Rosanna Smart

RAND Corporation Website

Rosanna Smart is Economist at the RAND Corporation and Affiliate Faculty at the Pardee RAND Graduate School. She holds a Ph.D. in Economics from the University of California, Los Angeles.

Voting History

Pretrial detention

Reducing the number of people detained pretrial will lead to a net increase in crime in the medium- to long-term.

Vote Confidence
Strongly Disagree 8
Median Survey Vote Median Survey Confidence
Disagree 7
Comments

Several rigorous studies have found pretrial detention to have negative consequences for those detained, and little evidence that pretrial release leads to higher likelihood of re-arrest, subsequent crime, etc. – with the exception of increased pretrial crime and failure to appear (e.g., Dobbie et al., 2018), which I am considering short-run effects.

Ending the use of cash bail will meaningfully reduce the number of people detained pretrial.

Vote Confidence
Agree 7
Median Survey Vote Median Survey Confidence
Agree 7
Comments

This would depend on what mechanism, if any, is put in places as a replacement to cash bail. However, looking at states and jurisdictions that have reduced or ended the use of cash bail (e.g., New Jersey, Philadelphia), these policies do seem to reduce the number of people detained pretrial (although I have some uncertainty around the magnitude that can be causally attributed to the end of cash bail).

Using risk assessment to inform detention decisions will meaningfully reduce the number of people detained pretrial.

Vote Confidence
Agree 4
Median Survey Vote Median Survey Confidence
Neutral/No Opinion 5
Comments

While the use of risk assessment tools to make pretrial detention decisions certainly has the potential to reduce the number of people detained, again these effects would depend on the algorithm/tool used (e.g., purpose, inputs, etc.), the extent to which directives for classifying risk levels are designed to reduce detention, the extent to which the application of such a tool actually affects pretrial decisions (eg, whether tool recommendations are consistently followed [Stevenson, 2018]), etc.