Look Ma, No Possum!

Well not quite, but I’m heading out to the boonies for a holiday so there will be less Possum over the next 2-3 weeks, depending on the joys of rural internet access.There will be updates once or twice a week.

Apologies for the slack comment responses over the last few days, I’ve been flat out but I’ll get to them in big wads over the holiday.

Pollycide Part 3 – The Verification.

Continuing on from Pollycide 1 where we used various weights derived from the quarterly Newspoll data to give us some experimental 2 party preferred results to show us the type of seats that would have likely changed hands in a July election, and following on from Pollycide 2 where we then used some linear programming to fit that seat pattern into the actual swings observed in the quarterly Newspoll – we now take it all back to basics, because sometimes the simple things in life are… well, simple and effective.

I noticed some dubious reaction to how these seat swings are possible (well, I noticed them once I managed to get past all the ALP supporters throwing confetti everywhere), so I thought we’d get back to the basics to show that I’m not making this up.

To do this, we’ll use the Newspoll quarterly data from March-June 2007, and compare that to the actual 2004 Election results to give us the TPP swings for each state. We can then compare these swings using the State Pendulums and we’ll get a basic, but realistic assessment of the seats that would have been likely to change hands were an election held in July.

To start off with, let’s take a squiz at the 2004 Election TPP results and the Q2 Newspoll data to find the state swings:

2004 Election	TOTAL	NSW	VIC	QLD	SA	WA
Coalition	52.8	51.2	51	57.1	54.4	55.4
ALP	47.2	48.8	49	42.9	45.6	44.6
2007Q2 Newspoll
Coalition	43	39	42	46	44	50
ALP	57	61	58	54	56	50
ALP Swing	9.8	12.2	9	11.1	10.4	5.4
Coalition Swing	-9.8	-12.2	-9	-11.1	-10.4	-5.4

That entry “ALP Swing” is the one you want to pay attention to here.

Now comparing these swings to these graphs we get the following numbers of seats changing hand for each state:

NSW: ALP gain of 17

Vic: ALP gain of 8

Qld: ALP gain of 14

SA: ALP gain of 5

WA: ALP gain of 2.

That’s a total gain of 46 seats.

However, lets look at Tassie and the NT where Solomon would fall with a 2.9% swing, Bass would fall with a 2.7% swing and Braddon would fall with a 1.2% swing.

With a national swing of 9.8% – it’s a safe bet that those three seats would all fall as well, giving us a total ALP gain of 49 seats.

The Pollycide 2 numbers and the seat list suggested 47 seats would have been lost in NSW, Vic, Qld, SA and WA. Remember, that seat list was derived from Pollycide 1, and adjusted to fit within the State based swings – although it was more luck than skill that the numbers matched up just about perfectly as it could have been 3 or 4 seats in either direction due to the nature of the maths and where I chose to draw the line in terms of leaving some unaccounted for swing in NSW.

So, from Pollycide Part 1 we can see the types of seats that would have been lost in a July election, from Pollycide Part 2 can see a good estimation of which seats and the margins involved were an election held in July, and with Pollycide 3 we have verified those numbers.

That’s the good news for the government. Yes, the GOOD news.

The bad news is that the above verification based on the state Newspoll swing estimations doesn’t account for the composition of marginal and safe seats. The ALP are only swinging 4.1% toward them in their safe seats, the marginals are swinging 9.3% toward the ALP and the safe government seats are swinging 14.6% toward the ALP for an overall national swing of 9.8%.

As there was residual swing left from the seat calculations in Pollycide 2 (which is where we tried to account for the marginal and safe seat movements), that means that the swings actually underestimate the actual seat numbers that would have been lost.

By how much? – up to 10-15 seats depending on where the residual swing fell.

So what we can feel fairly safe saying is that were an election held recently, according to 3 months worth of polling by Newspoll, the ALP would have gained 49 seats as a minimum, somewhere around 55-60 seats as a more likely possibility and a high of somewhere around 65.

That really is a Pollycide.

If you want to check out the details of any of the seats mentioned in the Pollycide series, Adam Carrs magnificent psephology site has a smorgasbord of info on all of them in his 2007 Federal Election Guide.

Small Update – I over counted NSW by 1 as a result of putting a seat in the wrong spot in my NSW swing graph.Its all fixed up now with NSW giving 17 seats rather than 18.

Thanks to the dozen or so people that pointed it out – nothing get’s past you folks.

Swings for Seats – National and State graphs

I’ve broken the Mackerras Pendulum down into national and state based swing graphs.The way to read them is like before, choose a swing on the vertical axis, trace it across until it hits the swing line, and then trace down to the horizontal axis to find the seats that the ALP would win or lose given that swing.

These are thumbnails so just click on them to blow them up to full size.

National Swing NSW Swing Qld Swing

Victorian Swing WA Swing SA Swing

For the remaining 9 seats in Tassie, NT and the ACT, I’ll give them as a table.

It shows the seat, the party that currently holds it, the State/Territory the seat is in and the swing needed for the ALP to win or lose it e.g a positive number represents the swing toward they ALP they need to take the seat, a negative number represents the swing away from the ALP they need to lose the seat.

Seat	Party	State	ALP Swing
Solomon	Lib	NT	2.9
Bass	Lib	Tas	2.7
Braddon	Lib	Tas	1.2
Lyons	ALP	Tas	-3.7
Franklin	ALP	Tas	-7.6
Lingiari	ALP	NT	-7.7
Canberra	ALP	ACT	-10.1
Denison	ALP	Tas	-13.3
Fraser	ALP	ACT	-13.4

UPDATE:

The outstanding Simon Jackman has up a lovely graph with state by state swings and the seats involved.And it looks good too!

http://jackman.stanford.edu/blog/?p=281

Possum Feeders Alert

This is for those folks that read the site by feeds.The two last articles Pollycide Part 1 and 2 have been rewritten to more accurately reflect what I was doing and what it means and how it all works.

For those reading this via a browser…

Move on, nothing to see here

Pollycide 2 – Draft Numbers

This article has been rewritten as well to hook more coherently into Part 1.

OK folks, we have the first draft of the numbers on the trends we highlighted earlier in Pollycide Part 1. If you haven’t read it, I suggest you do.

The way these were calculated was kind of complicated but used a type of linear programming, which is basically where glorious software figures out the value X (the size of the swing in each seat cohort) under the constraints of A, B, C etc, or in our case the various swings and proportions which make the numbers balance out to reflect the Newspoll data.

The table contains the seat, then the state, then the nominal TPP from the 2006 redistribution followed by the estimated ALP and Coalition TPP figures.These are all currently Coalition held seats.

 

Seat	State	2006 Redistribution			ALP Est	Coalition Est
		ALP TPP	Coal TPP
Macquarie	NSW	50.5	49.5		61.83	38.17
Wakefield	SA	49.3	50.7		59.24	40.76
Kingston	SA	49.9	50.1		59.01	40.99
Bonner	QLD	49.4	50.6		58.74	41.26
Makin	SA	49	51		58.11	41.89
Eden-Monaro	NSW	46.7	53.3		57.22	42.78
Wentworth	NSW	47.4	52.6		57.09	42.91
Lindsay	NSW	47.1	52.9		56.79	43.21
Moreton	QLD	47.2	52.8		56.54	43.46
Blair	QLD	44.3	55.7		56.11	43.89
Herbert	QLD	43.9	56.1		55.71	44.29
Bennelong	NSW	46	54		55.69	44.31
Hasluck	WA	48.1	51.9		55.62	44.38
Cowper	NSW	43.4	56.6		55.56	44.44
Stirling	WA	47.9	52.1		55.42	44.58
Longman	QLD	43.4	56.6		55.21	44.79
Page	NSW	44.5	55.5		55.02	44.98
Dobell	NSW	45.2	54.8		54.89	45.11
Sturt	SA	43.2	56.8		54.78	45.22
McEwen	VIC	43.5	56.5		54.64	45.36
Paterson	NSW	43.2	56.8		54.53	45.47
McMillan	VIC	45	55		54.50	45.50
Robertson	NSW	43.1	56.9		54.43	45.57
Corangamite	VIC	44.6	55.4		54.10	45.90
Petrie	QLD	42.1	57.9		53.91	46.09
Boothby	SA	44.6	55.4		53.71	46.29
Deakin	VIC	45	55		53.67	46.33
Kalgoorlie	WA	43.6	56.4		53.59	46.41
Gippsland	VIC	42.2	57.8		53.34	46.66
La Trobe	VIC	44.1	55.9		52.77	47.23
Dickson	QLD	40.9	59.1		52.71	47.29
Gilmore	NSW	40.5	59.5		52.66	47.34
Hughes	NSW	41.2	58.8		52.53	47.47
FLYNN	QLD	42.2	57.8		52.37	47.63
Bowman	QLD	41.1	58.9		52.08	47.92
Dawson	QLD	39.8	60.2		51.61	48.39
Higgins	VIC	41.2	58.8		51.51	48.49
Leichhardt	QLD	39.7	60.3		51.51	48.49
Hinkler	QLD	41.2	58.8		51.37	48.63
North Sydney	NSW	39.9	60.1		51.23	48.77
Dunkley	VIC	40.6	59.4		50.91	49.09
Kooyong	VIC	40.4	59.6		50.71	49.29
Ryan	QLD	39.5	60.5		50.48	49.52
Canning	WA	40.4	59.6		50.39	49.61
Macarthur	NSW	38.9	61.1		50.23	49.77
Goldstein	VIC	39.9	60.1		50.21	49.79
Warringah	NSW	38.7	61.3		50.03	49.97
Flinders	VIC	38.8	61.2		49.94	50.06
Menzies	VIC	39.3	60.7		49.61	50.39
Wide Bay	QLD	37.8	62.2		49.61	50.39
Forrest	WA	39.5	60.5		49.49	50.51
Hume	NSW	37.1	62.9		49.26	50.74
Casey	VIC	38.6	61.4		48.91	51.09
Forde	QLD	37	63		48.81	51.19
Fisher	QLD	37	63		48.81	51.19
Wannon	VIC	37.6	62.4		48.74	51.26
Greenway	NSW	39	61		48.69	51.31
Fairfax	QLD	36.7	63.3		48.51	51.49
Moore	WA	39.1	60.9		48.26	51.74
Berowra	NSW	36.9	63.1		48.23	51.77
Lyne	NSW	35.9	64.1		48.06	51.94

This is one alternative of what could explain the swings from quite a large number of mathematical possibilities, but it is the one probably closest to the ‘median possibility’ (which is not a particularly technical term BTW, but literal) , in that the others contained quite large swings in some seat cohorts and produced some pretty funny results.

That’s 47 seats falling in July according to the quarterly Newspoll data analysed to within an inch of its life, and we still have around a 2% swing to the ALP to allocate in NSW that can’t be extracted properly from the data without producing some silly things, as well as an extra 3% or so swing to the ALP in the government safe seats. Those two probably are a bit of a twin act in NSW somewhere, but the latter could well be spread around.

This is something approaching the most likely result of an election were it held in July based singularly on uniform swings by state, city type and safe/marginal grouping that is consistent with the Newspoll data. However, with a 2% swing to the ALP not really accounted for, there would have been some reallocation of the size of the vote in some of those seats, possibly including the loss of an extra 10-15 seats as a worse case scenario for the government, depending on where that extra swing went.

This is a little better for the Coalition than I had expected from the initial play around with the data in Pollycide Part 1, but only just. You might notice that the ALP TPP has calmed down a fair bit from the earlier post that used the Experimental TPP– that’s the result of removing as much feedback from the combined swing issues as was possible, but you also might notice that the seats, or rather the types of seats identified in Pollycide Part 1 as being in danger was actually pretty close to the mark.

That gigantic 14.6% swing happening in the safe government seats is the real, profound danger for the Coalition. I can’t remember anything like this happening before (maybe the living electoral encyclopaedia Adam could tell us) and only local factors would have saved the Coalition any marginals were an election held in July.

The other big killer is the fact that the safe ALP seats have only swung 4.1%.It’s not unusual for largish swings to happen, but in the wrong seats such as the ALP experienced in the 1998 election.

This, however, is a completely different ballgame. The swings are in all the right places for the ALP and all the wrong ones for the Coalition. It’s leading me to believe that if the polling behaviour holds, the ALP will win more seats, a lot more seats than their national swing would ordinarily suggest.

 

So the question becomes, if there are big swings happening in unusual places, which there must be for the Newspoll data to be right (which it surely would be – who doesnt worship the truthly godliness of the Newspoll? ) where are they?

Anyone on the ground in safe government seats, particularly in NSW and Qld that are picking up an anti-government vibe worth around, oh….. say, 14.6% ?

Continue on to Part 3 – The Verification

 

 

 

 

Pollycide Part 1

I’ve completely rewritten this article to make it more coherent and reflective of what I was trying to say rather than making it look like the government was going to be nailed with a 70 seat loss. Yesterdays attempt was appalling gibberish that confused everyone including myself.So if you’ve read this before – it’s now all new and improved!

If you havent, welcome aboard to a long argument

As many of you know, the ALP primary vote projections based on the combination of the recent quarterly demographic and Marginal/Safe seat Newspolls showed what can only be described as horrific results for the Coalition.

Well let me say from the outset, what you are about to witness cannot be described as anything other than absolute devastation. Coalition supporters might want to remove any sharp objects from their vicinity before reading on.

The previous projections were rough, they didn’t take into account the redistributions since the last election, and they didn’t take into account capital vs. non capital breakdowns which is really, really important in some States. But most importantly, as I stated at the beginning of that article, it didn’t include the most recent quarterly Newspoll because, frankly, I didn’t believe it’s contents. Not that it was wrong, but just that I COULD NOT BELIEVE what it was saying and what I was seeing.

You see, the data from the last quarterly Newspoll is even worse for the Coalition than the Q1 Newspoll. Much, much worse as the results in the states with the most Coalition seats, like NSW, tanked in the last quarter whereas the ground the Coalition made up were in States like WA ands SA where the seats they hold are relatively few.

This is the real danger that these latest Newspoll numbers hint at.

What we are going to do is create a set of weighted ratios out of that quarterly Newspoll data and apply them to Coalition seats – not to tell us what the vote necessarily is in those seats, but to highlight the Coalition seats that are in danger because of the way the vote is swinging.

To start with, we’ll use a breakdown of the redistributed two party preferred results since the last election based on the Mackerras Pendulum for the initial redistribution, and make experimental TPP projections and swing projections based on the last quarterly Newspoll, March to July 2007 including State, Capital City vs. Non Capital City and Marginal vs Safe swings.

I took the State based Newspoll figures and determined the TPP swing for the ALP since the last election by State. I then took the national TPP swing for the ALP since the last election from that same Newspoll.

I then divided each State swing by the national swing to give me a state weight. Let’s use NSW as an example. The NSW swing to the ALP was 12.2, and the national swing was 9.8.The state swing divided by the national swing is 1.245.

That means in NSW the swing towards the ALP is 1.245 times the national swing.

I then took the national capital city swing towards the ALP (9.5) and divided it by the national swing (9.8) to give me a capital city weight of 0.969.This means that the capital cities are swinging 0.969 times the national swing. I repeated this process for the non-capital cities to arrive at a non-capital city weight of 1.245 (Yes it is the same as the NSW State weight as both had a TPP swing to the ALP of 12.2).

Next I took the national marginal seat TPP swing to the ALP (9.3) and divided it by the national swing (9.8) to get a marginal seat weight of 0.949.I repeated this process to get a safe government seat weight of 1.4898. That, by the way, is huge.

To get an idea of how these weights stack up on a seat by seat basis is simple and best explained with an example. Let’s take the NSW seat of Bennelong (every psephs favourite seat)

At the 2004 election the TPP result was 54.33% Coalition and 45.67% ALP.

The effect of the 2006 redistribution on Bennelong was to reduce that TPP down to a notional Coalition 54% and ALP 46%.

Bennelong is a NSW, capital city, marginal seat. Hence, the estimated TPP vote for Bennelong based on the Newspoll quarterly breakdown, using the weights described above is:

Estimated Bennelong TPP for the ALP = Redistributed ALP TPP + (state weight * capital city weight * marginal seat weight * national swing).

Hence, the estimated ALP TPP for Bennelong = 46 +(1.245 * 0.969 * 0.949 *9.8)

= 46 + 11.22

= 57.22

These are the ALP swings and the ALP swing weights as determined by the Newspoll data for the March-July 2007 results.

The National Swing is 9.8

NSW, swing 12.2, weight 1.245

Vic , swing 9, weight 0.918

Qld, swing 11.1, weight 1.133

SA , swing 10.4, weight 1.061

WA, swing 5.4, weight 0.55

Marginal Seat swing = 9.3 weight = 0.949

Safe Government Seat swing = 14.6 weight = 1.490

Capital City swing = 9.5 weight = 0.969

Non Capital City swing = 12.2 weight = 1.245

However, we have a problem. There will be feedback between some of the weights, for instance if there is a large swing in non-capital city NSW, that will inflate both the non-capital city ALP swing AND the NSW ALP swing AND most likely the swing against the government safe seats. If we apply that weighted swing to most seats it will produce an inflated TPP result.

But it’s also a good thing.

The feedback built into the weights also allows us to identify the seats that are most likely to be suffering swings against them simply because of where those big swings broadly identified by Newspoll are occurring.

For instance, we know NSW has a 12.2% swing to the ALP, we know that there is a swing against the government of 14.6% in their safe seats and we know that non-capital city seats are swinging slightly more than capital city seats. So its probably a fair assumption to make that there are a number of NSW non capital city seats, that are considered safe government seats and which are experiencing very large swings.

What the following table attempts to do is to identify those seats by using an experimental, weighted TPP projection and a nominal swing projection for each seat derived from that TPP. These are based on the 2006 redistribution as the underlying, pre-existing TPP vote for the last election.All seats are nominally (as in post 2006 redistribution) Coalition held seats.

 

Seat	State	ALP TPP		Experimental ALP	Experimental ALP
		2006 Redistribution		TPP	Swing
Cowper	NSW	43.4		66	22.6
Paterson	NSW	43.2		65.8	22.6
Gilmore	NSW	40.5		63.1	22.6
Hume	NSW	37.1		59.7	22.6
Lyne	NSW	35.9		58.5	22.6
Farrer	NSW	34.6		57.2	22.6
Parkes	NSW	31.2		53.8	22.6
Riverina	NSW	29.3		51.9	22.6
Blair	QLD	44.3		64.9	20.6
Herbert	QLD	43.9		64.5	20.6
Longman	QLD	43.4		64	20.6
FLYNN	QLD	42.2		62.8	20.6
Petrie	QLD	42.1		62.7	20.6
Dickson	QLD	40.9		61.5	20.6
Dawson	QLD	39.8		60.4	20.6
Leichhardt	QLD	39.7		60.3	20.6
Wide Bay	QLD	37.8		58.4	20.6
Fisher	QLD	37		57.6	20.6
Forde	QLD	37		57.6	20.6
Fairfax	QLD	36.7		57.3	20.6
McPherson	QLD	36		56.6	20.6
Fadden	QLD	34.7		55.3	20.6
Groom	QLD	31		51.6	20.6
Moncrieff	QLD	30.1		50.7	20.6
Maranoa	QLD	29		49.6	20.6
Mayo	SA	36.4		55.7	19.3
Grey	SA	36.1		55.4	19.3
Barker	SA	30.1		49.4	19.3
Macquarie	NSW	50.5		68.1	17.6
Robertson	NSW	43.1		60.7	17.6
Hughes	NSW	41.2		58.8	17.6
North Sydney	NSW	39.9		57.5	17.6
Macarthur	NSW	38.9		56.5	17.6
Warringah	NSW	38.7		56.3	17.6
Berowra	NSW	36.9		54.5	17.6
Cook	NSW	36.3		53.9	17.6
Mackellar	NSW	34.5		52.1	17.6
Bradfield	NSW	32.5		50.1	17.6
Mitchell	NSW	29.3		46.9	17.6
McEwen	VIC	43.5		60.2	16.7
Gippsland	VIC	42.2		58.9	16.7
Flinders	VIC	38.8		55.5	16.7
Wannon	VIC	37.6		54.3	16.7
Indi	VIC	33.7		50.4	16.7
Murray	VIC	25.9		42.6	16.7
Mallee	VIC	25.2		41.9	16.7
Bowman	QLD	41.1		57.1	16
Ryan	QLD	39.5		55.5	16
Sturt	SA	43.2		58.2	15
Eden-Monaro	NSW	46.7		61.1	14.4
Page	NSW	44.5		58.9	14.4
Hinkler	QLD	41.2		54.3	13.1
Higgins	VIC	41.2		54.2	13
Dunkley	VIC	40.6		53.6	13
Kooyong	VIC	40.4		53.4	13
Goldstein	VIC	39.9		52.9	13
Menzies	VIC	39.3		52.3	13
Casey	VIC	38.6		51.6	13
Aston	VIC	36.8		49.8	13
Wakefield	SA	49.3		61.6	12.3
Wentworth	NSW	47.4		58.6	11.2
Lindsay	NSW	47.1		58.3	11.2
Bennelong	NSW	46		57.2	11.2
Dobell	NSW	45.2		56.4	11.2
Greenway	NSW	39		50.2	11.2
McMillan	VIC	45		55.6	10.6
Corangamite	VIC	44.6		55.2	10.6
Bonner	QLD	49.4		59.6	10.2
Moreton	QLD	47.2		57.4	10.2
Kalgoorlie	WA	43.6		53.6	10
Canning	WA	40.4		50.4	10
Forrest	WA	39.5		49.5	10
Pearce	WA	37		47	10
O’Connor	WA	29.6		39.6	10
Kingston	SA	49.9		59.5	9.6
Makin	SA	49		58.6	9.6
Boothby	SA	44.6		54.2	9.6
Deakin	VIC	45		53.3	8.3
La Trobe	VIC	44.1		52.4	8.3
Moore	WA	39.1		46.9	7.8
Tangney	WA	38.2		46	7.8
Curtin	WA	35.3		43.1	7.8
Hasluck	WA	48.1		53.1	5
Stirling	WA	47.9		52.9	5

Please note, this too is rough, and could be cleaned up even further with more time and more data. But that clean up process wouldn’t dramatically alter the nature of what is going on.

What the above table tells us is that the seats with the highest experimental swing are the best candidates to actually be experiencing swings against them that are necessary to balance out the Newspoll data.

The problem for the government comes from the 14.6% swing against their safe seats being an average swing. Now clearly some coalition safe seats wont be swinging much at all, which makes other safe coalition seats having larger than 14.6% swings against them to balance out the numbers. Some probably having much larger swings against them.

For the Newspoll figures to be right means that a lot of these seats will have fallen if an election were held in July. They wouldn’t have fallen with those margins (because, remember they are inflated), but still would have gone.I’ll put some more realistic numbers to these swings in Part 2.

The danger for the government here is if the swing against their safe seats is uniform. The more uniform it is, the most seats they will lose.The government holds 49 seats that are considered safe (i.e. are held with a buffer of more than 6%) but are held with a margin of less than this 14.6% swing against them.

This is why uniformity in this swing would wipe them out. When I was playing around with this a few days ago, I couldn’t believe what I was seeing – we all talk about elections being won or lost in the marginals, but if the polls hold, the marginals will be irrelevant. Not that the Coalition is doing particularly well in that area either, with a 9.3% swing against them there.

And nor can they be satisfied of the usual saviour of anti-government swings – the swings in the opposition safe seats. The ALP is only picking up an average of a 4.1% swing in their own safe seats. That too is extremely surprising.

In part 2, I’ll use some linear programming (or what the more pure end of mathematics calls matrix algebra) to whittle down the size of these swings to be consistent with the national, state, capital city vs. non capital city and safe vs. marginal seats swings.Then we might get a better idea of how an election may have looked like were it held in July.

One thing is clear however, these results are catastrophic for the government, suggesting that only a handful of seats are truly safe and that there will be massive swings against them in many seats they thought were untouchable. That simply has to be the case for the size of the swings to balance out.

 

CONTINUE on to Part 2 where some realistic numbers are produced to reflect the actual swings

What’s a swing worth?

Just a quick nifty little graph today based on the Mackerras Pendulum.Click on it to blow it up.

What it shows is how many seats the ALP would win or lose at the election for a given uniform swing.

The way to read the graph is simple, pick a swing on the left hand side, trace it across until it intersects the red line, then move down to the bottom axis to see how many seats the ALP would win or lose based on that swing.

For example, if the ALP had a swing against it of 1%, which would be a -1% swing, they would lose 4 seats.

If the ALP had a swing toward it of just over +2%, it would gain 8 seats, and a +5% swing would gain them 19 seats.

Play it, Sam. Play ‘As Polls Goes By’.

Yes, it’s another poll and another cheesy movie headline, but this time it’s best summed up by the famous misquote of this line from Casablanca:

“Play it again Sam”.

The latest ACNielson poll in the Fairfax papers is the same old song that’s been playing over and over again since December.

ALP primary vote stuck somewhere around 49, the Coalition primary vote stuck somewhere around 39 and the TPP stuck somewhere around the 58/42 mark.

The polls for the last 7 months are really starting to resemble that Goodies episode “Radio Goodies”, where the Goodies pirate radio station only had one song to play.

“And now, it’s ‘A Walk in a Black Forest’”

And now, its ALP 49, Coalition 39 and a TPP of 58/42.

This ACNielson poll like the stream Morgans before it, has simply verified the data that came out of the two big whopping Newspolls last week. The ACNielson voting intention by age cohort is consistent with the Newspoll quarterly data as is the ACNielson “city vs. rural” breakdown consistent with the Newspoll “capital city vs. non capital city” breakdown.

All of these polls suggest that the primary vote splits are no longer a trend, but a level and probably have been for quite some time.

But the interesting thing that came out of this ACNielson poll was on the issue of housing affordability.

Longer term readers here (if by longer you mean “have been reading for the last 2 months” since this blog is still wearing nappies) would know that I’ve been carping on about Interest Payments to Disposable Income having a long term relationship with the Oppositions primary vote. You can see the graphs and commentary about this HERE and HERE if you haven’t already.

One of the key arguments I’ve been making about how this arcane little figure is impacting upon Howards electoral prospects is that its not only the housing stress created by greater debt servicing burdens that is causing grief. I’d even argue that housing stress is the smallest of the electoral effects that IPDI is creating for the government.

The big one, the big elephant in the IPDI room is how higher debt servicing burdens (and higher rents trickling down from that) are reducing household discretionary income.

That discretionary income is what households use to fund the conspicuous parts of their standard of living. It’s not only the oft cited Plasma TVs, home electronics and new cars – but other types of consumption that frame the narrative for household living standards.

“How many kids can we send to that entry level private school this year? “

“Can we go on holidays to the Sunshine Coast in that unit like we usually do, or will we have to go camping instead?”

“I know all your friends have a new Playstation but we can’t afford one at the moment and you’ll just have to make do with your X-Box.”

“I know we usually order Pizza on a Friday night – but not this week”.

These aren’t profound issues when ranked against world poverty or the spread of HIV/AIDS in developing countries, and nor are they as trendy as global warming and climate change.

But they are important to many households because it is, to put it simply, their life.

Discretionary spending creates the visual yardstick by which large parts of households standard of living is judged against their neighbours, or against media archetypes of “what their standard of living should be”, or even against their own personal experience of the recent past.

The ACNielson poll asked “Which of the following statements best describes how you personally have been affected by the decline in housing affordability?”

20% of respondents agreed that “you have cut spending in other areas a lot” while a further 17% agreed that “You have cut spending in other areas a little”.

1 in 3 Australians are saying they have reduced their discretionary spending. “Housing Stress” might grab the headlines because its one of those sexy crisis things that sells copy, but it’s the little things that sometimes have the bigger effects.

When we add into this mix the policy formerly known as Workchoices , and the apprehension (real or imagined) that the policy itself and the campaigns against it have created over earnings, job security and working conditions – the behaviour in the suburban marginals does not surprise me at all. The Workchoices apprehension feeds into household discretionary spending by creating uncertainty over future discretionary spending budgets, which compounds the way housing affordability is impacting upon current discretionary spending budgets.

The IPDI/Housing stress/discretionary spending/Workchoices nexus is powerful, and for me it explains a very large chunk of the electorates behaviour in the polls.

On another note, three big cheers to Phillip Coorey over in the SMH.

In his article on this poll he stated at the end, “Mr Rudd’s approval rating fell two points, which is within the margin of error, but is still high at 61 per cent, while his disapproval rating stayed steady at 23 per cent.”

How refreshing it is to see some proper reporting of polls. “margin of error”… wow, there’s hope for the fourth estate yet.

Margin of Terror 2 – The Dismembering

Continuing on from Margin of Terror, let us combine these last two glorious Newspolls, the quarterly demographic Newspoll and the quarterly marginal and safe seat Newspoll to give us a very rough idea of the seats that would be in danger of falling with these enormous swings.

The following table was constructed by first dividing the individual state primary vote swings given in the demographic Newspoll by the national primary vote swing given by that demographic Newspoll. This gives us a weighting mechanism for each state that can be applied to the marginal and safe government seat swings to produce a rough estimate of the size of the marginal seat and safe government seat primary vote swings to the ALP that is occurring in each state. We can then use this measure and apply it to every Coalition held seat in the Federal Parliament to give us a very rough estimation of how the primary vote of the ALP would change in each of these seats.

For instance, the Victorian state swing to the ALP primary vote was 8.6%.If we divide this by the national swing of 11.4, we get 0.754 as the Victorian weight.If we then multiply the national ALP primary vote swing in marginal seats (9.2%) by this weight, we get a measure of the Victorian marginal seat swing of 6.9368%.

We can then add this 6.9368 to the primary vote of the ALP at the last election in any marginal Victorian seat (a seat held by the Coalition with a TPP vote of less than 6%) and get a very rough estimation of the size the ALP primary vote would be should these swings be uniform.

We can do this for every State, and for safe government seats as well.

Below is the list of all Coalition held seats after this primary vote swing was applied, where the projected ALP primary was over 40.It doesnt contain the Territory seats or the Tasmanian seats as these weren’t included in the Newspoll data.Note also Bennelong and Wentworth aren’t counted as being “in the mix” either as a result of the low primary ALP vote at the previous election in those seats, however these seats based on a TPP reading must also be considered up for grabs (big thanks to the eagle eye of Mr Cusack for pointing out that Bennelong was missing from the list).

The ALP>43 column shows an A (for ALP) if the projected ALP primary vote is greater than 43 or C (for Coalition) if it was below, signifying the most likely result.If the ALP gets 43 they usually win, with only Makin (ALP primary 43.02) and Braddon (ALP primary 43.05) having been lost at the last election where the ALP achieved over 43% of the primary vote.

The “Marginal/Safe” column tells us the whether the seat is currently marginal or safe.

The election data came from:

http://results.aec.gov.au/12246/results/HouseFirstPrefsTppByDivision-12246-NAT.htm

And the seats don’t take into account any redistributions since the last election.

This suggests 46 seats would fall.Now clearly this isn’t going to be the case but it highlights the depth of trouble the Coalition is in.

 

Seat	State	ALP Projected Primary	Win ALP>43	Marginal/Safe
Makin	SA	52.864	A	m
Bonner	QLD	52.594	A	m
Paterson	NSW	52.353	A	s
Wakefield	SA	52.214	A	m
Kingston	SA	52.134	A	m
Petrie	QLD	51.819	A	s
Robertson	NSW	51.063	A	s
Greenway	NSW	50.998	A	m
Herbert	QLD	50.899	A	s
Longman	QLD	50.389	A	s
Bowman	QLD	50.169	A	s
Dickson	QLD	50.029	A	s
Eden-Monaro	NSW	49.688	A	m
Macarthur	NSW	49.393	A	s
Sturt	SA	49.209	A	s
Hinkler	QLD	48.974	A	m
Moreton	QLD	48.964	A	m
Cowper	NSW	48.323	A	s
Lindsay	NSW	48.068	A	m
Parkes	NSW	47.583	A	s
Hughes	NSW	47.583	A	s
Dobell	NSW	47.308	A	m
Gilmore	NSW	47.303	A	s
Blair	QLD	46.099	A	s
Leichhardt	QLD	45.999	A	s
Dawson	QLD	45.979	A	s
Boothby	SA	45.684	A	m
Forde	QLD	45.659	A	s
Grey	SA	44.969	A	s
McEwen	VIC	44.8798	A	s
McMillan	VIC	44.7968	A	m
McPherson	QLD	44.789	A	s
Hume	NSW	44.543	A	s
Gippsland	VIC	44.4198	A	s
Ryan	QLD	44.069	A	s
Macquarie	NSW	44.063	A	s
Page	NSW	44.058	A	m
Cook	NSW	43.993	A	s
Dunkley	VIC	43.8998	A	s
Corangamite	VIC	43.6868	A	m
North Sydney	NSW	43.563	A	s
Fisher	QLD	43.499	A	s
Stirling	WA	43.218	A	m
Menzies	VIC	43.1198	A	s
Deakin	VIC	43.0868	A	m
Lyne	NSW	43.073	A	s
Fadden	QLD	42.689	C	s
Hasluck	WA	42.568	C	m
Berowra	NSW	42.463	C	s
Wannon	VIC	42.1798	C	s
Warringah	NSW	41.943	C	s
Fairfax	QLD	41.809	C	s
Aston	VIC	41.4998	C	s
Flinders	VIC	41.2198	C	s
Riverina	NSW	41.003	C	s
Higgins	VIC	40.8298	C	s
Goldstein	VIC	40.7898	C	s
La Trobe	VIC	40.7068	C	m
Wide Bay	QLD	40.349	C	s
Casey	VIC	40.0098	C	s

Margin of terror!

A new Newspoll today, this time marginal seat polling. You can see it over in The Oz.

To start off, Newspoll considers seats held with a margin of 6% or less as “marginal”.

I’m going to use 4 data points. The 2004 Election, Jul-Sep 2006, Oct-Nov 2006 and Jan-Mar 2007.

Let’s look at the most uninteresting aspect first (yes, the most uninteresting) – marginal seat polling using primary votes: The ALP and the Coalition primary votes can be read from the left hand side of the graph, the “others” as in, ‘everyone else’ can be read from the right hand side.

 

The ALP primary vote under Rudd has moved from 40 to 50%, while the Coalition primary vote has moved from 41 to 37 over the same period. As you can see, in the marginals the ALP has clipped the Coalitions wings, but they’ve decimated the “Others” vote. This is consistent with other data that suggested the Coalition lost 5% from their primary vote to the minor party +undecideds camp after the 2005 budget, which then moved across to the ALP under the Rudd leadership.

Since the last election, there has been an estimated swing away from the Coalition primary of 8.3% and a swing to the ALP of 9.2%.

This probably won’t hold as some homogenous pattern across all the marginals because there are a lot of local factors in play that protect some members against even large swings – but this looks like a blood bath. There are 23 Coalition held marginals according to the Mackerras Pendulum

Labor needs 16 seats.

Now let’s get onto the really interesting thing – safe government seats.

Below are two graphs, the first looks at the ALP vs Coalition vs Others primary votes in the safe government seats, the second looks at the ALP vs Liberal vs National primary votes in the safe government seats. These results have profound ramifications for the geographical distribution of where seats will likely fall at the next election if anything like this swing pattern holds.

As we can see from the first graph, the ALP has nearly completely closed the gap with the Coalition in those safe government seats in terms of the primary vote, and if you throw in the minor party preference flows, the two party preferred would be neck and neck. But if we look at the breakdown of that Coalition vote in the second graph, the Nats have dropped their vote by 2.8% since the last election (and the Nats do seem to get underestimated a bit in the polls – just my opinion), but the Libs vote has slumped by 8.6%!

The swing away from Coalition in the safe government seats has been 11.4%, and the swing to the ALP is 13.7%

Yes, that isn’t a joke, it says 13.7%.

If that pattern holds, you are talking about 40 Coalition seats being in play including some of the most blue ribbon Liberal seats in the country. These figures are absolutely disastrous for the Coalition.And taking into consideration the earlier quarterly newspoll data where the capital and non-capital city swings were roughly the same, the seats in question aren’t isolated to the capital cities, but are much much broader in geography.

 

This isn’t some “Oh yeah, we’re behind in the polls but we have a cunning plan” moment, this is staring down the barrel of an electoral execution.

 

Realistically, 40 seats wouldn’t change hands as swings aren’t homogenous, but those Liberal held outer suburban, mortgage belt seats where Interest Payments to Disposable Income are hitting (both in terms of housing stress and reducing discretionary spending budgets), and where Workchoices is impacting upon perceptions of job security and working conditions, especially for the second income earner – if anything like these swings persist through to the election, the capital cities will be surrounded by a sea of red Labor seats, all the way out until you run into a Nat.

Even the larger regional centers appear to be in danger of becoming red dots.Surely that cant be the case? Surely large parts of regional Australia wont return to their early history and support a Labor party?

These results aren’t just an ALP raid into the Coalition held marginals, these are an ALP assault on Coalition heartland. Or what has been since 1996 for suburbia, and decades for the regions.

I’ve been waiting for this poll for a few weeks now, expecting to see the marginals behaving the way they are, but I didn’t expect the safe government seats to be doing anything like this. A bit of movement, sure.

A 13.7% swing to Labor just wasn’t in the picture.

People are making comparisons between 2007 and 1996, and 2001 and 2004.But this is different in every aspect. The satisfaction rating differentials are different, the primary vote swings are different, the patterns of the swings are different.

Statistically, the relationship between the satisfaction levels of the government and their primary vote has completely broken down.

Over the last few days, the marginal seat patterns have shown themselves to be different, the State by State breakdown has been different. I’m convinced that insight wont be found for the forthcoming election by looking at the past – this is a different kettle of fish altogether.

That’s not to say that I’m writing the Coalition off (although it’s getting to point where I’m about too) nor do I expect these patterns to hold to their current level, but this is a completely different level of trouble they have found themselves in compared to the 1998 One Nation assault, the 2001 pre-Ryan by-election slump and the few months following the election of Latham to the ALP leadership.

No wonder the Coalition is creating a War Room , the question is whether it will become another Berlin bunker.

Continue on to Part 2

As God is my Newspoll, I’ll never be hungry again!

What a whopper of a Newspoll!

Quartely data broken down by State, capital city vs. non capital city, age and gender.

There’s plenty to look at here, but over the next few days I’ll try to throw up some of the more interesting things that don’t get the attention they probably deserve.

So for todays helping we’ll go first to the primary vote swings.

What the following graphs measure is the difference between the current quarterly measure of the primary vote for each party and their primary vote at the last election. These swings are broken down into various categories:

The most important thing here is the big swings for the ALP in Qld and SA, which is where large amounts of seats are up for grabs.That said, NSW isnt looking too shabby either if you’re an ALP type.A bit bleak if you happen to be a Lib or a Nat.

Even WA, which the commentariat insists isn’t looking that crash hot for the ALP is still showing a 5.3% primary vote swing to Labor. What is even more interesting is the non-capital city swing to the ALP being as high as the capital city swing.

In Qld particularly, there are swathes of non-capital city seats which aren’t yet really believed to be up for grabs – like Blair, Flynn, Forde (now a three cornered contest), Herbert, Hinkler, Leichhardt (3 cornered contest) and Longman which, if the non-capital city swing is roughly proportional to the capital city swing in Qld (as it is nationally), must make the sitting members in those seats nervous because these could well be in play. At the very least their large margins would be completely wiped out if these patterns hold.

But where are these votes coming from?

Moving onto the government:

It’s clear that in some states there are large movements toward the ALP that aren’t coming from the Coalition vote. So let us look at how the minor party vote is playing out since the last election in terms of the swing:

There’s a bit here and a bit there going across to the ALP

What’s interesting is how minor party votes have shifted to the Coalition in Victoria and WA, albeit not by much.

The other interesting thing that came out was more information on the voter movement that happened straight after the 2005 Budget. Below we compare the ALP and Coalition primary vote by gender over the period since the last election:

Some of that 5% of the Coalition primary vote that deserted the party after the 05 Budget and went ostensibly to the Minors+Undecideds camp reveals itself.

Women didn’t like the 05 budget, but the blokes seemed to absolutely hate it. Also notice how the slope of the male and female lines for the Opposition after Rudd was elected was steeper than the slope of the lines for the Coalition. This tells us that a fair chunk of these people were parking their vote with the minors before Rudd came along, and looking at the difference between the slope of the Male and Female post-Rudd opposition lines, it looks like more females than males were parking their vote in the minors until Rudd came along and gobbled them up.

Likewise, female voters deserted the Coalition more so than male voters after Rudd came along.

I’ll probably be playing with this data set for about a week or so. Who knows what else we’ll find.

 

UPDATE:

Bryan of Ozpolitics fame does a comprehensive analysis of the swings, the poll movements and all the pseph fit to print about this whopper of a Newspoll. Go and check it out here.

Go on. You know you want to.

 

Poll Wars: Episode 1 – The Phantom Metrics

Over in The Oz, Newspoll CEO Martin O’Shannessy writes:

“I was stimulated to consider whether Dennis Shanahan was right when he interpreted the turnaround in John Howard’s better prime minister rating as something more than merely encouraging. How could this be a turning point in the campaign if voter intention has not moved in spite of the Prime Minister’s improved ratings?

The question is whether the data supports the view that a turnaround in Mr Howard’s better PM rating presages an improvement in the Coalition’s electoral stocks. The short answer to this question has been yes in the past three elections.”

The fact that anyone could get stimulated over Baghdad Bob is concern enough – but I suppose that’s what Martin has got to say. But what is of real concern is that last paragraph. Let us focus on the phrase that is the glue of Martins piece; “whether the data supports…”.

We’ll get to some interesting regression models later on that completely tear this argument a new one, but first off let’s look at whether the Preferred PM ratings have clearly moved over the periods leading up to the last three elections: Below are the graphs of the Preferred PM rating for Howard and the Government Primary Vote over the 1998, 2001 and 2004 periods. The dashed vertical line is the election in each graph.

Well 2 out of three ain’t bad. Since we are looking at the effect of Preferred PM from June to July onwards for an election somewhere around November, and taking into account the margin or error on these polls which is plus or minus 3%, we be confident that in 1998 and 2001, Howards Preferred PM rating went up as the election approached, but in 2004 we simply cannot say that at all. In fact, in 2004 the only thing that can be said is that Howards PPM rating didn’t move in any way that could be classed as statistically significant between April and October of 2004.

Before we move on any further, I’d like to show you a couple of interesting things.

Firstly, what it is we are actually talking about:

 

In the 1993 election, Hewson was preferred PM over Keating in both February and March and lost. In the 1996 election, Keating was Preferred PM over Howard in February and March and lost. The theory of Preferred PM being the great deterministic force behind election victories isn’t looking too crash hot.

Second, have a squiz at Preferred MP ratings for the Opposition and Government overlaid against their respective satisfaction ratings (using Newspoll data):

 

From this we can see that PM satisfaction levels and his preferred PM status pretty much walk hand in hand, but the same cant be said for the Opposition. What is also worth taking a look at here is how the preferred PM rating from Howard became slightly decoupled from his satisfaction rating around mid 2005.Let’s take a closer look:

The reason for this is simple – there are more complex dynamics at play. The period directly after the 2005 budget saw around 5% of the Coalitions primary vote move over to the Minor Party+Undecideds camp.See Determining the Swinging Voter and Their Behaviour for a detailed explanation of this.

But what’s important here is to notice the consequences of that movement carry across into preferred PM ratings. Howards satisfaction ratings went down, people were shifting their primary vote to the minor parties and into the Undecideds camp but they still preferred Howard over Beazley. Howard kept the PPM status by a powerful lack of alternatives.Then along came Rudd and bang, PPM changed.

 

Now, keeping all that in your thought orbit, let us move on to test the other key argument made by Martin:

“The question is whether the data supports the view that a turnaround in Mr Howard’s better PM rating presages an improvement in the Coalition’s electoral stocks.”

Martin is saying that the Preferred PM rating is a leading indicator for the Coalitions primary vote.

This we can test – in depth. We’ll use the following variables:

C = a constant that is handy in these equations to measure the base value of the dependent variable (GOVPRIMARY in this case) that isn’t explained by movements in the values of the independent variables. This is automatically calculated by least squares regression software.

GOVPRIMARY = governments primary vote

PPMGOV = Preferred PM rating of Howard

And we’ll sometimes use (-1) or (-2) after those variables to represent a lagged variable i.e. PPMGOV(-1) is a variable that whose current value is last months value of PPMGOV.

And we’ll use the sample period of the Howard governments reign from April 1996 through to July 2007.

For those of you with a statistical bent – the following is a “cut down for a wider non-stats audience” affair, so please refrain from giving me grief. I’m sacrificing minutia for readability.

To start with lets run the obvious regression which explains how much of the government primary vote can be explained by changes in the previous months value of PPMGOV (which is Martins argument – what we are testing is whether PPMGOV(-1) is a leading indicator).

This suggests, very very superficially (I’ll demolish it in a minute), that Martin may be somewhat right.22.7% of the movement in the governments primary vote can be explained by movements in the previous months PPM rating.It suggests that a 1 point increase in this months PPM will lead to a 0.21555 increase in the governments primary vote next month.

Now let us run the obvious next regression which explains how much of the government primary vote can be explained by changes in the current months value of the PPM rating for Howard.

This suggests that 30% of the movement in the governments primary vote in any given month can be explained by changes in the value of the PPM for Howard in that month .So, superficially, this suggests that changes in PPM values walk hand in hand with changes in the governments primary vote rather than being leading indicators.

Now let’s run both variables together:

When we combine the dynamics of both the lagged value of PPM and the current value of PPM for Howard something stands out. The lagged value of PPMGOV called PPMGOV(-1) is not statistically significant. This means that is doesn’t actually explain any statistically significant movement in the government primary vote when taking into account the influence of the current value of the PPM rating for Howard.

 

BUT

There is another problem here tied into this. This equation is auto-correlated up the wazoo. This means that there is an awful lot of lagged explanatory power still laying around in the residuals of the equation that is messing with our explanatory variables. If we look at the correlogram for that last equation we get:

 

This means, in this particular case, that the previous value of the governments primary vote has a big influence on the current value of the governments primary vote (the inertia effect that you often find in voting estimation series) and is interfering with out modelling. We can make another regression which accounts for this using a one period lagged value of GOVPRIMARY:

This clears up all the autocorrelations and gives us a nice white noise process in the residuals which is exactly what we want.

What this new, clean equation tells us is that the 57-58% of the movement in the value of the governments primary vote can be explained by changes in the values of the 3 key variables of GOVPRIMARY(-1), PPMGOV and PPMGOV(-1).

However, far from Howards preferred PM rating being a leading indicator, it’s the opposite.

Put simply, what this means is that if we hold all the other independent variables constant, a one point increase in the last periods value of Howards preferred PM rating will lead to a 0.21 point reduction in the governments primary vote!

That is EXACTLY THE OPPOSITE of what Martins theory tells us should happen.

Preferred PM ratings for Howard are simply not a leading indicator for the governments primary vote, period. Their lagged effect is smaller than the effect of the PPM ratings current value, and that lagged effect is completely the wrong sign.

Howard’s PPM rating is covariant with the government primary vote, in that the two values move together through time, likewise the lagged values of the government primary vote are a leading indicator for futre values – but the lagged PPM values mean diddly squat.

I can appreciate that Martin tried to defend The Shanahans piffle over interpreting polls, but some piffle is simply not worth defending.

His last paragraph sums it up:

“So, will the PM drag the Coalition up with him on the strength of his personal rating and will that be enough to overcome Labor’s commanding lead? The next few Newspolls will answer these questions.”

Watch out for the next Newspoll! Buy the Australian to get it!

Same bat time, same bat channel!

Which is fair enough.