[EM] The Global Fight For Electoral Justice: A Primer

[Kristofer Munsterhjelm via Election-Methods]

On 12/27/2016 09:38 AM, Erik Moeller wrote:
> On Fri, Dec 23, 2016 at 9:53 PM, Michael Ossipoff
> <email9648742 at gmail.com> wrote:
> 
> Michael,
> 
> Thanks for your detailed thoughts.
> 
>> 2. Not Precinct-Summable:
>>
>> If we had pubic ballot-imaging, then any voting system would be secure against count-fraud.
>> But we don't, and verifiable vote-count is something that we'll probably never have. So, given the
>> ridiculously questionable vote-counting, the last thing that we need is a voting-system that makes
>> count-fraud even easiesr than it already is.
> 
> That's a good point, and I think anyone who's been paying attention to
> the ballot scanner chaos in Michigan and some of the other disasters
> of this election would take these concerns seriously.
> 
> I'm not familiar with the full set of rules governing ballot
> initiatives, but would it be feasible to make IRV ballot initiatives
> contingent on a third party voting equipment audit and overhaul?

One note here: any such ballot images should either have very tightly
regulated access, or be an aggregate that's large enough that
vote-buying becomes impractical.

Suppose there are 9 nobodies running in addition to the serious
contenders. Then if the full preferences are public, a vote-buyer could
tell a vote-seller to vote these in a particular order first on his
ballot to prove that he voted as instructed. There are 362880 possible
ways to rank these 9 first, so that could work as a fingerprint. But an
aggregate (summable) image would not be susceptible to that kind of attack.

> == IRV vs. other methods ==
> 
> There seem to be a few competing reform efforts in the US right now:
> 
> 1) the IRV->STV path, which appears to be Fairvote's main reform
> strategy, and which has plenty of precedent (positive and negative)
> 2) some limited advocacy for European-style PR models, mostly by academics
> 3) business-backed initiatives for nonpartisan blanket primaries
> ("jungle primaries")
> 4) some completely novel schemes, e.g. score runoff voting (which is a
> possible upgrade for 3)
> 
> Am I forgetting some?
> 
> A few comments on each of these reform projects.
> 
> === The IRV->STV path ====
> 
> I am skeptical about this path not because of the Condorcet Winner
> deficiencies of IRV, but because we've seen repeals of IRV and STV in
> the past. I think those repeals have almost always been politically
> motivated and used anti-intellectual arguments ("pinball voting",
> "lottery voting"), so we should not overestimate the role that IRV's
> failure to find the Condorcet Winner played in repeals. Jack Santucci
> argues convincingly that repeals generally are the result of changing
> party power constellations:
> 
> http://www.jacksantucci.com/docs/papers/repeal_dec2016.pdf

I gave it a quick read (I didn't look through the details of the data
points), and it seems like Jack is suggesting that a PR system is
unstable if/when the second largest party can't influence legislation,
or, to quote the paper, "permissive voting rules like RCV are stable
when the second-largest party shares power with the first on
the most salient dimension of conflict."

I live in Norway, and although it is parliamentary, there are no snap
elections. Legislative terms are fixed, and thus the electoral dynamics
should be within the scope of the paper, but the description doesn't
seem to match my experience.

The largest party here is the Labor party, and the second largest is the
Conservative party (see e.g.
https://en.wikipedia.org/wiki/Norwegian_parliamentary_election,_2013 ).
It's very uncommon for both of these parties to be in power at any given
time. What tends to happen is either:

- One of the two parties form a coalition towards the center and rule as
a majority bloc, or

- a minority government forms and has to seek support on a case-by-case
basis.

The model of the paper seems to suggest that, say, if the Conservatives
are in power, Labor should say that "we don't have a say anyway as we're
not in power, and if we repeal PR, all the minor parties will go away
and we'll get an increased seat share, so let's repeal PR". But trying
to repeal PR would be unheard-of.

So there has to be some additional assumption that holds for American
congressional/presidential politics but not parliamentary ones with
fixed terms; or that holds for STV-type PR but not party list PR. Or
perhaps the argument mainly holds for a transition period between
two-party and multiparty rule?

One could argue that the coalitions shift pretty often so that in the
longer term, Labor knows that it will regain power at some point. But
that leads to a continuity argument: how long an "often" is good enough
to keep PR? It's hard to tell. Also of note is that Norway, IIRC, used
to have single member districts up until the 1920s.

The paper also seems to assume that IRV will lead to STV all else equal
(i.e. in the absence of game-theoretical incentives to block PR), since
it uses the term RCV for both STV-type PR and for IRV as a single-winner
proposal. But IRV is not, in itself, very proportional. Consider
Australia: the IRV House of Representative is two-and-a-half party,
while the STV Senate is somewhat closer to multiparty. So the
introduction of IRV may not necessarily give more power to smaller
parties who could push harder for PR.

(I just stumbled across a possible explanation for European introduction
of PR - that it was a reaction to the increasing popularity of socialist
parties, where the old parties came to the conclusion that it's better
to be a smaller fish in a proportional pond than to be wiped out
entirely if the socialists were to gain enough support to become the
majority party. See e.g.
http://web.stanford.edu/~jrodden/wp/rodden_jan10_workshop_final.docx . I
don't know if this is the true reason, but if so, the US would be
different since there was no such great threat of socialist majority.)

> The story of IRV/STV's rejection suggests to me that simplicity is one
> of the greatest virtues of voting system reform (which relates to your
> point about voting security as well), and should lead us to question
> whether this is really the best path, regardless of near-term
> successes as in Maine.
> 
> At the same time, we should all be prepared to refute
> anti-intellectual arguments, and be ready to speak to IRV/STV's
> virtues relative to plurality.
> 
> === European-style PR models ===
> 
> As a European, I love open list PR, but I understand the desire to
> have local plurality winners that's a big part of US culture. One of
> the most interesting proposals in this regard is Single Ballot Mixed
> Member Proportional (SB-MMP), as explained here:
> 
> http://www.parl.gc.ca/Content/HOC/Committee/421/ERRE/Brief/BR8397882/br-external/HutcheonDavidA-TomekJennifer-e.pdf
> 
> And here:
> 
> http://www.parl.gc.ca/Content/HOC/Committee/421/ERRE/Brief/BR8538777/br-external/SlavenRobert-e.pdf
> 
> This system is easy to summarize: You vote for your preferred local
> candidate and party with the same vote. The plurality winner becomes a
> district representative. The overall party % is used to draft
> additional region-wide representatives, preferring the best-performing
> runners-up. The end result is a proportional legislature or council in
> which ever member has strong district ties, and where vigorous
> district-level campaigns are the norm.

MMP pays for this accuracy of district representatives by a greater
distance to the regional representatives. To use a metaphor, while STV
evenly spreads the distance from voters to representatives throughout
the district, MMP moves some of the distance away from the district
level (thus the district representatives seem more representative) onto
the regional level (where they seem less so).

If having local winners is more important, then shuffling the inaccuracy
over to top-up representatives might be a good idea, though. In any
event, it is better both locally and regionally than a nonproportional
method.

I'd probably prefer something closer to biproportional SNTV/party list,
myself, if local winners are important. Something like:

- Each voter votes for a candidate which also has a party assigned.
- The method uses party list PR to determine how many seats each party
should get.
- The method then reweights the parties (how many votes count as one
unweighted vote), reducing the weight of parties with too many seats and
increasing the weight of parties with too few seats, until the number of
seats per party matches the party list outcome.

A benefit to this method is that if party Y has too few seats, the
reweighting will make Y gain seats first where Y was close to winning
anyway. But the reweighting might be too complex to explain, or more
complex than MMP, and might also face criticism of the sort that
"candidate X got the most votes but didn't get a seat!"

> That last part, I think, is a pretty big deal. Imagine Green Party
> candidates in every district really fighting hard to get the most
> votes, because it may make the difference for _them_ to get drafted to
> the legislature. It would lead to a much richer, more diverse, more
> competitive democracy. And it would instantly obsolete gerrymandering,
> allowing districts to be drawn in a manner that reflects people's real
> neighborhoods.
> 
> Legal scholar Allan Ides similarly has expressed strong support for a
> (two vote) MMP style model for the California assembly, alongside
> transition to a unicameral assembly. ICYMI, his "Approximating
> Democracy" paper from 2011 is excellent, in my opinion:
> 
> http://digitalcommons.lmu.edu/cgi/viewcontent.cgi?article=1001&context=llr
> 
> Generally, I think for most reformers feel like MMP is a bridge too
> far in America. However, I think when we consider it in combination
> with other changes, it's perhaps not so crazy an idea.

The single-winner method could also be upgraded from Plurality to some
other method without making the MMP method fail, as long as that other
method is, in some extended way, monotone.

> 
> === Jungle primaries ===
> 
> The jungle primary (technical term: nonpartisan blanket primary -
> https://en.wikipedia.org/wiki/Nonpartisan_blanket_primary ),
> implemented by ballot initiative in WA and CA, is a pretty big change
> to elections and I don't think it's fully understood. It moves America
> closer to the French model of two-round voting, except that in France,
> every party still gets to hold its own primary, and both general
> election rounds are a big deal with 80% turnout.
> 
> In the jungle primary as implemented in CA, you may get weird
> intra-party spoiler effect, or you may end up with two candidates from
> the same party in the final round.
> 
> It's overall an odd reform; I'm frankly surprised it was implemented
> given the obvious adverse consequences for small parties (they don't
> have much of a chance to get to the general ballot anymore) and the
> seemingly insane decision to use plurality voting for the first round.
> Oregon soundly rejected it, but it's backed by business interests who
> want to see "less partisanship" so it will probably keep coming back
> onto the ballot in other states.
> 
> === Other reform initiatives ===
> 
> Range voting seems to have a lot of dedicated fans, and the Score
> Runoff Voting proposal is pretty elaborate:
> 
> http://www.equal.vote/
> 
> Intuitively, I'm skeptical of the strategic voting properties in the
> real-world, and the arguments on the site aren't wholly persuasive in
> that regard. But some alternative-voting first-round (perhaps approval
> rather than score) makes sense to me as an upgrade for jungle
> primaries.

One problem with both Approval and Range/Score in a runoff is that
cloning pays. Suppose every party fields two identical candidates rather
than one, and every voter ranks/rates them adjacent to each other. Then
the two runoff spots will be occupied by the two clones from the party
that won, which in essence bypasses the whole runoff.

> === Centrists vs. partisans ===
> 
> A big distinction between many reform efforts seems to be whether they
> want to elect centrists who appeal to everyone, or partisans who
> appeal to a large faction while being very off-putting to others. It
> doesn't make sense to me to aim for pure centrism in assemblies,
> Congress, etc. -- you lose the ability to _directly_ represent
> specific concerns, and instead aim for some kind of prototype
> politician-bot who can serve all constituencies. It's understandable
> why business interests would want such a system, since it has a
> consistent API for lobbyists.
> 
> But it's unlikely the general populace will be very satisfied with it
> in the long run, since many concerns and ideas will never be directly
> represented in such a system. And I doubt that it produces the best
> politics, because it lacks the tension/disruption that makes
> innovation possible.
> 
> At the same time, in the American presidential system, positions like
> Governor or President are arguably _least_ suited for strong
> partisanship, because these figures have to be able to speak to
> everyone, and help bridge divides in Congress or a state legislature.
> So this is where you might actually bias in favor of a system that
> elects centrists.

I tend to think of it like this:

- If you have only one winner, the best you can do is to have the winner
represent the people to the greatest degree possible.

- If you have many winners, it's better to have the compromise-seeking
happen out in the open than in the minds of the candidates. (This also
serves as protection against corruption by power, lobbying, etc.)

So you should have different winners with differing positions, but these
winners should be centrist candidates with respect to the dimensions
that don't fit due to too few seats.

Or, if there's only one dimension in opinion space, then a single-winner
method should find the candidate closest to the median voter, and a
two-winner method should find the candidates nearest to the voters at
the 33. and 67. percentiles.

> Hence my personal bias, so far, based on everything I've read about
> reform in the US and voting systems elsewhere:
> 
> 1) Transition to unicameral legislatures for the reasons outlined by Allan Ides,
> 2) Transition to SB-MMP or another MMP variant to elect members of
> these legislatures, as also well-argued by Ides; possibly extended to
> city councils, as well;
> 3) Replace the first-round voting system in jungle primaries and limit
> it to unique offices like governor, president.
> 
> Regarding 3), the use of approval voting, for example, would mean that
> the two candidates advancing into the final round would both be people
> with wide appeal (although they might still be from the same party,
> which is a bit bizarre for an office like governor).
> 
> Note that none of the above includes reference to IRV or STV. We may
> be able to do away entirely with any notion of ranking, or even
> scoring, by focusing on "single X" and "multiple X" methods. This is
> not a criticism of Fairvote's efforts, but simply a hypothesis that a
> reform platform more closely in line with the above steps would
> produce better long term outcomes. It's simple for voters, yields
> proportional representative bodies, and balances centrism with
> partisanship.
> 
> If you've made it this far, thank you -- would appreciate anyone
> poking holes in the above logic.
> 
> Warmly,
> 
> Erik
> ----
> Election-Methods mailing list - see http://electorama.com/em for list info
>