Liberal John Alexander defeated Labor’s Kristina Keneally in the Bennelong byelection by a 54.2-45.8 margin, a swing to Labor of 5.6 points since the 2016 election. Primary votes were 44.1% Alexander (down 6.3), 36.3% Keneally (up 7.8), 6.9% Greens (down 2.2), 4.5% for the Australian Conservatives, and 3.2% Christian Democrats (down 3.2).
Up to 16,000 postals are still to be counted, and these will further increase Alexander’s vote, probably pushing his lead out to 55-45.
The easy win for Alexander restores the Coalition’s 76 seats in the lower house, returning it to a two-seat majority (76 Coalition vs 74 for all others).
In Bennelong, Newspoll and Galaxy had Alexander respectively at 50% and 51% two-party-preferred in polls conducted in the final week, while ReachTEL gave Alexander a 53-47 lead. In this case, ReachTEL was better than Newspoll and Galaxy.
Before the byelection, I said that, given the inaccuracy of seat polls, Labor could win, or there could be a thumping Liberal victory. Unlike the Alabama Senate byelection, this time the vote of the right-wing candidate was understated.
In New England, there was a large swing to Barnaby Joyce following a Section 44 disqualification, so Labor’s consolation in Bennelong is that it received a swing that would have easily won it a general election. Nevertheless, given the polling that suggested a close contest, this is a disappointing result for Labor, and will boost Malcolm Turnbull’s standing within the Coalition.
At the 2016 election, the Christian Democrats won 6.4%, so the overall vote for the Christian right (Australian Conservatives and Christian Democrats) was 7.7%, up 1.3 points.
Bennelong voted marginally against same-sex marriage (50.2-49.8), but this result does not suggest a massive number of same-sex marriage opponents are turning to the Christian right. Alexander had supported same-sex marriage.
Queensland poll critique, preference flows, and how the LNP saved Jackie Trad
The table below shows the final three Queensland election polls, and how they compare with the election results.
Kevin Bonham estimated Labor won 51.2% of the two-party vote, virtually unchanged on 2015. A poll result within one point of the actual outcome is in bold.
ReachTEL asked for statewide One Nation support, while Newspoll and Galaxy only asked in the 61 (out of 93) seats One Nation contested. ReachTEL may have been close had One Nation contested all seats. Newspoll was very close on all primary votes, while Galaxy was a little high on the major parties, and a little low on the Greens and One Nation.
Tim Colebatch wrote in Inside Story that One Nation preferences flowed to the LNP at a 65% rate, while Greens preferences went to Labor at a 76% rate.
This data is based on the distribution of preferences, which includes preferences from other candidates in the One Nation and Greens totals. It is likely the flow from One Nation primary votes to the LNP was higher than 65%, and the flow from Greens primary votes to Labor was higher than 76%.
I believe Newspoll and Galaxy expected a One Nation flow to the LNP of about 60%, while ReachTEL used respondent-allocated preferences. The final ReachTEL poll was thus better than Newspoll or Galaxy on two-party-preferred terms. However, earlier ReachTEL polls consistently had the LNP ahead by 52-48, before the final poll became more in line with Newspoll and Galaxy.
31% of overall votes were won by parties other than the big two, but Colebatch says One Nation and Greens preferences effectively cancelled each other out.
84 of the 93 seats went to the primary vote leader. Of the other nine, Labor lost three it led on primary votes, but won four it trailed on. The LNP lost two seats to the Greens and Katter’s Australian Party that it led on primary votes.
Labor’s left-wing deputy premier, Jackie Trad, became treasurer after the election. She would almost certainly have lost her South Brisbane seat had the LNP recommended preferences to the Greens ahead of Trad.
Primary votes in South Brisbane were 36% Trad, 34% Greens, 24% LNP. Trad won 62% of LNP preferences, giving her a 53.6-46.4 win over the Greens. Had the LNP put the Greens ahead of Trad on its how-to-vote cards, rather than the reverse, the Greens would have very probably defeated Trad.
Belated Western Australian election poll critique
I was expecting a statewide two-party count in all Western Australian seats for the March 11 election, but this has not occurred.
Antony Green estimated Labor won 55.5% of the two-party vote, a swing to Labor of almost 13 points since the 2013 election. I have used this estimate in the table below.
All polls asked for One Nation support statewide, when One Nation did not contest many seats. This error led to the change in Queensland for Galaxy and Newspoll.
In WA, all polls underestimated Labor and the Greens, overstated One Nation, and had the combined Liberal and National vote about right. Labor performed better after preferences than expected.
As in Queensland, ReachTEL’s earlier polls in WA were worse for Labor, before its final poll fell into line with Newspoll and Galaxy.
Authors: Adrian Beaumont, Honorary Associate, School of Mathematics and Statistics, University of Melbourne