SPATIALMSM: The Australian Spatial Microsimulation Model
In recent years, the National Centre for Social and Economic Modelling at the University of Canberra has been developing a method of calculating estimates for small areas using survey data. This methodology has now been linked to our microsimulation model of the Tax/Transfer system STINMOD, allowing the regional impact of policy changes to be estimated. The SPATIALMSM model reweights the survey data to small area known reliable totals (benchmarks) from a Census. There are a number of benchmarks used, and these depend on the survey variable being estimated. For estimating poverty and housing stress, we use labour force status, age, sex, income, rent paid, mortgage paid, household type, Household size (number of people per household), dwelling tenure, dwelling structure, Non Private Dwelling, and total number of households. All these variables are correlated with poverty and housing stress, so help to provide the best estimates of these two variables. The weights for each small area are then applied to variables from the survey (equivalised disposable income for both poverty and housing stress, and housing costs for housing stress), and small area estimates for these variables can be calculated. The accuracy of these small area estimates depends on the correlation between the variable being estimated and the benchmarks used, so the benchmarks must be chosen with the final variable to be estimated in mind. This method is now being used in two grants to calculate small area estimates of needs based planning indicators (for instance, number of people aged 65 and over; number of people receiving rent assistance), and for estimating variables affecting children (for instance, children in households in poverty; children in overcrowded households). In future, we expect to derive small area estimates of variables affecting older people.