[EM] To Bill Lewis Clark re: stepping-stone

Adam H Tarr atarr at ecn.purdue.edu
Sat Jan 24 17:32:02 PST 2004

(Sometimes top-posting feels right.)

David's post is an example of one of the more novel arguments for IRV.  
Basically, it goes like this:

Axiom 1 - We're electing a legislature.  There are three (perhaps more) 
parties.  The "centrist" party is the weakest, in terms of first-place 
Axiom 2 - Truly proportional, multi-member districts are not possible(*) but 
we want the overall results to have some degree of proportionality.

Option 1 - If we use plurality, the centrist party has to out-poll both 
parties in first place support to win a district.  The centrist party will be 
under-represented, since they have poor first-place support.

Option 2 - If we use Condorcet or Approval-ish methods, the centrist party 
does not need to out-poll either party in first place support.  In fact, the 
centrist party will win each district unless one of the other parties has an 
absolute majority (50+%) there.  Therefore, the centrist party will be 

Option 3 - If we use IRV, the centrist party will win a district if they can 
outpoll at least one of the other two parties, and the other has less than an 
absolute majority.  It's hard to predict how often this will happen, but it 
obviously must be somewhere between the results of options 1 and 2.

Conclusion - use IRV, since it has the potential to produce good 
proportionality, whereas the other methods are almost assured of producing 
lousy proportionality.

Note the asterix (*) that I put in "Axiom 2".  This is the hole in the 
argument, in my opinion.  Condorcet and plurality are both lousy methods for 
producing proportionality, and IRV is only a potential improvement by virtue 
of being really erratic.  A method that actually gives proportionality (be it 
Proportional Approval Voting, Single Transferable Vote, or some form of party 
list voting) is far, far superior to all three.

Once again, IRV is "winning the day" simply by showing itself to be better 
than something that's not good.  And again, the improvement in proportionality 
that IRV shows here is only a possible one.  Why should we settle for IRV, 
when there are so many other methods (PAV, STV, open list, et cetera) that are 
much better?

I say "once again" because this harkens back to the single winner, IRV vs. 
Approval/Condorcet argument.  Why should we fight for IRV, when Approval is 
cheaper, easier, and better?  And when Condorcet is so much better?

Some argue that baby steps are better then no steps at all, but others (most 
notaby mr. Ossipoff) argue that bad reforms will get us stuck with the wrong 
system for a long time, and we're much better off arguing for a good system 
right away.


David Gamble wrote:

>Various people have said that IRV has no real advantage over Plurality.
>Consider a three party system like England ( or a four party system like 
>Wales or Scotland). The third party In England the Liberal Democrats is 
>consistently underrepresented in the House of Commons. In their book (The 
>General Election of 1997) analysing the 1997 General Election the authors 
>calculations of the number of seats the parties would have obtained under the 
>Alternative Vote (IRV). The Lib Dems actually obtained 46 seats out of 659, 
>IRV they would have obtained 91. 
>Whilst IRV is certainly inferior to any proportional system ( such as STV or 
>MMP) in probably all British General elections since 1960 the third party 
>would have gained more seats under IRV.
>For an example of how unfair single seats elections under Plurality consider 
>the 1983 General election result.
>Conservative  42.4% vote 397 seats
>SDP/Liberal Alliance  25.4% vote 23 seats
>Labour  27.6% vote 209 seats
>Others 4.6% vote 26 seats
>Any single seat system will give similar distorted representation but the IRV 
>result would have been less distorted. This is because the Alliance candidate 
>being the centre candidate would have gained the second preference support of 
>most Labour and most Conservative voters and therefore could have won any 
>seat where they gained second place and where the 1st place candidate lacked 
>David Gamble

More information about the Election-Methods mailing list