Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
EVALUATIONS OF MICROSIMULATION MODELS: LITERATURE REVIEW 267 KORMENDI AND MEGUIRE (1988) Kormendi and Meguire (1988) examined TRIM2's performance in estimating the number of households that participate in some form of welfare assistance program and the benefits they receive. These authors took two different approaches to their validation of TRIM2. The first was to randomly perturb parameters in TRIM2's participation module, according to reasonable estimates of the variability of the estimated parameters; then they examined the resulting variability in the estimates of participation and level of benefits. The participation module of TRIM2 uses a probit model with subjectively determined intercept dummies. Kormendi and Meguire multiplied each parameter of the probit model, including the coefficients of the dummy variables, by a random variable, uniformly distributed on the interval (0.8, 1.2). They characterized this as a conservative estimate of the true variability (an assessment they leave unsupported). This was repeated 15 times, running TRIM2 based on the 1985 CPS. They computed an approximation to the coefficient of variation for several outputs, including total AFDC units and benefits, AFDC units and benefits for households with and without earnings, and AFDC units and benefits for households with dependents between 0 and 2 years of age. The investigation was limited to 15 trials because of the high cost of running the simulations. The results of this first part of the TRIM2 validation indicated that the variability of simulated benefits was less than that of units. In nearly all cases the coefficients of variation of simulated benefits were less than 3 percent. However, the coefficients of variation of simulated units ranged from 3.6 to 18.2 percent. It should be pointed out that even in the case of units the variability of the results generally was somewhat less than the variability of the simulated parameters. In the second method, which Kormendi and Meguire call dynamic validation, TRIM2 was used to simulate changes in welfare participation and benefits from 1979 to 1985 and from 1983 to 1985, and the results were compared with those from the 1985 CPS. The first time period spanned the enactment of the Omnibus Budget Reconciliation Act (OBRA) in 1981. With 1979 as a baseline, TRIM2 was used to forecast units and benefits for 1985, making use of the actual changes in the law that occurred, which included OBRA. The second time interval represented a period when no major legislation was enacted. 5The ICF, Inc., comparison of HITSM and TRIM2 was carried out in the context of a potential contract competition between the two organizations. Michel, Webb, and Zedlewski (1987) critique the ICF study, noting that it was limited to aggregate comparisons and did not look at differences in distributions. They also note that the ICF study did not take account of errors in the administrative data used as the standard of truth.