NonresidentParent version 8.0The Urban Institute TRIM3 Reference Contact Us Version History Dictionary
|
NonresidentParent version 8.0 The Urban Institute TRIM3 Reference Contact Us Version History Dictionary |
The purpose of the Nonresident Parent (NRP) module is to impute nonresident parent status and child support payment information. A "nonresident parent" is defined here as a man or woman who has one or more "nonresident children" who do not live with him/her but who live with the other parent. Nonresident parent status and child support payment characteristics are not reported in the CPS, and so the Nonresident Parent module was developed to impute this information to men and women in the TRIM3/CPS data. The data generated by the Nonresident Parent module reflect all nonresident parents whose children live with the other parent, not just those who are participating in the federal/state Child Support Enforcement (CSE) program.
The imputations in the Nonresident Parent module represent NRPs with minor children, although no specific age cut-off for the child is assumed. Results of the nonresident parent module are aligned to targets derived from information reported by custodial parents in the Child Support Supplement to the Current Population Survey (CPS-CSS), using whatever age definition for "child" is desired (e.g., <18, <19, or <21). Because the CPS-CSS does not contain information about cases in which the child lives with someone other than the other parent (and because no other source for national targets is available) the Nonresident Parent module does not capture those cases in which the child lives with someone other than the other parent.
The Nonresident Parent module performs the following imputations:
The discussion proceeds as follows:
In order to be eligible to be imputed to be a nonresident parent, a person must be a man aged 15-65 or a woman aged 15-60 and must not be widowed or married with an absent spouse. Two logit models (one for men and one for women) are used to impute whether a person found eligible to be imputed to be a nonresident parent is assigned nonresident parent status.
The explanatory variables for men include dummy variables for race/ethnicity, marital status, age, educational attainment, whether the man's family receives CPS-reported TANF, whether the man's household receives CPS-reported food stamps, whether the man provides health insurance to someone outside the household, and the poverty level of the man's family (as reported on the CPS). When determining whether the man provides health insurance to someone outside the household, only unallocated responses are used. The imputation also includes three dummy variables that control for receipt of child support and presence of children under 18: (1) whether the man receives child support; (2) whether a man without child support has children in the household under the age of 18; and (3) whether a man without child support has children in the household who are 18 or older (and none who are under 18).
The explanatory variables for women are the same as those for men, except that they do not include whether the family has CPS-reported TANF. Rather than a single dummy variable indicating whether the woman receives child support, two dummy variables are used: (1) whether a woman receiving child support has children in the household under the age of 18; and (1b) whether a woman receiving child support does not have children in the household under the age of 18.
The results of the logit model are adjusted by additive adjustments specified through AdjManIsNRP and AdjWomanIsNRP (which provide adjustments by age, race, and ethnicity) and EarnAdjManIsNRP and EarnADjWomanIsNRP (which provide adjustments by earnings level). The adjusted logit result is converted to a probability and compared to the uniform random number variable RNNonresidentParent. If RnNonresidentParent is less than or equal to the probability that the man or woman is a nonresident parent, the man or woman is assigned to be a nonresident parent. The results of the imputation are stored in the results variable IsNonresidentParent.
If a man or woman is imputed to be a nonresident parent, then ordered logit models (estimated separately for men and women) are used to impute the number of nonresident children (one, two, three, or four+).
The explanatory variables for men include dummy variables for race/ethnicity, marital status, age, whether the man's household receives CPS-reported food stamps, and the poverty level of the man's family (as reported on the CPS). The imputation also includes dummy variables that control for the number of resident children of the nonresident father (one, two, three, or four or more). The explanatory variables for women include dummy variables for age, the poverty level of the woman's family (as reported on the CPS), and the number of resident children of the nonresident mother.
The results of the ordered logit model are adjusted by additive adjustments specified through AdjManNumAbsentChildren and AdjWomanNumAbsentChildren (which provide adjustments by age, race, and ethnicity) and EarnAdjManNumAbsentChildren and EarnAdjWomanNumAbsentChildren (which provide adjustments by earnings level). The adjusted results are then converted to cumulative probabilities that the nonresident parent has one, two, three, or four or more nonresident children. The resulting cumulative probabilities are compared to the uniform random number variable RNNumberAbsentKids to assign the number of nonresident children. The imputed number of nonresident children is stored in the result variable NumberAbsentKids.
The Nonresident Parent module uses logit models (estimated separately for men and women) to impute whether a nonresident parent is required to pay child support. When left unadjusted, this imputation tends to underestimate the number of nonresident parents required to pay child support, because many nonresident parents fail to report that they are required to pay child support in the SIPP data used to estimate the equation.
The explanatory variables for men include dummy variables for race/ethnicity, marital status, age, whether the man's household receives CPS-reported food stamps, whether the man provides unallocated health insurance to someone outside the household, earnings level, residence in a metropolitan area, and number of nonresident children. The explanatory variables for women include dummy variables for age, educational attainment, residence in a metropolitan area, and the number of nonresident children.
The results of the logit model are adjusted by additive adjustments specified through AdjManOwes and AdjWomanOwes (which provide adjustments by age, race, and ethnicity) and EarnAdjManOwes and EarnAdjWomanOwes (which provide adjustments by earnings level). The adjusted logit result is converted to a probability and compared to the uniform random number variable RnNRPOwesCS. If RnNRPOwesCS is less than or equal to the probability that the NRP is required to pay child support, the NRP is assigned to be required to pay child support. The results of the imputation are stored in the results variable NRPOwesCS.
If an NRP is required to pay child support, TRIM3 uses a logit model to impute whether he or she pays any child support during the year. Because NRPs substantially underreport the requirement to pay support, TRIM3 uses unadjusted results from the imputation of whether an NRP is required to pay support to determine whether to perform the imputation of whether the NRP pays support. NRPs who do not appear to be required to pay child support according to the unadjusted equation, but are simulated to be required to pay child support as a result of adjustments needed to bring the numbers up to target, represent those NRPs who fail to report the requirement to pay support, and are assumed to pay no support. If this approach yields too few NRPs simulated to pay child support for a particular gender, age, and race/ethnicity category, then AdditionalMenForPayImp and AdditionalWomenForPayImp can be used to allow a percentage of NRPs who are required to pay child support in the adjusted (but not unadjusted) results to be allowed the chance to be imputed to pay child support.
The explanatory variables include dummy variables for marital status, whether the NRP's household receives CPS-reported food stamps, whether the nonresident parent provides (unallocated) health insurance to someone outside the household, family poverty level (as reported on the CPS), and earnings level.
The results of the logit model are adjusted by additive adjustments specified through AdjManPays and AdjWomanPays (which provide adjustments by age, race, and ethnicity) and EarnAdjManPays and EarnAdjWomanPays (which provide adjustments by earnings level). The adjusted logit result is converted to a probability and compared to the uniform random number variable RnNRPPaysCS. If RnNRPPaysCS is less than or equal to the probability that the NRP pays child support, the NRP is assigned to pay child support.
Nonresident parents who are imputed to pay child support are assigned an annual amount of child support based on a highly detailed look-up table created from the 5-state administrative data. The look-up table varies by the NRP's age, number of nonresident children, and wage level. For NRPs with wages below $16,000 the table shows deciles for the amount of child support paid (for each $1,000 increment of wages). For NRPs with wages above $16,000, the table shows the deciles for the child support payment as a percentage of wages (for various ranges of wages). To avoid extreme outliers, the maximum child support payment amount in the look-up table for each age, number of nonresident children, and wage level category is set to the 99th percentile for that category in the administrative data.
The look-up table is based on wage level, because the administrative data on which the table is based do not include the nonresident parent's non-wage earnings or total income. When assigning a child support payment to a nonresident parent without wages, the Nonresident Parent module uses the absolute value of total income as a proxy for wages. For further details about the look-up table, click here.
Each nonresident parent who is simulated to pay child support is assigned a uniform random number, RnCSAmountPaid, representing the percentile for the amount to be assigned. When the random number falls between two deciles in the look-up table, interpolation is used to assign the child support amount.
The amount obtained from the look-up table is subject to various adjustments:
The Nonresident Parent module uses logit models (estimated separately for NRPs with income from wages and NRPs without income from wages) to impute whether an NRP who pays child support pays the full amount due.
The explanatory variables for NRPs with income from wages include the log of the amount paid and dummy variables for age, wage level, full-year worker status, receipt of unemployment compensation, and number of nonresident children. The explanatory variables for NRPs without income from wages include all of the above variables except for wage level and full-year worker status.
The results of the logit model are adjusted by additive adjustments specified through AdjManPaysAll and AdjWomanPaysAll (which provide adjustments by age, race, and ethnicity) and EarnAdjManPaysAll and EarnAdjWomanPaysAll (which provide adjustments by wage rather than earnings level). The adjusted logit result is converted to a probability and compared to the uniform random number variable RnPaidAllCSDue. If RnPaidAllCSDue is less than or equal to the probability that the NRP pays the full amount due, the NRP is assigned to pay the full amount due. The results of the imputation are stored in the results variable PaidAllCSDue.
The Nonresident Parent module does not currently contain any special requirements for performing alternative simulations. The number of nonresident parents, number of nonresident children, number required to pay support, number paying support, amount paid, and number paying the full amount due can be adjusted by rerunning the Nonresident Parent module and changing the relevant adjustment factors. No other changes are necessary to ensure consistency between the baseline and alternative simulation.