In a correctly constructed prediction scale, the more components considered, the greater the differentiation between low-rate and high-rate offenders. The prediction scale tested in this study had two, four, and seven components. Increasing the number of components in the prediction scale increases the effectiveness of the resulting sentencing policy in reducing either crime rates or the number of offenders incarcerated, or both. In any jurisdiction, there may be debate about what characteristics should be considered morally acceptable components of the prediction scale. Most who will accept the concept of selective incapacitation will also accept using prediction factors derived from the characteristics of the current offense and the defendant's prior criminal record. Unfortunately, the few characteristics of the current offense included in the study's data base did not distinguish between high- and low-rate offenders, except for the charged offense. Although most offenders who commit either robbery or burglary are also likely to have committed several other types of offenses, defendants charged with robbery are much more likely to be high-rate robbers than defendants charged with other crimes. Other characteristics of the current offense that might be tested for their association with offending rates are type of weapon, whether the offender was alone or working with a group, or the degree of planning for the offense. In determining which factors are acceptable, it should be remembered that excluding predictors that are statistically correlated with individual offense rates will increase the number of offenders that must be incarcerated to achieve a given level of crime. Four references are included, along with graphic and tabular data. (Author summary modified)