Gary Morgan has been polling marginal seats and finds a significant number would revert to the Coalition. What he doesn't highlight is that across all of his seats there is virtually no movement since the last election. This supports the thesis that there is no Gillard honeymoon. By only polling Labor-held marginals he also introduces a pro-Coalition bias into his sample.

There are a few problems with Morgan's analysis. The largest occurs because his sample for each seat is approximately 200. This gives a sample error in 95% of cases at the 50th percentile of just under 7%, which means that for statistical purposes 57% is indistinguishable from 43%. However, amalgamating the sample gives a total sample of 3,000, whereupon sampling error decreases to 1.79%. There are still problems, because the sample is not representative of Australia as a whole, and probably not of the demographics of marginal seats either, but it is a much more robust result.

The poll results can be downloaded from here, and I have summarised the two-party preferred results below.

Given my qualifications on sample size there is probably not much more useful to be said. However, some of the results that Morgan could have headlined include:

• The Macquarie result suggests that the result in Penrith, which panicked the ALP into replacing Rudd, appears to be purely state based as the swing to the ALP in that area since the last election could be 10.4%
• Major Lieutenant Colonel (H/T to @MiltonFriedmans) Mike Kelly appears to have generated a substantial personal vote in Eden-Monaro with an increase in his vote of 6.7%
• Ditching Belinda Neale may have cost the ALP 1.6% of the vote in Robertson
• Jodie Campbell possibly held the ALP vote back in Bass by 11.5%

Some of the state results derived by aggregating the seats together are also more robust than the individual results and they suggest that Labor is doing OK in NSW and WA, and badly in Queensland.

Share this article on your favourite social bookmarking sites: