STATISTICA Blogs

It is easy to find stories about government using data mining for counter terrorism. But predictive analytics can be used for so many other reasons in the government.

Exploring regulatory compliance?

Trying to understand the important factors for a social policy?

Sometimes the data miner is trying to model "human behavior". There is risk with this approach and it must be acknowledged. Human beings are inconsistent. The same human can respond very differently for no visable (measurable) reason.

• My favorite color is green today, but tommorrow it will be purple.
• I don't always answer survey questions truthfully.
• My favorite football team lost an important game. I am mad. I won't do the statistics report today.

Data mining is a tool. It can help you find meaningful patterns for complex problems.

Data mining is not a magic wand. It is critical to understand your data and problem space. Predictive analytics will not not automatically discover solutions without guidance. The relationships found via data mining are not necessarily causes of an action or behavior.

Here is a concrete example on how government could use data mining. It is a very human problem. 

The United States foster care systems are run by state (or county) governments, but there are federal laws which regulate these systems. This is the story of one federal law that is suppose to help vulnerable children.

In 1994, Congress passed the Multiethnic Placement Act (MEPA). The main goal of this law was to help black and other minority children. They were dramatically over-represented in foster care. 54% of foster children waiting to be adopted were black. They waited a long time to be matched to adoptive parents. A black child's likelihood of being adopted was 1/5 of a white child.

Prior to this law, the standard practice of social workers was to match foster children to families of the same race or ethnicity. Black foster children were only placed in black families. There weren't enough black families in the foster care system for all the children.

So MEPA tried to tackle the problem in two ways.

1. Increase the pool of possible parents: MEPA  prohibited foster care from using race, color or national orgin to delay or deny a foster child's placement or adoption. White parents could adopt black children. Black parents could adopt white children.
2. Increase the pool of possible parents: State governments were required to recruit additional foster and adoptive parents in different racial and ethic backgrounds.

September 2007, the U.S. Civil Rights Commission held a hearing on MEPA. Was this law working? Did this law make life better for black children in foster care? Was MEPA being enforced?

May 2008, Evan B Donaldson Adoption Institute published their response in a 58 page report. The table below is from this report.

Black children are still disproportionately represented in the foster care system. Black children are in foster care nine months longer than white children. State governments don't appear to have recruited many potential parents from different backgrounds.

This report made me think. State governments are struggling to recruit additional minority families for foster care.

What if the state governments gathered every descriptive statistic on its population and foster care system? 

Then they could mine the data to discover the important variables with adoption for black foster children.  They could use these factors to try and build recruitment campaigns.

Maybe they could uncover a strong relationship between "Percentage Workers  Who Traveled to Work by Public Transportation" and adoption.  The government could focus on purchasing advertisements on buses and trains. Or maybe they could post notices about foster training sessions in grocery stores near a bus stop.

/aw

Image Credit: D Sharon Pruitt / CC BY 2.0