Randomized Experiments on Crime and Justice

An experiment investigates the effect of changes in one factor (the independent variable) on another (the dependent variable). The independent variable is under the control of the researcher. A randomized experiment is one in which people (or other units) are assigned to conditions according to a table of random numbers, with every person having the same probability of being assigned to each condition. These experiments are especially useful for testing causal hypotheses. Their unique advantage over other methods is their high internal validity, or high ability to demonstrate the effect of one factor on another. The randomization ensures that people assigned to one condition are equivalent in every possible way to those assigned to another condition, within the limits of statistical fluctuation. Despite their methodological advantages, very few randomized experiments have been carried out on crime and justice topics. Most have investigated the effects of providing special help for offenders, and in most cases this help proved no more effective in reducing reoffending than did existing alternative treatment methods. It is difficult to arrange randomized experiments because program administrators are unwilling to relinquish control of assignment to experimenters, and because of ethical problems of denial of treatment. Randomized experiments are most feasible when the effects of a treatment are unknown and when it is impossible to treat everyone. Because of their high internal validity, hypotheses should be tested, and technologies should be evaluated, using randomized experiments whenever possible.