[EM] UK electoral systems "post mortem" discussion on radio

[Kristofer Munsterhjelm]

On 06/12/2015 10:47 PM, Fred Gohlke wrote:
> Thank you, James
>
> I particularly appreciate your link to the election results.
>
> I understand the point you are making, and, perhaps oddly, agree that
> the Frome event "doesn't do much for democratic representation of
> voters".  However, the "poke in the eye for party politics" should not
> be ignored.  I suspect the difference in our views is that I see
> proportionality as an internal quality, not an external one.

Ideally, proportionality *would* be internal. If it were, we could use 
single-winner methods to elect a large group of centrists, and each 
centrist would be internally proportional, being rather more like a 
judge than an advocate for his position. Each centrist would be able to 
consider all sides of an issue, weighing the different sides as the 
people would, and then come to a conclusion.

However, there are two problems. First, there's a matter of skill in 
itself. Being a balanced judge-representative requires considerable 
ability on part of the representative - moreso the fewer representatives 
there are on the council. They'd have to both be balanced, and to retain 
that balance in the face of corruption by power.

Second, election methods provide their own selection pressure. Plurality 
selects the largest sufficiently cohesive group (the representative with 
a plurality). IRV usually chooses from the strongest wing (the strongest 
individuals of the collectively strongest group by Plurality metrics); a 
similar logic holds for DAC and DSC.

To put it more briefly: the election method chooses the candidates that 
are electable according to a combination of its logic and the wishes of 
the people (possibly with imperfect information). Both must be up to the 
task in order to elect judge-representatives by ordinary means (that is, 
by using an election method as it would be used in an ordinary 
presidential or parliamentary system).

Perhaps the advanced systems (Condorcet etc) would be good enough that 
you would get internally proportional representatives as above. But I'm 
not sure the voters would be able to accurately point out such 
representatives to begin with, or would desire that sort of inclusive 
centrism. I might be wrong, but if I'm not, then external PR methods 
will at least ensure some measure of proportionality; getting internal 
proportionality in the sense I mentioned above would then require a more 
nonstandard system, not just a change of election methods.

(Note that standard PR methods don't "punish" per se. They simply limit 
the influence of those who aren't ranked or rated highly by some subset 
of the people. Indirectly, I can see that it would be considered a form 
of punishment, but it is not directly a form of punishment the way, say, 
impeachment or votes of no confidence are.)

As an example of a nonstandard system, consider a form of election by 
lot, where the resulting council then makes use of advisory and 
implementation organizations (somewhat like departments) to explore and 
implement the policies they think should be implemented. This 
nonstandard system is proportional if the council is large enough. Each 
particular member of the council might not be a particular good judge, 
but wisdom-of-crowds logic might make the council - as a whole - behave 
like a good judge. There's little recurring corruption, but there might 
be some corruption if the people in general are corrupt; this would take 
the shape of "I won the lottery, now I'm going to get as much money I 
can before I'm thrown out because I'm surely not going to win twice".

The nonstandard system works to the extent it does because it breaks the 
assumptions implied in the question "what voting system do we use?". But 
it isn't an election system in itself.

> When proportionality is seen as an external quality, it is
> confrontational; it sets the proportions at odds with each other.  It
> does nothing for the democratic representation of voters because it
> empowers a relatively radical portion of the electorate at the expense
> of the common-interest oriented portion of the electorate.

This, however, might be an artifact of the current systems. Or, rather, 
even if we want external proportionality, we can make it a lot less 
confrontational with the right methods.

First consider single member districts with Plurality. You either get 
the strongest radical (because the candidates with broad support have 
their votes split) or you get the lesser of two evils. The strongest 
radical might be proportional if each district has a different radical, 
but you still get a lot of radicals. The lesser evil might end up not 
really representing anybody.

Parties in party list PR are somewhat more balanced: they are subject to 
dual pressures. On the one hand, they have to be "radical" (clearly 
establish their identities and lean in a particular direction) to draw 
members and support. On the other hand, they are pulled towards the 
common position by the fact that there are more voters there. Parties in 
party list PR might be proportional when compared to other parties; that 
is, if party X has double the support of party Y, then twice as many 
voters are closer to party X than party Y. However, they might be 
unrepresentative *within* each block: the median party X-voter might be 
to the left (or right) of party X's official position, and the same 
thing holds for party Y.

But it might be possible to construct a candidate-based election method 
(like STV) that elects compromises within each slice. Like the good 
single-winner methods, it would pick a candidate with broad support in a 
single-winner election; and like the multiwinner methods, it would be 
proportional in that each slice of opinion space corresponding to a seat 
is given to an equally large share of the people.

My point is that while party list allocates slices to parties fairly, it 
doesn't ensure that the party is internally representative, and while 
something like STV allocates on a candidate basis in a way that 
represents groups well (due to the DPC), its reduction to IRV in the 
single-winner case means that it may not find the right compromise in 
each particular metaphorical slice. A Condorcet-PR mix could potentially 
do both. It's only unfortunate that these systems are hideously complex.

(The slice analogy here might be easier to understand if you think of a 
left-right axis as an example. Then, in a three-seat election, the first 
slice corresponds to the leftmost 33% of the voters, the second 
corresponds to the center 33%, and the rightmost corresponds to the 
right 33%, at least with Hare. The intuition for Droop is more complex.)