The analysis of behaviours of individuals or small groups of individuals has many applications in economic and social policy, in particular in study of impacts of alternative policy measures. Microsimulation models allow such microanalyses to be quantitative and empirical. Microsimulation may be viewed as an attempt to model and simulate whole distribution of policy target variables, not only their mean values. The microsimulation approach is thus primarily designed for studies of distributional effects of economic policy, and one of its main advantages is that it permits assumptions about heterogeneous behaviour. Troitzsch et al. (1996) describe social science microsimulation as the simulation of dynamic feedback (in both directions) between individual states and states of population as a whole or certain groups within a population. Microsimulation was introduced over four decades ago by Orcutt (1957), and has experienced something of a revival in social sciences over past decade (Merz, 1991; Clarke, 1996; Isard et al. 1998). It has been used to study various social phenomena such as population growth and development, effect of ageing and pension formulas on social insurance funding, and effect of various tax regimes on fiscal budgeting. The spatial dynamic microsimulation model (dubbed SVERIGE, or System for Visualising Economic and Regional Influences Governing Environment) built at Spatial Modelling Centre in Kiruna, Sweden, is unique (Vencatasawmy et al. 1999). It is first national-level spatial microsimulation model to be constructed, thereby permitting analysts to study spatial consequences of various national, regional and local-level public policies. Assisting model-building effort is an important database comprising longitudinal socio-economic information on every resident of Sweden for period 1985 to 1995. The locations of individuals in this database are given in co-ordinates accurate to 100 metres. This database may be augmented using data from other sources if required. It is therefore possible to estimate behavioural equations on various geographical scales and to describe complex dynamic spatial relationships. This paper is an investigation into what can be gained by replacing property value module of SVERIGE based on hedonic price regression equations with neural networks. It addresses design problems that have to be accounted for in both methods, and compares advantages and efficiencies of using one approach in favour of other.