[clug] Open Data Democracy

[Rodney Peters]

Michael,

This site gives the algorithm, in principle.  

http://www.aec.gov.au/Voting/counting/senate_count.htm

Does not appear to be too difficult to implement for any number transfer of 
surpluses followed by preference distributions.  The "decision tree" in 
determining whether or not a vote has been "exhausted" is as important as the 
algebra.

My feeling is that in ACT, where parties have only two candidates for two 
senate seats, ATL preference would go to the party's second candiate and 
likely be immediately "exhausted".  Much the same for BTL.

In the states, where 12 seats are contested, the algorithm is likely to be run 
for many iterations.  Consequently any error in the total program could 
produce erroneous results.
 
Rod

Disclosure:  I am a member of Sustainable Australia party.

On Wednesday, 29 June 2016 10:50:24 AEST Michael James wrote:
> > > The AEC publish a file with the raw vote tallies, this preference
> > > pattern scored this many votes. This file must have enough data to
> > > actually run the counting algorithm. Just first preferences (while very
> > > useful for regional analyses) is not enough.> 
> > The BTL files *should* allow you to construct all the preferences
> > (I haven't sat down and tried so I might be wrong).
> > For ATL, at the last election you wouldn't have anything beyond first
> > preferences - you would use the Group Voting Tickets to reconstruct the
> > full preference sequences.
> Dear Andrew,
> 
> Yes, you’re right, in the old senate system it’s enough to know
>  the count for each possible Above the Line vote
>  as they map to Below the Line using the party registered preference
> pattern.
> 
> I must look at the BTL files…
> 
> 
> Dear Bob,
> 
> Also yes, I am being very loose with the term computer.
> But it’s inevitable that  data entry operators are hammering many
> workstations (terminals? computers? thin-clients?) and short of filming
> them (not allowed) we can only keep the overall vote tally and trust the
> system is good. Personally I have faith in the AEC, but bugs do crop up.
> 
> I will be interested (while I scrutineer) to see how much
>  they sort, bundle and count common preference patterns,
>  and how much they rely on data entry and programmed aggregation.