Behavioral Response Simulations

ApplyBenefitResponse indicates whether the Child Support Module should perform a "behavioral response" simulation. When performing a behavioral response simulation, the Child Support module initially sets child support equal to the amount of child support from the baseline child support simulation (as specified through BRMonthlyCSBase), and then assigns child support to some additional families who receive TANF but no child support. Families potentially eligible for being assigned child support include custodial mother families, single father families, and "relative caretaker families" -- families in which someone other than a parent is caring for related children. In order for a family to be eligible to be assigned child support through the behavioral response simulation, the family must:

• Receive no child support during the year
• Receive TANF in at least one month of the year
• Have no children <21 receiving SSI during the year
• Have no children <21 without TANF in at least one month of the year

For this test, child support status is obtained through the variable specified for BRMonthlyCSBase, the family's TANF status is obtained through BehavRspMonUnitTANFBenBase, the child's TANF status is obtained through MonthlySimulatedTANFBenefits, and the child's SSI status is obtained through MonthlySimulatedSSIBenefits.

If a family is eligible to be assigned child support through behavioral response, then a logit model is used to impute the probability that the family receives child support. Since none of these families actually receive child support in the baseline, the family is assigned a random number that is greater than the probability that they receive child support and is less than or equal to one. The user then specifies multiplicative adjustments to the probability that the family receives child support. If the adjusted probability is greater than or equal to the random number assigned to the family, the family is selected to receive child support through behavioral response. Additional details are provided below.

Custodial Mothers and Single Fathers

Two logit models are used when simulating the behavioral response decision for custodial mothers and single fathers. Both models were estimated using data from the 2000 Current Population Survey April/March Matched File (CPS-CSS).

The first model imputes whether the family is due to receive child support under a formal order or agreement. Explanatory variables include the age, marital status, race/ethnicity, and educational attainment of the parent, region of residence, whether the family lives in a metropolitan area, the number of custodial children, and the age of the youngest child. The results of the logit model are converted to a probability, which is adjusted by the state-level multiplicative adjustment factor specified through ADJProbTANFDueSupport. The family is assigned a uniform random number between the values of 0 and 1. If the random number is less than or equal to the adjusted probability, the family is assigned to be due to receive child support.

If the family is simulated to be due to receive child support, then the second logit model is used to impute the likelihood that the family receives child support. Explanatory variables for this imputation include the age, marital status, race/ethnicity, and educational attainment of the custodial parent, region of residence, whether the family resides in a metropolitan area, the number of custodial children, and the age of the youngest child. The results of the logit model are converted to the probability that the family receives child support. Because the family does not actually receive child support, the family is assigned a random number that is greater than the imputed probability and less than or equal to 1. The probability is then adjusted by the multiplicative state-level adjustment specified through AdjProbTANFRecSupportParent. If the family's random number is less than or equal to the adjusted probability, the family is assigned to receive child support.

Relative Caretakers

A single logit model is used to impute the likelihood that a "relative caretaker" receives child support. Child support data for children cared for by non-parent caretakers is unavailable in the CPS-CSS data, so 2000 TANF Microdata were used to estimate the model. Explanatory variables in the model include the number of related children, the race/ethnicity of the family head, and the age of the youngest child. As with the custodial parent and single father imputation, the result of the logit model is converted to a probability, and the family is assigned a random number that is greater than the probability and less than or equal to 1. AdjProbTANFRecSupportOtherRel specifies multiplicative state-level adjustments to the probability. If the family's random number is less than or equal to the adjusted probability, the family is assigned to receive child support.

Assignment of Child Support Amount

If a family is assigned to receive child support through behavioral response, then the monthly amount of child support and the months in which child support is received is assigned using the same methodology as is used to correct for underreporting of child support recipiency by TANF recipients. Child support is assigned to some or all of the months in which the family receives TANF. The family's TANF status is obtained from the variable specified by BehavRspMonUnitTANFBen.

Results Variables and Summary Tables

The results of the behavioral response simulation are stored in MonthlyCSIncomeOfChild and AnnualCSIncomeOfChild. Other variables may be output by a behavioral response simulation, but are not necessarily reliable and should not be used. The behavioral response simulation does not produce any summary tables. Results of the simulation are obtained by viewing the child support summary tables in the TANF module.

Simulations in a Behavioral Response Run

A behavioral response simulation is typically performed as a single run containing four simulations, two child support simulations (CS1 and CS2) and two TANF simulations (TF1 and TF2). The purpose and function of each simulation is as follows:

1. The first child support simulation (CS1) assigns child support to additional TANF families using the behavioral response methodology described above.
2. The first TANF simulation (TF1) is like the baseline TANF simulation except that it uses the child support inputs from CS1 and simulates the TANF child support rules for the scenario of interest. The simulation is run in alternative mode and ParticipationBasedOnPreviousRun is set to "Units that were not eligible in the baseline simulation or did not participate in the baseline simulation continue to not participate." This ensures that there are no new participants and that families are allowed to stop participating as a result of the increased child support.
3. The second simulation (CS2) sets child support back to the original baseline amount for any family who stops participating in TANF in all months as a result of the additional child support assigned in CS1.
4. The second TANF simulation (TF2) is the final TANF simulation showing the effects of behavioral response. It is like TF1, but uses child support from CS2 as input.

The CS2 simulation is included because the methodology used to impute child support amounts to families may produce unrealistically high amounts for some TANF families, given the family's circumstances and benefit amount, and may cause an unrealistically high number of families to stop participating in TANF. The CS2 simulation prevents this from happening by setting child support back to the baseline amount for any family who would stop participating in TANF in all months as a result of the additional child support assigned in CS1. Although, in reality, a family might receive enough child support as a result of behavioral response to stop participating in TANF entirely, this possibility is not currently captured in TRIM3. The program rule TANFLossOption indicates what to do with a family during a behavioral response run, if there is any month of the year in which the family would stop participating in TANF as a result of child support obtained through a "behavioral response". The family can either be excluded from behavioral response, or the family is excluded unless there is at least one month of the year in which the family would continue to participate in TANF despite having been assigned child support through behavioral response.