Skip to main content
Open Access Publications from the University of California

Department of Statistics, UCLA

Department of Statistics Papers bannerUCLA

Conducting a Randomized Field Experiment for the California Department of Corrections: The Experience of the Inmate Classification Experiment


The California Department of Corrections (CDC) has big problems. It houses more prisoners than any other stateís corrections system: 160,000 inmates in 33 prisons and over 50 other facilities. The costs are enormous, including an average of about $30,000 per inmate per year and about $150,000 for each new cell built. The prisons are also very difficult to run. Each year about 25% of the inmates engage in some form of misconduct serious enough to document, and 2.5% commit an offense that would probably be a felony in the outside world.

One of the ways in which the CDC attempts to make the best use of its resources is to assign prisoners to facilities with varying levels of ìsecurity.î Higher levels of security place more restrictions on inmates because greater human and physical resources are brought to bear. There are higher staff to inmate ratios and physical surroundings that reduce the chances of serious infractions. For example, in some high security facilities, inmates are housed one to a cell and are only allowed into the exercise yard in small groups. However, higher security facilities are more costly to build and run. It is important, therefore, to place each inmate in the least restrictive setting necessary to insure the well-being of that inmate, other inmates, and CDC personnel.

This paper discusses the implementation of a very large, randomized field experiment testing two different procedures through which inmates could be assigned to facilities with different security levels. By most any measure, the experiment was implemented in a textbook fashion and led to useful results. The question addressed is how this success was achieved.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View