Oz Politics Blog

I have been meaning to look at George Megalogenis‘ data tables for some time. Rather than use the data to predict what might happen this year, I thought I would look at the extent to which electorate income explained the 2004 election outcome.

First, however, a couple of box-plots to get a feel for the data.

And now let’s look at the graph that surpised me. The next scatter-plot suggests absolutely no relationship between the median income wihtin in a seat and the Coalition’s two-party preferred vote from that seat. Just look at the R2. It is almost zero.

Why is it so? Well it appears that moderate correlations exist within inner metropolitan, outer metropolitan and provincial seats. That trend suggests that seats with a higher median income are more likely to favour the Coalition. When you get rural seats, many have very low median incomes, but almost all favour the Coalition.

Update: Andrew Leigh sent the following missive …

Dear Bryan,

I noticed your post yesterday on electorate income and voting. As you know, I’m a great fan of yours, and link to you regularly (it should be more often, in fact). However, I do think you’re making a bad mistake in ignoring the ecological fallacy here. The best evidence of this is the research by Glaeser & Sacerdote that finds that income is positively correlated with voting Republican on an individual level, but negatively correlated with Republican voting across US states (http://andrewleigh.com/?p=1419).

Given that we have very good information about the individual-level relationship between income and voting in Australia (eg. my own work, or that of Goot & Watson in a recent AJPS), I’m puzzled as to what you think we can learn from the aggregate-level stuff. You, of course, know all about the ecological fallacy, but I’m worried that some of your readers may be misled.

Politically, this plays into the muddle-headed argument that the Libs make from time to time: that they’re just as much the party of the poor as the ALP. That’s simply a lie.

Cheers,

Andrew.

I have been looking at Professor Roger Stimson’s 28 March 2007 Parliamentary Library presentation on his ecological analysis of the 2004 election using the data from the 7576 polling booths and the 2001 census. The presentation was based on research by Stimson, Dr T- K Shyy, and Dr Prem Chhetri.

Absolutely fascinating.

For their analysis the researchers categorised each polling booth on which major party (Labor, Liberal, National and Country Liberal) won the most votes, and whether a minor party (Greens, Democrats and Family First) won more than 20 per cent of the vote. Then taking 46 variables from the 2001 census, the researchers used multiple regression to see whether socioeconomic or demographic characteristics could explain the variability in voting patterns.

Their key finding was that three discriminant functions explained more than 96 per cent of the variability in terms of the seven-fold categorisation of polling booths. The first two functions were:

1. The monocultural/older - multicultural/younger discriminant function, which accounted for 54.7% of the variance [this function comprised 16 census variables]
2. The disadvantage - advantage discriminant function, which accounted for 28.9% of the variance [this function comprised 11 census variables]

If you take the first two functions, and map the centroid of the distribution of polling booths by the seven-fold category onto a Z-score Cartesian plane, you get some interesting observations on the core voting base of the major parties.

According to Stimson,

• The Labor Party is clearly separated from the other political parties, being located within the multicultural/younger – disadvantage quadrant of the graph.
• In contrast, the Liberal Party is located within the opposite monocultural/older - advantage quadrant of the graph.
• The National Party is located in the monocultural/older - disadvantage quadrant of the graph.
• The Australian Greens Party and the County Liberal Party are both located in the multicultural/younger - advantage quadrant of the graph.
• The widest separation is between the Nationals and the Greens.
• There is a wide separation between the Nationals and the Liberals within the Coalition, with the results from the discriminant analysis modelling demonstration just how much the voting constituencies for these Coalition partners are differentiated.

Using spatial mapping techniques, Stimson also created maps of the political landscape of Australia, including the following:

Stimson described the 2004 political heartlands as follows

Coalition

• The government Coalition parties have captured most of the settled rural and regional areas, and it is not just the Nationals but also the Liberals that have widespread ‘heartlands’ of political dominance.
• Within the big cities the, dominance of the Liberal Party is not only across wedges of the higher socio-economic areas of the mostly middle suburbs, but it has also extend across the large belts of the outer suburbs and even more widely across the outer fringe areas to capture many areas in what used to be Labor’s ‘heartland’ as transform them into Liberal ‘heartlands.’
• Much of the outer areas of the big cities where most of the population growth continues to occur are now more Liberal ‘heartland’ than domains of Labor dominance.
• These new outer suburban and urban fringe Coalition ‘heartlands’ are the places where many of the so-called ‘aspirational voters’ live with their families.

Labor

• The Labor ‘heartlands’ are predominantly found in the central city and inner suburbs of the big cities, and in suburbia the now more restricted belts of Labor dominance are clearly associated with populations characterised by immigration and multiculturalism as well as the traditional areas of disadvantage.
• In rural and regional Australia, Labor’s ‘heartland’ is spatially relatively confined to some of the old industrial regional centres and some of the mining towns as well as the areas with concentrations of Indigenous populations in remote areas of Australia.

Others

• The Greens have clearly defined and spatially restricted areas of concentration of higher levels of primary vote support, and those are typically found in the ‘café latte’ trendy inner suburbs of the big cities as well as selected more advantaged suburban areas, and in regional Australia the Greens vote tends to be concentrated in selected coastal and tourism-oriented regions.
• The Australian Democrats primary vote collapsed at the 2004 federal election, and it is very low and thin across just about all of the nation. The few pockets of concentration of primary vote support for the Democrats are a handful of outer suburban and fringe areas in Adelaide and Perth and in parts of regional South Australia.
• The Family First Party, contesting its first federal election in 2004, had its primary vote spread thin with the areas of concentration being in the more advantaged outer suburbs .and in some parts of regional Queensland.

Want to know more

• Visit the ARC Research Network on Spatially Integrated Social Sciences

Hat tip

Thanks to Mark Rodrigues for the copy of the presentation.

Andrew Norton noted on his blog that the Parliamentary Library has published a new monograph by Maurice Rickard called Principle and Pragmatism: A study of competition between Australia’s major parties at the 2004 election and other recent federal elections.

Andrew’s key observation:

The chart that most interested me (on p.68, for those who download the publication) was the division of issues into economic and non-economic. This shows that since 1998 the Liberals have moved to the right on economic issues and to the left on non-economic issues. Their campaign rhetoric is consistent with strong spending increases on health and education, and the overall philosophy of ‘big government conservatism’, with growth-oriented economic policies used to finance a large welfare state.

As I have argued before, the big question is how viable this is as a long-term political strategy. Despite the Liberals’ rhetorical and policy shifts on non-economic issues, public opinion still favours Labor on these matters. And that’s with the benefit of being in government and actually implementing big-spending policies. If the Coalition loses the 2007 poll, will voters believe Opposition promises, or fall back on long-standing stereotypes of the political parties? The danger, as has happened in the states, is that the Liberals will just look like a less sincere and less competent version of Labor.

Howard’s contention is that just as the Coalition won in 2001 and 2004, so it will win in 2007. It is a plausible argument. Early in 2001 and 2004, the Opposition was well ahead in the polls, but from around the fourth month of the year until the election, the Opposition slowly declined in the opinion polls, and ultimately lost the election. The trend can be seen clearest in the historically more reliable ACNielsen series, but it is also evident in the Newspoll series.

If Howard is correct, we can expect Labor’s polling to improve over the next four to eight weeks. Indeed, we may not be able to test the deja vu election hypothesis before May or June this year, by when the beginnings of a downward trend should have materialised. If it has not, we can kiss the hypothesis goodbye.

The plausibility of the deja vu election hypothesis makes it difficult calling this election in the first half of 2007. As the 2001 and 2004 elections demonstrated, the polling trends eight months out from an election are not necessarily a reliable indicator of the election outcome.

Commentators on this blog often make claims about the accuracy of various polling companies. I decided to take a look at the opinion polls within the final weeks of the last three Federal elections to see how the polling companies stacked up. I compared the predicted Coalition and Labor primary vote shares and the predicted two-party preferred (TPP) vote share with the final outcome. I also considered the predicted preference flows to the Coalition.

Where a prediction was within two percentage points of the final outcome, I scored it as a pass (shaded green in the table below). Where it was within three points I scored it as a possible pass (shaded yellow in the table below). Where it was off by three percentage points or more I scored it is a fail (shaded red in the table below).

Taking an average for the three elections, at a two per cent margin of error, ACNielsen was correct 74 per cent of the time. Newspoll was right 48 per cent of the time. And Morgan was correct 22 per cent of the time. If we take a three per cent margin, the scores improve to 96, 78 and 50 per cent respectively.

Is this conclusive? No. It is not conclusive. It was a small sample. And it was possible that three weeks out from an election actual voting behaviour may have been different to that which occurred on the election day. In addition, polling companies can change their methodology at any moment, and the methodologies used previously may not be used for the next election. Furthermore, there is always the risk that I have made a transcription or adding error. Still, it was interesting.

Update: as I had feared, there was a transcription error in my initial set of calculations. I had swapped the election primary votes for Labor and the Coalition in 1998. It has now been fixed. If there are any other errors, please let me know.