Posted by Possum Comitatus on October 30, 2007
An interesting little question popped up in Newspoll, and to fully chew it over properly, it needs to be put in the appropriate context:
It looks like most people are pretty locked in on this question, with even numbers of ALP and Coalition supporters suggesting that they may cross sides in the event of a rate rise. The alternative is that maybe people don’t realise how they will act until they actually see the money coming out of their bank account.
History gives us a good suggestion here of how people actually react to rate rises as opposed to whatever they may say about how they would react.
The other thing of note on the Newspoll vote was the low minor party support. This gets us on to how the small minor party vote estimates are highly volatile because they are so small.
To highlight this, we’ll go through a two step process. First, if we subtract the TPP vote of Coalition from the TPP vote of the ALP for every Newspoll since 2006, we’ll get a TPP spread. Then if we do the same for the primary votes, we’ll get a primary vote spread. These two together show us how much of the difference in the TPP vote is explained by primary votes. If we then subtract the primary vote spread from the TPP vote spread, we end up with a figure that represents how much of the TPP spread is caused by the minor party vote.
That might sound a little complicated, but it’s pretty easy once you chart it.
If we blow that minor party residual effect up to highlight its movement we get:
What this chart tells us is how much of the TPP spread is explained by the minor party vote. This minor party vote is relatively volatile as a result of it being a small number in a larger pool. For instance, if there was a survey of 100 voters and 40 were voting for party A, 40 for party B and 10 for party C, if in the next survey it turned out that an extra person was voting for party B, the vote for party P would become 41, which is a 2.5% increase, but a 1 person increase in Party C (from 10 to 11) would be a 10% increase in the Vote of Party C.
With the sizes of the polling samples we use in Australia, this effect never fully washes out of the system. So small party votes estimates are, proportionally to their vote size, more affected by sampling noise than the major party votes.
As a result of that, and because TPP estimates are based on the distribution of those noisy vote levels, a fair bit of the TPP spread is quite noisy. Hence the TPP estimates bounce around more than they would be moving in reality.