Monday, 23 September 2013

WA Senate: Follow-up posts in review of current results

WA Senate

National Party Ticket vote increases in value disproportional to original transfer value when allocated to Linda REYNOLDS and filtered through PUP
Count 27: Linda REYNOLDS (Liberal) elected #4

2,582 (0.20%) votes (64,034 ballot papers at 0.0403 transfer value) originally from The Nationals distributed to Palmer United Party (Zhenya WANG) via preference 15.

Count 29: Zhenya WANG (Palmer United Party) elected #5
2,911 (0.23%) votes (64,034 ballot papers at 0.0455 transfer value) originally from The Nationals distributed to Australian Labor Party (Louise PRATT) via preference 38.
The LNP vote also increased in value when Liberal (Linda REYNOLDS) was elected and her surplus transfer value calculated She absorbed part of the National party and SEP vote which were devalued in the calculation of the Liberal Party Surplus Transfer value. The Liberal Group Ticket increases in value disproportional to its contribution to REYNOLDS surplus
there are a numbeR of serious flaws in the way the AEC counts the Senate vote.
1 The calculation of the surplus transfer value
2. The segmentation distribution
Both methods were designed to facilitate a manual count
You could add a third if you wanted to remove the Droop quota (x/(y+1)) and adopt a pure proportional count (y/y)
If you count the current vote using the registered above-the-line group voting tickets you get the following results
As per current senate rules
LNP 3, ALP 2 PUP 1
Using a weighted Surplus Transfer value count
LNP 3, ALP 1 PUP 1, GRN 1
If you then remove the segmentation distribution and adopt a single transfer reiterative count which reflects more accurately the voters intentions kin the distribution of preferences. One single transaction per candidate only distributed weighted surpluses
The result is
LNP 3, ALP 2 PUP 1
What the above analysis shows is that you can not fix the system by just applying a weighted Gregory transfer value. You must also remove the segmentation distribution.
Segmentation allows the vote to skip and jump preferences even though an early preference indicates distribution to a candidate who has not been excluded.
Both non Weighted Surplus and segmentation are flaws in the method of counting that were introduced to facilitate a manual count.
With the use of computer aided counting we can and should adopt a reiterative weighted surplus counting system such as Meek or Wright that excludes segmentation.

A reiterative count would ensure that the vote is redistributed as if the excluded candidate had not stood. It would form part of the Groups surplus value and be transferred as part of that groups overall surplus at a recalculated transfer value
The point of segmentation plays a significant role in the distortion and allocation of the count
We Should only be distributing surpluses on each iteration/distribution. If a candidate is excluded from the count, the count should be reset and restated ignoring preferences for excluded candidates. This way a vote at full value is always transferred to the first available continuing candidate. Surplus transfers should be weighted to the value of the vote and transferred as a single transaction.
A liner distribution model

No comments: