Criminal JusticeNo RCT exists
Does incarceration reduce crime?
The deterrence and incapacitation effects of incarceration have never been tested in a randomized trial. All estimates come from regression discontinuity designs at sentencing guidelines thresholds, which apply only to marginal offenders near the cutoff. For the general prison population, causal estimates are essentially unavailable. Given that the US incarcerates 2.1 million people, this is arguably the largest unanswered question in social science.
Housing PolicyNo RCT exists
Does inclusionary zoning increase or decrease overall housing supply?
Inclusionary zoning mandates that new developments include affordable units, but critics argue this raises development costs and reduces total construction. Multiple observational studies reach opposite conclusions. No randomized experiment has ever been conducted, and the policy is difficult to randomize at the level at which it operates (entire jurisdictions). The debate continues without a credible causal answer.
EducationContested evidence
Do school vouchers improve long-run outcomes for students who use them?
Several lottery-based RCTs of voucher programs (DC Opportunity Scholarship, Milwaukee) find positive effects on graduation and some academic outcomes; others (Louisiana, Indiana) find negative effects on test scores. The heterogeneity across contexts is large and unexplained. Long-run adult outcomes have been measured in only one program (DC). The weight of evidence is genuinely contested rather than settled.
Mental HealthFollow-up too short
Does community mental health treatment reduce homelessness among people with serious mental illness?
Housing First programs (which provide housing without treatment preconditions) have strong RCT evidence for housing stability, but weaker evidence for mental health recovery. The question of whether treatment-first or housing-first sequencing produces better long-run outcomes — across the full distribution of severity, not just stable subjects — remains unresolved. Most evaluations follow participants for 12–24 months, too short for long-run psychiatric outcomes.
International DevelopmentFailed replication
Do microfinance loans improve household welfare for the very poor?
Six simultaneous RCTs published in 2015 (American Economic Journal: Applied Economics) across six countries found consistently null or small effects of microcredit on consumption, business profits, and women's empowerment — overturning two decades of enthusiastic claims. Subsequent debate has focused on whether specific program features (group lending, longer terms, savings linkages) can produce positive effects. The evidence remains fragmented with no clear answer for which populations and conditions microfinance helps.
Public HealthNo RCT exists
Does paid sick leave reduce infectious disease transmission in workplaces?
Economic theory and survey evidence both suggest workers without paid sick leave come to work sick, spreading disease. But no RCT has ever tested mandatory paid sick leave's effect on actual transmission outcomes. Studies using state policy variation find mixed results depending on the outcome measured (cases, hospitalizations, or days sick) and comparison period. COVID-19 generated enormous natural experiment data that has not been fully exploited.
Criminal JusticeContested evidence
Does cash bail reform reduce crime or court appearance rates?
Multiple jurisdictions have eliminated or reduced cash bail; early observational studies show minimal effect on court appearance rates and crime, but all rely on pre-post designs without adequate controls. New Jersey's bail reform is the best quasi-experimental evaluation available and suggests null effects on crime, but methodological disputes persist. No RCT of bail policy has been conducted because the constitutional stakes make randomization politically infeasible.
EducationFollow-up too short
Does universal pre-K improve long-run outcomes, not just test scores?
Universal pre-K programs in Boston and Tulsa show large test score gains at kindergarten entry that are well-documented. But there is a longstanding puzzle in early childhood research — the 'fadeout' problem: test score gains from pre-K often disappear by 3rd grade. Whether long-run benefits persist (as they did in Perry Preschool and Abecedarian, which were far more intensive) is unresolved for the lighter-touch universal programs states are actually implementing.
Urban PolicyNo RCT exists
Does public transit investment change land use patterns and reduce car dependence?
Cities spend billions on rail and bus rapid transit based on the assumption that infrastructure shapes travel behavior. But most evidence on mode-shift comes from before-after studies of new lines, which cannot separate the infrastructure effect from simultaneous development. Ridership on new lines often underperforms forecasts. No RCT of transit investment exists; the unit of analysis (an entire urban region) makes randomization impractical.
Public HealthFollow-up too short
Do sugar taxes reduce obesity and diet-related disease?
Philadelphia, Berkeley, and the UK have implemented sugar-sweetened beverage taxes; observational evidence consistently shows reduced soda consumption. But the downstream question — whether reduced soda consumption translates to lower caloric intake (rather than substitution), reduced BMI, or reduced diabetes incidence — has not been established. Weight and metabolic outcomes require decade-long follow-up, and no jurisdiction has sustained a tax long enough to measure health endpoints in a rigorous design.
Labor PolicyContested evidence
Does raising the minimum wage reduce employment?
This is the most contested question in empirical economics. Card & Krueger (1994) found no employment loss from New Jersey's minimum wage increase; subsequent work using county-pair designs (Dube et al.) consistently finds null employment effects; CBO analyses estimate significant job losses from federal increases. The disagreement is not primarily about data but about research design — which comparison group best represents the counterfactual. No jurisdiction has ever randomized the minimum wage.
Voter EngagementContested evidence
Do automatic voter registration laws increase turnout among newly registered voters?
Automatic voter registration (AVR) has been adopted by 20+ states and substantially increases voter registration rates — this much is established. Whether it increases actual turnout among newly registered voters (who differ systematically from those who self-register) is less clear. Studies of Oregon and Georgia find mixed effects on turnout, and the mechanism — whether AVR changes civic engagement or simply formalizes the registration status of already-engaged citizens — is unresolved.