Determining the Swinging Voter and Their Behaviour
Posted by Possum Comitatus on June 28, 2007
If we look at the primary votes of the parties over the last 20 years, it becomes pretty clear that there are a certain amount of voters that support a given party regardless of how that party is behaving. The Coalition for instance never gets below 35 in Newspoll for the primary vote. It has received Newspoll estimated primaries of 35.5 in June 98, 36 in March 01 and 35 in May 07 – but never below 35. This suggests that there is about 35 percentage points of the Coalition primary vote that is rusted on given the ordinary course of politics.
Likewise with the ALP there is a bottom limit of 33% of the primary that they never go below. In April 04 they had 33, January and October 2002 they had 34 and for 8 consecutive months between October 1990 and May 1991 they received 33 to 35.This suggests there is about 33 percentage points that the ALP primary vote that is rusted on given the ordinary course of politics.
The minors and undecideds are a more complicated kettle of fish.They have scored as low as 6.25 back in 1987, but this segment has been growing over the last 20 years as highlighted in Time Trends and Primary Votes .
However in may 1996 they scored 9% and December 1999 scored 10%.So let us take a conservative estimate here and assume that there is 9 percentage points of rusted on minor party and undecided votes. As this group includes both parties and those that haven’t decided, the variation of this group over time is large with a maximum Newspoll estimation of 20% or greater occurring in 28 months over the last 20 years. But for the issues we are going to explore here this variance really doesn’t matter.
These minimums combined add up to 77%.This suggests that only 23% of the electorate are swinging voters, and this 23% of the electorate determines governments.
So now that we know the rusted on component of each groups primary vote, let us produce 3 new modified series that represents the Newspoll primary vote for just the swinging voters.
This is easy to do – we simply subtract the rusted on vote of each group from their estimated Newspoll primary each month. When we do this, the swinging voter primary vote series for each party become:
The above series tell us how much of that 23% of swinging voters are supporting each group.
For example, Junes Newspoll primary vote results had the ALP 46%, Coalition 39% and Minors+Undecideds 15%.
Once we subtract the rusted on component for each group, this becomes ALP 13%, Coalition 4% and Minors+Undecideds 6%. This means that of the 23% of swinging votes, the ALP has 13 percentage points of it, the Coalition 4 percentage points of it and the Minors +Undecideds have 6 percentage points of it.
Now we have this, we can do some snazzy analysis of where those votes are moving, and where they’re moving from, and to where.
Firstly, let’s look at this over the period of October 2006 through to today. If we take the first difference of these three series, which means measuring the change that occurs in each of these three series each month, we get the following:
A word of warning – as Newspoll has a margin of error, taking this graph as gospel is a silly thing to do (and we’ll get onto a means to get around that margin of error problem a little later on). So, acknowledging FIRST that the above observations could be out by a few a points – let us make an initial assumption that they’re true, even though they probably aren’t exactly true (there’s a self-deprecating joke about economists to be had somewhere here).
Now we can read the graph. This graph shows where the swinging voters votes are going between the three parties each month by looking at the distance of each observation from zero. Pay no attention yet to the coloured lines that join the observations – at this stage they’re irrelevant.
So if we take the November 2006 observations, they show that the Minors and Undecideds primary vote increased by 3.5 points from the previous month (the distance between the observation and zero), the Coalition primary vote fell by 0.5 points and the ALP primary vote fell by 3 points. This tells us that in November 2006, the Minors and Undecideds gained 3.5 points on their primary vote completely at the expense of both major parties (3 points from the ALP and 0.5 points from the Coalition).
If we look at the last Newspoll (June 2007) it tells us that the Coalition primary increased by 4 points, the Minors +Undecideds increased by 2 points and the ALP primary decreased by 6 points. This tells us that in June the Coalition and the Minors+Undecideds gained totally at the expense of the ALP primary.
That’s all interesting enough, but we can do something far more interesting still. If, for each series we take the cumulative sum of the values of the first differences (i.e. cumulatively add the observations that we observed above for each specific series), two things happen:
First – we end up with a series that shows the size of the swinging voter movement between parties over time.
Second – we remove most of the margin of error problem associated with political polls. The estimated Newspoll value of the primary vote for each group will fall randomly above or below the true value of the primary vote for that group within a couple of points each side. By cumulatively adding our above series, the influence of the margin of error will approach zero as the number of observations approach infinity.
For what we are about to do, we don’t quite have an infinite number of observations, but we do 34 monthly observations based on averaging nearly 70 Newspoll observations. That means we can safely use the below graph to draw conclusions generally free from the problems of error margins, or at the very least, we have seriously diminished their effects.
So let us now graph these three new series over the period from the last election through to today:
What this shows us is how the swinging voter has changed their votes over the period since the last election, and to where they have changed them.
Over the first 7 months of Howards 4th term, swinging voters didn’t change their voting intention a great deal. There was a small bit of movement between the three groups.
Then came the 2005 Budget. After that budget (maybe even because of that budget), swinging voters started to move away from the Coalition and parked their votes with the Minors+Undecideds.5% of the governments primary vote was lost to the Minors and Undecideds and has NEVER RETURNED. The ALP didn’t benefit at all with their primary vote staying around what it was at the 2004 election.
Between April and December of 2006, there was a bit of voting intention change with the ALP and the Minors+Undecideds swapping supporters, and the Coalition and the Minors+Undecideds swapping supporters and a handful of Coalition swinging voters moving to the ALP.But only is small amounts.
That 5% block of swinging voters that moved away from Howard after the 2005 Budget didn’t come back to him. They might not have wanted to vote for ALP, but they weren’t going to vote for Howard. Then along came Rudd in December 2006 and three interesting things happened.
Firstly, that entire group of AWOL Coalition supporters that abandoned Howard after the 2005 Budget went straight across to Rudd lock stock and barrel – let’s call them Group 1.
Secondly, a few percentage points of the Minors+Undecideds vote that wasn’t the ex-Coalition voting block went across as well – lets call them Group 2
Thirdly, another 5% of the Coalitions primary vote, a completely separate group of swinging voters, went across to Rudd – lets call them Group 3.
At Rudds height in May, he had the entirety of the 23% of swinging voters except for the 4% that were in the Minors+Undecideds camp. The Coalition in May had zero swinging voter support.
Group 1 that abandoned Howard after the 2005 Budget 2 years ago isn’t coming back, they’ve had repeated chances and they haven’t moved anywhere except to the ALP and a few of them back to the Minors+Undecideds. None seem to have gone back to Howard.
With the June decline of Rudd, Howard seemed to only get half of Group 2 back, while the Minors+Undecideds picked up a few points of Group 1.
If we look at the 23% of primary votes available that are cast by swinging voters which ultimately determine the election result – things look bleak for the Coalition.
First lets take the Minors. Remember that the rusted on Minors+Undecideds was a conservative estimate at 9%.During the last 4 elections, the minors have never received below 14%. If that continues, this leaves only 18 percentage points of the primary votes left for to be grabbed. (The 23% of swinging voters minus the 5 percentage points minimum that has gone to the minor parties in the last 4 elections)
For the Coalition to win, they need 44% of the primary to be sure. As they have 35% already rusted on, that leaves them needing to get 9 points. Yet Group 1 have abandoned the Coalition, and have placed their vote over the last 2 years anywhere but with the Coalition. That means that Howard has to get 9 points of primary vote from 13 points available to him (the 18 left after the Minor Party vote minus the 5% points from Group 1).
That leaves Howard needing 9 from 13, or 69% of those votes.
The ALP however can easily win with 42.That means they too need 9 points of primary vote (coming from a rusted on base of 33), but they can get it from the 18 available to them.
That leaves the ALP needing to get only 50% of the votes available to them. And this is the Coalitions best case scenario.
I actually believe that the ALP has already picked up 5%, boosting their rusted on vote to 38.This is made up of 3-4 points from Group 1 and 1-2 points from the Minors+Undecideds, particularly in Qld.
Under this scenario, that leaves the Coalition needing to get 9 points from the 11 or 12 available to them to win, which is between 75% and 82% of the votes available to them.
The ALP on the other hand only needs to get 4 points from the 13 available to them. That means the ALP needs to get 30% of the votes available to them to win.
Just a note – when I posted this yesterday, somehow I ended up with 23-5=19… ick!
This looks better.