Public Policy Experiments without Equipoise: When Is Randomization Fair?

Area of Science:

• Public Policy Evaluation
• Randomized Controlled Trials (RCTs)
• Ethics in Research

Background:

• Government agencies and nonprofits increasingly use RCTs for policy evaluation.
• Random assignment is typically considered fair when there is equipoise (uncertainty about intervention effectiveness).
• Debate exists on the fairness of randomization when an intervention is expected to be superior.

Purpose of the Study:

• To investigate the ethical permissibility of randomization in public policy RCTs when equipoise is absent.
• To identify conditions under which random assignment to a superior intervention is ethically fair.
• To critically evaluate alternative study designs for allocating beneficial interventions.

Main Methods:

• Theoretical investigation of ethical principles in randomization.
• Analysis of conditions for fair allocation of scarce resources.
• Rejection of claims regarding the fairness of stepped-wedge and uneven randomization designs.

Main Results:

• Two sets of conditions are identified where random assignment to a superior intervention is deemed fair.
• The fairness argument is linked to the allocation of scarce, beneficial interventions via lottery.
• Alternative designs like stepped-wedge and uneven randomization are found not to be fair alternatives for allocating beneficial interventions.

Conclusions:

• Randomization in public policy RCTs can be ethically justified even without equipoise, under specific conditions.
• Fairness in allocation hinges on the potential benefits and scarcity of the intervention.
• Alternative study designs do not provide a fair substitute for justified randomization when allocating superior interventions.