Why Did One Nation Win 11 Seats In Queensland 1998 But None In 1998 Federal?

The One Nation party (which I am on the verge of restyling Wuss Nation after its sobbing about its poor supporters being traumatised by filling out ballot papers) has been attacking compulsory preferential voting in the leadup to the South Australian election.

Linked into this I came across a narrative from Pauline Hanson which I thought deserved detailed examination.  Interviewed on Sky (and yet again, where else) Hanson told the tale of how in 1998 her party won eleven seats from scratch in the 1998 Queensland election and noted that it was optional preferential voting.  Then she moved on to the 1998 federal election where although her party won over a million votes, all the other parties recommended preferences against One Nation and they didn't win any seats.  Famously she lost Blair where the major parties cross-recommended against her.  

She doesn't in this excerpt mention what became of those eleven Queensland seats.  Every single one of those MPs quit the party or the parliament by the end of the 1998-2001 term, though One Nation did retain two of those seats and win a third with different candidates.    But my interest here is, is there really any causal link between the current OPV/CPV debate and what happened in those two elections?  Or are the explanations different?  A warning that this article is very numbery and has been graded Wonk Factor 4/5.

Preference flows in Queensland 1998

In Queensland 1998 One Nation won eleven seats and made the final two but lost in 22 others.  The Liberal and National parties recommended preferences to One Nation and Labor recommended preferences against them.  In general, One Nation therefore gained on preferences in seats where their final-two opponent was Labor and lost ground on preferences where Labor was eliminated.

Had the election had compulsory preferencing, the preference flows would have been stronger.  Seats that might have had different results under compulsory preferencing, therefore, are seats where Labor was eliminated and One Nation won narrowly, and seats where Labor was One Nation's final two opponent and One Nation were competitive but lost.

There were three seats that finished as One Nation vs National with One Nation winning - Barambah, Lockyer and Tablelands.  The standard method for estimating a result under CPV for these cases is to assume that the votes that were exhausted would flow at the same rate as the votes that flowed as preferences.  On that basis One Nation would still have won Barambah and (by less than 1%) Lockyer but would have very clearly lost in Tablelands.

Fourteen seats finished Labor vs One Nation with Labor winning, but most of these were safe Labor seats where Labor won on primaries or nearly so and would have clearly won in any system.  One interesting exception is Bundaberg, which Labor won 52-48.  My compulsory preferences estimate closes this to 50.4-49.6.  The other vaguely close ones are Cairns (52.3 in to 51.7) and Kallangur (54.0 to 51.9).

So on the surface, with compulsory preferencing One Nation would have won ten seats in Queensland 1998 and probably not got eleven, which is hardly a big difference.  But even this is missing something very important.

That is that One Nation were only able to win as many seats under OPV as they did because the Coalition and One Nation cross-recommended preferences in (at least tacitly) exactly the kind of preference deal that One Nation 2026 model is now whinging about.  One Nation used Coalition preferences to overtake and beat Labor in Caboolture, Hervey Bay, Ipswich West, Mulgrave, Thuringowa and Whitsunday.  Had the Coalition parties run an open (just vote 1) how to vote card in these seats and had that caused even a quarter of their voters to switch from preferencing One Nation to exhaust, One Nation would have not won any of these seats - and the impact of such a card would most likely have been higher.  

The strategy of recommending preferences to One Nation dogged the Borbidge government on the campaign trail and contributed to its loss.  While the Coalition has recently been recommending preferences to One Nation in CPV elections as the political costs of doing so have reduced, would it really do so in an OPV election today knowing that it would cop flak for giving preferences to One Nation when it doesn't even need to?  And knowing that One Nation voters follow cards at a far lower rate than Coalition voters so it wouldn't even get very much back?  

1998 federal

Given that One Nation won 12.4% of seats in the 1998 Queensland election, why did it not win even a few seats at the 1998 federal election?  All and sundry recommending preferences against it didn't help, but there was a much simpler reason - its primary vote in Queensland crashed.

One Nation polled 22.7% in the 1998 Queensland election, which would have been in the high 23s had it contested every seat.  In 1998 Queensland federal, One Nation contested every seat and polled 14.3%.

As a consequence of this, One Nation only made the final two in one Queensland seat (Blair), which it lost, to be discussed further below.  But had One Nation polled as strongly as in the Queensland state election then even by uniform swing it would have won Blair very easily, and made a bunch of final-twos including leading on primaries in Wide Bay.  On the uniform swing model it still would have only won Blair because of unfriendly preferences.  But the pattern is that when One Nation has a good election, it surges more in its good seats than its bad ones.  Based on the modelling I am doing for South Australia, which is in turn largely derived from the vote spread in Queensland 1998, it's very likely that had One Nation matched their state vote in the 1998 federal election, they would have polled something like 40.8% in Wide Bay, which probably would have just got them over the line (not by very much), while Hanson would have won Blair on first preferences.

So even with compulsory preferences running against them, had One Nation held their state election vote they may well have won two federal seats.

What about if they had held their state vote and the system had been OPV and the Coalition had run "just vote 1" instead of preferencing against them?  In that case they would probably have gained one more, Hinkler - where on my model the three parties would have finished up pretty level with the Nationals most likely excluded in third.  In almost all the other candidate cases Labor would be eliminated in third which would mean One Nation loses under either OPV or CPV unless they have a hefty primary lead.  

Blair 1998 - Would OPV Have Saved Hanson? Dubious!

It's been generally received wisdom that Pauline Hanson probably would have won Blair in 1998 if only preferences were optional.  Eyeballing it from the perspective of later OPV preference flows this looks obvious.  There's a simple argument that Hanson would have won - the winner needed a 66.9% preference flow and got 74.9%.  Assuming the same flow of non-exhausting votes then an exhaust rate of anything over 32.1% would cause Hanson to win, and with what we know about exhaust and OPV it would surely have been higher?  But there are two things wrong with this.  Firstly 32.1% is not greatly different to the exhaust rates in Queensland 1998 in the days before "just vote 1".  Secondly there is one candidate in Blair 1998 who can't be modelled with anything near that flow: the National.

In Blair 1998 Pauline Hanson polled 35.97%, Labor 25.29%, the Liberals' Cameron Thompson 21.69% and the Nationals 10.25%.  Hanson's large primary vote lead here is an illusion, because the Coalition had two candidates.  By the three-candidate stage it was Hanson (ON) 38.91 Thompson (Lib) 31.78 Clarke (ALP) 29.30.  73.8% of preferences from the Clarke exclusion flowed to Thompson and Thompson won 53.4-46.6.

To look at what might have happened assuming OPV and that Labor recommended preferences against One Nation (as they did in the state election where it worked fine for them so they would have done it again) I can look at the 1998 Queensland election for cases where the final two was Coalition vs One Nation and the One Nation primary was similarly large.  There were twelve of these, conveniently and spookily with an average ON primary of ... 35.97%.

Less conveniently in Barambah the ALP candidate wasn't thrown, so I've kicked out Maroochydore (with the weakest One Nation primary and as it happens a very strong preference flow from Labor to National) from the sample to balance that out.  

On that basis, the average Labor 3CP exclusion flow in the remaining ten seats was 45.9% National, 18.7% One Nation, 35.4% exhaust.  The 71.1% flow of those preferences that did flow is pretty similar to the actual 73.8% flow in Blair.

If this OPV preference flow is applied to the 3CP point in the federal election, Hanson loses 49.53-50.47.  However, the primary vote numbers at the 3CP point wouldn't have been quite the same, because of exhaust at earlier stages of the count.  And the crucial difference with the state counts here is that the federal contest had both a Liberal and a National running.  If the federal seat was OPV the National wouldn't have run.  The National's voters would generally have voted so we can treat their preference flow as essentially zero exhaust, barring the preferences they received from other parties.  For other minor candidates, in Queensland state 1998 the exhaust rate without reaching the top three in such seats with multiple minor candidates was about 35%.

Bumping the minor candidate exhaust rate up to 40% because of the larger number of minor candidates but applying no exhaust to the Nationals primary votes, I estimate the 3CP point under OPV in this seat would have been Hanson 39.40 Thompson 31.58 Clarke 29.02.  And Hanson still loses in my model ... by 49 votes.

That's way too close for a reliable conclusion and any number of things could be out in such a model.  The exhaust rate off minor candidates might be higher (there were six of them), some Nats voters might have not bothered voting, and so on.  On the other hand, applying a small exhaust rate off the Nats preference flow for those preferences that reached them from elsewhere ignores the fact that the 1 Nats votes were probably stronger for the Liberals.  And it may well be that Thompson as a Liberal would have got a better Labor preference flow than a generic National did in Queensland state 1998.  Indeed the evidence on both the 2CP and 3CP flows suggests rather strongly that he did, so I think that he would have won under OPV anyway.  Him winning also follows from the observations about overall exhaust flow and the impact of the Nationals candidate from that at the top of this section.

In any case the whole idea that Hanson wuz robbed by compulsory preferencing combined with a major party gangup is actually much shakier than I previously thought that it was. 

In summary:

1. All but one of One Nation's wins in Queensland 1998 (state) had nothing to do with the preferencing system.

2. Several of One Nation's wins under OPV in the state election only happened because the Coalition recommended preferences to them, and under a just-vote-1 strategy informed by this election these wins wouldn't have happened.

3. One Nation would have won at least one seat with a good chance of two in the 1998 federal election had their vote been as high as the state election.

4. Even without compulsory preferences, it's not clear Pauline Hanson would have won Blair anyway.