Fri May 17

the social and demographic factors that influence your vote

Written by Rob J Hyndman, Professor of Statistics, Monash University

Australia has changed in many ways over the past two decades. Rising house prices, country-wide improvements in education, an ageing population, and a decline in religious affiliation, are just some of the ways it has changed. At the same time, political power has moved back and forth between the two major parties. How much can we attribute changes in political power to changes in who we are?

Quite a lot, as it turns out.

Finding the ‘average’ electorate

We analysed election results from 2001 to 2016 and mapped them against data from the census to see how socio-demographic characteristics influence voting patterns, and how this has changed over time.

A simple way to measure voting patterns is to consider the two-party preferred (2PP) vote, looking at only the Coalition and the Labor party.

More than 30 socio-demographic characteristics were considered, and an “average” electorate was created using the national electoral average for these characteristics. The influence of each characteristic is then measured by how much the two-party preferred vote differs from the average electorate due to that particular socio-demographic characteristic.

So, which factors strongly influence how we vote?

Income, unemployment and education

Successive Labor leaders accuse the Coalition of only caring about the “top end of town”. The Labor party typically campaigns on more progressive policies, which often include tax policies that adversely affect higher income earners. Conversely, the Coalition tend to favour policies that reduce taxes.

So it is no surprise that wealthier electorates are more likely to support the Coalition, with incomes having a strong positive effect on the Coalition’s two-party-preferred vote. Unemployment however, is not as influential.

And since 2007, electorates with higher education levels are associated with supporting the Labor party, although this effect is significant only in 2016. Before 2007, education had a negligible effect.

Industry and type of work

Despite the Construction, Forestry, Mining and Energy Union (CFMEU) traditionally supporting Labor, electorates with higher proportions of workers in “extractive” industries (mining, gas, water, agriculture, waste and electricity) and “transformative” industries (construction or manufacturing) are consistently linked with higher support for the Coalition, with the impact of this effect slightly increasing over the years.

This is not surprising. The Coalition is seen as the party with closer ties to traditional energy industries, which still see a role for fossil fuels in Australia’s energy mix. Labor, on the other hand, introduced the mining tax in 2012 (which was first floated by Kevin Rudd in 2010), and has a renewable energy target of 50% electricity generation by 2030 .

Similarly, electorates with proportionally more workers in managerial, administrative or sales roles are also more likely to support the Coalition.

Diversity

Larger migrant populations from the Middle East and South-Eastern Europe are associated with Labor support. Whereas the number of people born in Asia, the United Kingdom and elsewhere have no discernible effect.

However, speaking languages other than English appears to have a far stronger effect. Electorates with more diverse languages are associated with higher support for the Coalition from 2004 onwards.

In 2016, an electorate with a high proportion of people who speak a language other than English favour the Coalition by more than 12% when compared to the average electorate (on a 2PP % basis).

Other influencing factors: household mobility, relationship types and age

In each of the six elections, electorates with a higher proportion of people that have recently (in the past five years) moved house were more likely to favour the Coalition.

Our analysis controls for characteristics of home ownership and rental prices, so this effect is not simply due to electorates having low rates of home ownership, or due to electorates having high rental prices. Instead, it suggests people who are more transient are also more likely to be conservative voters, regardless of their home ownership or rental status. (This would need further study, as we do not have individual level voting data.)

De facto relationships, but not marriages, are also found to be an important (and significant) predictor of the two-party preferred vote in all six elections, with more de facto relationships associated with higher support for the Labor party.

Older people are often believed to be more conservative, and indeed we found that electorates with a higher median age are more likely to support the Coalition party.

Against the tide

When does an electorate vote very differently from what their socio-demographics would suggest?

The ten electorates with the largest difference between actual and predicted results in 2016 are shown below:

This suggests something beyond socio-demographic characteristics is impacting the results. For example, the Coalition had a much higher vote in Wentworth than predicted in 2016 (and also in 2013), probably due to the popularity of Malcolm Turnbull.

Jeremy Forbes, a former Monash University Honours student in econometrics, coauthored this analysis with Rob Hyndman and Di Cook.

The full analysis is available here and the code used for the analysis can be found in the github repository.

Authors: Rob J Hyndman, Professor of Statistics, Monash University