EvalCompassEvalCompass
Causal inference & experimentsCausal inferenceRandom assignmentExperimentDevelopment economicsEffect investigation

Randomized Controlled Trial

Randomized Controlled Trial
Also known as: Randomized controlled trial、RCT、Experimental design

Randomly assigns targets to intervention/comparison groups and compares outcomes to estimate effects. Check attrition, non-compliance and between-group influences after assignment. [1]

RCT design checklist — openReferences and checked sections
Difficulty
Advanced
Cost
High
Duration
Illustrative planning durations, not common standards: design 2–6 months; implementation depends on intervention, from several months to multiple years
Content updated: 2026-10-07Source checked: 2026-10-07 (includes AI cross-checking)No documented human reviewIncludes fictional examples

How it works at a glance

Random assignment avoids systematic links between intervention status and participant characteristics. It does not guarantee identical characteristics in finite samples. [1]

Predefine effects and populations, considering assignment units, power, consent and delivery. Even ITT estimates of assignment effects require separate bias examination when outcomes are missing. [1]

Target populationCheck eligibility and participationRandomassignmentIntervention groupAssigned to interventionControl groupAssigned to comparison conditionGroup differenceAssignment effect: ITT
A conceptual diagram of random assignment and comparison of outcomes between assigned groups. Check noncompliance, attrition, and spillovers, and specify the target population and effect being estimated. Random assignment differs from random sampling.

When to use it and when it is unsuitable

When to use it

  • Plan investigation of new social interventions or policies before implementation
  • Design assignment and comparisons beforehand to examine causal effects
  • Sufficient participants and technically/ethically feasible randomization
  • Strong evidence required for development aid or social-policy resource allocation

When it is unsuitable

  • Past interventions cannot be randomized retrospectively. Separately consider new participants or additional measures in existing programs.
  • Withholding benefits is ethically unacceptable, such as emergency humanitarian support
  • Whole-community policies cannot provide controls; consider phased quasi-experimental designs

Participants, roles, and deliverables

Participants and roles

Lead evaluator, statistician, field coordinator and ethics committee. Specialist organizations such as J-PAL may be practical at scale.

Main deliverables

  • Report with ITT, effect sizes and confidence intervals
  • Baseline and endline datasets
  • Preregistered trial plan
  • External-validity and policy discussion

Implementation steps

  1. Set questions and effect hypotheses
    Predefine targets, intervention, comparison, primary outcomes, follow-up and estimand. Record multiple-indicator treatment and analysis plans.
  2. Sample size and power analysis
    Design numbers from meaningful effects, variance, significance, power and expected attrition. For clusters, consider cluster numbers, sizes and intracluster correlation. Explain numerical choices.
  3. Ethics review and preregistration
    Examine rights, benefits/burdens, consent and comparisons; check applicable ethics and registration. Choose suitable registries or analysis-plan publication. Do not treat UMIN-CTR as mandatory for all social experiments.
  4. Baseline and random assignment
    Record methods and implementation and prevent prediction/manipulation of assignment. Record initial traits, but neither identical means nor non-significant baseline tests prove successful randomization.
  5. Implement and follow up
    Faithfully record intervention and control treatments. Monitor attrition and examine differential loss.
  6. Endline and analysis
    Conduct endline after intervention. Base analysis on intention-to-treat, comparing assigned groups regardless of implementation compliance.
  7. Report and examine external validity
    Report effects and confidence intervals. State conditions and populations and honestly discuss generalization limits.

Strengths and limitations

Strengths

  • Proper randomization prevents assignment bias from initial traits and supports causal group comparisons
  • Observed/unobserved traits balance in expectation, not necessarily within each trial
  • Publishing protocols, analysis plans and changes supports transparency and critique
  • Accumulated RCT findings can enter meta-analysis/systematic reviews and cumulative evidence

Limitations

  • High internal validity does not establish applicability elsewhere
  • It is often difficult to secure enough participants, conduct long-term follow-up, or withhold the intervention from the control group.
  • Average effects alone may not explain mechanisms; combine process evaluation
  • Completed assignment cannot change. Ongoing programs require newly randomizable interventions or targets.

Common pitfalls

  • Differential attrition distorts comparisons if lost participants differ. Always compare rates and characteristics in both groups.
  • Trial awareness may change behavior: Hawthorne effects. Take care in social experiments where blinding is difficult.
  • Do not equate significance with meaningful effect sizes. Large samples detect small effects; always report effects and confidence intervals.
  • Neglecting ethical accountability for exclusion risks refusal and reputational harm.

Conditions of use and evidence

  • After randomization, attrition, non-compliance and spillovers threaten internal validity; generalization requires further examination. [1]
  • Check pre-implementation randomization feasibility, sufficient power, no influence on comparison groups and ethical risks. [2]
Fictional example; not actual results

Randomize information methods

A fictional calculation: randomly assign 100 each to ordinary or additional individual information and establish all application status. 100 is not a recommended sample size.

Assigned groupApplicants / assigned peopleApplication rate
Additional information60／10060%
Ordinary information45／10045%
ITT difference60%−45%15 percentage points

Common errors in judgment

Common mistake

Compare only readers of the information

How to revise

Predefine ITT by assigned groups and separately show actual use

ReasonSelecting users loses randomized comparability.

Common mistake

Assume RCTs have no bias after dropping missing outcomes

How to revise

Examine group-specific missingness/reasons and conduct sensitivity analysis

ReasonRandomization does not guarantee equal follow-up probability.

Organize your thinking by writing

RCT design checklist

A practical template is available to fill in and print on A4 paper.

Open worksheet

Related practical guides

Evaluation methods for EBPMWhat is impact evaluation?

Related methods

Difference-in-DifferencesRegression Discontinuity DesignInterrupted Time Series AnalysisSynthetic Control MethodContribution Analysis

References and checked sections

  1. The elements of a randomized evaluation
    Michael Gibson・Anja Sautmann／J-PAL
    Section consulted
    3. Designing the evaluation / 4. Randomization / 5. Data analysis
    How it relates to the text
    After randomization, attrition, non-compliance and spillovers threaten internal validity; generalization requires further examination.
    Source checked: (includes AI cross-checking)
  2. Magenta Book Annex A: Analytical methods for use within an evaluation
    HM Treasury / Evaluation Task Force · 2026
    Section consulted
    A2.1 pp.20–21, How is it used? / Pros and cons: design before implementation, statistical power, control group exposure, ethics
    How it relates to the text
    Check pre-implementation randomization feasibility, sufficient power, no influence on comparison groups and ethical risks.
    Source checked: (includes AI cross-checking)

Editorial and checking information

Content updated: Source checked: (includes AI cross-checking)No documented human reviewIncludes fictional examplesEditorial policy and scope of checkingCite this pageReport a correction

How to cite this page

EvalCompass Editorial Team. “Randomized Controlled Trial”. EvalCompass, updated 2026-10-07. https://evalcompass.jun-nakatani.com/en/methods/rct/
Add the date you accessed the page. Text created by this site is available under CC BY 4.0. Quotations, external materials, and bundled fonts have their own terms of use. When reusing content, credit the source, link to the license, and indicate any changes.

Suggest a correction or improvement to this page

Change history
  • — Checked sources for definitions and conditions and added fictional examples, judgment rationales and completed entries for every field. Revised unsupported historical or uptake claims. Reviewed procedures and cautions, adjusting fixed counts, thresholds and institutional generalizations. Preserved field positions, types and counts while revising fixed conditions and thresholds.
  • — Added Magenta Book Annex A (May 2026) A2.1 solely for checks of pre-intervention randomization, power, comparison-group influences and ethics. Retained definitions, finite-sample/missingness cautions, J-PAL sources, worksheet and examples.