Exploring Microsimulation methodologies for the estimation of household attributes

Microsimulation is a rapidly expanding area of spatial modelling, which seems to offer great potential for applied policy analysis. However, currently there is considerable debate on the most appropriate methodology for estimating micro-data. Household or individual attribute data can be represented both as lists and/or as tabulations. It has long been argued (Birkin and Clarke, 1995; Clarke, 1996; Williamson et al, 1998) that the representation of information on households and individuals in the form of lists offers greater efficiency of storage and spatial flexibility as well as an ability to update and forecast. This paper reviews the possibilities and methodologies of building list-based population micro-data for small areas. First, it evaluates the methods, which have been developed and employed so far for the estimation of population micro-data, outlining the advantages and drawbacks of each one of them. Then the paper investigates the comparison of methods for generating conditional probabilities by statistical matching techniques or by using probabilities directly from household data sets such as the Samples of Anonymised Records (SARs) and the Small Area Statistics (SAS) from the UK Census of population. In addition, it explores the combination of these methods in a microsimulation framework and presents micro-data outputs from a local labour market microsimulation model for Leeds. Finally, it highlights the difficulties of calibrating this kind of model and of validating the model outputs, given the absence of suitable observed statistics.