“The time has come”, the Walrus said, “to talk of many things” – but we’ll start with our new polling trend measures.
You may remember that last term we had two trend measures – the all pollster trend (which was an aggregation of the results of every pollster) and our phone pollster trend (which did the same thing, but just with phone polls). The all pollster trend line was consistently more pro-Labor than the phone pollster trend line because of the relatively pro-ALP lean of Morgan’s face to face polling.
This term, rather than run two separate measures, we’re just going to run one. As a result, we won’t be using Morgan’s face to face polls in our vote estimate measures at all – though we’ll still take a look at any issue based polling that might come out of them as they are still valuable.
So the pollsters we’ll this term for our trend estimates will be Newspoll, Nielsen, Galaxy, Morgan’s phone polls, Essential Report and any other phone poll that might pop up.
The trend line mechanics run much as they did last term. First we take the raw polling data for each pollster.
Second we create a new series where each observation is made up of the average of the most recent raw polling result from each pollster and weighted by sample size. So if we had 2 polls where one had a sample size of 2000 and the other had a sample size of 1000, the poll result with a sample size of 2000 would have twice as much say as the poll result which only had a sample of 1000.
Finally, we run a locally weighted polynomial regression through this new series to get our trend line.
However, the one small difference this year is our treatment of Essential Report.
As their voting intention results are already averaged over two weeks, and because they run off relatively large samples often over 2000 in size, we don’t want them to dominate the aggregated series – so we are going to give the Essential Report results a weight that is equivalent to a constant sample size of 1000.
To see how the mechanics of it all play out, we can chart the raw two party preferred poll results for the ALP since October (the red circles), the aggregated poll series (the dark blue dots) and the trend line that we run through our aggregated series (the black line).
It took a while for enough polls to come in from enough pollsters to enable us to start our aggregated series mid October. You can see how the variation of the aggregated series is much smaller than the variation in the raw poll results. The reason for this is that the aggregated series knocks out a lot of the noise of the sampling error.
When we run the trend algorithm through it, it attempts to look through even more of the residual noise.
The other thing we will do when we get a few more months of polling observations, is not only weight the aggregated series by sample size, but also by time – treating a poll that is, for example, released today with more importance than a poll that was released 2 weeks ago.
Effectively, we’ll be weighting information not only on the basis of its sampling uncertainty (sample size), but also on the basis of its temporal uncertainty (newer polling information is more valuable than older polling information).
Over on the sidebar – these Pollytrend measures have been done not only for the two party preferred, but also the primary vote results of the ALP, the Coalition and the Greens.
Also in the sidebar are the current point estimates of the trend measures – the basic “this is the best estimate of where we are today” – as well as a graphic showing the swing in each state at the two party preferred level where blue represents a swing to the Coalition and red a swing to Labor.
Those swing results for now are based on the last Newspoll quarterly breakdown. As more polling data comes in such as new Nielsen polls, we’ll combine the results of the pollsters to get our state breakdowns.
On the primary vote results, what is interesting is the absence of movement since the election:
Apart from a little bit of action in November, not a lot has happened. The Coalition are currently sitting on the same exact primary vote result they achieved last election – 43.6%.
The ALP are currently sitting on 34.9% – a 3.1% drop from the last election. The Greens are on 12.5% – a 0.8% increase from the last election.
This looks a little bit more exciting, but most of that is just a function of the axis only being 12 points high rather than the 50 points in the first chart.
Just before Christmas, the Coalition pulled close to 2 points of two party preferred into their bosom and kept it there until the last week of January, where they’ve since lost nearly a point of that Christmas gain back to the Labor party – with the Coalition currently sitting on 51.1% to the Labor Party’s 48.9%.
So we’re talking pretty small numbers here that don’t really mean a great lot at the moment.
The other interesting piece of recent polling came last week via Essential report.
Even though we have the Coalition leading on the two party preferred were a new election held now, the Labor Party are 13 points clear on whether the government should change without an election.
To put that in perspective, the voting intention results are derived after the undecided respondents – those who tell pollsters they don’t know who they’ll vote for or lean towards – are removed from the sample (it’s usually between the 4% and 8% mark). If we do the same with this result, we end up with a 59/41 split of the decided population believing the Independents should support the status quo, even though 51% of the decided population would vote for a change of government were an election held tomorrow. It’s a peculiar but insightful little result.
If Tony Abbott wants to get public support to push the Independents into switching sides, that would require just over half of the undecideds identified here to fall Tony’s way without any moving to Labor – and that’s just to get a plurality.
To get a majority, Abbott would need 3 out of 4 undecideds to move directly across to the Coalition – which is a bit of a steep ask by any yardstick since these things don’t ever happen in a vacuum.
With little public support for the Independents changing which side of politics occupies the Treasury benches, it appears that the government we have is best described as “tolerated”.
If the government is to change, it appears the population would much rather it happen at the ballot box from their own actions rather than have it done for them in the back rooms by others.