Monday, March 11, 2013

Updated aggregation

The House of Morgan surprised me last week with a new multi-mode poll based on an Australia-wide sample of 9,101 electors aged 18 years and over. That is a big sample. 

The new multi-mode technique combines web-based surveying with face-to-face surveying. This is the first time I have seen it from Morgan. At this stage I don't know what to make of Morgan's new polling methodology. I will need some months of data before I have a sense of how affected it might be. Nonetheless, I have dropped it into my models.

Let's look at where this latest Morgan data-point sits against the combined localised regression. In terms of the population estimate for two-party preferred (TPP) voting intention, the latest Morgan poll is roughly where a lot of polls have been in recent weeks.

Like the Morgan face-to-face series, the new multi-mode poll has the Coalition primary vote share down on the other polling houses.

It has the Labor primary vote a touch up on other recent polls.

Consistent with the Morgan face-to-face series, the primary voting intention for other parties and independents is above the LOESS line.

Turning to the Bayesian aggregation, the latest data support my contention of two weeks ago: the decline in Labor's TPP vote share in January/February 2013 appears to have found a floor.


The floor effect is even more evident when today's (Monday 11 Mar 2013) Essential poll is added into the mix:

No comments:

Post a Comment