Steps in Simulating Tax and Transfer ProgramsThe Urban Institute TRIM3 Reference Contact Us |
Despite the great variety of tax and transfer programs, the same basic steps are involved in simulating most programs. Those steps are:
The level of complexity of each step and the specific procedures involved vary across programs. However, the conceptual issues and problems are very similar across the simulation modules.
To understand filing units, it is helpful to understand the definitions of households and families used by the U.S. Bureau of the Census. A household simply consists of all the persons who occupy a housing unit, whether or not they are related. A family includes only related persons; there may be more than one family in a household. The family containing the household head is the householder family or primary family. A related subfamily includes a head or spouse who is related to the primary family; it consists of a married couple with or without children, or one parent with one or more never married children under age 18. An unrelated subfamily does not include the household head or any relatives of the household head.
A filing unit for a program may be as broad as the entire household, or as narrow as a single individual. Family-level filing units could be defined to include related subfamilies in the primary family, or to consider related subfamilies separately. Individuals may be included or excluded from a filing unit on the basis of age, family relationship, or other criteria.
Once a household has been divided into filing units, each filing unit is taken through the simulation steps described next.
Although some questions on a survey may use an annual reference period, household composition and characteristics are generally collected only for the point at which the survey is conducted. On March Current Population Survey (CPS) files, the members of a household and their ages and other characteristics are collected as of the March survey date. Those characteristics must be used to determine categorical eligibility for the prior calendar year. The approach taken by TRIM3 is to use the March characteristics as a "snapshot" of the population and to implicitly assume that they have been unchanged since January of the prior year.
One characteristic reported as of March is age. TRIM3 analysts have generally rejected the options of subtracting one year from each age, or having age vary month by month. To maintain the appropriate age distribution with either of those options, infants would have to disappear from the population, and the recently deceased would have to reappear. To avoid these complexities, TRIM3 typically treats a person's age in March as his or her age for the entire calendar year.
Disability status is used in determining categorical eligibility for SSI, Medicaid, and the elderly and disabled tax credit that may be applied against federal income taxes, but it is not well reported on the March CPS files. The only disability information on the March CPS is indirect. Persons who are not working in the survey week, or who did not work in the prior calendar year, are asked why they did not work; one possible response is "illness or disability." Use of the labor force data may be supplemented by examining whether a potentially disabled person reported receiving a type of income associated with disability. Because neither labor force information nor income data is reported for children, no disabled children can be identified from the March CPS.
Citizenship status affects categorical eligibility for many programs. Illegal aliens are ineligible for assistance under most programs. Depending on the program, other non-citizens may only be allowed categorical eligibility if they meet certain requirements, such as those related to age, disability status, immigrant/refugee status, work history, and length of time in the United States. Since there is only limited citizenship information on the March CPS, Urban Institute demographers and TRIM3 staff have developed sophisticated methodologies to assign immigrant and citizenship status to individuals for most years of the TRIM3 input data. Other personal characteristics used in some categorical eligibility definitions, such as whether or not a woman is pregnant, are not reported at all on the March CPS files. The designer of a simulation module must compare the actual categorical eligibility definition used by a program to the available data to arrive at the operational definition to be used in the simulation.
The March CPS files include no data on asset holdings of any kind. The amount of financial assets can be imputed from asset income: interest, dividends, and rental income. Asset income is converted to asset holdings by assuming that all assets generate income and that all assets have the same rate of return. That rate of return is a program rule in each of the TRIM3 modules performing the assets imputation. Nonfinancial assets tests, such as limits on the value of an automobile, are not simulated in TRIM3.
The exact definition of income varies by program. Most simulation modules include an input variable list program rule that lists the types of income counted by that program. Some programs use different income concepts for different purposes. For example, the Food Stamp Program computes gross income by adding up the earned and unearned income variables specified in input variable list program rules, but then takes deductions from gross income to compute net income.
Three problems arise in determining income for simulation purposes. First, income may be unavailable for some persons. On the March CPS files, income is unavailable for persons under 15 years o1d. Although children under age 15 are very unlikely to have earnings, they may have unearned income, such as Social Security survivor's benefits or SSI payments. Presumably, those dollar amounts are reported on the record of an adult in the household. Second, income may not be reported in sufficient detail for purposes of a particular program. For example, the dollar amount reported as "other" income may include a number of items, some of which are included in a program's income definition and others of which are not. Third, some kinds of income may not be reported in any form. For example, there are no data on capital gains income on the March CPS files. TRIM3 must statistically match data on capital gains income amounts onto the March CPS in order to run the federal income tax simulation.
When a program uses an income concept in which deductions are taken from gross income, additional problems arise. Deductions are generally based on expenditures. For example, home mortgage interest expenses are deductible from income for federal tax purposes, and child-care expenses are a deduction from TANF income. The March CPS does not include expenditure data. Therefore, TRIM3 must impute the expense information needed by a particular simulation. Child care expenses are imputed by the Child Care Expenses module, housing expenses are imputed by the Public and Subsidized Housing module, and expenditures that may be used as deductions in the Federal Income Tax module are statistically matched to the March CPS using data from the Internal Revenue Service's Statistics of Income Public Use File (SOI). Simple imputations of additional expenses are included in some of the modules.
A transfer program may include more than one income test. The Food Stamp Program includes both a gross income test, comparing gross income to a gross income cutoff, and a net income test, comparing income after certain deductions to a lower cutoff. The income cutoffs used in eligibility tests are entered through TRIM3 program rules.
A tax program may include an explicit tax-filing test based on income. For example, self-employed persons are not required to pay Social Security self-employment tax unless their earnings are over a certain cutoff. The cutoffs used for income-based tests in tax programs are also entered through program rules.
Technically, transfer programs may use "retrospective budgeting," in which this month's benefit is determined based on last month's income. TRIM3 bases a particular month's benefit on that same month's income. Several of the simulated transfer programs have the ability to perform "income smoothing"-in which the variability in month-to-month earnings is reduced by calculating an average monthly paycheck rather than by adding up weekly or bi-weekly paychecks to determine the actual amount of income received in each month.
In each transfer program module, participating filing units are chosen from among those units found to be eligible by the simulation. During each year's baseline simulation, a module's participants are selected in such a way as to meet targets for that year. Targets used in various modules include the overall number of program recipients, different types of participating units, and the numbers of units in different states. When an alternative law is simulated, the number and type of participants will differ owing to changes in the population of eligible units.
The specific methods used to determine participation vary across the transfer program simulations. In general, each eligible unit is given a probability of participation based on its characteristics. The probability may be obtained from a look-up table of probabilities based on only a few characteristics or may be generated by a complex multivariate model. A random number between 0 and 1 is compared to the probability, and if the random number is less than the probability, the unit will participate. Because TRIM3 uses the same random number for each run, the same participation results are produced whenever the same program rules are applied to the same input file.
The simulations can use participation information reported on the March CPS, but cannot rely exclusively on that information. In some modules, the fact that an eligible unit reported receiving the benefit gives that unit automatic participation in the simulation. However, additional participants must be chosen to achieve program targets. Some units that report participating in a particular program are found ineligible by the simulation. Possible reasons for the presence of "ineligible reporters" include reporting errors, coding errors, and the mismatch between March demographic data and annual income data. For example, a single woman with no children in March might have received TANF benefits in the prior year because she had a teenage child living with her at that time. TRIM3 never assigns participation to ineligible reporters.