I keep hearing ditties about Labor getting a bounce in the polls when Parliament sits, with Jack the Insider over at The Oz being the latest, to wit:
One of the quirkier patterns in the results of the past 17 months is that Labor gets a bounce during Parliamentary sittings while the Coalition does better during the recesses. It was thus for the Christmas period of 2007-08 when the Coalition vote stayed, more or less at its election level, only to head south once the new parliament began.
And this summer, the Coalition increased its primary vote but when the parliament resumed in February of this year, the gains disappeared.
But is this really true?
Firstly, we need to change our Polls series. If we take the first difference of that series with respect to each pollster, where we measure the change in polling values for each poll rather than its nominal value, we end up with a series that measures change in political support. So, for instance, if a Newspoll result came out that had the ALP two party preferred on 58 when last Newspoll it was 57, we’ll give that latest Newspoll result a value of +1. If an Essential poll result came out that had the ALP on 58, down from the 60 that Essential reported in the previous poll, we’d give that result a value of -2.
We can do that for both the ALP and Coalition primaries as well as the ALP two party preferred (we don’t need to do the Coalition TPP because it’s simply a mirror image of the ALP).
We can also create a dummy variable that has a value of 1 when the pollster was either in the field when Parliament was sitting, or was in the field the weekend immediately after Parliament sat – and for all those other times when the pollsters were in the field but Parliament wasn’t sitting, we’ll give those observations a value of 0. The Parliament schedule data comes from here.
This gives us 136 observations from January 2008 onwards.
If we regress the differenced series for the ALP primary against a constant and the Parliament Sitting dummy variable, the result will tell us how big a bounce, on average, the ALP primary vote receives when a pollster is in the field during a Parliamentary sitting. We can do the same for the Coalition primary and the ALP two party preferred.
These are our results:
There are a number of problems we have here when trying to measure this – the obvious being sampling error possibly interfering with our calculations. However, even taking that into account, and with 136 observations, if this was a meaningful relationship we would expect to see it be significant at maybe the 10 or 15% level (we could widen our confidence to accommodate the expected variability).
On the two party preferred, there is, on average, a one tenth of one percent increase in the ALP TPP vote when parliament sits compared to when it doesn’t. Yet that number is not statistically significant in anything remotely approaching a meaningful way.
On the Coalition primary vote, when parliament sits it is, on average, 0.005 points lower than when Parliament isn’t sitting – yet that too isn’t statistically significant at any meaningful level. It’s actually so statistically insignificant, we could plug random numbers into the series instead of the Coalition primary vote and we would expect to see exactly the same outcome.
The ALP Primary vote is the only case which comes within a bulls roar of being real, suggesting that the ALP primary vote increases, on average, by nearly half a point when Parliament is sitting compared to when it’s not.
However, the level of significance here is extremely low. Normally we’ll accept a 5% level of confidence with these things, a 10% level if we’re looking for relationships with uncertain data, or even a 20% level of confidence if the data is highly uncertain – but we’d have to accept a 30% level of confidence to accept this relationship as being true, which is far, far too wide for us to acknowledge any meaningful statistical relationship between Parliament sitting and ALP primary vote bounce.
So we can conclude that there is absolutely no observable relationship between the Coalition primary vote and polling bounces or slumps when Parliament sits, we can say the same with the Two Party Preferred vote as well.
We can also say that while the data suggests that the ALP primary vote shows a very, very loose bounce when Parliament sits – not quite half a point on average – we cannot be confident that this relationship is meaningful in any way, it might simply be random noise. The relationship as it exists is statistically insignificant at all accepted levels of confidence.
So to answer the question “Does Labor get a bounce when Parliament sits?”
No, they don’t.