[EM] Why I think IRV isn't a serious alternative 2

[Kristofer Munsterhjelm]

Abd ul-Rahman Lomax wrote:
> At 10:36 AM 12/28/2008, James Gilmour wrote:
>> Kristofer Munsterhjelm  > Sent: Sunday, December 28, 2008 9:45 AM
>> > The UK is also parliamentary, so I suppose there would be few places
>> > where you could actually have a runoff.
>>
>> Given that all members of the UK Parliament are elected from 
>> single-member districts (UK "constituencies") and that all districts
>> were contested by at least three candidates (max 15 in 2005), it would 
>> be theoretically possible to have run-off elections in all
>> 645 districts.  In the 2005 general election, 425 of the districts 
>> were "won" with a plurality of the votes not a majority, so that
>> could have been 425 run-offs.  Quite a thought!
> 
> Sure. Consider the implications. Most of those who voted, in those 
> districts, did not support the winner. Odd, don't you think, that you 
> imagine an outcry over a "weak Condorcet winner," when what is described 
> is, quite possibly, an ongoing outrage.
> 
> Is it actually an outrage? It's hard to tell. It's quite possible that 
> the majority was willing to accept the winner; that is normally the 
> case, in fact. Bucklin would have found some majorities there. IRV 
> probably -- in spite of the theories of some -- probably a bit fewer. In 
> nonpartisan elections, IRV almost never finds a majority when one is not 
> found in the first round, but those were, I presume, partisan elections, 
> where finding a majority is more common.)
> 
> However, consider this: the Plurality voting system (FPTP) encourages 
> compromise already. There would have been more sincere first preference 
> votes. My guess, though, is that the use of, say, Bucklin, would have 
> resulted in *at least* half of those pluralities becoming a majority, 
> possibly more. However, this is the real effect of the system described:
> 
> In maybe one election out of 10, were it top two runoff, the result 
> would shift, which, I contend, is clearly a more democratic result. 
> There might be a slightly increased improvement if the primary method 
> weren't top two Plurality, majority win, but a method which would find a 
> Condorcet winner or at least include that winner in a runoff. How much 
> is it worth to improve the result -- it could be a very significant 
> improvement -- in 10% of elections?
> 
> I'd say it's worth a lot!

I'd also say that even if you had a magic "best utility" single-winner 
method, it wouldn't be the right method to use in a parliamentary 
context. If a majority of the voters agree with some position (and we 
discard gerrymander-type effects, intentional or not), then the 
parliament will agree with that position, unanimously. Hence... PR is 
needed.