Development and Annual Updates
Each year, TRIM3 is updated by converting the new March Current Population Survey (CPS) to TRIM3 format and performing baseline simulations of the major tax and transfer programs. This may require modifying computer code to capture new program rules. Periodic revisions are made to participation functions and imputation functions. New modules may be added to the system in order to simulate new transfer programs or to address new areas of interest.
Annual Updates
The goal of the annual update is to produce a set of baseline simulations on the most recently available input file. This involves converting a new survey file to TRIM3 format, changing the TRIM3 program rules or computer code to capture the rules in the year of that survey, and running and aligning baseline simulations.Conversion
Each year, the new March CPS file is converted to TRIM3 format. The raw March file is generally available in the fall of the survey year. (For example, the March 2002 survey was available in the fall of 2001.) The conversion process is more or less complex depending on how much the survey has been changed since the prior year's survey was converted. For example, if a variable used in TRIM3 has been coded in a different way on the new file, the conversion programs must recreate the old coding to avoid making multiple changes in the computer code that simulates the tax and transfer programs.Program Rule Revision and Module Updating
As the raw survey file is being converted, the model is prepared to simulate the program laws in effect in the calendar year represented by the survey. For example, the model is prepared to simulate 2001 program rules on the March 2002 CPS, since that file includes income data for 2001. In some years, all program changes can be captured via TRIM3 program rules. The new program rule values are stored as part of the baseline simulation for that year, and do not "overwrite" any existing program rule values for prior years. Often, a change in the law cannot be simulated with the existing TRIM3 program rules. In such cases, changes are made in the computer code, and new TRIM3 program rules may be added.Alignment and Validation of Baseline Simulations
The baseline simulations are run when the new TRIM3 microdata file is created, the baseline program rules are entered into the simulation, and any necessary changes have been made to the code. The cash transfer programs are simulated first, so that the in-kind transfer programs can use the simulated cash transfer payments in determining program eligibility. The tax modules are generally run after the transfer program modules, to allow the use of simulated transfer program income in imputing unavailable tax-related variables. In general, when a simulation uses as input a variable that is both reported and simulated, the simulated amount is used if it has already been created.Most of the transfer program modules are "aligned" to target data. By "alignment", we mean that adjustments are made that bring simulated aggregate results closer to actual program data for the year being simulated. Although the exact alignment procedures vary by module, the same general principles guide alignment of all the modules. Simulated participation decisions and imputations of variables not available on the raw survey are adjusted as necessary to reach targets. But variables reported on the original survey are not altered as a way to reach program targets, and no changes are made to the results of mathematical tax and benefit calculations.
Reported variables are not changed as a way to reach program targets, for three reasons. First, it may be unclear how to perform such an adjustment. An underestimate of Social Security taxes might indicate underreporting of wages on the original survey, but it would not necessarily be clear what types of persons misreported their wages by what percentage. Second, if variable modification were used as a way to reach program targets, one type of change might be necessary to align the TANF module while a change in the opposite direction might be required to align the Food Stamp module. Third, even when discrepancies between simulated and actual figures are apparently due to reported variables, it is not necessarily the case that the reported variables are incorrect. For example, if reported self-employment figures produce too many persons paying self-employment tax, the discrepancy could be due to nonpayment of taxes that are owed as well as to misreported data on the CPS.
No changes are made to the results of mathematical tax and benefit calculations. Such changes would create logical inconsistency in the information about a particular person, family, or household. For example, if aggregate TANF benefits fell short of the target by 3 percent, it might be tempting to increase every unit's benefit by the same percentage to reach the target. But an analyst examining all the variables for a particular TANF family would then see a TANF benefit that was too high for that unit's size, income, and state of residence.
The baseline simulation of each transfer program is aligned so that the number of simulated program participants matches actual program participants for that year, along whatever dimensions are considered important by program analysts. Each transfer program simulation includes some sort of participation function that assigns each eligible unit a probability of participation. A random number is compared to that probability to determine if the unit will participate. The alignment of the baseline simulation to the targets involves adjustments to the participation function or to the resulting probabilities. These adjustments are entered through TRIM3 program rules and are used in any alternative simulations performed on the same microdata input file. Conceptually, the participation function is modified to produce the correct results for the current year. The function is then held constant across alternative simulations.
As each program module is run, its results are validated. Simulated data are compared to any actual data that are available. The baseline simulations are examined closely because they will he used throughout the year. Each alternative law simulation run on the input file will be compared against the baseline run to assess the effects of the change.
Periodic Changes and Additions to TRIM3
Periodically, a variety of other developments and improvements are made to the TRIM3 system. The imputation functions used to assign values to missing variables and the participation functions used in the transfer program modules are updated as time and resources permit. Program rules that were not included when a module was first specified may be added to the simulation at a later date. If a series of legislative proposals focuses on the same issue, changes may be made to facilitate the simulation of those proposals. Periodic improvements are made to increase the efficiency of TRIM3 from a programming or computing standpoint.Occasionally, new modules are added to the system, or existing modules are greatly expanded. During the late 1990s, much of the TRIM3 development work was in response to the sweeping changes introduced by the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA). PRWORA replaced the Aid to Families with Dependent Children (AFDC) program with the Temporary Assistance for Needy Families (TANF) block grant, created a Child Care Development Fund (CCDF) to provide child care subsidies to low-income families, and restricted non-citizens' eligibility for assistance under many programs. To capture these changes, the TRIM3 AFDC module was extended to handle the detailed rules of state TANF programs, a Child Care Expense module was created in order to simulate eligibility for CCDF subsidies, and changes were made to all of the major transfer program modules in order to handle the new rules related to citizenship status. Also during this time period, the Medicaid module was expanded to allow simulation of the State Children's Health Insurance Program (SCHIP), which was enacted by the Balanced Budget Act of 1997.
TRIM3 ©2012