Sensitivity of Error and the Burgess Method

• In Event: The Role of Risk/Needs Assessment in the Correctional System

Abstract

In a recent study, “The Impact of Error in Offender Risk Classification”, Aaron Ho (2013) employed a novel method of measuring error in offender risk classification instruments. The study, which sets the theoretical foundation for the current study, found that classification outcomes in risk assessment instruments are highly sensitive to human mistakes.
In constructing a risk assessment, Burgess employed a simple method that assigned a “1” to an individual’s risk score for exhibiting a certain risk factor, and a “0” if this risk factor was absent. Such method was criticized by its successors for failing to account for the differential weights of each factor, and for having too many risk factors. Glueck then revised the Burgess method so that the individual weight of each risk factor is calculated, and so that the number or risk factors in these instruments is reduced. While the Burgess method is often discredited as being unscientific and primitive, it has nonetheless proven itself to be an effective classification instrument.
The current study argues that shrinkage alone does not explain the divergence in validity for these two methods. The sensitivity of error has been neglected.

Author

• Aaron Ho, New Jersey City University