The Usefulness of a General Risk Assessment, the Static Risk Assessment (SRA), in Predicting Pretrial Failure: Examining Predictive Ability across Gender and Race

Abstract

Approximately 60% of those in U.S. jails are awaiting trial without having been convicted of the crime for which they are being detained. We examine whether the Static Risk Assessment, a tool not specifically designed for pretrial outcomes, can predict pretrial success. As implemented in Tarrant County, Texas, the tool is predictive of failure to appear and new offending within 90 days of release pretrial using Logistic regression and area under curve analyses. Results hold across gender and race.

Keywords:

• Pretrial risk assessment
• failure to appear
• pretrial release decisions

Disclosure Statement

No potential conflict of interest was reported by the authors.

Notes

1 An anonymous reviewer noted the recent suggestions that electronic monitoring may have socioeconomic bias as well in instances where defendants are required to pay fees for such monitoring.

2 Code was developed to search the text for words and phrases indicative of failure to appear (such as “failure,” “fta,” “fail to appear,” “failed,” “did not appear,” “no appearance”). Extensive “hand checks” followed to ensure the code accurately captured warrants for which two co-authors jointly were confident were FTA warrants. As such, the current study may underestimate the number of warrants for FTA, in the event the text was not detailed enough to be deemed an FTA, making the current study’s findings conservative; meaning the tool may perform slightly better than indicated herein with respect to predicting FTA.

3 Additionally, we conducted survival analysis models using Cox Regression, censoring all those who did not exhibit pretrial failure (FTA or new offending) by 90 days. Results of those models were substantively identically to the 90-day models presented within the text, and therefore not included for brevity.

4 Due to extremely small cell counts this analysis was not possible for the 5-category SRA risk measure. As mentioned before, only one female in the sample was classified as high-risk violent.

5 Of note, personal communication with Tarrant County Criminal Courts Administration indicated 90 days to be adequate follow-up to assess pretrial failure for all cases. Nonetheless, the limitation of not having the court date information exists.

6 Unfortunately, data do not permit us to calculate the actual cost savings as we do not have release dates for all defendants that remained housed beyond our 4-day window. Nonetheless, the cost savings of relying on the SRA would be substantial.