[EM] Student government - what voting system to recommend?

[Juho]

On Apr 22, 2007, at 20:58 , Tim Hull wrote:

> Regarding the constituencies, the 19-seat one is elected 10 seats  
> one semester, 9 seats the other.  The other multi-seat  
> constituencies are similarly divided.  I would say that none of  
> these can be combined for a simple reason - they do represent a  
> clear group (each individual school/college within the University)  
> as opposed to being a territorial district.  Additionally, each  
> such group has its OWN student government - which makes them  
> somewhat resemble "states".  Thus, combining the single-seat  
> districts would make about as much sense as combining several of  
> the one seat at-large Congressional districts for small U.S. states  
> for STV purposes.  Likewise, there is no logical subdivision for  
> the 19-seat grouping - any such division would be an arbitrary new  
> construction.  One might be able to split based on class status or  
> on off-campus/on-campus residency, but such designations tend to  
> change much more than school/college, leaving some students who run  
> for the seat they are eligible for becoming ineligible to hold it  
> the next semester.  Regarding major party domination of such  
> districts - often these seats are not even contested by the  
> "parties", and half of them are won with a few votes by independent  
> write-ins.
>
> This does present a somewhat weird situation as far as PR and  
> elections, though it seems as if the best solution would be to  
> leave the division of representatives alone.  However, the division  
> between two elections is something to consider.  According to what  
> people think in here, it seems that this may be good for the 19- 
> seat constituency.  However, it seems like it may not be for the  
> others (especially the 2 and 3-seat constituencies, but also the 6  
> and 7 seat).  The problem, though, with doing this (combining some  
> multi-seat elections and dividing others) is that each election is  
> contested by only half the campus (whereas now, each election is  
> contested by 90% of students - everyone minus the 1-seaters not up  
> for election).  Thus, advertising and getting turnout becomes more  
> of a problem.
>
> Any comments on this?  As far as single-winner goes, I see IRV as  
> being the likely choice with STV used in multi-winner due to the  
> fact that it would reduce the amount of explaining (as opposed to  
> doing something like Condorcet).

Both IRV and Condorcet are based on rankings => equally complex to  
voters. IRV is a "single winner STV" so you save in words when  
explaining them to the decision makers, but simplest Condorcet  
methods are easy too (and complex ones more or less explainable too).

The only reason favouring IRV I have seen in this stream is the  
simple explanation. I this is crucial, then that's maybe the way  
forward. Note that IRV and Condorcet differ also on their behaviour.  
IRV favours large parties. For example in the case of three parties  
the candidate of the smallest of them (in first place support) will  
be eliminated in all elections first, even if he/she would be a good  
compromise for all (would e.g. beat both others in pairwise  
comparisons).

>   As far as approval, I really don't see that working very well -  
> only voters who think their favorite has NO CHANCE to win would  
> vote for more than one.  In this case, it seems like IRV is better.
>
> Tim
>
> P.S. Under my "pro wrestler" example, I was assuming that the voter  
> would, under a range system, give the pro wrestler a 3 or a 2 out  
> of 10, except for those who prefer them first.  In this case, both  
> IRV and Range would not elect this candidate.

Range is expressive and it is able to treat these two different types  
of "Pro Wrestlers" differently. Its problem is that it in practice  
easily becomes Approval (only min and max values used) in competitive  
elections. IRV and Condorcet pay no attention to the "numeric utility  
values", only to the relative preferences, and therefore can't make a  
difference between these two Pro Wrestler cases.

(The reason why Condorcet (a ranking based method) is good despite of  
this is that it is not easy to get sincere "numeric utility values"  
from the voters, and it may be better not even try to use that  
information in the calculation process than to try and fail.)

Juho



		
___________________________________________________________ 
To help you stay safe and secure online, we've developed the all new Yahoo! Security Centre. http://uk.security.yahoo.com