What the evidence says
The best-identified estimate on this page comes from a single study of "add-on" gun laws — statutory sentence enhancements applied when a gun is used in a crime — using the addition of extra prison time as the policy lever and gun robbery rates as the outcome (Abrams, 2012). In its preferred specification, the law is associated with a 6% drop in gun robbery rates within three years of the law's effective date, but that specific estimate is statistically insignificant. Across a broader set of specifications, most coefficients are negative and significant at the 10% level (though not at 5%), the pattern holds whether the analysis uses state-level or agency-level data, and the estimated deterrent effect grows in magnitude over the first three years after adoption. The same study finds no evidence of a gradual trend break in crime after the law takes effect — the shift appears as a level change rather than a slope change, consistent with an immediate rather than accumulating response.
A separate strand of evidence addresses mandatory minimums generally rather than a specific gun-law enhancement. A review characterizes mandatory minimum sentencing as intended to work through deterrence and incapacitation, but reports that both the studies available when the policies were adopted and those published since found the crime-reduction effect to be minimal, non-existent, or counterproductive (Martínez-Garay, 2023). This is a broader claim about mandatory minimums as a class of policy, not a re-analysis of the gun-law estimate above.
On drug sentencing specifically, a simulation model finds a cost-effectiveness result rather than a null result: a million dollars spent extending sentences to mandatory-minimum lengths does reduce cocaine consumption, but less than a million dollars spent on the pre-mandatory-minimum mix of arrests, prosecution, and sentencing, and less than a million dollars spent on treatment for heavy cocaine users. The same treatment comparison holds for cocaine-related crime — treatment spending reduces it more per dollar than mandatory-minimum sentence extension (Caulkins, 1997). No magnitude is quantified beyond the relative ranking of the three spending options.
One randomized comparison bears on substance-specific behavior under a mandatory-minimum alternative: assignment to a DUI court, compared with mandatory minimum sentencing, produced no measured effect on the number of days in the prior 30 that participants reported drinking five or more alcoholic beverages (Campbell, 2026). This rests on one RCT with self-reported outcomes rather than a biochemical measure.
A separate body of evidence looks at public response to mandatory minimums rather than crime outcomes. Surveyed on a joint-criminal-enterprise murder scenario carrying South Australia's 20-year mandatory non-parole minimum, the public's own recommended sentences for secondary offenders often departed from that statutory floor, and respondents more often preferred desert-based, proportional sentencing over the mandatory minimum for secondary offenders (Deegan, 2026). A focus group examining the same scenario almost universally rejected a mandatory sentence in favor of a sliding scale tied to individual culpability, and the same source reports that current mandatory sentencing arrangements for this offense are seen to undermine public confidence and create a perception of injustice.
Where the evidence splits
A general review of mandatory minimum sentencing describes crime-reduction effects across the literature as minimal, non-existent, or counterproductive. Set against that, the add-on gun law study finds a consistently negative (crime-reducing) direction across most specifications, and the drug-sentencing simulation finds mandatory minimums do reduce cocaine consumption and cocaine-related crime, just by less per dollar than the alternatives. These are not estimates of the same effect: one is a general characterization of mandatory minimum sentencing as a class, the other two are estimates for a specific enhancement (add-on gun laws) and a specific substance (cocaine) under a specific model. The claims on this page do not state why the general characterization diverges from the specific negative-direction estimates — the split is unexplained by the sources on this page.