Making sense of the polls

Making sense of the polls

Final Victorian Lower House Results

At the Victorian state election held on 29 November, Labor won 47 of the 88 lower house seats, a comfortable 6 seat majority. The Coalition won 38 seats (Liberals 30, Nationals 8), the Greens won 2 seats and an Independent won Shepparton from the Nationals. Compared with the Coalition’s 48-40 post-redistribution notional majority, Labor gained 8 seats from the Coalition, but lost Melbourne to the Greens, with the Greens also gaining Prahran from the Liberals.

The final primary votes were 42.0% for the Coalition (down 2.8% from the 2010 election), 38.1% for Labor (up 1.8%) and 11.5% for the Greens (up 0.3%). Antony Green has tweeted that the final statewide two party preferred (2PP) result is a Labor win of 52.0-48.0, a 3.6% swing to Labor from the 2010 election. In Labor vs Coalition two party terms, Labor won 48 seats to 40. Although the Greens won Prahran, Labor lost it vs the Liberals by a narrow 25 votes or 0.03%. Turnout for the election was 93.0%.

Four of Labor’s gains (Bellarine, Monbulk, Wendouree and Yan Yean) came in seats that Labor already held, but which had been redistributed into notional Coalition seats. Labor made a clean sweep of the sandbelt marginals, gaining Bentleigh, Carrum, Frankston and Mordialloc.

Antony Green has published a post-election pendulum. This shows that Labor won Frankston, Carrum and Bentleigh by slender margins between 0.5% and 0.8%. However, Labor could have afforded to lose all three, and there would still have been a Labor/Greens majority (44 Labor + 1 Green). The clinch seat for a Labor/Greens majority was Mordialloc, which Labor won by 2.1%. This means that, on this election result, Labor actually required only 49.9% of the statewide 2PP to win a majority with the Greens.

This actual requirement for a Labor majority is less than the 50.5% estimated by both Kevin Bonham and Peter Brent, but it is greater than the 49.3% if you use the raw pre-election pendulum. The major reason for why Labor needed a little less than what was estimated is that there were some small swings to the Coalition, but most of these swings happened in safe Labor seats; none occurred in Labor or Coalition marginals.

The Greens only increased their statewide vote by 0.3%, but they performed very well in the inner city, increasing their vote by 9.5% in Brunswick, 8.9% in Melbourne, 4.4% in Northcote, 5.0% in Prahran and 2.9% in Richmond. The Greens would very probably have won Brunswick, Northcote and Richmond if the Liberals had preferenced them ahead of Labor in these seats. Despite the relatively low Labor prijmary vote in the four Labor vs Greens seats, Labor’s 2PP vs the Liberals is in the 70’s in all four seats; in Northcote and Brunswick, Labor had over 79% of the 2PP vs the Liberals, more than in their traditional heartland seats of Broadmeadows and Thomastown.

With all upper house primary vote counting completed, two seats remain in doubt: either Labor or the Country Alliance will win the last seat in North Victoria, and either the Sex Party or the Greens will win the last seat in South East Metro. Vote 1 Local Jobs has a clear path to victory in West Victoria. I will post an article next weekend on the final upper house results.

Poll Performance

Below is a table comparing the final polls’ primary votes and Labor 2PP estimates against the actual election results. A poll estimate that was within 1% of the actual result is bolded.

final Vic polls vs elec.

The clear loser was the final Morgan SMS poll, which was overly favourable to the Coalition. Morgan’s previous Victorian SMS polls had the Greens on an unrealistically high 18%, so I think his methodology needs work. In the end, all polls overestimated the Greens vote, with Newspoll erring the least.

Other than Morgan, other polls were right on the 2PP, but for the wrong reasons. Newspoll, Galaxy and ReachTEL all used 2010 preference flows to estimate the 2PP from their primaries. The actual preference flows were more favourable for Labor than in 2010; Labor would only have won 50.8% 2PP on the actual primaries under 2010 preferences. So these polls overestimated Labor’s primary vote performance, but their 2PP looks good because of changes in preference flows.

Ipsos was accurate on the Coalition primary, but missed on the Labor and Greens primaries by over 3%. Their headline 2PP figure used respondent allocation of preferences, and that is fortunate for them, as their previous election figure was a 50-50 tie.

Overall, there was no clear winner from the final Victorian polls, but Morgan is a clear loser.

Individual seat polls for Bentleigh and Buninyong by Galaxy taken over two weeks before the election were too pro-Coalition. The Liberals led by 52-48 in Bentleigh, compared with a Labor win of 50.8-49.2. In Buninyong, Labor led by 54-46, but actually won by 56.4-43.6. A Galaxy exit poll of the early voters in the four sandbelt marginals that had Labor leading 52-48 was inaccurate; the Liberals won the early vote in all four sandbelt marginals.

Labor Wins Fisher By-Election in SA

Labor has won the Fisher by-election, defeating the Liberals by 23 votes after eliminating Independent Dan Woodyatt by 226 votes at the point where either Labor or Woodyatt was excluded. The seat had been previously held by deceased Independent Bob Such. The final postals and rechecking increased Labor’s lead by a net two votes from what they held last Wednesday. Final primary votes were 36.0% for the Liberals (up 0.9), 26.7% for Labor (up 9.0) and 23.3% for Woodyatt (down 15.2% on Such’s performance). Labor won with a 7.3% two party swing in its favour.

The win returns Labor to majority government status in SA, after they had lost their majority at the March election. Kevin Bonham has researched state by-elections thoroughly, and he says that state governments beyond their first terms have not won a by-election from another party or Independent since 1973.