It was tested. Secure Communities checked the immigration status of everyone local police arrested, across more than 3,000 counties, and produced no observable effect on crime.

This is the rare policy question with a clean natural experiment. Secure Communities rolled out county by county over several years, which let researchers compare places that had it against places that did not yet. Thomas Miles and Adam Cox studied the rollout across more than 3,000 counties, using detainee counts obtained by FOIA, and published the result in the Journal of Law and Economics: no observable effect on crime rates. Not on homicide, not on rape, not on robbery, not on aggravated assault.

It has been replicated. A separate team — Treyger, Chalfin and Loeffler — found the same nothing. Studies of 287(g), the program that deputizes local police as immigration agents, reach the same place: a Cato working paper and an NIJ-funded study both found no statistically significant reduction in violent or property crime.

The mechanism failure is not mysterious. These programs mostly catch people stopped for misdemeanors and traffic offenses, not the violent offenders they are sold on. And they make immigrant witnesses and victims afraid to call police at all, which subtracts from public safety exactly as fast as enforcement is supposed to add to it.