Tuesday, June 10, 2014

Denison Ballot Papers Damage Report Released

In the first few months of this year, this site provided extensive coverage of the Tasmanian state elections. In my Denison post-count thread I touched on an unusual incident that happened during the counting of postal ballot papers, in which the incorrect operation of a letter-opening machine damaged over 2000 ballots.  Announcements at the time said that 163 ballots (a number later reduced slightly, then increased again) were "irreparably damaged and have had to be treated as informal."

At the time there was a clear danger of this damage affecting the close contest between Madeleine Ogilvie and Julian Amos for the final seat (ultimately won by Ogilvie after two lead changes during the distribution of preferences).  In the end the margin of 331 votes was large enough that it was clear the Tasmanian Electoral Commission had dodged the 2013 WA Senate count bullet.  But this is not only clear because the final margin was of that size, but also because of some advantages of Hare-Clark over the current Senate system.


Firstly, because all preferences in Hare-Clark are controlled by the voter, the situation that arose in WA in 2013 (in which the exclusion order of very minor candidates determined two positions) does not apply.  In Hare-Clark candidates cannot get elected on preference snowballs from very tiny vote shares.

Secondly, because of the way Hare-Clark does surpluses (by last bundle rather than by the elected candidate's whole vote history), Hare-Clark doesn't have tipping points based around the exact time of a successful candidate's election.  In the 2014 WA rerun, I discovered a tipping point (first discussed in the section marked "Sunday 8 pm (6 pm): Tipping Point" in which a difference of one vote in the PUP total could potentially affect the final outcome by a few thousand votes.  This happens because of the way the Senate system calculates surpluses.  The votes that flow in to put a candidate over the line and the votes that flow out as their surplus are not the same votes.  In Hare-Clark, they are the same votes, making the system much less sensitive to errors.

(Also, it was lucky that it was the Liberal column that was affected rather than some of the others, since most Liberal votes were not only not for the candidates in question, but also would have either never reached them or done so only at reduced value.)

Although it was clear that the outcome had not been affected, I was commissioned to conduct a review of the potential of the incident to affect the outcome, in part to provide recognition of the votes that had been disenfranchised.  It's been a fascinating task.  I spent parts of about five days at the TEC office in April examining ballots and files and then more recently did a large amount of analysis and writing at home - which largely explains the scarcity of new articles here lately.

The following links have now appeared on the TEC site:


If any of the last three links are taken down in future I will rehost them here.

I found not only that the incident did not alter the outcome of the election for the division of Denison, but that it had minimal impact on the margin and no impact at all on about half of the candidates, and the only thing that might have been different in terms of the order of count events is that Elise Archer might have been elected slightly earlier.  In a simulated rerun, the difference to the eventual Ogilvie-Amos margin of all that confetti came down to a mere eight votes.  It's mathematically possible that the impact could have been as high as about 35 votes, but in practice, it's not realistically likely to have even been half that.

Even by my standards some of the work involved in assessing the impact of the incident was challenging, and I can't guarantee my reconstructions of what the election could have looked like are completely error-free. However, any mistakes not eliminated even after I went through the process for the fourth time (!) won't change the overall picture.

There is one recommendation that I wish to highlight on this site, because it's the only one that wasn't a no-brainer.  I've recommended:

"That appropriate means - legislative and/or procedural - are explored to increase the chances of damaged ballots being ruled formal in similar situations in future.  Damaged ballots should be assumed to have been originally formal, and interpreted on that basis, unless in the returning officer's view there is reason to believe they were originally informal.  Also, it should be assumed the elector has numbered boxes in sequence and without mistakes unless there is evidence otherwise".

What this means, in short, is that I reckon too many of the damaged ballot papers were ruled informal.  I had no trouble at all following the original voter intention of dozens of those ruled out (provided that I assumed they were formal to begin with) and think that about half of them would, in an ideal world, still have been counted.  Whether their non-counting resulted from shortfalls in legislation, in procedure, or both, is something I (not being a lawyer) deliberately made no judgement on.  I also made no attempt to get into the inner workings of the TEC in view of decisions made by particular people involved in producing the count - my aim was to assess the output.

But even had every discernable vote been saved, there are still dozens that would have to have been ruled informal, so there would still have been potential for a micro-close final seat result to have been affected, which could in theory have led to a full Denison by-election.  

The TEC has accepted all my recommendations and is also taking actions on its own initiative to prevent a repeat.  I'd like to thank them very much for picking me out of the thousands of Hare-Clark authorities hanging around on every street corner, but the fact is that there are precious few people outside electoral agencies who are able to conduct reviews of this kind.  All the more reason to make sure this uncharacteristic mistake never happens again.

Questions welcome.  

No comments:

Post a Comment