Extralegal variables and arrest
Risk dimensions used in guidelines systems have been implicated as contributing to racial (and gender) disproportionalities in America's prison and jail populations. Developers of some systems dealt with invidious risk predictors by purposely ignoring them, resulting in misspecification while n...
| Main Author: | |
|---|---|
| Format: | Electronic/Print Article |
| Language: | English |
| Published: |
1996
|
| In: |
Journal of research in crime and delinquency
|
| Online Access: |
Volltext (doi) |
| Journals Online & Print: | Drawer...
|
| Availability in Tübingen: | Present in Tübingen. IFK: In: Z 31 |
| Check availability: | HBZ Gateway |
| Keywords: |
| Summary: | Risk dimensions used in guidelines systems have been implicated as contributing to racial (and gender) disproportionalities in America's prison and jail populations. Developers of some systems dealt with invidious risk predictors by purposely ignoring them, resulting in misspecification while not eliminating their effects. Eliminating all variables correlated with suspect factors would greatly attenuate power, rendering practical decision-making tools useless. This article shows that one risk-prediction device forming the basis of an operational guidelines system is, in fact, correlated with race and gender but that the approach suggested by some critics does not overcome this: Control factors will remain correlated with the final model even after second-order policy controls are implemented. Further, although the suggested approach is agnostic with respect to the nature of policy controls, these will have considerable practical importance. Illustration is provided. Unbiased models can be estimated with little appreciable loss of predictive utility, and this is demonstrated |
|---|---|
| ISSN: | 0022-4278 |
| DOI: | 10.1177/0022427896033003006 |