State child support enforcement agencies are charged with recovering financial support from non-custodial parents to help cover the costs of raising children. Payment amounts are established via court order, owed to the parent who has primary physical custody of the child.
Federal guildelines require that states secure at least 80 percent of child support owed by non-custodial parents, and the Pennsylvania Bureau of Child Support Enforcement has a solid track record of meeting these guidelines.
In fact, Pennsylvania is the only state currently meeting or exceeding all federal guidelines, said Dan Richard, bureau director, in an interview with Government Technology. But despite that distinction, officials felt there was room for improvement. They wondered if data they already had access to could potentially help bring about more reliable support for families.
”Families need a reliable, consistent source of income, particularly when it comes to child support," Richard said. "You want to know that the payments will be steady, so that you can use them to meet the ongoing daily needs of the child."
The bureau began looking for variables to consider to help determine how likely the non-custodial parent is to pay, he said, "and what things can we do to help them overcome various barriers to making reliable, consistent payments?”
Over a period of a year, the bureau developed a payment score calculator that factors in several pieces of data on the non-custodial parent and arrives at a “score” that predicts how likely that person is to satisfy their child support obligations. Included in the algorithm are factors like the parent’s age, employment status and history, residential stability and number of current child support cases. The state uses about 20 such demographic variables to arrive at a score in each case.
While predictive analytics have long been used in private industry, there weren’t many examples of comparable public-sector agencies using these kinds of tools. But in Pennsylvania, the process was fairly simple.
“Once the basic information about a case is put on the computer, you can press a button and it will automatically calculate a score that projects the likelihood of payment,” Richard said. “From a policy point of view, and a programmatic point of view, it was a major initiative. But from a technological point of view, it was not a major leap, and it was implemented at a modest cost.”
The state was able to use existing technical resources to bring the system online; and the data sets, a robust data warehouse and data mining capabilities were already available as well. Other states, as well as private companies, have contacted Richard’s office to inquire about these back-end requirements for putting predictive analytics to work for them.
Using the system allows the bureau to adopt a more hands-on, proactive stance when it comes to collections. Rather than waiting until a non-custodial parent fails to make payments and amasses large past due amounts, a lower score prompts earlier intervention and tailored outreach. The bureau can help connect the parent with job placement and training services provided by other agencies, for example.
Richard reports that their predictive method has proven very accurate in the first year of use, helping the bureau to use its resources more efficiently. They are now using information they gather on the most effective collection methods for certain groups to further inform collection efforts.
While the benefits to custodial parents and children are self-evident, the program also benefits taxpayers as a whole, Richard says. Higher rates of compliance with child support orders translate to fewer demands on public assistance programs that single parents often turn to absent court-ordered child support.