Imputation of Amount Due for Custodial Mothers

Details concerning the imputation of Amount Due for custodial mother families are provided below. See section 6 for discussion of an alternative imputation for whether a non-TANF family without child support is due support.

Imputation of Amount Due for Custodial Mothers With Reported Child Support

There are two steps to the imputation of the amount due for custodial mothers with reported child support. First, a logit model is used to impute whether or not the custodial mother received the full amount due. The imputed probability that a custodial mother who receives childsupport receives the full amount due may be adjusted by the multiplicative factors in ReporterFullAmountDueAdj, ReporterFullAmountDueAdj2 and ReporterFullAmountDueAdj3. ReporterFullAmountDueAdj factors vary by race/ethnicity of the mother, ReporterFullAmountDueAdj2 by marital status, and ReporterFullAmountDueAdj3 by number of custodial children. These adjustments also apply to cases where child support is assigned through correction for underreporting. For those imputed to receive less than the amount due, an OLS model imputes the percent of the amount due that was received.

Imputation of Whether Full Amount Due is Received

Custodial mothers who receive high amounts of child support (per custodial child) are more likely to have received the full amount due than similar custodial mothers who receive lower amounts of child support per child. To impute whether a custodial mother with reported child support receives the full amount due, a set of dummy variables is defined indicating the ratio of the amount of child support reported received (per custodial child) to the median amount due per custodial child for that family type (where family type is defined by whether the family reports receiving TANF/AFDC, the number of custodial children, and adjusted family income level).

Adjusted family income is defined as the absolute value of the family's income, excluding child support, TANF/AFDC, and SSI. The income variables used in determining adjusted family income are specified through the child support module program rule FamilyIncome.

The amount of child support received per custodial child is equal to the amount of reported child support income, divided by the number of custodial children.

The median child support due per custodial child for different family types (based on tabulations of the 1998 CPS-CSS) is shown below:

The amount of child support received per custodial child is divided by the median amount due per custodial child as obtained from the above table. The logit model contains a set of dummy variables indicating the ratio of the child support received per custodial child to the median amount due per custodial child: < .1, .1 < .2, .2 < .4, .4 < .5, .5 < .6, .6 < .7, .7 < .9, .9 < 1.1, 1.1 < 1.5, 1.5 < 2. The omitted category is 2+.

The program rule ReporterFullAmountDueAdj can be used to specify multiplicative adjustments (by race/ethnicity) to the imputed probability that a custodial mother with reported child support receives the full amount due.

Imputation of Percent Received, for Custodial Mothers Receiving Less than Full Amount Due

If a custodial mother with reported child support is imputed to have received less child support than she is due, the child support module imputes the amount due. An OLS model imputes the log of the ratio of the amount of child support received to the amount due. Variation is restored to the predicted estimate, and the ratio of the amount of child support received to the amount due is obtained. The amount received is then divided by the ratio of the amount received to the amount due in order to obtain the estimate of the amount due.

Four OLS models are used in the imputation. Each has two explanatory variables: the number of custodial children and the log of the amount received. The choice of model depends on the ratio of the amount of child support received per custodial child to the median amount due per custodial child (< .2, .2 < .45, .45 < .8, and .8+).

The median amount due per custodial child comes from tabulations of the 1998 CPS-CSS for custodial mothers who reported receiving less than the full amount due. The median child support due per custodial child, by family type, is shown below:

Imputation of Amount Due for Custodial Mothers without Reported Child Support

If a custodial mother does not have reported child support, then the child support module imputes whether she is due child support, and if so, the amount. First, a logit model is used to impute whether or not the custodial mother is due child support. For those imputed to receive less than the amount due, an OLS model is used to impute the amount due.

The logit model that is used to impute whether a custodial mother without child support is due support relies on the following explanatory variables: dummy variables for age, marital status, race/ethnicity, region, educational attainment, TANF/AFDC recipiency/eligibility, urban/rural status, family poverty level, and number of custodial children.

Program rule IsDueAdjustment provides multiplicative adjustments (by race/ethnicity) to the probabilities that a custodial mother without child support is due support.

If a custodial mother without child support is imputed to be due child support, then an OLS model imputes the log of the amount due. Variation is restored to the predicted estimate, and the estimate of the amount due is obtained. The OLS model includes the following explanatory variables: a continuous variable for the number of custodial children, and dummy variables for age, marital status, racial/ethnic minority status, educational attainment, TANF/AFDC recipiency/eligibility, urban/rural status, and family poverty level.