Monday, September 24, 2018

Poll Roundup: The Current Polls Aren't All That Meaningful

2PP Aggregate: 54.2 to Labor (-0.6 since last week) by 2016 election preferences
53.6 to Labor with One Nation preference adjustment 
Labor would win election "held now" with a large majority
(scores and text, but not graph, updated for Essential)

This week Newspoll, which has so far produced the worst readings for the Coalition since Malcolm Turnbull was replaced by Scott Morrison, came down two points on the two-party preferred vote from 56-44 to Labor to 54-46.  Taking into account the primary votes, the Coalition's gain was probably slightly greater.  Indeed this Newspoll had a slightly smaller gap (0.7 points) between the expected last-election preferences off the primaries and the published 2PP with Newspoll's adjustments of One Nation preferences than has usually been the case lately.



The limited evidence offered by a one-point 2PP improvement in last fortnight's Essential (edit: and another one this week), and also a mere 53-47 to Labor from the volatile Ipsos series last week, are also consistent with a degree of recovery from the worst of the post-spill polling.  My aggregate shows this recovery as being worth 1.9 points so far.  (Other aggregates generally do not show the worst of the blowout as being as bad as mine or the recovery as being so much - the main difference being that mine discards all previous data whenever there is a change of Prime Minister.)  Here's what it looks like on the smoothed tracking graph:


We are also seeing some personal ratings for Scott Morrison.  In the last two Newspolls his net satisfaction scores have come in at +2 and +5.  That might not sound like much, but these are the first positive netsats for a leader of either party since Turnbull's last positive rating in March 2016.  Bill Shorten has now been in minusland for over four and a half years, and slipped back into the -20s again this week, dropping from -14 (a score he last saw just after the 2016 election) to -22.

Scott Morrison is also getting handy leads as Better PM.  After trailing Shorten in his first poll as PM (33-39) he has since led 42-36 and now 45-32.  These leads are modest given that Better PM typically skews to incumbent PMs by around 16 points, but they are also large in the context of the terrible 2PP polling.  Indeed, no PM has ever led by six points while their party trailed as badly as the Coalition did last fortnight, though Julia Gillard once led Tony Abbott by five while trailing 43-57.

Moreover, Morrison's 19-point increase in Better PM lead over two consecutive Newspolls against the same opponent almost broke the all-time record of 20 points, which has occurred three times: Paul Keating against John Hewson after winning the 1993 election, John Howard against Mark Latham after winning the 2004 election, and Howard against Kim Beazley in March 2006 during Labor leadership tensions.

The horrid 2PPs continue to inspire claims that the Coalition is doomed to lose the next election badly, but the 2PP recovery combined with Morrison's personal ratings are spawning plenty of the reverse - claims that Morrison is an instant winner and that Labor will have trouble beating him.

All this poll-spinning in either direction is nonsense.  In fact, it is hard to think of a time when polls could possibly tell us less about the next election than they do at the moment.

The reason for this is that current polling is unusual for two different reasons at once.  The Coalition is both in recovery mode from a poll-crash event and also in the expected honeymoon period for a new PM.  Both these factors point in the same direction (up) from where they were four weeks ago, but we don't know how much more of that there is to go, since neither honeymoon bounces nor crash recoveries necessarily reach their full strength overnight.  On the other hand, the benefit of a honeymoon bounce for a new leader has tended not to fully last (with the exception of Keating's from an extremely low base.)

So on that basis, one might argue that both the honeymoon bounce for Morrison and the effect of putting time between the spill crash and the present could power the Coalition to a much greater recovery than we have seen so far, perhaps even to the lead and hence to victory.  One could also argue that Morrison's personal honeymoon effect is artificially inflating the Coalition's 2PP, that the damage from the spill is lasting, and that before too long we will see Morrison's ratings crash and the Coalition slump back to its immediate post-spill polling or even worse.

More likely the truth will be somewhere in between these extremes, but polling will be easier to read sense into when it stops moving than while the wheel's still in spin.

PMs Poll Best Ratings Early

Scott Morrison's early personal ratings are being taken as vindication of the leadership change, but at this stage they are not.  The reason is that almost all Prime Ministers poll their best ratings (especially net satisfaction ratings) early in their careers.  These early-career peaks have no predictive value concerning how the PM will be received later in their career.

The following table shows when PMs polled their best ratings - using Morgan for net ratings for PMs up to Hawke, and Newspoll thereafter:


Some of the data are a bit spotty here, eg Abbott may well have peaked earlier, but we don't know because it took so long for Newspoll to resurface after the election.  Also I'm not certain the rating shown for Whitlam was the highest Morgan had early in his term.  Nonetheless, all the PMs except Keating polled their best net ratings during their first six months in office, in most cases within about two months.  Keating's best rating came after unexpectedly winning the 1993 election.  The reason for all this is simple: voters tend to need evidence before they will express dissatisfaction with someone's performance, so most PMs poll very low dissatisfaction scores early in their careers.

When it comes to Better PM it's a similar story for five of the seven Prime Ministerships for which Newspoll records exist.  Keating and Howard polled their largest Better PM leads much later in their careers, but not everyone can be up against Alexander Downer and Simon Crean.  If we look specifically at Howard's record vs Kim Beazley in 1996-2001, Howard's biggest lead (42) again came after just two months.

Ipsos Issues

The Ipsos poll continues to be very problematic.  Last week's primaries were Coalition 34 Labor 31 Greens 15 others 20 (One Nation's share of Others is not yet known to me, but tends to surface eventually, so I now have a record of it for all polls in the present term except this one.)  The 31% for Labor, especially, is ridiculous - Labor's vote has no reason to be falling following recent events on the Coalition side.  The combined third-party vote (35%) is way above the mid-20s scores Essential and Newspoll are getting.

Ipsos over-predicted the Green vote at all previous elections it sampled (federal 2016, Victoria 2014 and NSW 2015).  It continues to produce obviously inflated readings for the Greens and in this case a clearly spurious primary vote for Labor.  So far this seems to cancel out, on average, on the 2PP front, so I am not tempted to drop Ipsos from my 2PP aggregate.  However, in general, if you see an article in a Fairfax publication that draws conclusions from Ipsos without mentioning any other pollster, the safer course is simply to ignore it.

Updates to follow through the week when convenient.


No comments:

Post a Comment