[EM] the intrinsic value of the metric of *strength* of personalpreference (was: Re: Compatibility)

[Juho]

On May 5, 2010, at 10:24 PM, James Gilmour wrote:

> At present all 650 MPs are elected from single-member districts  
> (here called constituencies).  It is impossible to have a PR voting
> system that is based only on any voting system exclusively within  
> singe-member districts.

Yes, if the single-member districts are independent. In principle one  
could also have a PR method where the votes are counted at national  
level and then a proportional result is forced on the single-member  
districts so that in some districts the FPTP winner does not win but  
that seat is given to some other party. The smallest parties would get  
seats in districts where they are strongest (although not in a  
majority position). This method may not be very nice in the sense that  
in some district the voters will be represented by a person that  
represents only a minority view in that district.

> The main thrust for reform is for STV-PR with sensibly sized multi- 
> member electoral districts.  For example, Edinburgh presently
> elects 5 MPs from 5 single-member constituencies.  The City of  
> Edinburgh should be ONE 5-member STV-PR electoral district.
> Similarly, the City of Glasgow should be a 7-member electoral  
> district.  In rural areas the district magnitude could be less, with
> even one or two single-member districts reflecting remoteness and  
> long-standing political "realities".

Larger districts allow better proportionality. Variation in the size  
of the districts means that some districts are more proportional than  
others. And at national level small parties may suffer since they can  
get representatives only from districts that have good enough  
proportionality / many enough members.

Note that in Finland the planned electoral reform aims at balancing  
the different and unfair treatment of the smallest parties in the  
different size districts (from 6 to 32) using this very same  
technique, i.e. the proportions are to be counted at national level  
and only then one finds the best fit to allocate the seats to the  
districts. As a result it should be possible and sensible to vote for  
the smallest parties also in the smallest districts.

One benefit of single-member or small districts is that  
representatives are closer to the voters (less population to listen to  
and communicate with). If one wants to maintain some of this property  
(also for reasons of tradition) one could use some relatively small  
district size. If one determines the proportionality at national level  
and then allocates those seats to the districts there will be some  
distortion in what parties will get seats in which regions. But  
already with district size of 5 (or even smaller) the distorting  
effect may be quite marginal. Even if some small party would on  
average get e.g. (votes worth of) 0.2 seats per district, it is  
probable that there are districts where that party got more than 0.5  
seats. Giving them a seat in that district could already be considered  
not a violation against the will of the voters but just a fair way to  
allocate the seats. Some large party that managed to get only 3.7  
seats could well get its fractional (0.7) seat in some other district  
and satisfy with the idea that the small party got exceptionally good  
result in this district and therefore deserves to get that seat.

So, my point is that one could well have both very accurate  
proportionality and small districts, and also only marginal distortion  
with respect to rearranging the allocation of seats between the  
districts. If one determines the proportionality at national level the  
optimum district size will be smaller that it would otherwise be (if  
there is some interest to have smallish districts). If UK will make a  
reform one option could be to aim at keeping both small districts and  
aim at very accurate proportionality (and same level of  
proportionality and viability of the smallest parties across all the  
districts).

Juho