Morgan Poll: Combined poll produces funny results

Morgan Poll has published its latest poll which combines information collated disproportionally in a recent inner city poll with a more statewide phone poll. (Statistically you should not mix concentrated polling areas with less concentrated data - like the way they count the Senate it serisouly distorts the data presented. - So why did Morgan publish it?)


It has the Dr Bludger stumped , William (proGreen) Bowe, in his bludger blog states:

Morgan has produced another poll on the Victorian state election, and this one’s a lot more credible than the last – a statewide phone survey conducted from a sample of 943 over the past three nights, with a margin of error of a bit over 3 per cent. Despite nervous talk emanating from the Labor camp, the poll gives them a comfortable two-party lead of 52.5-47.5, from primary votes of 39 per cent for Labor, 41.5 per cent for the Coalition and 15.5 per cent for the Greens. I gather the poll consists of the inner-city results from Tuesday topped up with a further 667 responses from elsewhere, with the former presumably weighted downwards to reflect the fact that they only account for 4.5 per cent of statewide enrolment. However, I’m not entirely sure what to make of Gary Morgan’s accompanying spiel in which he says the headline two-party figure is “slightly ‘misleading’ as it includes a very high ALP Two-party preferred vote (72.5%) cf. L-NP (27.5%) in the marginal Inner Melbourne seats of Richmond, Northcote, Brunswick and Melbourne”. It isn’t clear to me why the predictable weakness of the Liberals in this area would contribute to a “misleading” total any more than would Labor’s corresponding weakness in rural areas and wealthier parts of the city. The poll also has John Brumby leading Ted Baillieu as preferred premier 47.5 per cent to 35.5 per cent, although Baillieu has better personal ratings: Brumby is on 38 per cent approval and 43 per cent disapproval, while Baillieu’s approval and disapproval are both 40 per cent.