This paper explored the feasibility of constructing an experimental socio-economic index of disadvantage at the household level using General Social Survey (GSS) data. The interest in finer level indexes arises from the need for detailed disadvantage information to complement broad measures such as area based indexes. The GSS covers a broad range of socio-economic variables which enables the incorporation of many dimensions of disadvantage. The 2010 and 2014 GSS datasets were analysed separately to construct a socio-economic index of disadvantage for each period.
The paper discussed the concept of socio-economic disadvantage and how it has evolved over time from a narrow focus on resource and income based indicators to a more broad-based multidimensional concept encompassing both economic and non-economic factors. It also discussed the distinction between area level and individual level measures of disadvantage and provided a rationale as to why the household level was the more appropriate level at which to construct the index of disadvantage compared to an individual or family level.
Both simple and complex measures of disadvantage were explored. The simple measures consisted of counts of indicators and domains of disadvantage while the complex method involved using weights derived from principal component analysis (PCA) to combine the variables to derive a summary or composite measure of disadvantage.
Results from the simple measures showed that a large majority of households experienced few counts of disadvantage and a small proportion experienced severe levels of disadvantage for both 2010 and 2014. The composite method of index construction, which overcomes the limitation of equal weighting of the simple methods, involves the use of an explicit weighting scheme to combine different variables of disadvantage to construct a summary measure of disadvantage. Principal component analysis was used to derive the weights for the compilation of the composite index. The steps used to derive the final set of variables and their corresponding weights in this paper are similar to the approach used for Socio-Economic Indexes for Areas (SEIFA).
An analysis of the results from the composite index showed that the majority of the final set of variables used to construct the index and the distribution of the created index was similar across both periods. The most highly influential variables for both the 2010 and 2014 indexes were from the domains of the health and economics including, financial stress, income and wealth. The results at this stage are experimental and the caveats and limitations discussed in the paper should be kept in mind when interpreting the results. Further work could include additional validation of the methodology and the results and investigation into alternative methods to calculate scores for those records with missing index scores.