[EM] Delegate cascade and proportional representation

Juho juho4880 at yahoo.co.uk
Wed Aug 20 14:18:27 PDT 2008


On Aug 19, 2008, at 20:11 , Michael Allan wrote:

> Juho wrote, in thread PR favoring racialminorities:
>>
>> ... I was also thinking about trees that offer more detailed
>> grouping of the candidates.
>
> I just spoke with someone at Texas Tech.  We were discussing how
> cascade voting might be used to elect a proportional assembly.
> Basically, you just take the roots and branches of the trees (straight
> from the election results) and that's your assembly.  Is this roughly
> what you guys are proposing, in this sub-thread?

Basic tree based methods could be maybe easiest to see as extensions  
of basic open list systems. Instead of having just flat lists those  
lists could consist of more detailed groupings. And those groups  
could in turn be optionally split in groups. Each candidate belongs  
to one of the groups (let's say only to the last layer of the groups  
(allowing individual candidates also in the groups closer to the root  
could be arranged)).

The trees are based on candidate opinions and are constructed already  
before the actual election (so the voters will know the initial set- 
up) (assuming non-continuous elections here). The votes of the  
candidates are just regular votes that do not impact the destiny of  
the votes of those voters that voted this candidate.

Seat allocation should be as well straight forward in this basic tree  
method. First allocate seats to the top level lists/groups based on  
the number of votes of each group. The number of votes of each group  
is the sum of votes of its subgroups+candidates. Then allocate the  
seats using the same method to the subgroups of each top level group.  
And so on until we reach the candidates. Those candidates that have  
highest number votes within their group will be elected.

The tree structure could also change after the initial round (e.g.  
due to some representative moving to a different party). That could  
lead to changes in the votes (now assuming continuous elections).

This description of the basic tree method assumed that the  
intermediate nodes would not be candidates but just "groups". If one  
wants to see the nodes of the tree as "key candidates" then the  
election method could first elect the candidate(s) sitting at the  
node and only then distribute the remaining seats to the branches.  
This would make the method one step closer to closed lists (where the  
party determines the order in which candidates are elected). (This  
explanation of the various combinations of the cascade method and  
basic tree method is getting complex. But I hope you got the idea.  
Basic trees are simple. Could be extended in various ways.)

Juho


>
> ===== quote from private email =====
>
> [This] solution depends on the candidate rankings, as revealed by the
> election results.  For example, here are the current rankings in a
> demo election (in this case, for a Mayor):
>
>   http://zelea.com:8080/votodemo/w/?wicket:bookmarkablePage=% 
> 3Avotorola.a.election.WP_Count&s=demo-mayor
>
>     Or see the screen shot:
>
>   http://zelea.com/project/votorola/a/web/_/guide/results.png
>
>   (These results are unrealistic.  The voters are mostly computer
>    drones, and there's only a single cascade - 100% of the votes
>    flowing to the leading candidate.  Normally there'd be many
>    cascades.)
>
> What you see in the rankings (above) corresponds to the roots, and to
> the thickest branches of the cascades, as they exist at a particular
> moment in time.  For illustration, in the smaller election below, the
> top-ranked candidates are X, W, N and M (left cascade), and J and H
> (right).
>
>
>                         (I)  (K)  (L)
>                           \ 1 | 1 /
>                            \  |  / 1    (A)   (B)
>                   (P)  (O)  \ | /        | 1  /
>          (R)        \ 1 |    \|/         |   / 1
>            \ 1       \  | 1  (M)         |  /
>             \         \ |     |          | /  (E)  (F)
>              \         \|     | 4        |/    | 1 /
>           1   \        (Q)    |         (C)    |  / 1
>       (S)-----(T)        \ 3  |          |     | /
>                 \ 3       \   |          | 3   |/
>                  \         \  |          |    (H)-----(G)
>                   \         \ |    (D)   |    /     1
>        1       2   \         \|      \ 1 |   /
>    (U)-----(V)-----(W)       (N)      \  |  / 4
>                      \ 6     /         \ | /
>                       \     / 8         \|/
>                        \   /            (J)
>                         \ /
>                         (X)              8
>                                         ---
>                          14
>                         ----
>
> The solution is to call an election for a proportional assembly... and
> use the ranked results to populate the seats (N seats).  If we open
> the voting to the entire jurisdiction... and take the top N candidates
> (rank 1 to N) from the results, enrolling them in the assembly, then
> the membership will accurately reflect the structure of the
> electorate.  If N=6, for instance, the current members of the assembly
> are:
>
>   X, W, N, M  (left side)   J, H  (right)
>
> We could re-enroll the assembly at frequent intervals.  People will be
> continually shifting their votes as new information becomes available
> to them, and the rankings may shift as a result.
>
> ===== end quote =====
>
> My own work is aimed at using a delegate cascade to open actual
> legislation, policies, and so forth, to direct voting.  But it's
> interesting to consider how it might interplay with PR elections.
>
> -- 
> Michael Allan
>
> Toronto, 647-436-4521
> http://zelea.com/
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for  
> list info


		
___________________________________________________________ 
The all-new Yahoo! Mail goes wherever you go - free your email address from your Internet provider. http://uk.docs.yahoo.com/nowyoucan.html




More information about the Election-Methods mailing list