We’ve all heard of the Electoral Pendulum – where seats are sorted from top to bottom according to their two party preferred margin. The beauty of the pendulum is that it provides an excellent approximation of how many seats would change hands for any given two party preferred swing – something that Antony Green’s spiffy pendulum calculator utilises. If we want to know what would probably happen in an election with, say, a swing of 2% toward the government, then we can count up all the Opposition seats that have a margin of less than 2% – on today’s pendulum that is 12 seats in total – and give that number of seats to the government. But we only give that number of seats, not necessarily those actual seats.
This is because swings are never uniform across Australia – some seats swing more than others. What this means in practice though is that for every Opposition seat on a margin of less than 2% that doesn’t swing enough to put it in the government’s column, there will generally be an Opposition seat on a margin of above 2% that will swing enough to make it a government gain.
The theoretical basis for the pendulum is derived from the assumption of swings taking an approximately normal distribution – a distribution approximating a bell curve shape. Yet, if we look back on election the history of federal elections, we can generally be a little more accurate than this by throwing the average standard deviations of federal election results into the mix.
The standard deviation is a measure of the variability around a uniform swing, and in Australia over the last few elections it has averaged 2.5%. With our assumption that swings follow a normal distribution, and with a standard deviation of 2.5% – that means that approximately 68% of all seats swing by an amount within 1 standard deviation of the mean swing, and approximately 95% of all seats swing by an amount within 2 standard deviations of the mean swing.
So if we had a uniform swing of 6% and a standard deviation of 2.5, then we would expect 68% of seats to swing between 3.5% and 8.5%, and 95% of seats to swing between 1% and 11%.
Currently we have a phone poll average showing a swing of 3.8% toward Labor. If we assume a standard deviation of 2.5% (the average standard deviation over the last 4 elections) and apply these qualities of a normal distribution to such a swing, we end up with something looking like this:
Around 68% of Coalition seats would swing between 1.3% and 6.3% (one standard deviation from the mean swing) toward Labor, while 95% of Coalition seats would experience a swing between 1.2% toward the Coalition and 8.8% toward Labor.
But there’s another thing that happens with Federal Elections – when we have swings toward a party larger than the value of the standard deviation of the swing – nearly all of the seats that sit on a margin less than the difference between the uniform swing and the standard deviation, fall.
Using our current figures – with a swing of 3.8% and a standard deviation of 2.5, we would expect that nearly all of the Coalition held seats that sit on a margin of less than 1.3% (the uniform swing of 3.8 minus the standard deviation of 2.5) would fall to Labor. We’ll call these seats the Red Zone.
Similarly, if we look at seats within 1 standard deviation of the mean swing – Coalition seats that sit on margins of between 1.3% and 6.3% – and apply the pendulum principles, we can get some broad probabilities for that middle group of seats. We’ll call this the Orange Zone. Because of the large number of Coalition seats that sit in this Orange Zone, we would expect around 1 in 3 of these seats to fall to Labor in an election that had a 3.8% swing.
Finally, we would expect a couple of seats further up the pendulum to experience swings much larger than the mean swing of 3.8%. So those seats between 1 and 2 standard deviations greater than the mean swing will be called the Yellow Zone.
If we redraw our first chart, but add the Zones, this is what it looks like:
Finally, if we then list the actual seats themselves into these Zones (including the seat on the 3.8% margin), this is what we end up with:
The thing to remember here is that for each seat that has a rock solid partisan history (for example, Goldstein and Menzies in the Orange Zone), because their probability of falling to Labor is virtually zilch, other seats around them will take the weight of probability – making in more likely that other seats within the Zone will fall.
All up, this is a pretty handy anxiety thermometer for Coalition MPs.
15 Comments
Nice to see Hume in the Orange Zone! I recently moved to Yass from Surry Hills (Sydney), where Tanya Plibersek is more at risk from the Greens than the Liberals.
My brain’s in neutral today, and it took me about 15 minutes to figure out that’s not all the seats!
Good analysis Poss. I think the Libs are in trouble.
In time Poss, the Libs will learn to love you.
But that looks to be a long way off.
Oh no, we could lose Bob Baldwin! (repressed giggle). The Labour party played dead last time. I can’t imagine them not giving it a go this time.
And after that the most satisfying seats would be Wentworth, North Sydney and Goldstein. Shame the mad monk seems untouchable.
More seriously, you gotta wonder how this factors in to the Liberal Party’s own game playing. Those who are not in denial probably know it is better to wait for Turnbull to crash and burn. But the rest of them?
Can’t wait for the fireworks over Wentworth
inverting the last table – and ordering the zones as you would increasing temperatures, in a thermometer shaped graphic would look really spiffy – especially if you could animate it according to current swing data so it acted like a thermometer.
I note the Mincing Poodle is a dead dog.
Possum,
Might be worthwhile to put the Members names who’s seats have bubbled to the top of your anxiety thermometer. Suggest their will be a collection of big swinging dicks, good local members passed their use by date and a few old goats ready for the knackery.
Perhaps you can set up a scream watch for us spectator types.
Possum,
If you could convince me that Higgins deserves to be in the yellow zone I would be your slave for ever. As one of citizens of Higgins I can assure you the conservative thinking of these hide-bound old dears is deadly. Many of them love Peter Costello but when I ask them what he has ever done for the electorate. Many of them will start a rave on his period as the Liberal Party Treasurer. I then interject, no, “I mean for the electorate of Higgins?” Mouths fall open as they try to remember any single good thing he did anyone anytime? In Higgins. Silence.
Disclosure: I’m a Liberal supporter from Wentworth.
This is all based on a current phone poll average swing of 3.8% towards Labor. Between now and the next federal election, the swing is likely to diminish or turn around the other way. The seats in the orange zone are unlikely to be seriously under threat, although they should not be taken for granted.
Having said that, Poss I have been reading some of your other analysis about the future of the Coalition. I have no doubt that there is an enormous amount of work for the Coalition to do to remain viable in the long term. However, looking at David Cameron’s success with changing the Conservative Party in the UK gives me hope that there is a contemporary model to use.
Based on the same arguements that dictates that around 1 in 3 seats in the orange zone would fall, you get around 13% of seats in the yellow zone falling, which equatues to an expected 1.3 seats (so you might expect 1 possibly 2, though 0 is quite possible). That means it is still quite unlikely that Higgins would fall given this analysis, although in reality obviously each electorates own specific profile would change the expectations if put through a more detailed seat by seat analysis (which would be very difficult, requiring consideration of the voting history of every booth in every electorate in the country).
Very nice bit of work her though. For those that are interested, I put together this list of names of members under threat, given the above results (note that this ignores whether or not the sitting memebers are re-contesting):
Red zone:
Fran Bailey
Andrew Laming
Peter Dutton (Shadow Minister for Health and Ageing)
Peter Lindsay (Shadow Parliamentary Secretary for Defence)
Jason Wood (Shadow Parliamentary Secretary for Justice and Public Security)
Pat Farmer
Luke Simpkins
Michael Keenan (Shadow Minister for Employment and Workplace Relations)
Luke Hartsuyker (Shadow Minister for Competition Policy and Consumer Affairs and Deputy Manager of Opposition Business in the House)
Christopher Pyne (Shadow Minister for Education, Apprenticeships and Training
Manager of Opposition Business in the House)
Top few from the orange zone:
Bob Baldwin (Shadow Minister for Defence Science and Personnel and Assisting Shadow Minister for Defence)
Paul Nevil
Donna Veil
Andrew Southcott (Shadow Minister for Employment Participation, Training and Sport)
Alexander Somlyay
Peter Slipper
Michael Johnson
In the seats under serious threat there are 4 shadow ministers (2 from shadow cabinet) and 2 parliamentary secretaries. The most marginal of the orange zone has only 2 outer shadow ministry members. In any case this really highlights (I think) that the Liberals really need to get rid of the dead wood in safe seats and get some talent in there (Wilson Tuckey and Phillip Ruddock, I’m looking at you…). The likelyhood is that the Liberals will lose some of the talent it percieves itself to have at present, and will be unlikely to win any new seats. Without getting the talent into the safe seats they hold it’s hard to see them lifting themselves out of their current troubles any time soon.
I think early reports of the pre-selections aren’t promising for the Liberals…
I love this analysis you do… but who really cares about the lower house at the minute? Sure it is what changes government but at the moment doing an analysis on this is a bit redundant if the swing is towards the government – the only interest is to local members and if, as described above, a big swinging dick is in trouble.
On that point, it seems that all the polls are mainly for the lower house. I know heaps of people who vote differently in the senate than the lower house.
What would be a much more interesting article is something like this for the senate.
Taking these numbers and doing your monte carlo magic on them to make the talk:
http://blogs.abc.net.au/antonygreen/2009/07/double-dissolution-versus-halfsenate-election-which-would-be-better-for-labor-in-the-senate.html#more
Obviously this is much more complex from a maths perspective – unless you can actually make assumptions about a uniform swing (I think you might have to look at a skewed curve locking in the unlikely to move base support in each state for labor, coalition and the greens) with might not be as valid in the senate… I don’t know?
Anyway – looking at this most likely situations in the senate is likely to be more important to the next election than then house of reps.
I don’t think you should necessarily be giving the election to Labor as a forgone conclusion. However, your messing with the low probability here.
Also, whay would be amazing to include is an increase in the margin of error when comparing a DD election and a regular election… I’d assume that the extra time to a regular election would, necessarily, have a larger margin of error in votes changing.
Something I have also missed in doing this sort of thing is the seat of Swan.
Can you give Steve Irons some love?
Tri$tan,
It’s almost impossible to do the same for the Senate because, firstly, the final Senate spot for every election is nearly always decided by arcane preference deals – and we don’t know what those deals are because they arent decided yet.
Secondly, we have no decent polling on the Senate (and even if we did, it would be highly dubious – Senate polling is rarely worth it’s weight in phone calls)
Thirdly, a number of Senate spots are dependent on how small parties fare, parties that only attract 1-1.5% of the vote (because those prefs then flow to bulk up other parties). But it’s hard to get accurate readings for parties that poll 1-1.5% in surveys. The level of their support is just too low to poll properly.
Thanks for the response Poss.
I get what you are saying. That’s a shame really. I mean, you could technically overcome the polling issue if somebody actually did a poll large enough… but the preference deals. Hmm, this would include so many assumptions that, you are probably right: it’s not analysis. it’s turning guesses into numbers a bit too much.
I guess Anthony Greens analysis is the best you can do at this stage.
However, if you ever manage to do anything valid with the senate it would be very interesting to read.