Population microdata (survey data) comprise a list of individuals, normally nested into families and households, with each individual having an associated set of personal demographic and socio-economic characteristics.  However, to protect respondent confidentiality the spatial coding attached to publicly available population microdata is often severely degraded, leaving tabular data (aggregate counts) as the main source of spatially detailed information. The creation of 'synthetic small-area microdata' represents one approach to combining these tabular and survey data to provide estimate of unknown local area distributions, particularly unknown multivariate local area distributions.

There are two main alternative approaches to creating synthetic small-area microdata: synthetic reconstruction and combinatorial optimisation.  A thorough assessment of the statistical reliability of both approaches has been undertaken (see Huang and Williamson, 2001 ).  This evaluation identified Combinatorial Optimisation (CO) as the preferred approach.