A microsimulation model is a tool used to evaluate changes in a hypothetical system that replicates the 'real world'. Microsimulation models can be designed to describe the impact of an interest rate rise on unemployment, predict the size of India's population in 2030, evaulate if it worth converting a highway intersection into a cloverleaf or guess what the weather will be like tomorrow.

Microsimulation models originated in the United States in the late 1950s, but the limited performance of computers at the time has meant that only in the 1990s would models be widely available. They are now increasingly being used in most OECD countries; indeed in the United States Congress will not consider any social security or tax legislation without closely examining the distributional outcomes predicted by microsimulation models.

A model is developed with many dependant and independant variables with algorithms that define their relationships. Within a model on a data file will be thousands of units representing independent actors with different characteristics, the number being proportionate to the real population (1.1 billion Indian citizens could be summarised down to 1000 units, with perhaps half being female). In modelling population sub-groups by age, gender, occupation, ethnicity, incomes etc one could end up with thousands of categories, requiring extensive computing resources. Then the analyst may fiddle around with some variables to replicate a change in circumstances, run the program and check the output to see the effects. They may look at what how each sub-group fares as a result of the change; policy analysts might then see who are the losers and winners of a new tax or welfare benefit.

Static or dynamic models give different results. A static model only looks at the instant effects of a change in circumstances. When the model on India's population runs, these 1000 pseudo-Indians will give birth, immigrate, emigrate and die at the same rate as their real life equivalents. Those mortgage calculators you see on home loan websites are essentially static models. Dynamic models consider how variables may change and compound trends over time. If the model predicts that more Indian women will be university graduates in 2020, then the model might predict a baby bust (assuming a relationship exists between education and fertility). Dynamic models are considerably more complicated and rely on more variables to be effective.

Some notable socio-economic microsimulation models that replicate taxation and transfer systems include EUROMOD (developed by Cambridge University, to look at the European Union), CORSIM (developed by Cornel University to look at the US), STINMOD (developed by the University of Canberra to look at Australia), DYNACAN (developed by Statistics Canada to look at Canada) and PENSIM (developed by UK's Department of Social Security to look at Britain)