Wednesday, August 26, 2015

Poll Roundup: Behind Longer Than Howard

2PP Aggregate: 53.1 to ALP (-0.3 in a week) 
Labor would comfortably win election held "right now"

Time for another federal polling roundup, with three polls out this week.  This week Morgan (which leans to Labor) moved from 54.5% 2PP in Labor's favour by last-election preferences to 53.5, Newspoll stayed at 54 and Essential curiously moved to 51.

This was the first time Essential has had a reading other than 52 or 53 since a cluster of 54s in January through March.  It's the first time it's made it down to 51 in one year and two weeks.  The reason is that last week's half of the sample was very poor for Labor, most likely as a result of random sample noise.  As a result of this, it's possible my aggregate is behind a few tenths of a point unkind to Labor at the moment; the picture should be clearer after next week's data.

After considering the primary votes and house effects I aggregated the Morgan at 52.7 to Labor, the Newspoll at 54.3 and Essential at 51.1.  With two polls saying things are getting better for the Coalition versus one influential one saying they are getting slightly worse, the net impact of all this is a 0.3 point recovery for the government, but one that I would not read anything into yet.  Here's the smoothed tracking graph:

Expect some varying estimates from other aggregators again this week: so far I've seen Phantom Trend at 51.8 and Mark the Ballot (which ignores Essential) at 53.5.  My own system is having an interesting time of things with the behaviour of the new Galaxy-run Newspoll, which so far seems to be quite kind to Labor.  My "global house effect" correction is now giving the Coalition 0.1 of a point to account for an apparent slight Labor lean of pollsters generally (even after snipping a point from Morgan). There's a case for adding even another 0.1 to that, but I am not doing that because I believe the strange Essential result is making Newspoll look stronger for Labor than it is.

20.5 Months Behind

The Abbott government has now trailed on smoothed 2PP aggregate for twenty and a half months in a row.  What's significant about that is that it is longer than twenty months, twenty months being the period the second Howard Government trailed for between the start of January 2000 and the end of August 2001.  Not only has the Abbott government now been continuously behind for slightly longer than Howard's second government prior to the latter's famous recovery, but it has also been behind by slightly more (I make the Coalition's average while behind 47.6 this time compared to 47.9 for Howard).  While the 2000-01 case is remembered for the spectacular recovery in the last few months prior to the election, it was only briefly during March to early May 2001 that Howard's government was polling really badly.

Only one government is now known to have trailed for longer and won, and that was the Hawke-Keating Labor government (1990-1993), which trailed for 28 months.  In that time it averaged 46.5, but more of the blame for that belonged to the Hawke part (45.8) than the Keating part (47.1).   In the pre-rich-data era, one of the winning Menzies governments (1951-1954) effectively trailed in every poll taken over 31 months of its term (averaging a feeble 45.3 in that time), but as I've only been able to find ten old Morgan Gallups taken in all that time, it's not clear it was really behind for all of that.

However, while it sounds bad for the current government that there is only either one or two precedents for recovery after trailing for so long, it's not as predictively dire as it sounds.  The reason is that while there are so few cases of governments trailing for so long and then winning, there are also very few cases of governments trailing continually for so long and then losing.  About two, in fact (Fraser 1980-83 and Gillard/Rudd 2010-3).


Not a great amount to add in this department.  Newspoll replicated Ipsos' finding of a 10-point improvement for Opposition Leader Bill Shorten, who rises to the giddy heights of a -18 netsat (35-53).  Meanwhile PM Abbott has polled another shocker at -33 (30-63, his worst since March) and Shorten has a five-point better-PM lead (40-35), his best since March.

Other Polling

Essential made a proper attempt at the same-sex marriage popular-vote-vs-parliamentary-vote question (compare the unsuitable wordings mentioned last week), and found 66% supporting a national vote to 22% a parliamentary vote.  A second question showed that there was overwhelming support (including among Liberal supporters and SSM opponents) for the timing of such a vote (if held) being either before or at the next election.  So the PM wants a people's vote to be held at a time when the people don't want it.  Essential also found 60-31 support for same-sex marriage, a result very similar to its others in the last few years but with a lower don't-know rate than usual.

On the Trade Union Royal Commission, 39% thought it was a "legitimate investigation into union practices" while 27% thought it was a "political attack on Labor and the unions".  This question had a degree of partisan splitting which was strongest among Liberal supporters (70-6).  However 38% to 25%, including all non-Coalition voter groups, believed Commissioner Dyson Heydon has a conflict of interest and should step down.

Meanwhile a question on carbon emissions targets might have got a less polarised result had the question not specified which policy was the Government's.

An article in The Australian entitled "Greens to fight on two fronts" saw the Greens raiding inner-city Sydney ALP seats if preferenced by the Liberals based on state election results (always a dubious comparison), but running from raids on their vote by the Nick Xenophon Team.  The support for the last was a ReachTEL conducted across "six federal seats" in April, and various figures are given for percentages of voters who say they would consider voting for NXT.  These figures are given as 54% in Sturt (SA) and between 14.6 and 27.6% in four NSW and one Queensland seat.

The article points out that if realised these vote shares would see several NXT Senators win, but it's irrelevant, since they won't be.  When a voter says they would consider voting for a party, it frequently doesn't mean they'll vote for that party.  It most likely means they've heard of the party and don't totally hate it.  Many voters will consider voting for several parties at a given election.  Some voters would consider voting for anyone.  Really this kind of question has very little benefit and has been no use at all in predicting minor party vote shares in the past. This is not to say NXT will be a non-issue beyond South Australia, just that this particular poll proves nothing about the sort of support level they might expect.

Coverage of Canning continues in a separate post.

No comments:

Post a Comment