Issues in spatial microsimulation estimation: a case study of child poverty

Spatial microsimulation techniques have become an increasingly popular way to fulfil the need for generating small area data estimates. Nevertheless, this technique poses numerous methodological challenges, including those that relate to fundamental differences between the multiple data sources which spatial microsimulation techniques seek to combine. Using two different databases simultaneously to produce estimates of population characteristics may come up against problems related to different distributions of key variables within the two databases. Such differences can make it difficult to adequately validate small area estimates, as it can be hard to assess whether differences between synthetic and original data are due to failures or inaccuracies within the estimation procedure, or simply to the differences within the underlying data. This study presents a case study of this problem using a very important small area estimate – child poverty rates. We compare how income distributions for children are different in two Australian databases being combined within a spatial microsimulation model. We then assess the extent to which this affects our estimates of child poverty, and gauge its impact on the apparent validity of these synthetic small area poverty rates.