The Pitfalls of Prediction

• NCJ Number: 240702
• Journal: NIJ Journal Issue: 271 Dated: February 2013 Pages: 34-40
• Author(s): Greg Ridgeway
• Date Published: February 2013
• Length: 7 pages
• Agencies: NIJ
• Publication Type: Issue Overview

Annotation

This article discusses the advantages of using the latest scientific developments to make reliable predictions about crime.

Abstract

Research has shown that criminal justice researchers, practitioners, analysts, and other stakeholders often make common mistakes in when making predictions about crime and criminals. These common mistakes are 1) trusting expert predictions too much; 2) clinging to what was learned in Statistics 101; 3) assuming one method works best for all problems; 4) trying to interpret too much; 5) forsaking model simplicity for predictive strength - or vice versa; 6) expecting perfect predictions; and 7) failing to consider the unintended consequences of predictions. This article discusses each of these mistakes and how the use of recent scientific developments could be used to avoid them. Some of the ways to avoid these mistake include constructing and using predictive models, be willing to use different prediction methods based on the particular prediction problem, focus on the accuracy of the predictions, and use models that are simple to understand and simple to implement. Figures, table, and notes