Using Aggregated Demographic Data To Inform Electoral Boundary Redistributions: 2010 South Australian Election

Abstract

Electoral boundaries in South Australia are currently a contentious issue in politics, with allegations that the current boundaries are unfair. South Australia has fairness provisions that are unique in Australia governing the boundaries of electoral districts. However, in three of the last six state elections, the objective of fairness as characterised by these provisions has not been met. Boundaries are drawn by the independent Electoral Districts Boundaries Commission, and are revised after every general election in South Australia. The Commission's method uses estimates for the voting behaviours in small areas to inform the decisions about boundary changes. The objective of this thesis is to develop an alternative method for calculating these estimates, and test the credibility of the resultant estimates from our new method. We develop a series of gradually refined regression models that use demographic data in South Australia to predict voting behaviour. The demographic data is sourced from the periodical Census of Population and Housing. In this research we also test the proposition that income, education level, and the language people speak at home are significant factors in their voting behaviour, at an aggregated group level. We contend that the predictions calculated under the preferred model in this thesis are credible, and that the techniques used warrant further exploration.