Polls & predictions

Elections encourage an outbreak of ‘poll fetishism’ every poll or hint of one is racked over obsessively. But polls are not a magic time tunnel to the future but a summation of frequently unclear voter responses that reflect views held with varying degrees of intensity. Too often poll watchers fail to see the forest for the trees polls may be more useful to indicate broad patterns than specific results. All too often results are reported as gospel truth regardless as to their plausibility.A good example of that would be the JWS/Fairfax poll of 22000 votes in 54 seats. It sounds impressive and an article today takes it as accurate but this poll predicts in Victoria alone an 11% swing to Labor in the Frankston-based Dunkley and a 6% swing in Bendigo and a 10% swing in Deakin but no swing in Corangamite (mostly Geelong), their Brisbane estimates vary almost as much. This poll also predicts an extraordinary variation from 2007 across the electorates surveyed the R-Sq of the labor vote from 2007 to this poll is only 0.12 (see diagram). My guess is that the R-Sq between this poll and Saturday’s result will be under 0.5.

Is there any pattern we can discern? it is clear that Labor is traveling badly in NSW and Queensland but this swing seems to be concerted in Brisbane and Sydney. Labor’s vote is holding up better in provincial areas, its a fascinating reverse of past voting patterns, rural conservatism has declined along with ‘countrymindiness’. The election may be a landmark in the decline of the Nationals but more generally of rural conservatism, if Labor wins they may have a much more rural and provincial caucus than for previous Labor governments. Low-income rural areas have moved towards Labor whilst the more affluent cities have shifted the other way. So I see few rural seats at risk, they are the Queensland coastal belt of Leichhardt, Dawson and Flynn, Herbert is nominally Labor after the redistribution and the the Liberal victory in 2007 was clearly due to a personal vote for the Liberal incumbent as Labor out polled the Coalition in the Senate and Labor held Herbert 1983-96. Out of these four I think Labor can hold at least two. However I expect Brisbane to a disaster and suggest Labor could lose everything up to Moreton, but will probably hold one seat in this range. The other problem is WA I expect Labor to lose Swan and probably Hasluck. Analysis of NSW has to distinguish between greater Sydney (which includes Gilmore) and the rest. Labor to lose everything up to Greenway, Bennelong is ethnic and not outer suburban but I fear Labor’s margin is too narrow to resist the general swing. Outside greater Sydney could Labor make any gains? Cowper, Paterson and Calare are possibilities, but all would be a stretch.  So a total loss of 14 seats but I expect Labor to gain McEwen and La Trobe and one out of Dunkley and rural NSW. SA perhaps Sturt but unlikely. Greens to win Melbourne. So Labor will end with 75. This looks worse than I thought but there are a lot of Labor marginals at risk and in Sydney and Brisbane it will be a ‘wave’ election. The seats Labor has to hold for majority are the Brisbane marginals over 4%.

One constant theme of media commentary has been that this will be an election with widely divergent patterns between the states, the historical evidence is that 2007 was an election where voting patterns were uniform across the states. One indicator of variability is the ratio of the standard deviation of the ALP 2PP vote across the states divided by the mean vote. I have excluded the territories to enable comparison across 1949 to 2007 and because they are highly volatile and multiplied by 100 for ease of convenience. These figures revel that it would not be that unprecedented for there to be much larger variations in levels of ALP support across the states on Saturday than in 2007. The high points of electoral variability have been driven by Queensland’s conservatism and the now long past impact of the Labor split in Victoria.

1949 6.520945
1951 6.809481
1954 6.168959
1955 7.621668
1958 10.01337
1961 10.72638
1963 9.994547
1966 10.93443
1969 7.551779
1972 8.594396
1974 7.527009
1975 5.84762
1977 7.474423
1980 3.840326
1983 8.466554
1984 5.400743
1987 3.615436
1990 3.906188
1993 7.345184
1996 9.975673
1998 7.952581
2001 9.413469
2004 8.471456
2007 6.409441

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>