Background

Microsimulation is a form of modelling that operates at an individual unit level, and then aggregates up to get results for higher levels, either geographic or demographic (eg results for couple families). By aggregating the individual-level data taken from surveys or administrative records, microsimulation modelling can identify the results of a tax policy change on incomes by family type, and which income groups are most affected by a policy change. For policy analysis, this provides a very powerful tool that cannot be replicated with any other type of model.

The method was first developed by Guy Orcutt (1957). This forward-looking article predicted models that incorporated individual behaviour, using inputs and outputs based on operating characteristics (equations, graphs or tables that determine outputs or the probabilities of possible outputs from the unit). The model was further developed by Orcutt (Orcutt et al, 1961) and others (Hoschka, 1986; Morrison, 1990; Harding and Polette, 1995).

Microsimulation models now have a firm place in the social sciences, being used for policy modelling (Bourguignon and Spadaro, 2006; Percival et al, 2007), population projections (Van Imhoff and Post, 1998; Booth, 2006), demographic modelling (Booth, 2006), small-area modelling (Ballas et al, 2005; Tanton et al, 2011) and data imputation, for example, when income estimates are not available (Figari et al, 2011).

Many countries now use microsimulation models for analysing tax/transfer policies and this is one of the main applications of microsimulation models. Euromod in Europe (Imervoll et al, 1999) and STINMOD in Australia (Percival et al, 2007) are examples of tax/transfer microsimulation models.

The primary advantage of microsimulation models over other models used in the economic and social sciences like Computable General Equilibrium (CGE) or cluster analysis (see Chapter Nine) is that because they operate at the individual level, they can be used to investigate distributional results, so distributions of income can be calculated, rather than just a summary measure like a mean. This was used to powerful effect by Tanton (2011), who was able to use a spatial microsimulation model to compare the frequency distribution of incomes in two suburbs in Sydney, Australia, to identify the reason for differences in poverty rates. This analysis pointed to a high proportion of older people on pensions in one suburb compared to another, leading to a higher poverty rate.