiency gap is simply based on the idea of "wasted votes." Now, what it actually does if accepted is create a scenario where partisan gerrymandering would be required.

Imagine if the courts accept the 7% EG. That would mean that if a state had a substantial third party that gathered enough votes but didn't win any seats, it might be required to redistrict in a way to ensure 3rd party representation. So imagine if a solid blue city in a red state starts to see a big DSA presence. Imagine, for example, if the metro Atlanta districts start to see DSA contesting democratic seats from the left. And there's enough DSA voters that they gather enough votes in those districts while still losing by a lot. So that citywide they gather some 30% of the votes while still losing by 20 points to a democrat, and statewide they gather 8% of the vote, but no seats. The EG measure would still be valid, and in this case would trigger the proposed 7% EG discussed in the case even if the republican/democrat divide was very efficient.

So the math makes no assumptions about bipartisanship. But a decision based on that test would potentially lead to unintended consequences in the case of substantial historical alterations in third parties.

esquilax
Jan 3, 2003

Subjunctive posted:

How much does this model bake in the assumption of only two parties? In Canada it's pretty common to form a majority government with 40% of the popular vote.

The efficiency gap metric is based on the assumption of two parties.

It is likely possible to use the baseline "wasted votes" metric to develop a single efficiency gap number that takes into account multiple parties, but I don't think anyone has attempted to do that yet.

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

Rygar201 posted:

Almost every election here is first past the post in a single member district. Splitting a sympathetic electorate just loses you a seat someone ideologically inclined to you could win. Over the centuries, this has lead to two main parties becoming a facet of American politics. Our parties are more like governing coalitions of parties in parliamentary democracies. Not every Democrat or republican believes the same things, or has the same priorities. They work together because that is how you get into power though.

Sure, but people still vote Green and Libertarian, albeit currently in small numbers. Others run as independents. Does the model have to ignore them to work? What breaks if the Greens take a seat in the legislature?

joepinetree
Apr 5, 2012
Again, there's nothing in the metric itself that requires only two parties.

In fact, the example for wasted votes (which is the core of the EG) in the wikipedia page uses a 3 party system:

https://en.wikipedia.org/wiki/Wasted_vote

And the efficiency gap number is just net wasted votes divided by total votes.

Now, in the particular case before the supreme court, what is considered extreme in terms of EG and what they are suggesting as thresholds is based on historical US series. But there is nothing in the math itself that precludes multiparty systems.

joepinetree fucked around with this message at 21:15 on Sep 6, 2017

DACK FAYDEN
Feb 25, 2013

Bear Witness

Subjunctive posted:

What breaks if the Greens take a seat in the legislature?
Causality, slightly before election day :v:

e: also the thing you're looking for might (but might not) be Duverger's Law

evilweasel
Aug 24, 2002

BirdOfPlay posted:

Would you say that, if Kennedy is convinced, it might move this into a 6-3 decision with Roberts coming over because of legacy concerns/getting a voice on the governing opinion? I know little about Roberts' jurisprudence save for his concerns of legacy and valuing state's rights.

Because 5 of the Justices were not on the court for Vieth v. Jubelirer, are they expected to fall in along "party" lines?

Entirely possible, gerrymandering is such indefensible garbage he may not feel like going to bat for it.

evilweasel
Aug 24, 2002

Dead Reckoning posted:

That's the big step though. The Supreme Court would have to say, "this equation is objective, mathematical proof of unconstitutional unfairness." It's an academic's wet dream: the most vexing problems of government reduced to a simple formula.

no they don't. the equation is evidence considered by the court in reaching its determination but was not conclusive. the mccain/whitehouse amicus brief lays out quite clearly what the supreme court can do: say yep, on these facts gerrymandering is justicible and the plaintiffs have met their burden to show intent and invidious discrimination.

Subjunctive
Sep 12, 2006

✨sparkle and shine✨

DACK FAYDEN posted:

e: also the thing you're looking for might (but might not) be Duverger's Law

I like that the page says that it's a principle of political science, but also that most evidence is against it being true.

esquilax
Jan 3, 2003

joepinetree posted:

Again, there's nothing in the metric itself that requires only two parties.

In fact, the example for wasted votes (which is the core of the EG) in the wikipedia page uses a 3 party system:

https://en.wikipedia.org/wiki/Wasted_vote

Now, in the particular case before the supreme court, what is considered extreme in terms of EG is based on historical US series. But there is nothing in the math itself that precludes multiparty systems.

Wasted votes is tracked for individual parties and can track any number of them. To generate an efficiency gap to determine if things are biased, you need to compare exactly two numbers.

If you set an efficiency gap threshold of 10% based on typical results of two parties, it is not smart to take every pairwise comparison and check it against your two-party threshold.

evilweasel
Aug 24, 2002

Dead Reckoning posted:

What other violations of rights can be proven by mathematical modeling?

disparate impact

Evil Fluffy
Jul 13, 2009

Scholars are some of the most pompous and pedantic people I've ever had the joy of meeting.

Dead Reckoning posted:

I'm not a lawyer, but wouldn't finding in favor of the plaintiffs set a rather sweeping precedent? If the Supreme Court adopts a test involving efficiency gap and similar metrics, they basically take a formula created by a group of academics and substitute it for the power of elected state governments to draw their districts.

It'd only matter in situations where those elected officials refuse to do their job (redistricting) in a proper fashion. Some states could draw districts that aren't (heavily) Gerrymandered but for those who don't, like NC and WI, others would do it for them to ensure the rights of the voters are being upheld. The judiciary would be deciding the metrics and not the means, though. Ideally the metrics would simply force states to use systems like California, but lots of GOP-held states would fight that to the death because they'll lose power in a lot of states they currently hold via gerrymandering and/or suppression. The right of the people to have their votes matter in a more equal and even district should outweigh any claim that elected officials would have to a 'right' to Gerrymander.

Gerrymandering replaces "the voters choose their government" with "the government chooses their voters." In the end, I simply do not trust Kennedy.

twodot posted:

Literally any violation of rights that affects large amounts of people? If a police station has a database of arrests and it turns out that 100% of their arrests are left handed people, we don't need a political progress to figure out they are being discriminatory.

For example: The NYPD's Stop and Frisk policy which was blatantly discriminatory.

Rigel posted:

Its definitely possible to get Roberts, especially if Kennedy comes over. However Roberts is a very "defer to politics/states" guy so he might come down on the side of "gosh gerrymandering sure sounds bad, but courts are poorly equipped to deal with it, its a political issue, etc". Thomas is a firm no, and I don't think there's any chance in hell that Alito can be persuaded, he's more deferential on this sort of thing than Roberts. I have no idea on Gorsuch, we could cynically call him a no, but I'll wait and see while also not counting on him.

Gorsuch: The Constitution doesn't say Gerrymandering is bad therefore it's ok and good just like this paint I'm drinking.

DACK FAYDEN
Feb 25, 2013

Bear Witness

Subjunctive posted:

I like that the page says that it's a principle of political science, but also that most evidence is against it being true.
It's the thing with "Law" in the name that is least true out of everything with Law in the name that I've ever encountered. And I have a master's in economics.

joepinetree
Apr 5, 2012

esquilax posted:

Wasted votes is tracked for individual parties and can track any number of them. To generate an efficiency gap to determine if things are biased, you need to compare exactly two numbers.

If you set an efficiency gap threshold of 10% based on typical results of two parties, it is not smart to take every pairwise comparison and check it against your two-party threshold.

The formula is simply the difference between votes wasted by the winning party and votes wasted by the losing parties divided by total votes. Nowhere in the formula does it require that votes wasted by the losing party all come from one party. Since we can clearly calculate votes wasted by the winning party and votes wasted by multiple losing parties (as the wikipedia example clearly demonstrates), there is absolutely nothing, from a mathematical point of view, that requires bipartisanship.

In fact, the efficiency gap in the wikipedia example with 3 parties the efficiency gap would be 9.2%.

Again, there is absolutely no requirement of two parties in the formula. It's only the difference in wasted votes between winner and loser(s) divided by total votes.

The only way bipartisanship plays a role in this discussion is in terms of whether the court should use historical numbers to determine what counts as extreme partisanship. But nothing in the formula requires bipartisanship.

esquilax
Jan 3, 2003

joepinetree posted:

The formula is simply the difference between votes wasted by the winning party and votes wasted by the losing parties divided by total votes. Nowhere in the formula does it require that votes wasted by the losing party all come from one party. Since we can clearly calculate votes wasted by the winning party and votes wasted by multiple losing parties (as the wikipedia example clearly demonstrates), there is absolutely nothing, from a mathematical point of view, that requires bipartisanship.

In fact, the efficiency gap in the wikipedia example with 3 parties the efficiency gap would be 9.2%.

Again, there is absolutely no requirement of two parties in the formula. It's only the difference in wasted votes between winner and loser(s) divided by total votes.

The only way bipartisanship plays a role in this discussion is in terms of whether the court should use historical numbers to determine what counts as extreme partisanship. But nothing in the formula requires bipartisanship.

We're talking past each other a bit. I'm talking about the specific implementation of the efficiency gap analysis as developed, and not how it can be modified to accommodate N-party systems. The intent is obviously to apply it in the US (which does technically have N-parties), but the simplifying assumption of two parties makes it difficult to apply it to systems where there are substantial votes for third parties without major modifications.

The actual paper describing efficiency gap analysis assumes two parties, and the math does not work for a three-dominant-party system for a few reasons:

1. The calculation of wasted votes in the wikipedia article is not consistent with how wasted votes are calculated in the Stephanopoulos-McGhee efficiency gap paper. In the wiki wasted votes example, vote surplus is calculated using the smallest plurality number. In the efficiency gap example from the wiki (which is consistent with the paper), it is based off of a 50% of the vote instead of a plurality. These two methodologies produce different numbers for wasted surplus votes. In a 55-45 two party race, the paper methodology would say 5 votes are wasted surplus, another would say 10. Results would be similar in a three-party 55-45-0 race. In a 45-35-10 race, if you naively applied the paper methodology, it says (-5) votes are wasted surplus, and the other says 10.

2. Efficiency gap is directional - a 5% efficiency gap for R's is different than a 5% efficiency gap for D's. Your 9.2% +A in that calculation isn't directional in a way that can be added across multiple districts, where other districts might have +B or +C bias. Simply adding the wasted votes together and comparing at the end would have the problem of a huge correlation between the number of wasted votes and the total number of votes.

3. The efficiency gap is agnostic over who actually wins any individual seat, which is a really big strength. Introducing "winner and loser(s)" into the calculation reduces it's value, and has the major risk of significantly understating bias against the winning party when there is a substantial third party vote.

4. On a practical level, developing norms around acceptable efficiency gaps is far easier in a two-party system than it is in a three-party system. A district that packs the Liberals with the New Dems and no Conservatives is significantly different than one that packs the Liberals with a mix of New Dems and Conservatives, but we haven't developed the tools to differentiate the two.


Again, I don't have doubts that you could develop a methodology to determine a general efficiency-gap-like metric for N-party systems. However, without significant changes, the best and most obvious way to apply the Stephanopoulos-McGhee efficiency gap it in the real world is to completely ignore third parties, which is perfectly feasible in the US but not really feasible in Canada. Is there a paper I'm unaware of that has attempted to include third parties at all?

Platystemon
Feb 13, 2012

BREADS
I think that the formula is a fine piece of mathematics. It’s good enough for government work.

But it wouldn’t be hard for Kennedy to dismiss it on philosophical or other grounds.

Talking about “efficiency” and “wasted votes” is opening a can of worms, and the formula itself is almost too simple. It certainly didn’t take ten years to develop.

A geographical metric might go over better.

Like, U.S. state boundaries create a lot of wasted votes* but they’re reasonably geographically compact.

*Anyone have an EG table of the states? I’m curious.

joepinetree
Apr 5, 2012

esquilax posted:

We're talking past each other a bit. I'm talking about the specific implementation of the efficiency gap analysis as developed, and not how it can be modified to accommodate N-party systems. The intent is obviously to apply it in the US (which does technically have N-parties), but the simplifying assumption of two parties makes it difficult to apply it to systems where there are substantial votes for third parties without major modifications.

The actual paper describing efficiency gap analysis assumes two parties, and the math does not work for a three-dominant-party system for a few reasons:

1. The calculation of wasted votes in the wikipedia article is not consistent with how wasted votes are calculated in the Stephanopoulos-McGhee efficiency gap paper. In the wiki wasted votes example, vote surplus is calculated using the smallest plurality number. In the efficiency gap example from the wiki (which is consistent with the paper), it is based off of a 50% of the vote instead of a plurality. These two methodologies produce different numbers for wasted surplus votes. In a 55-45 two party race, the paper methodology would say 5 votes are wasted surplus, another would say 10. Results would be similar in a three-party 55-45-0 race. In a 45-35-10 race, if you naively applied the paper methodology, it says (-5) votes are wasted surplus, and the other says 10.

2. Efficiency gap is directional - a 5% efficiency gap for R's is different than a 5% efficiency gap for D's. Your 9.2% +A in that calculation isn't directional in a way that can be added across multiple districts, where other districts might have +B or +C bias. Simply adding the wasted votes together and comparing at the end would have the problem of a huge correlation between the number of wasted votes and the total number of votes.

3. The efficiency gap is agnostic over who actually wins any individual seat, which is a really big strength. Introducing "winner and loser(s)" into the calculation reduces it's value, and has the major risk of significantly understating bias against the winning party when there is a substantial third party vote.

4. On a practical level, developing norms around acceptable efficiency gaps is far easier in a two-party system than it is in a three-party system. A district that packs the Liberals with the New Dems and no Conservatives is significantly different than one that packs the Liberals with a mix of New Dems and Conservatives, but we haven't developed the tools to differentiate the two.


Again, I don't have doubts that you could develop a methodology to determine a general efficiency-gap-like metric for N-party systems. However, without significant changes, the best and most obvious way to apply the Stephanopoulos-McGhee efficiency gap it in the real world is to completely ignore third parties, which is perfectly feasible in the US but not really feasible in Canada. Is there a paper I'm unaware of that has attempted to include third parties at all?

No, the paper where the formula actually comes from is:
McGhee, Eric. "Measuring Partisan Bias in Single‐Member District Electoral Systems." Legislative Studies Quarterly 39.1 (2014): 55-85.

Of course, what counts as a wasted vote is going to be different in a two party election versus a multiparty one, but that is absolutely meaningless in terms of employing the efficiency gap analysis. In the very same wikipedia page there is a calculation of wasted votes in a 12 seat election. Again, efficiency gap is just the difference in terms of wasted votes between the ruling party and the opposition, regardless of how many opposition parties there are. In the very same page on wikipedia there is a calculation of wasted votes for multi 3 party races.
In neither the paper that applies it to the US or the original that develops it you will see it stated that it requires a two party system. They will use 2 parties in the demonstration for the sake of simplicity, but there is literally no difference in the efficiency gap measure for multiparty systems. Wasted votes are simply counted a little bit different, in mathematically obvious ways.

FAUXTON
Jun 2, 2005

spero che tu stia bene

evilweasel posted:

disparate impact

and other mythical notions

esquilax
Jan 3, 2003

joepinetree posted:

No, the paper where the formula actually comes from is:
McGhee, Eric. "Measuring Partisan Bias in Single‐Member District Electoral Systems." Legislative Studies Quarterly 39.1 (2014): 55-85.

Of course, what counts as a wasted vote is going to be different in a two party election versus a multiparty one, but that is absolutely meaningless in terms of employing the efficiency gap analysis. In the very same wikipedia page there is a calculation of wasted votes in a 12 seat election. Again, efficiency gap is just the difference in terms of wasted votes between the ruling party and the opposition, regardless of how many opposition parties there are. In the very same page on wikipedia there is a calculation of wasted votes for multi 3 party races.
In neither the paper that applies it to the US or the original that develops it you will see it stated that it requires a two party system. They will use 2 parties in the demonstration for the sake of simplicity, but there is literally no difference in the efficiency gap measure for multiparty systems. Wasted votes are simply counted a little bit different, in mathematically obvious ways.

The 12 seat example in the wiki is based on a form of proportional representation, which is not at all helpful if you are looking at FPTP. If you're not talking about FPTP then I concede that it can be a useful measure.

The McGhee paper is also based on the assumption of a two party system, and does not attempt to apply the efficiency formula in an N-party system. This isn't a fault of the author, it's very reasonable to make that assumption since it's focused on American elections.

Using 50% as the wasted surplus reference point (as both papers do) versus plurality as a reference point (as the wasteful vote calc in the wiki does) is a fundamental difference. A 30%/70% split is considered biased pro-winner using one reference point and biased pro-loser in another. You can't just make that change and call it "mathematically obvious" without actually doing the math.

All of the math was based on two-party systems, and all of the testing they've done to actually verify that it's legit has been done on two-party systems. If anyone has attempted to apply the reasoning to multiparty systems I'd like to see it. As I've been saying, it might be relevant for places like Canada, but all of the work is being done on a two-party system basis and they need to actually go through the motions to find out.

joepinetree
Apr 5, 2012

esquilax posted:

The 12 seat example in the wiki is based on a form of proportional representation, which is not at all helpful if you are looking at FPTP. If you're not talking about FPTP then I concede that it can be a useful measure.

The McGhee paper is also based on the assumption of a two party system, and does not attempt to apply the efficiency formula in an N-party system. This isn't a fault of the author, it's very reasonable to make that assumption since it's focused on American elections.

Using 50% as the wasted surplus reference point (as both papers do) versus plurality as a reference point (as the wasteful vote calc in the wiki does) is a fundamental difference. A 30%/70% split is considered biased pro-winner using one reference point and biased pro-loser in another. You can't just make that change and call it "mathematically obvious" without actually doing the math.

All of the math was based on two-party systems, and all of the testing they've done to actually verify that it's legit has been done on two-party systems. If anyone has attempted to apply the reasoning to multiparty systems I'd like to see it. As I've been saying, it might be relevant for places like Canada, but all of the work is being done on a two-party system basis and they need to actually go through the motions to find out.

Ok, you either fail to grasp basic math, didn't read the paper, or both. Your 30-70 example is a scenario that is contemplated in the paper itself. As the paper itself mentions, there is this intercept around 75% in a bipartisan system. In a 70-30 split where the governing party wins all the seats, the efficiency gap formula will get you an EG showing a gap against the smaller party. At 75%-25% (using the paper's assumption of 50% to win to avoid having to use decimal points), the EG will show a gap of 0. And at 76-24 it will show a gap in favor of the opposing party. So your example is simply the result of the fact that in a system with three parties the intercept will be lower than 75%. Around a 60-20-20 split that leads to the majority party holding all seats would lead to an EG of 0, with the same issues of being an intercept point as the ~75% mark for bipartisan systems. In fact, here:

code:
                     Party 1          Party 2          Party 3
Seat 1           50                      30                20
Seat 2           44                      40                16
Seat 3           20                      45                 35

Surplus/wasted votes
                     Party 1          Party 2          Party 3
Seat 1           19                      30                20
Seat 2           3                       40                16
Seat 3           20                      9                 35

Total             42                       79                71
Difference in wasted votes between the ruling party (42) and the combined opposition parties (150) is equal to 108. 108/300= 36% efficiency gap in favor of the party with the most seats. Calculated without having to make any massive mathematical changes. The only thing that assuming bipartisanship allows is skipping the process of going district by district because he derives a simpler formula for the entire state in appendix B of the paper. But that is only a matter of simplifying the calculations, not of altering the formula itself. In fact, the framing of the paper from the very first paragraph is about how gerrymandering can allow parties to win seats with bare pluralities, something that wouldn't be a concern if the argument only applied to bipartisan elections. Yes, there are points in a multiparty system where the formula can lead to weird or counterintuitive results, especially in lopsided elections, but the same can be said for the formula in bipartisan systems.

Again, the formula can be simply read as (Governing party wasted votes-Opposition wasted votes regardless of how many parties)/total votes.

joepinetree fucked around with this message at 06:07 on Sep 7, 2017

esquilax
Jan 3, 2003

joepinetree posted:

Ok, you either fail to grasp basic math, didn't read the paper, or both. Your 30-70 example is a scenario that is contemplated in the paper itself. As the paper itself mentions, there is this intercept around 75% in a bipartisan system. In a 70-30 split where the governing party wins all the seats, the efficiency gap formula will get you an EG showing a gap against the smaller party. At 75%-25% (using the paper's assumption of 50% to win to avoid having to use decimal points), the EG will show a gap of 0. And at 76-24 it will show a gap in favor of the opposing party. So your example is simply the result of the fact that in a system with three parties the intercept will be lower than 75%. Around a 60-20-20 split that leads to the majority party holding all seats would lead to an EG of 0, with the same issues of being an intercept point as the ~75% mark for bipartisan systems. In fact, here:

code:
                     Party 1          Party 2          Party 3
Seat 1           50                      30                20
Seat 2           44                      40                16
Seat 3           20                      45                 35

Surplus/wasted votes
                     Party 1          Party 2          Party 3
Seat 1           19                      30                20
Seat 2           3                       40                16
Seat 3           20                      9                 35

Total             42                       79                71
Difference in wasted votes between the ruling party (42) and the combined opposition parties (150) is equal to 108. 108/300= 36% efficiency gap in favor of the party with the most seats. Calculated without having to make any massive mathematical changes. The only thing that assuming bipartisanship allows is skipping the process of going district by district because he derives a simpler formula for the entire state in appendix B of the paper. But that is only a matter of simplifying the calculations, not of altering the formula itself. In fact, the framing of the paper from the very first paragraph is about how gerrymandering can allow parties to win seats with bare pluralities, something that wouldn't be a concern if the argument only applied to bipartisan elections. Yes, there are points in a multiparty system where the formula can lead to weird or counterintuitive results, especially in lopsided elections, but the same can be said for the formula in bipartisan systems.

Again, the formula can be simply read as (Governing party wasted votes-Opposition wasted votes regardless of how many parties)/total votes.

The fact that a 70R-30D election is treated as 10% +R in the paper methodology and 70R-30D-0G is treated as 10% +D in your methodology should tell you that changing how wasted votes are calculated is a fundamental difference. Both papers consistently use a definition of wasted votes that is different from the one that you are using, neither paper attempts to discuss yours.

As I said earlier, using the "government minus opposition" metric as efficiency gap has the serious risk of understating bias against the winning party. The use of your formula in a three-relevant party system means that there is a big "dead zone" where a very small change (i.e. which party is "in government") has a huge effect on the calculated bias. Similarly, it also has the serious risk of finding bias where there is none. If used in 2015 Canada, there is a major danger of your formula falsely identifying bias in favor of the Liberal party due to combining Conservatives and NDP - even if the underlying districting is still biased in reality in favor of Conservatives after years of rule. It is a sensitive model when comparing multiparty systems, whereas when applied to two parties the paper methodology is a robust model - small changes in # of seats lead to small changes in the calculated efficiency gap.

Efficiency gaps near 0% are possible and likely in hypothetical unbiased two party systems, but may not be likely under hypothetical unbiased three party systems . I don't think it's likely that an analysis of Canada or an unbiased Canada-like system using your methodology will result in a chart looking like page Figure 4 of Stephanopolous-McGhee, and might not even be centered at 0%. Which leads to significant problems in application when trying to identify where districting is biased.

So in short, just because you can modify the formula to take into account multiple parties doesn't mean that it is useful to apply it in the real world. The two papers developed and tested the formula on American elections and a two-party system. Taking a methodology that has been developed under one set of assumptions and tested under conditions that match those assumptions, and then naively applying it to conditions that break those assumptions, is not wise. It's possible that they can go through the work to determine how your methodology actually looks when applied to Canada, but it has not been tested under those conditions and has the potential for the major problems that I described. Those problems being the inability to identify bias where it does exist, and finding bias in cases where it doesn't exist.


I'll give you the final post on the topic so please play nice

vyelkin
Jan 2, 2011

evilweasel posted:

surprising development in the redistricting case:

I think McCain's brain tumour is one of those ones that changes your entire personality and worldview.

Imagine, the course of American history changed because a shithead got a brain tumour in his lovely head that magically turned him into a less lovely person.

evilweasel
Aug 24, 2002

vyelkin posted:

I think McCain's brain tumour is one of those ones that changes your entire personality and worldview.

Imagine, the course of American history changed because a shithead got a brain tumour in his lovely head that magically turned him into a less lovely person.

It appears he's not the only one. A bunch of Republican governors, representatives, and former representatives also signed amicus briefs calling for the court to strike down gerrymandering (including, oddly, some representatives in specifically gerrymandered for their benefit districts).

quote:

Breaking ranks with many of their fellow Republicans, a group of prominent politicians filed briefs on Tuesday urging the Supreme Court to rule that extreme political gerrymandering — the drawing of voting districts to give lopsided advantages to the party in power — violates the Constitution.

The briefs were signed by Republicans including Senator John McCain of Arizona; Gov. John R. Kasich of Ohio; Bob Dole, the former Republican Senate leader from Kansas and the party’s 1996 presidential nominee; the former senators John C. Danforth of Missouri, Richard G. Lugar of Indiana and Alan K. Simpson of Wyoming; and Arnold Schwarzenegger, a former governor of California.

“Partisan gerrymandering has become a tool for powerful interests to distort the democratic process,” reads a brief filed by Mr. McCain and Senator Sheldon Whitehouse, Democrat of Rhode Island.

The Supreme Court will hear arguments in the case, Gill v. Whitford, No. 16-1161, on Oct. 3.

The Republican National Committee, the National Republican Congressional Committee and the Republican State Leadership Committee all filed briefs on the other side. They urged the Supreme Court to reject a challenge to State Assembly districts in Wisconsin that, by some measures, gave Republicans outsize political power unjustified by the overall vote.
...
Representative Brian Fitzpatrick, Republican of Pennsylvania, joined a brief filed by current and former members of the House of Representatives urging the Supreme Court to reject the Wisconsin maps. He said partisan gerrymandering had contributed to toxic polarization in the House of Representatives.

“You have 435 districts in the nation, and there’s probably only 20 or so that are legitimate swing districts,” he said. “For the 415 safe seats, their main election is in the primary, not the general. When the main election is in the primary, you legislate accordingly. The result has been a growing cavernous divide, which has created a Hatfield v. McCoy environment in the legislature, and it’s hurting the American people.”
...
The brief from current and former House members noted that Mr. Reagan had referred to gerrymandering in 1987 as “a national scandal” and called for “an end to the anti-democratic and un-American practice of gerrymandering congressional districts.”

https://www.nytimes.com/2017/09/06/...itics&smtyp=cur

I'm actually starting to think this has a real chance.

Rigel
Nov 11, 2016

evilweasel posted:

It appears he's not the only one. A bunch of Republican governors, representatives, and former representatives also signed amicus briefs calling for the court to strike down gerrymandering (including, oddly, some representatives in specifically gerrymandered for their benefit districts).


https://www.nytimes.com/2017/09/06/...itics&smtyp=cur

I'm actually starting to think this has a real chance.

They may be staring into the abyss of the 2020 elections, terrified that the Dems will be able to gerrymander the poo poo out of them as payback.

joepinetree
Apr 5, 2012

esquilax posted:

The fact that a 70R-30D election is treated as 10% +R in the paper methodology and 70R-30D-0G is treated as 10% +D in your methodology should tell you that changing how wasted votes are calculated is a fundamental difference. Both papers consistently use a definition of wasted votes that is different from the one that you are using, neither paper attempts to discuss yours.

As I said earlier, using the "government minus opposition" metric as efficiency gap has the serious risk of understating bias against the winning party. The use of your formula in a three-relevant party system means that there is a big "dead zone" where a very small change (i.e. which party is "in government") has a huge effect on the calculated bias. Similarly, it also has the serious risk of finding bias where there is none. If used in 2015 Canada, there is a major danger of your formula falsely identifying bias in favor of the Liberal party due to combining Conservatives and NDP - even if the underlying districting is still biased in reality in favor of Conservatives after years of rule. It is a sensitive model when comparing multiparty systems, whereas when applied to two parties the paper methodology is a robust model - small changes in # of seats lead to small changes in the calculated efficiency gap.

Efficiency gaps near 0% are possible and likely in hypothetical unbiased two party systems, but may not be likely under hypothetical unbiased three party systems . I don't think it's likely that an analysis of Canada or an unbiased Canada-like system using your methodology will result in a chart looking like page Figure 4 of Stephanopolous-McGhee, and might not even be centered at 0%. Which leads to significant problems in application when trying to identify where districting is biased.

So in short, just because you can modify the formula to take into account multiple parties doesn't mean that it is useful to apply it in the real world. The two papers developed and tested the formula on American elections and a two-party system. Taking a methodology that has been developed under one set of assumptions and tested under conditions that match those assumptions, and then naively applying it to conditions that break those assumptions, is not wise. It's possible that they can go through the work to determine how your methodology actually looks when applied to Canada, but it has not been tested under those conditions and has the potential for the major problems that I described. Those problems being the inability to identify bias where it does exist, and finding bias in cases where it doesn't exist.


I'll give you the final post on the topic so please play nice

Again, it's not a fundamental difference. It is a feature of the formula that is present even in two party systems as the paper itself explicitly discusses. There is no modification of the formula. To the extent that the original paper focuses on the formula for district by district calculations, and only derives a more general formula for calculating entire states at once for the special case of only 2 parties in one of the appendices (and if the paper has to present a general formula to be used district by district, and a specific state wide formula to be used in the special case of 2 parties, it is absolutely obvious that the general formula doesn't require 2 parties). The argument for using it despite the fact that it can lead to counter intuitive conclusions in lopsided scenarios exists in both a scenario where you have 2 parties and where you have 3 parties. The scenario where the winning party has more wasted votes than the losing party despite winning all seats exists in both bipartisan and multipartisan situations. Whether those scenarios happen frequently enough to warrant not using the formula is an empirical consideration that has nothing to do with the formula. This much is absolutely clear to anyone who actually read the paper and understands basic math. All you need for the formula is to work is to be able to tell what are the wasted votes due to being more than required to win a seat and what are the wasted votes because they didn't lead to winning a seat.

vyelkin
Jan 2, 2011

Rigel posted:

They may be staring into the abyss of the 2020 elections, terrified that the Dems will be able to gerrymander the poo poo out of them as payback.

Maybe this is payback for the last decade of Tea Party primaries to moderate Republicans, fuelled by gerrymandering meaning there's no need for the eventual candidate to appeal to the other party or the nonexistent centre.

Rigel
Nov 11, 2016

vyelkin posted:

Maybe this is payback for the last decade of Tea Party primaries to moderate Republicans, fuelled by gerrymandering meaning there's no need for the eventual candidate to appeal to the other party or the nonexistent centre.

This is also why moderate republicans have been signing on to these briefs, gerrymandering forces Republicans in hopelessly red districts to be crazier than poo poo.

Sulphagnist
Oct 10, 2006

WARNING! INTRUDERS DETECTED

It's not nearly as common or problematic in the big picture but there's some partisan gerrymandering favouring Dems, too.

Evil Fluffy
Jul 13, 2009

Scholars are some of the most pompous and pedantic people I've ever had the joy of meeting.

evilweasel posted:

It appears he's not the only one. A bunch of Republican governors, representatives, and former representatives also signed amicus briefs calling for the court to strike down gerrymandering (including, oddly, some representatives in specifically gerrymandered for their benefit districts).


https://www.nytimes.com/2017/09/06/...itics&smtyp=cur

I'm actually starting to think this has a real chance.

Maybe they think they can get a SCOTUS ruling that'll allow them to go in and gently caress California by claiming it's heavily gerrymandered to be a liberal super majority or something equally stupid?

I refuse to accept that a bunch of GOPers have actually found a shred of decency and are simply acknowledging that Gerrymandering is BS (unless they think 2018 and 2020 are going to be blue tidal waves and geryrmandering would let Dems turn the entire country blue).

e: The reality is that even if the Dems don't gently caress up again and manage to sweep in 2018 and/or 2020 they'll "govern responsibly" and not do that because they're loving idiots.

Evil Fluffy fucked around with this message at 19:29 on Sep 7, 2017

Platystemon
Feb 13, 2012

BREADS

Evil Fluffy posted:

I refuse to accept that a bunch of GOPers have actually found a shred of decency and are simply acknowledging that Gerrymandering is BS (unless they think 2018 and 2020 are going to be blue tidal waves and geryrmandering would let Dems turn the entire country blue).

:same:

I can believe that a tumor hosed with McCain’s brain, but Kasich having honorable motives? LOL no.

evilweasel
Aug 24, 2002

Evil Fluffy posted:

e: The reality is that even if the Dems don't gently caress up again and manage to sweep in 2018 and/or 2020 they'll "govern responsibly" and not do that because they're loving idiots.

Democrats gerrymandered to the hilt when they got the chance to in 2010, it's just they got the chance to in, like, Maryland and that's it.

Rigel
Nov 11, 2016

Evil Fluffy posted:

e: The reality is that even if the Dems don't gently caress up again and manage to sweep in 2018 and/or 2020 they'll "govern responsibly" and not do that because they're loving idiots.

Dems gerrymandered the gently caress out of the GOP in the few states that they had the opportunity. They tried to defeat the independent redistricting commission ballot measure in CA because they wanted to gently caress the GOP in CA too.

Adbot
ADBOT LOVES YOU

joepinetree
Apr 5, 2012
There's also the fact that it is very unlikely that the Republicans will be in a position in 2020 that will be as strong as it was in 2010. If you are going to grow a conscience about this, right before you are about to lose some power anyways is the time to do it.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply