Child support agency uses analytics to provide better options for parents

Orange County’s data-driven approach empowers caseworkers to help parents make decisions that benefit their children

When parents don’t – or are unable to – meet their child support obligations, social service agencies often have little recourse. Jailing delinquent parents won’t help them pay their bills, and the threat of jail could lead them to further avoid child support staff – and their children. Rather than help the child, these efforts can send a parent and their children into a poverty spiral.

California’s Orange County Child Support Services is trying something different. It uses analytics to predict who is at high risk of failing to pay support and identifies the factors that could get the parent into compliance more effectively. This new approach is data-driven – based on the county’s extensive research into the factors that help parents fulfill their obligations – and sensitive. The ultimate goal is to reduce the number of children living in poverty.

Steve Eldred, Director of Orange County Child Support Services, says that predictive analytics, used in conjunction with social safety net support tailored to the individual, can help the agency find long-term solutions that help parents support their children. It is an approach that he believes works particularly well in areas like Orange County, where jobs are plentiful and wages are decent – but the cost of living is high. Eldred’s team focuses on finding ways to help noncustodial parents earn enough to pay their child support.

“The model of ‘People won’t support their children without coercion or threat of jail’ isn’t effective or accurate. We start from the premise that almost all parents want to support their children to the best of their ability. So let’s tailor social support to enhance that ability. Our approach is to concentrate on multigenerational poverty and look at the long term rather than the short term,” Eldred says. The county’s caseworkers manage 69,000 active child support cases annually, with 12,000 new cases opened each year.

Looking long-term

Prior to using predictive analytics, caseworkers tried to figure out how to help parents be in a better position to pay, but the data was not easy to access. Staffers would poke through dozens of screens to scrape together details on the parent that might point to a solution. Could they use more education? Did they need social services support themselves? “They don’t really have the luxury to go through 100 screens to decipher what is pertinent to the case,” Eldred says.

With better analytics, the agency gathers everything it knows about the parent, analyzes it and assigns something it calls an iScore, which details information concerning ability to pay and what factors might encourage full or timely payments. Parents with significant issues – unemployment, criminal records or homelessness – receive a “seedling” iScore classification that lets caseworkers know that this parent needs more time and help to be able to pay. Others are classified as “saplings” or “young trees” if they have one deficit that could make a significant difference in their ability to pay.

“Our research shows that a person with a high school diploma or GED pays 44 percent more support over the life of that child than a high school dropout,” says Eldred. With that knowledge, a caseworker might decide to encourage night school. A parent with an old felony conviction hindering employment might be sent to a public defender to get the conviction expunged. If the conviction is recent, the parent might be referred to employers who hire those recently released from prison.

In each case, analytics surfaces the best course of action, sparing the caseworker from trying to guess the best option for recourse. “We started to develop the model using 450 variables and slimmed it down to about 30 of the most significant variables,” Eldred explains, noting that the models were built using information from 30,000 cases. A big benefit: Rookie caseworkers are nearly as productive as their most experienced colleagues.

“iScore is an attempt to mimic the knowledge of a 20-year caseworker,” Eldred says. And while the score draws in some demographic data like ZIP code and age in its model, those are only two of 30 variables because demographic data alone can mislead.

Early results show promise

Within six months of implementation, the system has been embraced by the county’s 60 caseworkers who use it daily. Among the successes:

• Using the iScore tool, a caseworker was prompted to find additional government benefits for a disabled parent that enabled him to pay $100 a month that he owed in child support.
• Another parent was referred to low-cost medical help for a nagging health problem that was interfering with work.
• Caseworkers can review iScore with a customer, then collaboratively design an earnings improvement plan that meets the customer’s needs, interests and skills. “Building a success plan with parents that is tailored to them specifically, where they can see the possible results, makes all the difference for customer engagement,” Eldred says.

According to Eldred, the analytic approach saves time and empowers caseworkers to help people make changes that truly benefit their children. “I had a caseworker manager tell me years ago that cases are like snowflakes,” he says. “From a distance, they are all identical. But every single one is unique. Instead of saying most 22-year-old single fathers from this ZIP code have these characteristics, we use analytics to truly understand what social, economic or structural barriers keep this particular 22-year-old from succeeding. And then we figure out how to help.”