Towards a new research field: tax analytics using Australian Taxation Office data

Abstract

Tax datasets are becoming increasingly available in Australia; however, discussions on how such information can be utilised to benefit the tax literature are scarce. This thesis aimed to develop a general approach to explore tax datasets and form expectations of data patterns based on understandings of tax. Three substantial and independent research projects were used to illustrate how this approach can lead to tangible research outputs that benefit the tax literature. The first project developed an income tax deduction monitoring system in Australia using a linear mixed model, due to the concern of potential changes in individual tax compliance behaviour This system tracked the changes in the average income deductions of many demographic groups. The model suggested that most groups exhibited gradual changes, except for the consultant group that showed erratic swings. The second project combined the philosophy underpinning robust statistics with the understanding in current individual tax compliance and applied the minimum covariant determination method to individual taxpayers who were most likely to be compliant with tax laws compared to those who were most likely to not be compliant. This method could complement the tax crime detection model used by tax administrators and assist in addressing the influence of outlier problems. The third project estimated the aggregate tax compliance costs in Australia using a bivariate state-space model. This was an extension of the method used in the study by Wu and Tran-Nam (2017) . This bivariate approach addressed and extended the applicability of this general approach to other tax jurisdictions. In addition, it provided an update on the aggregate tax compliance costs in Australia.