[EM] STV - the transferrable part is OK (fair), the sequential round elimination is not

[Jonathan Lundell]

On Nov 1, 2009, at 10:51 PM, Juho wrote:

> I wouldn't be as strict as saying that Droop proportionality is an  
> absolute requirement. I'd be happy to classify all methods that  
> approximate the principle of x% of votes means x% of seats as  
> "acceptable PR".

I'd like to see a definition of what that really means.

To harp on California again: we have 53 Congressional districts, all  
(of course) single-seat FPTP. The distribution of Democratic and  
Republican seats is surprisingly close to representing state party  
registration.

Is this acceptable PR? I hope your answer is "of course not" (if it  
isn't, we can have that discussion).

The important thing about DPC is that it guarantees proportional  
representation to solid coalitions. The PR isn't dependent on  
strategic nomination or voting, on segregated or gerrymandered  
districts, or on fortunate accident.

If we didn't have DPC methods, then we'd certainly be justified in  
finding alternative "acceptable" methods. But since we do, it seems to  
me that alternative methods have a high bar to meet.

(I'd class party lists as at least potentially meeting the DPC, within  
whatever nomination and threshold constraints they have.)

>
> Note that even if some method strictly follows e.g. Droop  
> proportionality there may be other factors that distort the picture.  
> It is for example typical that the size of electoral districts  
> causes bigger deviation from proportionality than the method that is  
> used within each district. In the extreme case single member  
> districts may give disproportional power to few (e.g. two) parties  
> (even if the actual method would be proportional (like plurality in  
> a way is for single member districts :-)). Also e.g. 10 districts of  
> 10 seats each typically means considerable bias in proportionality  
> in favour of the large parties.
>
> If the votes (and proportionality) are counted at national level  
> that fixes the (district fragmentation related) problem. STV is at  
> its best in small districts with small number of candidates and  
> seats, so it typically leaves some space to distortion in  
> proportionality as caused by the district structure. List based  
> methods have also similar problems but in them it is easier to have  
> the whole country as one district (=> better proportionality but  
> weaker local representation (and as a result weaker "regional  
> proportionality")), or they can be easily extended to count the  
> "political proportionality" at national level but still allocate the  
> seats in the districts (and thereby maintain also "regional  
> proportionality" and more local representation).

Certainly if we had national PR in the US (or even statewide PR in the  
larger states), we'd have a degree of locality--STV within multi-seat  
"superdistricts", say, or some variation of MMP.

>
> My point thus is that proportionality should be observed at the  
> "national level", taking into account also factors like districts  
> and number of available candidates and parties, cutoffs,  
> restrictions in nomination etc.