Improving Information for Social Policy Decisions -- The Uses of Microsimulation Modeling: Volume II, Technical Papers (1991)
Chapter: PARTICIPATION FUNCTIONS FOR SSI, AFDC, AND FOOD STAMPS IN TRIM2, MATH, AND HITSM
of simulations with all three models; compare the patterns of employment and income variability produced by TRIM2, MATH, and HITSM; and relate differing patterns to differences in model outputs.
However, such comparisons are hampered by other differences in the models. For example, correction of income for underreporting is not required in MATH or TRIM2, but it is a feature of HITSM that apparently cannot be turned off and that precedes the annual to monthly conversion. There are also differences in the databases used by the models. For example, the TRIM2 modelers each year produce an updated baseline file from the latest March CPS, while typical practice for the MATH modelers has been to develop, every 2 or 3 years, a new baseline file by aging a March CPS forward several years.
Differences in model components and databases obviously impede valid cross-model evaluations of particular routines. Moreover, the costs of modifying current models and databases for enhanced comparability are likely to be high, as are the costs of performing an evaluation within one model by programming alternative routines based on other models. Nevertheless, evaluations of the effects of particular model components are important to pursue, as is the development of microsimulation models designed to facilitate validation exercises.
PARTICIPATION FUNCTIONS FOR SSI, AFDC, AND FOOD STAMPS IN TRIM2, MATH, AND HITSM
A critical aspect of the simulation of means-tested transfer programs such as AFDC, SSI, and food stamps in microsimulation models involves the participation decision. Not all families or individuals eligible for public assistance benefits choose to apply for them, whether because they are ignorant of the programs, believe that participation would be unacceptably stigmatizing, or calculate that the benefits are not high enough to offset the hassles and costs of application. The end result is that an estimated 40 to 50 percent of individuals and couples eligible for SSI do not participate in the program (Giannarelli, 1989b), while an estimated 20 to 25 percent of families eligible for AFDC do not participate (Giannarelli, 1989b; see also Ruggles and Michel, 1987), and an estimated 40 percent of households eligible for food stamps do not participate (Doyle, 1990).
The procedures used by TRIM2, MATH, and HITSM to simulate participation in major income support programs represent one of the few behavioral elements in these cross-sectional static models as they are currently implemented. Generally, participation probabilities are determined by each eligible unit’s expected benefit level, as economic theory would predict, and also by the unit’s demographic characteristics. Often, but not always, participation probabilities take into account whether the eligible unit reported participation in the March CPS.
