Thought bubble - ACT Senate BTL preferences favouring Zed Seselja

There has been a lot of discussion about the ACT Senate race recently. Others have recently reported that this is now a clear 1-1 split.

On current numbers, the total Liberal party group vote is 33.37%. This is effectively 64 votes above quota. However, it is likely the LIB #2 candidate got >64 BTL votes, so at least the first few stages of a preference distribution will be required. At this stage, Zed Seselja does NOT have quota by himself - his party does, but not him.



I have conducted analysis of preference flows based on the download of all BTLs from the 2010 ACT election. Just like this election, the race to 33.333 is effectively a two horse race between the 1st Liberal candidate (LIB1) and the first Green candidate (GRN1), and unlike other states the order of election hence doesn't matter. All we need to do is see, for all votes, whether GRN1 or LIB1 was placed higher. In 2010, there were only nine candidates for the ACT Senate. Oh, those were the good old days.

In the interest of simplicity, and to avoid wasting time on a hypothetical, let's make a couple of heroic assumptions:
- the BTL% in 2010 will drop from 24% to 12% in 2013
- the composition of BTL voters will remain the same in 2013 as 2010
- there is no swing in 2013 from 2010, roughly consistent with ACT HoR votes.
- the addition of different candidates in 2013 does not affect the likelihood of a BTL voter preferencing GRN1 over LIB1.

From each candidate, here is the percentage likelihood of a BTL vote preferencing LIB1 over GRN1.

ALP1	19%
ALP2	28%
DEM1	39%
DEM2	35%
GRN1	0%
GRN2	6%
LIB1	100%
LIB2	91%
UG1	48%

Surprisingly, the ALP numbers are high. Yes, a weighted average 20% of Labor BTL voters preferenced LIB1 ahead of GRN1!

Overall in 2010, of the 55,186 BTL votes, there were 16,353 for ALP1 (who won) and 38,833 for others. So applying the 2013 ALP1 Transfer Value of approx 0.032 to the ALP1 ballots gives us 39361 votes remaining.

Of these:

19004 voted GRN1 (these preferences don't get distributed)

11531 voted LIB1 (these preferences don't get distributed)

887 voted LIB2

1946 voted GRN2

5993 voted someone else.

Next we can calculate the quantum of those votes that "leaked" from the assumption that GRN1 would get 100% of non-LIB2 preferences, and LIB1 would get 100% of LIB2 preferences.

2426 votes leaked to LIB1 from others and GRN2

78 votes leaked to GRN1 from LIB2

Net result: 2348 votes leaked to LIB1, or 0.71% of the total vote

As the total BTL% is likely to be half this time, BTL leakage is equivalent to an effective "swing" of 0.36% that Zed Seselja can anticipate. While not massive, it may have been significant.

Hence any ABC Senate calculation that showed Zed Seselja's final vote as being 33.00%+ (a losing position) would actually be enough to allow him to win after a quantitatively reasonable expectation of BTL leakage.