The Mathematics of Risk Classification: Changing Data into Valid Instruments for Juvenile Courts

The primary goal of a risk scale is to predict—not
to explain, but to predict. Individual items are placed
on the risk instrument because they are believed to be
better at predicting the criterion (which in the work
that follows is recidivism to juvenile court) than are
other available measures. Some risk instruments are
developed subjectively, by asking knowledgeable professionals
to select items their experience tells them
are predictive. Other instruments are developed empirically,
using statistical analyses to select among a set
of available data elements that capture various attributes
of the youth and his/her juvenile court history.
Statistically, when an item is a significant predictor of
the criterion, it improves the power of the instrument
to predict and should be used in the scale.

This report is meant to help juvenile courts develop
practical risk-screening instruments. Courts
increasingly are using some method of risk classification
to assist in assignment of youth to differential
service/supervision programs. A comparison of commonly
used or advocated risk classification methods
may provide courts with guidance in selecting a
method for developing a screening instrument. This
report assesses the strengths and weaknesses of several
prediction methods in the context of juvenile courts'
risk classification needs, based on analyses of one
court's data that were submitted to the National
Juvenile Court Data Archive maintained by the
National Center for Juvenile Justice.