NCJ Number
183346
Journal
Crime and Delinquency Volume: 46 Issue: 3 Dated: July 2000 Pages: 401-424
Date Published
July 2000
Length
24 pages
Annotation
Randomized field experiments can provide trustworthy evidence about the effects of interventions, but investigators have limited control over important features of the field experiment, e.g., program implementation, receipt of intended and unintended services, and retention of study participants.
Abstract
The introduction to this article suggests that many of the best instances of field experiments and quasi-experiments were products of persistence and creative exploitation of the circumstances surrounding a particular study. The introduction also highlights the notion that the best instances of quasi-experimentation apply the logic of experimentation rather than simply laying a static design over a program. The main claim advanced in the body of the article is that experimental methods have been transported into field settings as if the central feature of the experiment is randomization of units to groups. There are several other critical features that are controlled in true experiments. These features (e.g., specification and delivery of the independent variable) are usually not controlled in field settings. As such, even when randomization is successfully implemented, other sources of variation are left uncontrolled. To overcome the influence of these uncontrolled sources of variation, this article describes an enhancement of the measurement process. 57 references