Correction for Underreporting of Child Support

Correction for Underreporting of Child Support

CorrectForUnderreporting indicates whether TRIM3 should correct for underreporting of child support by TANF recipients. For convenience, the following discussion refers to child support "received" by TANF families, although what is actually being imputed is the total amount of support collected on each family's behalf. The TANF module determines what portion of the collected child support is actually transferred to the family.

Identifying Families Eligible For Correction for Underreporting

A family without child support must meet the following requirements in order to be assigned child support through the correction for underreporting methodology. The family must:

Receive TANF in at least one month of the year.
Not receive child support in any month in which TANF is received
Not have a child receiving SSI
Not have a child who is without TANF for the entire year
Be one of the following family types:
- Single custodial mother
- Married custodial mother
- Single father
- Relative caretaker

The values in variablelist rule UnitTANFBenefits indicate whether a family receives TANF in a certain month. Additional requirements can be imposed by setting variablelist rules UnitTANFBenefits2 and UnitTANFBenefits3. If a variable is specified for UnitTANFBenefits2 (and UnitTANFBenefits3), then the module corrects for underreporting only if UnitTANFBenefits and UnitTANFBenefits2 (and UnitTANFBenefits3) agree as to whether a family participates.

A single father is defined as a man who is separated, divorced, or never married, and has a never-married child under the age of 21 living in the household.

A "relative caretaker" family is defined as a primary family with no "own" children (i.e., no children with HhFamilyRelation in (4, 7, 9, or 12) and age < 21) that has at least one related child (i.e., at least one person in the primary family with HhFamilyRelation in (16, 19, 21, 24) and age < 21).

Selecting Additional TANF Families to Receive Child Support

For each family found eligible for correction for underreporting of child support, the Child Support module compares the probability that the TANF family actually has child support to a random number to determine whether or not the family will be assigned child support. The probabilities vary by family type: married couple, single mother, single father, and relative caretaker. Single mothers are further disaggregated into Hispanic, non-Hispanic black, and non-Hispanic white/other. The probabilities were calculated using 2000 data from the TRIM3 TANF baseline and TANF Microdata.

AdjustProbTANFFamHasCS provides multiplicative adjustments, by state, to the probability that a TANF family actually has child support. If a random number assigned to an eligible family is less than or equal to the family's adjusted probability, then the family is selected for correction for underreporting, and child support is assigned to the family.

Assigning the Number of Months of Support

If a family is selected for a correction to underreporting, then the Child Support module assigns child support to some or all of the months that the family receives TANF; no new child support is assigned to non-TANF months. If a family receives TANF in only one month of the year, child support is assigned to that month. Otherwise, the Child Support module uses probabilities in look-up tables derived from Survey of Income and Program Participation (SIPP) data to impute the number of months a TANF family receives child support. Program rule MonthlyAllocationOption specifies which set of cumulative probabilites should be used--either the original 1998-1999 probabilities or ones calculated from 2005-2006, 2009-2010 SIPP data.

The imputation relies on three look-up tables -- one for families receiving TANF for all 12 months of the year, one for families receiving TANF in 7 to 11 months of the year, and one for families receiving TANF in 1 to 6 months of the year. The look-up table for families receiving TANF for all 12 months of the year shows the cumulative probabilities that a family receives child support in 1, 2, …12 months. The look-up tables for families receiving TANF in 1 to 6 months and in 7 to 11 months show the cumulative probabilities that a family receives child support in various "percentages" of its TANF months. AdjMonthsTANFFamHasCS provides multiplicative adjustment factors, by state, to the cumulative probabilities obtained from the look-up tables. The Child Support module assigns a random number to each family and compares the random number to the adjusted cumulative probabilities in order to determine the number of TANF months in which the family will be assigned child support.

If a family is imputed to receive child support in some, but not all, of its TANF months then the Child Support module randomly selects a "starting month" to which to assign child support. Beginning with the "starting month" the family is assigned to receive child support in each month in which TANF is received until the imputed number of months has been assigned. If not all of the months have been assigned by December then assignment continues with January until all months have been assigned.

Imputing the Total Monthly Amount of Child Support

If a TANF family is selected for correction for underreporting, or reports receipt of child support but only appears to be reporting the amount of child support transferred to the family, then the Child Support module imputes the total monthly amount of child support collected on behalf of the family. The TANF module determines whether the family is reporting just the transferred amount of child support, and passes this information to the Child Support module through variable UnitCS_ReporterType (which is specified through the Child Support rule MonthlyTypeCSReporter). If the amount imputed to the family is less than the reported amount, then the Child Support module assigns the reported amount to the family.

The Child Support module uses four OLS models to impute the log of the monthly amount (per child) of child support collected on the family's behalf: one for single mothers, one for married couples, one for single fathers, and one for families with relative caretakers. The models were estimated using the 2000 TANF Microdata. TRIM3 restores variation to the predicted amount, and the resulting estimate is converted to a dollar amount representing the total monthly amount of child support collected on behalf of the family. The program rule AdjCSAmtTANFFamilies provides multiplicative state-level adjustments to the resulting estimate.

The OLS model for single mothers includes dummy variables for number of children, marital status, and race and ethnicity. The OLS models for married couples and relative caretakers include explanatory variables for number of children and race and ethnicity. The OLS model for single fathers includes just one explanatory variable -- a dummy variable for whether there is one child (two+ children is the omitted category).

Imputing Current Support from Total Support

The results of the OLS imputation represent total child support, including arrears. The Child Support module splits the total imputed child support into arrears and current support based on patterns observed in administrative data from Illinois, Massachusetts, New Jersey, and Texas. The Child Support module does not currently save the arrears data because arrears are not available for families other than those affected by the correction for underreporting methodology. The Child Support module assigns the portion of the total collection imputed to be current support to each of the months the family has been assigned to receive child support. If all of the child support is imputed to be arrears, then no child support is assigned to the family.

Two sets of state-level program rules control the allocation of total support into arrears and current support. The first set of program rules contains cumulative probabilities that a TANF child support recipient falls into the following "ranges" for the percent of child support going to arrears: 0%, 1-19%, 20-39%, 40-59%, 60-79%, 80-99%, 100%. The second set of program rules specifies the percent of child support to be assigned to arrears for child support recipients in each of the specified ranges. The program rules vary by the number of custodial children (one, two, and three or more) and the monthly amount of imputed child support ($1-$200, $201-$400, $401-$600, and >$600).

Consider the following example: a TANF child support recipient in Illinois has one custodial child, monthly imputed child support in the $201 to $400 range, and a uniform random number of .65. The Child Support module would use the program rules OneKid201to400Range and OneKid201to400Pct for this family, which have the following values for Illinois:

OneKid201to400Range (Illinois)

0% 1-19% 20-39% 40-59% 60-79% 80-99% 100%

.32 .5 .75 .9 .92 .93 1

OneKid201to400Pct (Illinois) (Illinois)

0% 1-19% 20-39% 40-59% 60-79% 80-99% 100%

0 .121 .28 .48 .67 .87 1

The Child Support module would assign a "range" to the child support recipient by finding the first value in the OneKid201to400Range program rule that is greater than or equal to the child support recipient's uniform random number (.65). In this example, that is the 20-39% range, which has a value of .75. The Child Support module would then obtain the percentage of child support to be assigned as arrears for child support recipients in the 20-39% range (28 percent) from the program rule OneKid201to400Pct. Twenty-eight percent of the child support recipient's imputed monthly child support would be considered arrears, and the remaining 72 percent would be considered to be current support and would be saved in the Child Support module's results variables.

Results Variables and Effect on Other Variables

The Child Support module incorporates the results from correction for underreporting into the variables AmountReceived, MonthlyAmountReceived, AnnualCSIncomeOfChild, and MonthlyCSIncomeOfChild. The imputation of AmountDue (in correction for underreporting runs) is based on child support income after correction for underreporting. AmountDue is not imputed for single father and relative caretaker families, so remains zero for these families. Child support received during non-TANF months is unaffected by the correction for underreporting methodology.

If a family has reported child support (i.e., the family appears to be reporting the amount of child support transferred to the family under the rules of the TANF program) then the newly assigned child support is assigned to the children in proportion to their original monthly amounts. If the family does not report receiving child support, then the imputed child support is allocated equally among the custodial children in the family (for custodial mother families) or among all children in the family (for single father and relative caretaker families).

Simulations in a Correction for Underreporting Run

A correction for underreporting run typically consists of two child support simulations and two TANF simulations:

The first child support simulation (CS1) performs correction for underreporting for families simulated to receive TANF in a preliminary baseline TANF simulation.
The second TANF simulation (TF1) is identical to the preliminary TANF simulation, but uses child support income from simulation CS1. The TF1 TANF simulation is run in alternative mode so that families assigned a high enough amount of child support in CS1 that it would be unlikely that they would participate in TANF stop participating in TANF.
The second child support simulation (CS2) assigns child support to some families who had TANF in simulation TF1. All program rules in CS2 are exactly the same as CS1, except that CS2 uses TANF input from TF1 (rather than from the preliminary TANF baseline) to identify TANF families. If the child support assigned to a family in CS1 was so high that the family stopped participating in TANF in all months of the year (in TF1) then the family does not appear to be a TANF family in CS2 and so is not assigned child support.
The second TANF simulation (TF2) is like the preliminary TANF baseline and is run in baseline mode using child support variables from CS2. The summary tables in TF2 show the effect of using child support corrected for underreporting in the TANF simulation.