The principal purpose of this thesis is to construct a spatial microsimulation model (SMM) for obtaining small area housing stress estimation in Australia and then to develop some methodological capacities in spatial microsimulation modelling, particularly for overcoming the lacunas in validation techniques as well as measures of statistical reliability of the SMM estimates. The thesis starts with providing an extensive review of the small area estimation (SAE) methodologies to see where the SMM techniques stand in SAE, and whether or not better results can be identified. Findings of the appraisal reveal that spatial microsimulation modelling is a geographic approach of indirect SAE that is quite robust and has advantages over other approaches such as direct and indirect statistical methods of small area estimation. The thesis then systematically describes the construction process of the SMM and generates housing stress estimates at statistical local area (SLA) level in Australia. Discussions of various results derived from the model outputs include numbers and percentage estimates of housing stress by households’ tenure types into a range of spatial scales such as SLAs, major capital cities, different states, and overall Australia. Findings demonstrate that the housing stress estimates vary significantly by tenure types and with geography. In particular, private renter households are more likely to be in housing stress in most of the SLAs located in major capital cities and costal centres across Australia. Moreover, this research study makes several advancements in the SMM methodologies. These include the development of statistical theory based new types of validation tools to test the statistical significance of small area estimates produced by the SMM, and the creation of confidence intervals to establish the measures of statistical reliability of the SMM estimates. The empirical results of these validation tools are generally satisfactory in terms of scientific notion and statistical interpretation. The key findings suggest the following: the model produces statistically accurate housing stress estimates for a significantly large number of SLAs in Australia, and the confidence interval measures for these SLAs are also reasonably narrow, which confirm high level of statistical reliability of the model estimates with good precision. Finally, the thesis proposed the Bayesian prediction theory based an alternative methodology for small area microdata simulation in the spatial microsimulation modelling.
|Date of Award||2011|
|Supervisor||Ann Harding (Supervisor) & Shuangzhe Liu (Supervisor)|
Small area housing stress estimation in Australia : microsimulation modelling and statistical reliability
Rahman, A. (Author). 2011
Student thesis: Doctoral Thesis