Socially disaggregated and spatially explicit models, like spatial micro-simulation or agent-based modelling, allow for flexible testing of policy interventions at household level and analyzing of their consequences at relevant levels of socio-economic or spatial aggregation. However, the approach requires demographic data both at household and small area levels, information that is not publicly available in Australia. The authors propose to use a synthetic census to generate this information and couple it with a spatial micro-simulator. This paper describes the approach and provides an example of application to taxation in Australia. The application examines the impact of making mortgage interest payments tax deductible, similar to the way that investment property mortgage payments are tax deductible. Results show that, overall, the proposed change in taxation regime was nearly cost neutral by increasing tax rates for top income earners. The impact of this change is looked at by family type, household income decile and the area that the person is living in. the model shows that this housing policy benefits low income households most, and the impact is relatively similar across all Australian States and Territories. Households with a household head aged 16 – 44 benefit more than those with an older household head, although households who lose most have a household head aged 35 – 44. Single parents with dependent children win most, and couples with dependent children lose most. These results also show that households in regional and remote areas were mainly in the middle of the distribution in terms of losses, but experienced significant gains due to the rental policy change. Most households in richer areas lost, while households in poorer areas won, due to a mix of lower incomes and greater benefits from being able to claim mortgage payments as a deduction against their income. From a modelling perspective, the study shows that bringing together a synthetic census and a spatial microsimulation model is possible using direct linking of the household ID, as long as the two models use the same base dataset. Where this is not the case, then other methods can be used. However, these methods will not provide the same amount of variability as a direct linking.
|Number of pages||7|
|Publication status||Published - 1 Jan 2017|
|Event||22nd International Congress on Modelling and Simulation: Managing Cumulative Risks through Model-Based Processes, MODSIM 2017 - Held jointly with the 25th National Conference of the Australian Society for Operations Research and the DST Group led Defence Operations Research Symposium, DORS 2017 - Hobart, Australia|
Duration: 3 Dec 2017 → 8 Dec 2017
|Conference||22nd International Congress on Modelling and Simulation: Managing Cumulative Risks through Model-Based Processes, MODSIM 2017 - Held jointly with the 25th National Conference of the Australian Society for Operations Research and the DST Group led Defence Operations Research Symposium, DORS 2017|
|Period||3/12/17 → 8/12/17|