To Marginals and Beyond!

In the Pantheon of Polls, Galaxy has consistently been the most favourable to the Coalitions fortunes, showing them to be two to three points better off in their primary and TPP vote than the other polling outfits. Subsequently, Galaxy became the poll de jour for Coalition members, supporters and assorted hangers on – their last refuge in the storm of polling annihilation.

Yet, todays Galaxy poll tears away their last remaining veil of delusion. It has the Coalition primary down 2 on 39, the ALP primary up 3 on 47 for a TPP split of 57/43 to the ALP.

This puts the Galaxy results right between the last Morgan, Newspoll and ACNielson polls -perhaps it’s the start of a larger polling convergence, but only time will tell.

One thing though has now become undeniable; these polls now all reflect the type of Pollycide that Newspolls last quarterly breakdown suggested, with maybe only a slight reduction in its magnitude since the last quarter. Large numbers of seats are looking to change hands – the only remaining question is “where?”

Today we’ll look at the nature of marginal vs safe seat voting movement over the last few years.

To start off, we’ll have a squiz at the marginal seat movement. The most regular marginal seat polling results are done by Newspoll, however there is not a complete series, with large gaps in the polling over the electoral cycle, particularly between elections. So to account for this, I used the actual Newspoll polling in marginal seats, and modelled the missing parts using data from Newspoll, ACN and Morgan. For each series, I simply specced out the best fitting, forecasting regression model (some were complex, some were quite simple) using all of the available data I have (my polling database currently consists of about 55000 datapoints from over 100 series of raw, aggregated and derivated data).I’m pretty confident that the modelling that fills the gaps between the actual Newspoll marginal seat polling is only a point or two off the actual results – but it gives us a useful longer term marginal seat pattern to work with.

Below are the results graphed quarterly – note that “marginal” is defined by Newspoll as being any seat held with less than a 6 point two party preferred margin, which is what I also used in some of the modelling.

 

 

The last entry is for the 2nd quarter of 2007.I expect that the primary vote for the Opposition has dropped from its 51 to somewhere around the 47 mark at the moment, while the government primary vote is currently up somewhere around the 38/39 mark…… well that’s what the modeling suggests anyway. Let’s also take a look at the TPP breakdown of the marginal seats according to Newspoll. I didn’t fill in the gaps in this series because there is too much uncertainty to accurately model the TPP results between the gaps:

Over the history of the series, it’s interesting to see the effect of Latham (what will be a recurring theme in this post). When he was elected to the party leadership, he dragged the ALP primary up 6 points in the marginals to put them in a position to dominate the marginal seats at the beginning of 2004.But a few interesting things happened after that. Firstly, while he continued to increase the ALP primary into the 2^nd quarter of 2004, it came at the expense of the minor party vote rather than the Coalition primary vote. In fact, both major parties enjoyed primary vote growth in the 2^nd quarter of 2004.That should have been a warning sign to the ALP at the time, but apparently it wasn’t.

By the end of the Latham implosion, the Coalition had picked up a 2.5% primary vote swing to them in the marginal seats compared to the 2001 election results and a 0.7% TPP preferred swing. The swing in the marginals was extremely non-uniform in both numbers and composition. You might also notice that the Latham implosion didn’t fully wash out of the system until after the 2005 Budget, but instead of marginal seat voters changing their vote from the Coalition to the ALP, they changed from the Coalition to the minor parties – a long running theme in all of the Newspoll data.

The end result of the 2004 election was to turn 14 marginal seats into safe Coalition seats by pushing them through the 6% safety margin.

These seats were:

NSW: Paterson, Cowper

WA: Kalgoorlie, Canning, Moore

Vic: McEwen, Gippsland, Dunkley

QLD: Herbert, Longman, Petrie, Hinkler, Dickson, Bowman.

Note the number of Qld seats there, they’ll become extremely important in explaining a few things a little further down.

The blowout of the ALP vote in the marginal seats suggests that even if a non-uniform swing should play out in the marginals at the impending election, the size of the overall swing will lead to nearly all of the Coalition held marginals changing hands – Bennelong and Wentworth probably being the safest two Coalition marginals outside of WA because of the local issues involved, with Stirling and Hasluck in WA being the next least likely to fall because of the smaller swing in WA that seems to be happening (that said, I think the WA twins will fall anyway, as will Bennelong – but not Wentworth).

But while the marginal seats are interesting – let’s get to the really interesting movements; the safe government seats, those government seats held by a TPP buffer of more than 6%. Again, the Newspoll quarterly data left gaps to fill, so I modelled the safe government seat vote over the last few years in the same way as I did for the marginals.

And we’ll throw in the Newspoll TPP safe government seat vote while we’re at it:

 

A few things stand out here. The long slow decline in the Coalition vote between the last quarter of 2002 through to the third quarter of 2004 being one. Through that period, the Coalition lost 5 points from their primary vote and somewhere around 6 to 7 points from their TPP vote. When Latham imploded, that blew the Coalition out from 47 to 55.4 in their primary vote, and from 54 to 62.6 in their TPP. If Latham had kept momentum, or at least kept the level he was at before the implosion, over a dozen Coalition safe seats would have become marginal and a handful would probably have fallen. However, as a result of the implosion, barely safe seats like Pearce (6.9% margin) blew out to a over a 13% margin, Leichardt (6.4%) to 10.2%, Macarthur (7%) to 11% and things like Parkes (8.7%) out to over 17%.

Yet since then, there has been what can only be described as dramatic movement in the Coalition safe seats back to the ALP, to the point where the ALP is now ahead in these seats in both primary vote and TPP terms. That has probably slipped back slightly in the Coalitions favour since the 2^nd quarter of 2007, but only just. However what is really important here are those ex-marginal seats that Latham pushed back into the Coalition safe seat column, especially in Qld.

The swings in the safe government seats are simply not going to be uniform. Seats like Murray, Riverina and O’Conner with margins of 20% or more are safe Coalition territory. At the last election their margins changed from 21.9->24.08, 19.9->20.85 and 19.1->20.39 respectively. They aren’t moving like the 14.6% average safe government seat swing away from the Coalition since the last election. Likewise, seats like Lyne and Berowra would fall with such a uniform swing, but who can honestly say the denizens of the Manning Valley would be turfing out Mark Vaile, or that the folks of Galston and Wisemans Ferry will be kicking out Ruddock?

For these swings to be averaging 14.6% in the safe government seats, and with there being a fairly large number of seats where a swing of that size is demographically bordering on the impossible, that means that there must be a fair number of seats taking up the slack with swings approaching the 18-20% mark or more. This is where Coalition members should be completely shitting themselves, and its where those marginal seats that Latham pushed into the safe seat column are exerting their influence.

But even if they are reverting back to their pre-Latham levels as a starting point, and then swinging like the marginal seats they were before Latham – those swings simply don’t add up to the size of the swing required to balance out the Newspoll polling in conjunction with the assumption that many of those ultra-solid Coalition seats wont swing anywhere near the average.

What would make those swings balance out though is if the Qld seats are swinging, but the marginals are swinging a lot less than the safe government seats. For instance, a marginal seat swing in Qld of only 6 to 7%, but a safe government seat swing of 16-18% would start to balance the numbers out. And there is a decent narrative to support such an argument in Qld. The anti-Latham factor in Qld was big – it pushed a lot of people away from the ALP more so than just about anywhere else. The margin on nearly all government seats be they marginal or safe is significantly inflated by that Latham factor. Add to that the fact that nearly all of the coastal, non Brisbane safe seats have undergone serious demographic and economic change over the last 3 years and Qld has a long standing habit of swinging large when the swing is on, particularly in the regional seats and such an argument starts to look a serious possibility

The other key player influencing the safe government seat swing is NSW.

The last Newspoll suggested that NSW had the largest average swing of all states at 12.2%. In the process of speccing out some of my models, I found myself playing around with all sorts of data. What caught my eye was the results of regressing the estimated safe government and marginal seat series against the various primary votes of the states by quarter, as well as the reverse, regressing the various state primaries on the estimated safe government and marginal seat votes.

What the results started to hint at, and it really is only a hint as there wasn’t enough observations for me to be as confident I as usually like to be with these things, was that NSW and Qld safe government seats were taking most of the weight of explanatory power in any safe government seat analysis and had been doing so since 2000.Likewise, SA, Vic and WA movements were mostly in the marginal rather than the safe seats, SA particularly so – which is bad luck to the large number of commenters on this site that seem to have a long standing dream of Downer losing his seat – bad luck with that guys 😉

I can come up with an arguable explanation of how Qld could be shouldering a large amount of the weight of the swing against the government in their safe seats, but I cannot come up with how it seems to be playing out in NSW.

The data suggests NSW is swinging big in the safe government seats, all of the data points to that and it has to be happening for the Newspoll figures to even begin to balance out (even accounting for Qld).

If any one reading this has any idea at all about how the dynamics of the big swings in NSW safe government seats are working, or where they’re likely to be happening, or how the Latham effect might have played out in seats, or how other issues like housing and whatnot could have some role in explaining the hows and why’s of the big NSW safe government seat swings that must be there- I’d love to hear from you.

Because at the moment – I’m baffled.

On a slightly related issue, since the election is approaching I suppose its time to start putting up some predictions.

Over the last few weeks, especially after churning over the ever increasing amount of data coming out of not only the polling organisations, but the last ABS census – I’m starting to see a few things that I think are becoming pretty solid relationships. Being a Qld’er, I have a better sense of how the data relationships are playing out on the ground in Qld than I do in the other states.

So, at this stage, I’m becoming pretty confident that these 5 safe government seats in Qld are gone:
Herbert, Longman, Petrie, Hinkler, Bowman.

These 2 are probably gone:
Ryan and McPherson (big call on McPherson – I know)

These 6 will have their margins slashed, and one or two may even go down to the wire:
Fisher, Dickson, Forde, Leichardt, Widebay and Fairfax (watch Fairfax)

And the three marginals of Blair, Moreton and Bonner are gone.

That’s 8-10 seats gone in Qld, with maybe 1 or 2 others if it gets really nasty for the Coalition.