Microsimulation

A microsimulation model differs from other types of models in that it operates on individual units rather than on aggregate information. In the social sciences, those units are individual economic units. The database used as input to a microsimulation model contains records describing persons, households, or businesses. The simulation model applies a set of rules to each individual record. The result of the computations might be the amount of taxes owed by the unit or the dollar amount of benefits to which the unit is entitled under a government program. Each individual result is multiplied by whatever weight is associated with the unit in the microdata file. The weighted individual results are then added together to obtain the aggregate result.

As an example, consider the federal income tax simulation in TRIM3. The computer program applies the detailed federal tax rules to each family on the microdata file. In effect, the program fills out the tax forms for each family. TRIM3 counts dependents, adds up income, subtracts adjustments to income, subtracts the larger of itemized or standard deductions, subtracts personal exemptions, computes taxes on available income, and computes and subtracts tax credits to arrive at the final tax liability. To obtain aggregate tax liability, each family's tax liability is multiplied by its weight and then added to obtain the total. If a family's tax liability is $5,000 and its weight on the microdata file indicates that it represents 1000 families, $5 million is added to the aggregate simulated tax liability.

The TRIM3 federal income tax simulation can be contrasted with other sorts of tax models. In a macroeconomic model, for instance, total federal income tax liability might be estimated as a function of the tax rates and brackets, and of aggregate estimates of income and deduction amounts. A cell-based model would divide observations of individual tax units into a matrix of cells, the dimensions of which might include type of return, level of adjusted gross income, and so on. Averages would then be computed within each cell for other relevant variables. Taxes would be estimated by applying the tax rules to the cell averages and multiplying the result by the weighted number of units in the cell. Because cell-based models operate on cell averages rather than individual economic units, they are not pure microsimulation models, but they do offer a less aggregated unit of observation than macro models, and thus provide more group detail.

Microsimulation models can be divided into two broad classes-"static" and "dynamic." The terms refer primarily to the methods used to "age" a microdata file to create a synthetic file for a future year. TRIM3 is an example of a "static" microsimulation model-best suited for performing detailed simulations of the past, the present, and the near future. Dynamic microsimulation models are capable of performing simulations into the distant future, but typically do not capture as much detail as static models.

Microsimulation models such as TRIM3 offer key advantages relative to other types of models. By operating at the unit level, microsimulation models are able to capture interactions between simulated programs. For instance, the model can capture the fact that a change in SSI benefits would also have impacts on SNAP benefits, since SSI income is used in determining SNAP eligibility and benefits. Further, the model can produce tabulations of results by a wide variety of socio-economic characteristics (subject to the sample size constraints of the input data). Since results are calculated according to the actual rules of the simulated program, a virtually unlimited number of policy scenarios can be modeled simply by changing the rules of the simulated program. Additional advantages of TRIM3 are its modularity, parameterization, historical continuity, and web-interface access. TRIM3 system configuration and technical policies provide a secure and stable platform for simulation work.