Cost Avoidance v 8.2 Version History

The CostAvoidance module tabulates child support "cost avoidance"-total savings to means-tested transfer programs resulting from receipt of child support income, and calculates the cost avoidance ratio-average savings per dollar of child support received. The model imputes whether child support income is received through the federal/state IV-D program or is paid privately. Results are displayed by IV-D status, income level relative to the poverty threshold, and TANF status.

Since child support income is underreported on the CPS, the initial cost avoidance estimates are likely to be too low. However, simple out-of-model adjustments can be made to generate IV-D cost avoidance estimates that reflect the actual level of child support distributed by the IV-D program, as reported by the federal Office of Child Support Enforcement (OCSE). Because there is no external target for the total amount of child support paid through private transactions, no similar adjustment is possible for privately paid child support.

Additional information about the CostAvoidance module is presented in the following sections:

1. Sources of Cost Avoidance
2. Data Set Preparation
3. Monthly Accounting Period
4. Filing Unit
5. Change in Transfer Benefits
6. The Cost Avoidance Ratio
7. IV-D Status
8. Adjusting Results Using OCSE Financial Data
9. References

Sources of Cost Avoidance

The CostAvoidance module tabulates cost avoidance for the following means-tested transfer programs: SSI, TANF, public and subsidized housing, SNAP, and Medicaid. TRIM3 simulates eligibility for SCHIP and assistance under the Child Care Development Fund. However, at the time the cost avoidance module was developed, participation functions for these programs were not yet in place, and so they are not included in the cost avoidance estimates. Further information about cost avoidance in each simulated program is provided below.

SSI

The SSI program treats child support income as the income of the child, and so affects the eligibility and benefits of disabled children, but not of disabled or aged adults. Thirty-three percent of the child's child support income is disregarded in determining eligibility and the benefit amount.

Cost avoidance to the SSI program occurs when child support income lowers the benefit amount for a participating child, renders an otherwise eligible child ineligible, or is sufficient to cause an eligible child to decide not to participate. TRIM3 captures the first source of SSI cost avoidance-a lower benefit resulting from receipt of child support income. Due to limited disability information for children on the CPS, TRIM3 captures little of the cost avoidance arising from changes in eligibility and participation.

TANF

The CostAvoidance module displays results separately for TANF and non-TANF families. Results are entered into the "baseline TANF" row for each month in which the family receives TANF in the baseline simulation. Otherwise, the results are entered into the relevant "non-baseline TANF" row corresponding to the family's poverty level.

For families not receiving TANF in the baseline simulation, TANF cost avoidance occurs when child support renders a family ineligible for assistance or makes an eligible family decide not to participate. TRIM3 captures both of these sources of cost avoidance.

Child support collections made on behalf of TANF recipients are often referred to as "cost recovery" rather than "cost avoidance." Families are required to assign their child support to the government in order to receive TANF benefits. The government then retains child support collected on behalf of the family in order to help offset the family's TANF benefits. Some states "pass through" a portion of the child support and/or use child support to help "fill the gap" between the family's benefit and the state need standard.

TRIM3 approximates the treatment of child support by the TANF program by counting child support as income and treating the "pass through" as a disregard. This lowers the family's TANF benefit such that the sum of the family's TANF benefit and child support income (for most cases) is equal to the sum of the TANF benefit and child support pass-through that a family would receive under "real world" rules.

The TANF cost avoidance results displayed in the "baseline TANF" row should be interpreted with caution. TANF families are likely to be unaware of the total amount of child support collected on their behalf, and many may just report the pass-through amount. Thus, total TANF savings from child support paid to TANF families are likely to be understated. For information about the extent of TANF cost recovery, it is best to refer to OCSE published reports.

Cost avoidance estimates for the other programs in the "baseline TANF" row should also be interpreted with caution. Cost avoidance ratios are calculated using total child support income as the denominator. However, for some purposes, it is more appropriate to use a denominator representing the actual child support paid to families (i.e., one that excludes child support retained by the government to offset the family's welfare expenditures). A special summary table (Basic Table 7) provides cost avoidance ratios calculated under this definition.

Depending on the number of states with a pass-through, the sample size in the "baseline TANF" row may be insufficient to produce reliable estimates. If so, a strategy for developing a rough estimate of cost avoidance ratios for baseline TANF families is to do a special simulation where all states have at least a $50 pass-through, and to then perform a set of public and subsidized housing, SNAP, and Medicaid baselines using the special TANF simulation as input. This set of runs can then be substituted for the "baseline" runs in the cost avoidance simulation, and cost avoidance ratios for baseline TANF families can be obtained.

Housing

TRIM3 simulates the value of rental subsidies for CPS-reported public housing tenants and subsidized renters. This includes households receiving either project-based assistance (public housing and Section 8 project-based) or household-based assistance (Section 8 certificates and vouchers). Cost avoidance arises when child support income increases the household's rental payments-which are typically set at the larger of 10 percent of gross income or 30 percent of net income.

TRIM3 does not capture cost avoidance that arises when child support renders a family ineligible for housing assistance. However, since housing assistance is not an entitlement and is often of limited availability, this is not a key source of cost avoidance.

SNAP

Cost avoidance to SNAP occurs when child support lowers the amount of benefit a participating household receives, renders an otherwise eligible household ineligible, or is sufficient to cause an eligible household to decide not to participate. TRIM3 captures each of these sources of cost avoidance.

Cost avoidance is offset to a certain extent because households that pay child support are able to deduct the child support payment from income when calculating the SNAP benefit amount. TRIM3 does not capture this effect. Analysts wishing to account for the effect of the child support deduction can develop out-of-model adjustments using data from the SNAP Quality Control File.

Medicaid

The child support system provides two sources of cost avoidance to the Medicaid program. Health insurance and/or medical support payments provided by the nonresident parent may eliminate the need for Medicaid coverage or reduce the costs of providing coverage. Child support income may render a person financially ineligible for Medicaid assistance, or may be sufficient to cause the person to decide not to enroll.

TRIM3 captures cost avoidance that arises when child support renders a person ineligible for assistance. Cost avoidance from enrollment decisions is captured to a certain extent. If child support is sufficient to keep a person from receiving SSI or TANF, and as a result, the person chooses not to enroll in Medicaid, then the resulting Medicaid savings are counted as cost avoidance. TRIM3 does not capture the cost avoidance that arises when health insurance and/or medical support payments provided by the nonresident parent eliminate the need for Medicaid coverage or reduce the costs of providing coverage.

The CostAvoidance module was developed at a time when the Medicaid module simulated eligibility for SCHIP, but not enrollment. As a result, the CostAvoidance module counts "savings" to the Medicaid program that occur when a person shifts from SCHIP eligibility to Medicaid enrollment. If, in reality, the person was enrolled in SCHIP, then it is not clear that this represents savings to the government. At least one prior analysis has used a SAS program to "back out" these savings (Wheaton 2002). When resources permit, an option to "back out" these savings may be added to the CostAvoidance module itself.

Data Set Preparation

The CostAvoidance module requires two sets of simulations for input: the "baseline" simulation, showing transfer program benefits calculated under the rules in effect in the year being simulated, and the "alternative" simulation, showing transfer program benefits under a scenario in which no child support is paid. The CostAvoidance module also requires input from the Poverty and ChildSupport modules. Additional details are provided below.

ChildSupport Module

The IV-D imputation requires certain variables created by the ChildSupport module. These include the amount of child support due (for custodial mothers) and variables identifying custodial mothers, custodial fathers, and children with a nonresident parent.

The ChildSupport module also produces the variable MonthlyCSIncomeOfChild, in which child support reported on a parent or other adult's record is allocated to the appropriate children in the household. The SSI program treats child support as the income of the child, and the TANF program does not count child support received on behalf of children outside the TANF unit (such as an SSI child). By assigning child support to the child's record, these provisions of SSI and TANF can be captured. A corresponding variable-ChildSupportRcvdForHhldKids is created for persons aged 15 and older. For households containing children, ChildSupportRcvdForHhldKids is equal to the child support income reported by the person on the CPS. If a household does not contain children, then ChildSupportRcvdForHhldKids is set to 0 for all persons in the household.

At the time of writing, MonthlyCSIncomeOfChild and the associated SSI and TANF provisions have not yet been incorporated into the standard versions of the TRIM3 baselines. Should this still be the case upon the next use of the CostAvoidance module, users are advised to create their own version of the baseline simulation that uses MonthlyCSIncomeOfChild as the child support input variable for the transfer program simulations, and ChildSupportRcvdForHhldKids as the child support input variable for the CostAvoidance simulation.

It is important that ChildSupportRcvdForHhldKids be used as input to the CostAvoidance module whenever MonthlyCSIncomeOfChild is used in the transfer program simulations. The persons whose reported child support is "zeroed out" in ChildSupportRcvdForHhldKids are predominantly middle-aged women who most likely had a child present in the household during the prior calendar year who is no longer present at the time of the CPS survey. Since there is no child in the household to whom child support can be assigned, no program savings are incurred when MonthlyCSIncomeOfChild is used in the baseline simulation. To be consistent (and to avoid underestimating the cost avoidance ratio) ChildSupportRcvdForHhldKids should be used as the input variable to the CostAvoidance simulation.

Transfer Program Simulations

The "baseline" simulation can either be the official baseline for the simulated year, or a user-specified "baseline" that better captures key provisions related to child support. For example, if the MonthlyCSIncomeOfChild and associated SSI and TANF provisions described above have not been incorporated into the official "baseline" for the year, the user is advised to create their own "baseline" to capture these provisions.

The "alternative" simulation should be exactly the same as the "baseline" simulation, except that child support income is "zeroed out" and the other program rules are set as appropriate for an "alternative" simulation. In addition, if a particular module uses input variables from another program for which there is an "alternative" simulation, then the inputs should be taken from the "alternative" rather than the "baseline" simulation. See the documentation for each module for information about how to specify program rules in an "alternative" simulation.

Poverty Module

The variable StandardPovertyRatio, created by the Poverty module, is used to define table rows by poverty status and to assign IV-D status to custodial fathers. The Poverty module simulation should use ChildSupportRcvdForHhldKids and the "baseline" SSI and TANF variables in the income definition. The simulation should be performed using version 4_1 or later, since CostAvoidance relies on the family's poverty ratio being stored on each person's record.

Monthly Accounting Period

The CostAvoidance module tabulates child support and transfer program savings on a month-by-month basis. A family's child support and transfer program savings are added into the "baseline TANF" row of the CostAvoidance summary tables for each month that the family receives TANF in the baseline simulation.

The IV-D imputation is performed on an annual basis. A family's IV-D status is assumed to remain constant throughout the months of the year.

Filing Unit

Transfer program savings are tabulated at the household level and are then allocated to families in proportion to their relative levels of child support income. "Family" is defined as family with related subfamilies separate.

Change in Transfer Benefits

CostAvoidance calculates the change in benefits for each transfer program by subtracting the benefits received in the baseline simulation in a given month from those received in the alternative simulation. For Medicaid, CostAvoidance can either use the value of coverage simulated in the baseline and alternative Medicaid simulations, or use an alternative approach to estimating the value of coverage.

Depending on the program, a change in child support income for one family may affect the benefits received by other families in the household. Therefore, CostAvoidance calculates the change in transfer program benefits for the entire household, and then allocates that change to the various families in the household in proportion to their relative levels of child support.

Further details about the methods for calculating Medicaid cost avoidance and allocating the change in transfer program benefits to families in the household are provided below.

Medicaid

Two methods are available for estimating the value of Medicaid coverage. The first is to use the simulated values of Medicaid benefits (MedicaidValueOption = 0). The second is to use values derived from targets in the baseline Medicaid run (MedicaidValueOption = 1). Both options were implemented at a time when the Medicaid module made a single annual enrollment decision (either the person enrolled in all or none of the months in which he or she was eligible). The code will need to be reviewed and revised in order to accurately compute Medicaid cost avoidance for future simulations in which the Medicaid enrollment decision is made on a monthly basis.

As mentioned previously, the CostAvoidance module counts "savings" to the Medicaid program that occur when a person shifts from SCHIP eligibility to Medicaid enrollment. If, in reality, the person was enrolled in SCHIP, then it is not clear that this represents savings to the government. When resources permit, an option to "back out" these savings may be added to the CostAvoidance module. In the meantime, such savings can be "backed out" through external tabulations conducted using SAS or some other statistical software package.

Further details on the two methods for estimating the value of Medicaid coverage are provided below.

1. Using Simulated Medicaid Values

This method is similar to the method used to calculate the change in benefits for the other transfer benefit programs, but with the added complication that the TRIM3 generated benefit for Medicaid is an annual value.

In the Medicaid simulation, a person who is enrolled in Medicaid is assumed to be enrolled in all months in which he or she is eligible. In the CostAvoidance module, there are four variables that must be specified for both the baseline and the alternative simulations: BaseAnnualMedicaidCost, BaseMonthsOfMedicaidEligibility, BaseEligType, BaseMedicaidEnrollment, AltAnnualMedicaidCost, AltMonthsOfMedicaidEligibility, AltEligType and AltMedicaidEnrollment.

The benefit is calculated on a monthly basis. If a person is enrolled and eligible in a month (i.e., BaseMedicaidEnrollment !=0 and BaseEligType[month] != 0) then the value of the baseline Medicaid benefit for that month is equal to BaseAnnualMedicaidCost divided by BaseMonthsOfMedicaidEligibility. The same process is repeated for the alternative scenario. The household's total Medicaid change is equal to the value of the Medicaid benefit in the alternative simulation, minus the value in the baseline simulation, summed over all persons in the household.

2. Using Values Based on Targets

The value of Medicaid coverage imputed by the Medicaid module is an annual value based on number of months of coverage, demographic characteristics, and eligibility type. The regression used to impute the value results in negative values for some cases with only a few months of coverage, and is not sensitive to each incremental change in the number of months of coverage. Therefore, an alternative method for imputing the Medicaid insurance value was introduced into CostAvoidance. The alternative method does not capture all of the demographic and eligibility-type variation of the Medicaid imputation, but does produce insurance value estimates that are always positive and are sensitive to incremental changes in the number of months of coverage.

The monthly value of Medicaid coverage is calculated for each state and user group (children, adults, disabled, and the elderly) and entered into the CostAvoidance module through the state array program rule MonthlyMedicaidValue. For each state and user group, the monthly value of Medicaid coverage is calculated by dividing the annual Medicaid expenditure target by twelve times the average monthly Medicaid enrollment target.

If, in a given month, a person is not enrolled in Medicaid in the baseline simulation but is enrolled in the alternative simulation, then the change in the Medicaid value for that person for that month is equal to the MonthlyMedicaidValue corresponding to the person's state and user group. In the unlikely event that a person is enrolled in the baseline, but not in the alternative simulation, then the change in the Medicaid value would be -1 times the MonthlyMedicaidValue. If the person is enrolled in both the baseline and the alternative, then there is no change in the Medicaid value. The household's total Medicaid change is calculated by summing up the change in Medicaid value over all persons in the household.

Allocation by Child Support

Once the household's total change in transfer program benefits has been calculated, it is credited to the family or families with child support income. The weight of the family head (where family is defined as families with related subfamilies separate) is used in tabulating the cost avoidance results. If anyone in the family receives TANF in the baseline in a given month, the family is tabulated in the "baseline row." If any child support recipient in the family is imputed to be receiving IV-D services, then the family is tabulated as a IV-D family. The poverty level used in the summary tables reflects the poverty level for the family as defined in the Poverty module simulation used as input to the CostAvoidance run.

There are a handful of households where more than one family receives child support income. The household's change in transfer program benefits is allocated to families in proportion to their relative levels of child support income. For example, if the first family in a household had $100 in child support in a given month, and the second family had $200, then the first family would be assigned credit for 1/3 of the total change in transfer program benefits for the household.

For some programs, a change in child support in one month of the year might affect benefits received in another month. If nobody in the household has child support in a month in which there is a change in benefits, then credit for the change in benefits is assigned to families in the household in proportion to their relative levels of annual child support income.

The proportion used to assign a family a share of the household's change in transfer program benefits is output as variable FamilyProportionOfChildSupport. If no allocation is required, then the value of FamilyProportionOfChildSupport is set equal to 99. The result variables output by CostAvoidance are equal to the monthly change in the household's transfer program benefits multiplied by the value of FamilyProportionOfChildSupport for the family in question.

The CostAvoidance Ratio

The CostAvoidance summary tables tabulate total child support income and program savings by baseline TANF status, IV-D status, and (for non-TANF families) poverty level. Cost avoidance ratios are then calculated for each of these groups by dividing the total change in transfer program benefits by the total amount of child support income.

IV-D Status

The imputation of IV-D status varies by whether the child support recipient is a custodial mother, custodial father, or non-parent. Custodial mothers, custodial fathers, and non-parents are identified using output variables from the ChildSupport module. A family is considered to be receiving IV-D services if anyone in the family is assigned IV-D status.

A child support recipient is identified as a custodial mother if the ChildSupport result variable CustodialMotherType (specified through variable list rule CustodialMom) is > 0. The recipient is identified as a custodial father if the ChildSupport result variable ChildSupportAllocatedTo (specified through variable list rule CustodialDad) is equal to 3 and RaceAndSex is equal to 1, 2, or 3. If the child support recipient is not a custodial mother or a custodial father, then he or she is identified as a non-parent.

Due to limited data, the IV-D imputations for fathers and non-parents are very rough. The program rules IncludeFathersInIVD and IncludeNonParentsInIVD allow users the option to include or exclude fathers and non-parents from the IV-D estimates. Further details concerning the imputation of IV-D status for custodial mothers, fathers, and non-parents is provided below.

Custodial Mothers

If anyone in a custodial mother's family is simulated to receive TANF or Medicaid in any month of the year in the baseline simulation, then the custodial mother is assumed to receive IV-D services (IVDStatus=1). Otherwise, the mother's IV-D status is imputed using a logit model developed using child support data from the 1998 CPS April/March Matched File (CPS-CSS). The imputation is performed for child support recipients only, since TRIM3 does not need to know the IV-D status of nonrecipients in order to simulate cost avoidance.

The logit model imputes the log of the odds ratio that a custodial mother (who receives child support and is not in a family with TANF or Medicaid) receives IV-D services. The log of the odds ratio is converted to a probability, which is then compared to a uniform random number. If the mother's random number is less than or equal to the imputed probability, then she is flagged as receiving IV-D services (IVDStatus=2). Otherwise, she is flagged as receiving child support privately (IVDStatus=4).

The logit model controls for a variety of demographic and child support characteristics. Demographic characteristics include the mother's age, marital status, race and ethnicity, region of residence, educational attainment, urban/rural status, poverty level, number of children with an absent parent, age of oldest child with an absent parent, and household receipt of SNAP benefits or housing assistance. Child support characteristics include whether or not child support was due under a formal arrangement, whether or not the full amount due was received, and the level of child support income relative to other family income.

Most of the explanatory variables are available on the raw March CPS. However, three variables are imputed by the ChildSupport module using data from the CPS-CSS. The imputed variables include the number of children with an absent parent, whether child support was due under a formal arrangement, and whether the full amount due was received.

Custodial Fathers

Four options are available for assigning IV-D status to custodial fathers. The default option for IncludeFathersInIVD is "TANF/Medicaid and random dads included." If anyone in a custodial father's family is simulated to receive TANF or Medicaid in any month of the year in the baseline simulation, then the custodial father is assumed to receive IV-D services (IVDStatus = 1). Otherwise, the father is randomly assigned IV-D status based on probabilities entered through the program rule PctOfRecipientDadsIVD. PctOfRecipientDadsIVD varies by poverty level (< 200 percent of poverty, 200-< 400 percent of poverty, and 400% of poverty and above). If a custodial father's uniform random number is less than or equal to the probability obtained from PctOfRecipientDadsIVD, then he is assigned IV-D status (IVDStatus = 5). Otherwise, he is considered to be receiving child support privately (IVDStatus = 6).

Other options for IncludeFathersInIVD:

• "No dads included": No custodial fathers are flagged as IV-D. IVDStatus is set to "0" for all custodial fathers.
• "Medicaid/TANF dads only": Custodial fathers in families receiving Medicaid or TANF in the baseline are flagged as IV-D (IVDStatus = 1). Otherwise, they are treated as receiving child support privately (IVDStatus = 3).
• "All dads included": All custodial fathers are flagged as IV-D and are assigned IVDStatus = 5.

Non-Parents Receiving Child Support

Three options are available for assigning IV-D status to non-parents. The default option for IncludeNonParentsInIVD is "Medicaid/TANF non-parents included." If anyone in the non-parent's family receives TANF or Medicaid in any month of the year in the baseline simulation, then the non-parent is assumed to receive IV-D services (IVDStatus = 1). Otherwise, he or she is considered to receive child support privately (IVDStatus = 3).

Other options for IncludeNonParentsIVD:

• "No non-parents included": No non-parents are flagged as IV-D. IVDStatus is set to 0 for all non-parents.
• "All non-parents included": All non-parents are flagged as IV-D. IVDStatus is set to 3 for all non-parents, and the CostAvoidance code interprets this as indicating that all non-parents are receiving IV-D services.

Adjusting Results Using OCSE Financial Data

Because there is no external target for total child support income that captures private payments as well as IV-D collections, it is impossible to know the full extent of underreporting of child support income on the March CPS. However, there is little doubt that underreporting exists. With the exception of income from wages and salaries, income amounts are typically underreported in household survey data. The March CPS captured an estimated 71 percent of property income, 88 percent of transfer income, and 77 percent of pension income in 1996 (Roemer 2000). Wheaton (1997) estimates that child support on the March 1992 CPS is underreported by as much as 21 to 27 percent relative to child support reported on the 1992 CPS-CSS.

Child support is particularly likely to be underreported for TANF recipients, who are unlikely to report child support payments retained by the government. Therefore, unadjusted TRIM3 estimates will fall far short of the actual level of TANF cost recovery. To the extent that persons who do not receive TANF underreport child support, unadjusted TRIM3 estimates of cost avoidance will be understated as well.

Data on the actual amount of child support income distributed by the IV-D system can be obtained from OCSE publications such as the Child Support Enforcement FY 1999 Preliminary Data Report (OCSE 2000). OCSE data on TANF cost recovery can be substituted for the unadjusted TRIM3 estimates. Revised IV-D cost avoidance estimates can be calculated by multiplying the TRIM3 cost avoidance ratio (savings per dollar of child support) by the total amount of payments to families as reported by OCSE. No revised estimate is possible for persons not receiving IV-D services, since there is no external target for the total amount of child support paid privately.

For further discussion of the issues involved in applying TRIM3 cost avoidance ratios to OCSE data, see Wheaton (2002).

References

Office of Child Support Enforcement (September 2000). Child Support Enforcement FY 1999 Preliminary Data Report. U.S. Department of Health and Human Services.

Roemer, Marc I. (June 2000). Assessing the Quality of the March Current Population Survey and the Survey of Income and Program Participation Income Estimates, 1990-1996. U.S. Census Bureau, Income Surveys Branch, Housing and Household Economic Statistics Division.

Wheaton, Laura (1997). "Child Support Underreporting: A Tale of Two Supplements." American Statistical Association 1997 Proceedings of the Section on Government Statistics and Section on Social Statistics, p. 370-375.

Wheaton, Laura (2002). "Child Support Cost Avoidance in 1999." Final report prepared for the U.S. Department of Health & Human Services Office of Child Support Enforcement, IDIQ Contract No. 105-00-8303, Task Order No. 5.