[EM] Re to Kristof M

[Kristofer Munsterhjelm]

David L Wetzell wrote:
> 
> 
> On Sat, Nov 26, 2011 at 3:20 PM, David L Wetzell <wetzelld at gmail.com 
> <mailto:wetzelld at gmail.com>> wrote:
> 
>>    Here's a bunch of responses 
> 
>  
> dlw: SL may be more proportional than LR Hare, but since I'm advocating 
> for the use of a mix of single-winner and multi-winner election rules, I 
> have no problems with the former being biased towards bigger parties and 
> the latter being biased somewhat towards smaller parties.  For there's 
> no need to nail PR if PR itself does not nail what we really want PP, 
> proportionality in power.  This is also part of why I prefer 
> small-numbered PR rules (less proportional) that increase the no. of 
> competitive elections and maintain the legislator-constituent relationship. 

Proportionality in power is quite well approximated by proportionality
in representation, however. The degree of fit depends on strategy and
coordination: if every member of the assembly votes for himself, it's
near-perfect (within per-issue variance that gets evened out); if
everybody colludes into one or two superparties, then it may diverge
greatly.

At that point, the question is how far you should take PR. From my own
observation, PR as approximation of PP has problems in certain edge
cases (kingmaker parties), but these are rare. Therefore I think that as
long as you patch up the edge cases or make them unlikely - say, by an
explicit threshold or an implicit one such as STV's - you can optimize
for PR within those bounds.

Even if you don't think PR approximates PP, you can use the same
advanced PR method to give seats fairly according to PP instead. Poland
has proposed something like this be done in the Council of the EU, the
proposal saying that each country should have a weight proportional to
the square root of the number of people in the country.

My preference for integrating single-winner and multiwinner (if you're
going to do both in the same context) is then that whatever you decide
to apportion on (power or votes), the single-winner method can take it
into account. It knows about it, and so you play it a bit more safe. If
the multiwinner rule is bad, the single-winner rule can compensate, and
if the single-winner rule is bad, the multiwinner rule can compensate.

If you're risk averse, as you said you might be, that's a good property!

The flipside is that you won't get the optimal result if it turns out
that the real thing you should optimize is whatever either the
single-winner or multiwinner method optimizes. If that is the case, then
the multiwinner or single-winner method (respectively) will only be a
burden and pull you away from your goal.

>>    KM:You might be able to get something more easily understood yet
>>    retaining some of the compensation part of the first version, by
>>    doing something like this: first elect the single winner/s. Then
>>    start STV with the single winner/s marked as elected (and thus with
>>    vote transfers already done).
>      
> 
>  dlw:The rub here is the desirability of guaranteeing that the Condorcet 
> winner is elected.  In "more local" elections that attract less 
> attention, I put less emph on the usefulness of rankings and thereby the 
> Condorcet winner.

The single-winner doesn't *have* to be the CW (although I would prefer
the system to ensure it is). Even if you for some reason thought the
Plurality winner was the best one, and wanted to design the integration
accordingly, you could do STV with the Plurality winner already elected.


>>>    dlw:   1. While all forms of PR fall short of proportionality in
>>>    representation, the best predictor of proportionality is the number
>>>    of contested seats. 
> 
>      
>>     KM:The Hix-Johnston-MacLean document states that these effects are
>>    weak. To quote:
>>    "Turnout is usually higher at elections in countries with PR than in
>>    countries without, It also tends to be even higher in PR systems
>>    with smaller multi-member constituencies, and also tends to be
>>    higher where citizens can express preferential votes between
>>    individual politicians from the same political party rather than
>>    simply choosing between pre-ordered party lists. In general, the
>>    more choice electors are offered, the greater the likelihood that
>>    they will turn out and exercise it. However these effects are not
>>    particularly strong, there is some evidence that highly complex
>>    electoral systems suppress turnout, and turnout levels may partly
>>    reflect influences other than the electoral system, for instance in
>>    some countries voting is compulsory."
>>    So I don't think you can necessarily draw that conclusion. The
>>    apparent competitiveness between seats may be lesser (because of
>>    what I mentioned above in that single-member districts are much more
>>    win-all/lose-all), but that doesn't mean the real change in voter
>>    opinion from term to term is any greater in SMD countries.
> 
> 
> dlw: I interpret what they're saying is that other factors also come 
> into play that impact the competitiveness of elections.  So my 
> conclusion could still be"useful", even if it abstracts from a lot of 
> real-world stuff that also affects voter-turnout.  The election rules 
> that best guarantee proportionality tend to reduce voter interest in 
> elections, thereby making PR not the key criteria for choosing an 
> election system.

The whole point is that you can't say that it tends to do so. There are 
so many other parameters, and the conclusion could just as well align in 
the favor of "more choices is higher turnout", as they themselves say. 
If you wanted to make that conclusion rather than yours, you could point 
at that PR in the first place gives higher turnout than non-PR; and that 
countries with ranked balloting types of PR has higher turnout than 
countries with party list, all else held equal. Sure, ranked balloting 
type PR tend to have fewer seats, but then the weak "fewer seats -> more 
turnout" is a correlation, not a causal relationship.

>>>   2. Proportionality in representation does not entail proportionality
>>>   in power and the latter is desired more than the former. As such, it
>>>   seems that minority dissenters will need to use extra-political
>>>   methods (not unlike #OWS) to move the center, regardless of whether
>>>   PR or another mixed system is used.
>      
>>    Proportionality in representation is correlated with proportionality
>>    in power. The correlation isn't perfect, as Banzhaf and
>>    Shapley-Shubik's measures make apparent, but to leap on that and
>>    conclude that proportionality isn't proportional... that's unwarranted.
> 
> dlw: But it waters down the desirability of nailing PR even further and 
> opens the door to a greater valuation of other conflicting criteria.  

See above. If you think you should optimize proportionality of something 
else, feel free to use the fine-honed machinery of PR with a different 
target than proportionality in numbers. Use square root laws, dynamic 
programming or simulated annealing 
(http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.34.6640) to get 
proportionality by power (under assumptions of great amounts of 
collusion) if you wish. In any case, you'll get that for which you 
optimize, instead of a fuzzy perhaps-what-you-want, perhaps-what-you-don't.

>>    KM:If anything, when proportional representation disagrees with
>>    proportionality in power, the power favors the minority parties.
>>    Minor party kingmakers can make themselves costly if they know there
>>    won't be any coalition without them. Hence the presence of
>>    thresholds in most PR systems: these keep too minor parties from
>>    becoming potential kingmakers.
>>    Over here, the threshold of 4% keeps most "swing parties" (as one
>>    may call them) out of power. Yet the threshold is soft - even
>>    parties below 4% of the total vote can get representatives, they
>>    just don't get MMP-esque compensation on the national level. (Our PR
>>    system is a bit unusual in this respect: parties get additional
>>    seats if their per-region seats reflect their national share of the
>>    vote too badly.) Perhaps you'd want a hard threshold for a less
>>    homogenous country, but my point is that the problem can be managed.
> 
> 
> dlw: Since PR->PP, we deny PR.  My wider point is that American forms of 
> PR takes a different approach to the problem, one that presumes both PR 
> and single-seat elections are in use so that as long as the latter 
> favors bigger parties, PR may be biased somewhat in favor of smaller 
> parties.

I don't know what you mean by "PR->PP". "PR leads to proportionality of 
power"? Anyway, if the inaccuracy in PR wrt proportionality of power 
favors the small parties, there's the small-party bias you want. Use it, 
then integrate your single-winner method with the PR so that if you 
against expectations are wrong, the method can still work.

>>>   dlwThese seem to imply that we need not strive for proportionality
>>>   in representation as the gold standard for electoral reform.  If the
>>>   two major parties, with a somewhat disproportionate amount of
>>>   representation, are more dynamic then they'd tend to represent well
>>>   the majority of the population and heed minorities that frame their
>>>   issues respectfully.
>      
>>    KM:Do note, though, that the same Lijphart as you referenced on your
>>    page, said:
>>    "If partisan conflict is multidimensional, a two-party system must
>>    be regarded as an electoral straitjacket that can hardly be regarded
>>    as democratically superior to a multiparty system reflecting all the
>>    major issue dimensions." ("Democracies: Patterns of Majoritarian and
>>    Consensus Government in Twenty-One Countries", 1984, page 114.)
> 
> 
> dlw:What if there's a dialectic between multi-dimensionality and 
> single-dimensionality that gets worked out in an ongoing process?  If so 
> then a 2-party system isn't so bad if it need not be the same two 
> parties de jure and de facto and the two major parties together serve as 
> melding(not melting) pot with significant inputs from dissenters/third 
> parties who raise up new dimensions of conflict into our political 
> systems that lead to a repositioning of the de facto two major parties.

I'd imagine Lijphart knows that parties are not static fixed-in-stone 
things. Beyond this, I don't know what arguments Lijphart employed. 
Myself, I would say that the difference is like that of a proposal to 
use market pricing in a corporation versus doing logistics calculations 
directly: the first gets there in a slower, more roundabout way; the 
latter goes right to what is needed.

Of course, since I'm not Lijphart, I'm not the same kind of authority 
and so you could counter more easily. But also note that Lijphart is in 
favor of multiparty democracy for other reasons, too. To quote Wikipedia 
( https://en.wikipedia.org/wiki/Arend_Lijphart#Major_works ):

"While Lijphart advocated consociationalism primarily for societies 
deeply divided along ethnic, religious, ideological, or other cleavages, 
he sees consensus democracy as appropriate for any society. In contrast 
to majoritarian democracies, consensus democracies have multiparty 
systems, parliamentarism with oversized (and therefore inclusive) 
cabinet coalitions, proportional electoral systems, corporatist 
(hierarchical) interest group structures, federal structures, 
bicameralism, rigid constitutions protected by judicial review, and 
independent central banks. These institutions ensure, firstly, that only 
a broad supermajority can control policy and, secondly, that once a 
coalition takes power, its ability to infringe on minority rights is 
limited."

Now you could say that "it'll be different". Okay, but then your 
particular construction would have to give significantly different 
outcomes than the range of majoritarian democracies that Lijphart studied.