Firstly, the one place I definitely was not wrong was that preference deals were going to get Family First a Victorian senate seat despite receiving less than two per cent of the primary vote.
The first and most important place I was wrong was underestimating the strength of the swing to Liberals in the senate from 39.61% in 2001 to 44.05% on current figures in 2004. While I had allowed a 1.4% swing to them in my predictions below, I still don't see where a swing three times that size could have come from, especially given the smaller swing in the Reps.
So the Liberals get the fifth seat without needing to go to preferences and even have 1.2% left over to contribute to the battle between the ALP and the minors for the last place.
I was also wrong about the donkey vote. Liberals for Forests managed to carry an almost 2% vote down to column C while Citizens Electoral Council stayed a minnow even in column A, so the Greens did not benefit as I had expected from CEC preferences.
Where I was close enough to right was that there were four tickets to watch below the Greens, each with just under 2%. I had been a bit kind to the Democrats suggesting they might hold onto 4%, but their collapse put them down on level terms with LFF, the DLP and Family First whose actual first preference votes were still less than half the 4% needed for them to get public campaign funding.
But the flow of preferences was irresistible. The minnows, One Nation and the Liberal surplus pushed Family First clearly ahead of the other sub two percenters, whose preferences then all went to them too, leaving three candidates at the next to final stage:
(based on Sunday afternoon figures and minor simplifying assumptions). Of course Labor did a deal with Family First which would have given it 1.9% of the vote if Labor had been one of the top two at this stage. But they aren't. So at least 98.5% of the Labor residue vote is sure to follow their below the line preference deal and see Family First's Steve Fielding get in courtesy of the 8.5% that voters had intended to help reelect Jacinta Collins.
One other place I may have been wrong, but which turned out not to matter this time around, was my interpretation of how votes are distributed once you get beyond a quota. I had assumed, maybe wrongly, that the proportional allocation of surplus votes would only apply to the last batch of preferences which took a candidate over quota. But I now suspect, although I still have not seen it spelt out unambiguously, that all the votes in the quota and a bit flow on proportionately, irrespective of whether the candidate who received the quota was the first or the third to last preference of the voter.
The Age writer Tim Colebach provides even more detail. (You may need to register to see his report.) I'm leaving my original predictive analysis unaltered below for the record, at least aside from some belated spelling corrections.
Since the idea of placing a single vote "above the line" on the senate ballot seduced 98.5% of major party voters and 91% of other voters (2001 Victorian figures), what the parties (and other grouped candidates) might be doing below the line has become more and more of a mystery, especially given the appalling bandwidth the Australian Electoral Commission provide to anyone trying to download the 2MB PDF of the parties' tickets (mirror).
To try to make some sense of way too much data, I grant that the 2 major party tickets will each get in excess of the 28.57% of the formal vote that they need to each win 2 of the 6 places and focus my attention on their respective 3rd candidates plus the most credible of the minor parties. In the case of the 2004 Victorian senate election that boils down to the Greens, the Democrats and, after taking into account the general shape of preference allocations, a newcomer in Family First who have managed to make themselves a lot of other parties' second (significant) preference.
Of course Family First does have to get ahead of enough of the minors before those preferences become of any use to it, but given that in 2001 the donkey vote, the DLP and One Nation evenly split 7% of the vote and an expectation that Family First are the most likely heirs to much of the One Nation rump in Victoria, it becomes possible that they start from a very small first preference base and accumulate in turn the minnows, whatever is left of One Nation and the DLP to sneak ahead of the Democrats.
If that happens, with Democrats, Labor and Liberals all preferencing them, Family First would find themselves with the 5th seat before any preferences get a chance to flow in other directions.
One important thing to remember in this discussion is that there is no preference drift from that vast majority of votes recorded above the line. They all follow exactly the allocation set down by the party or group.
In 2001 the major parties shared 76.4% of the formal vote, the Greens and Democrats shared 13.8%, leaving all others and the donkey vote to share 9.8%.
Taking on board the current state of the polls, we could conservatively estimate 2004 percentages as:
Allocating the minnows evenly here could be a slightly problematic assumption, but it is hard to get an accurate line on the prospects of individual parties whose very identities often chop and change between elections.
The donkey vote has moved form "Liberals for Forests" to "Citizens Electoral Council", both seeming minnow candidates in 2001 and 2004 and we can safely assume it will be confirmed at around 2% and this time flow to the Greens who, if they actually get the 12% that opinion polls are forecasting (or better) would then certainly grab the 5th seat without needing to use any Democrats or major party preferences. And if the Greens do grab the 5th seat through gaining first preferences from the majors and/or the Democrats (relative to the figures above) that just makes the following argument for Family First grabbing the other place even stronger.
From the first preferences above, remove 2 quotas each from the majors (a total of 57%) then the 12 "others" go 6 to Family First, 4 to Greens and 1 each to Labor and the Democrats, after which the donkey vote goes to the Greens, and the DLP goes to Family first, leaving my 5 candidates of interest at:
The Democrats then goes to Family First taking it past Labor after which Labor then goes to Family First until it has a quota, leaving for the final race:
Labor then goes to the Greens.
Given all that, my real question is how to make my vote as effective as possible. If enough others are going to vote for my most preferred of the 5 to get them over the line, then I would like my vote to have a chance of helping my 2nd choice, or my 3rd or even my 4th. Clearly this requires me to vote below the line, which is the whole reason I looked at all this information in the first place.
The first step is to make sure that the right party (or group) gets my $1.95 in public funding, but I can do that by giving my first preference to an unelectable candidate from that party. I then try to make sure that my vote is only counted for my first choice of the 5 if it is needed to get them over the line for the 6th seat. In theory I might be able to just vote 61-65 for the 5 in my chosen order and randomise votes 2-60 knowing they won't count, but I might err just a little on the side on conservatism and put the top 2 from the 2 majors at 62-65 to make sure there is no way my vote can be counted with the sheep. I will probably also be careful with one or three others who I have assumed cannot win, but who I could well be wrong about.
Having done all this research I have found several contenders for any award for the most disingenuous allocation of preferences:
Family First itself preferences Labor ahead of Democrats and Liberal, so if it were to fail to overhaul the Democrats in early redistribution Labor would be a lot closer to bridging any gap in first preferences between it and Liberal, with the above mentioned assistance of the DLP, but not of One Nation.
 Compounding the bandwidth problem the AEC site manages to convince both my browsers that what are in fact much larger PDFs are actually 865K. When the 865K has been downloaded Mozilla decides to stop and that the file in invalid. Fortunately Safari just kept going to the point but it too failed after it had downloaded "5.1 MB of 865 KB" of a second AEC PDF which I had hoped would provide more details of how preferences are actually allocated than are readily found on the AEC web site.
 Yes I know it could have been easier to just donate them $2 and spend the time on something more profitable, but it is the principle. Now I just have to decide how widely to reveal the existence of this web page. If you do find it and feel pressed to respond, a post in the announcement thread at TransForum would be welcome.