In CGE models exogenously-specified elasticity parameters describing the behaviour of consumers, producers and trade patterns calibrate the model equations to analyse a wide array of socio-economic issues. Since CGE models give no insight into how aggregate changes in the economy affect individuals, a tendency of linking microsimulation and CGE models has emerged in recent studies to highlight the distributional effects of policy simulations by using the complementary advantages of both approaches. Therefore, it is desirable to represent disaggregated households in a CGE model to evaluate the distributional effects of policy simulations. We estimate the behavioural parameters essential for the calibration of the demand side of a CGE model for 5 household types in Australia, grouped according to income. We utilise the 2009-10 Household Expenditure Survey (HES) to estimate the LES parameters and expenditure elasticities for Australian households. The Seemingly Unrelated Regression Equations (SURE) model is used to solve a 21-good demand system simultaneously. It is worth noting that the econometric estimation is done in a model whose structure is consistent with the CGE model. A comparison of the results for household income quintiles suggests that except for the highest quintile the estimated subsistence consumption parameters for some commodities such as Agricultural Products, Processed Food, Electricity, Gas and Water are relatively higher for lower income quintiles. The estimated expenditure elasticities for the 21 aggregated commodities indicate that Agriculture, Forestry and Fishing, Electricity, Gas, Water and Waste, Transport and Postal Services, Administrative Services and Other Services are necessities for all income groups. Results also reveal that most of the remaining commodities are luxuries for lower income quintiles, while they appear to be necessities for the wealthier households.