The Australian has published another set of geographic and demographic breakdowns, combining two weeks of polling (the 52-48 from yesterday and last week’s 50-50) to produce samples of about 670 per state. The results thus include half the polling which contributed to Newspoll’s geographic and demographic results from last week.
The table below provides an artist’s impression of how state-level polling has tracked through the campaign week-by-week, based on an aggregage of Newspoll and Nielsen results. The results appear to suggest that the swing to Labor has faded in Victoria and that Western Australia is weaker for Labor than generally supposed, but the margins of error is high enough that this should be treated with caution. Samples for any given observation were 765 for NSW, 665 for Victoria, 585 for Queensland, 465 for WA (865 in week three, achieved by throwing in the Westpoll result) and 445 in SA, producing margins of error ranging from 4.6 in South Australia’s case to 3.6 for New South Wales.
Perhaps the greatest point of interest is an implausible Labor collapse in New South Wales in week two. Most likely what this tells us is that unfavourable samples for Labor there dragged down their overall results that week.
As well as that, Roy Morgan has produced one of its quite useless Senate polls. This draws on 5000 face-to-face interviews conducted over the last two months, but for all its massive sample is of far less use in predicting the Senate result that an ordinary lower house poll would be. Of greater interest is Morgan’s Polligraph worm results for the treasurers’ debate. Amusingly, the pattern for Labor-supporting and Coalition-supporting participants forms a perfect mirror image. The Greens line is consistently quite close to Labor’s, but a gap emerges when Wayne Swan spruiks Labor initiatives to assist housing affordability.