What is Randomized Encouragement Design?
Randomized Encouragement Design (RED) is an experimental approach that randomises invitations or nudges (encouragements) to take up a program when you cannot directly randomise the program itself or when people may not comply. It lets you estimate the causal effect of actually using the service by comparing groups that were more or less likely to take it up.
A Randomized Encouragement Design randomly assigns people to receive an encouragement (for example a phone reminder, personalised message, or extra outreach) that raises the probability they will use a service or attend a programme. Because the encouragement is random, differences in outcomes between encouraged and non‑encouraged groups are causally attributable to the encouragement. Under additional assumptions, researchers can use that random variation as an instrument to estimate the effect of actually taking up the treatment (the Local Average Treatment Effect, often called the complier average causal effect). RED is useful when you can't force participation, when uptake is imperfect, or when ethical or practical issues prevent direct random assignment of the treatment.
Usage example
A council wants to know whether attending a community information session improves vaccine uptake but cannot force residents to attend. They randomly send a subset of residents a translated SMS with personalised encouragement and easy sign-up instructions. Later, they compare vaccination rates and use the random assignment of the SMS as an instrument to estimate the effect of attendance on vaccination among those who attended because they were encouraged.
Practical application
Why it matters: many real-world programmes depend on voluntary participation and face imperfect compliance. RED preserves the benefits of randomisation without needing to withhold a potentially useful service. Practical uses include testing outreach strategies (e.g., multilingual invites), estimating the impact of using a service when you can’t assign use directly, and comparing different encouragement approaches to boost participation. In community engagement and surveys, RED can show whether translated reminders, phone calls, or extra support actually increase response rates and whether those increased responses change measured outcomes — helping teams choose effective, inclusive outreach strategies without creating multiple full-scale rollouts.
Key practical points: design a clear encouragement that plausibly affects uptake, measure who actually takes up the service, ensure sample sizes are large enough (encouragement typically produces partial uptake), and plan analysis using intent-to-treat (effect of assignment) and instrumental-variable methods (to recover the effect of uptake among compliers).
FAQ
How is RED different from a standard randomised controlled trial (RCT)?
In an RCT you randomise who receives the treatment itself. In RED you randomise an encouragement to take up the treatment. RED is used when you cannot or do not want to force treatment assignment, or when uptake will be imperfect—encouragement preserves causal identification through random assignment while accepting that only some people will comply.
What assumptions are needed for RED to give a valid estimate of the effect of treatment?
The main assumptions are: (1) random assignment of encouragement, (2) exclusion restriction — the encouragement affects outcomes only through its effect on uptake (no other direct effects), and (3) monotonicity — encouragement does not make anyone less likely to take the treatment. If these hold, you can estimate the Local Average Treatment Effect for compliers.
What results do I get and how do I interpret them?
You typically report an intent-to-treat (ITT) effect (the impact of being assigned encouragement) and, when possible, an instrumental-variable estimate of the complier average causal effect (the effect of actually taking up the treatment among those who complied because they were encouraged). ITT shows the real-world impact of the outreach itself; the IV/CACE estimate shows the impact of the treatment for people who respond to that outreach.
When should I not use RED?
Avoid RED if the encouragement will directly change the outcome independently of treatment (violating exclusion), if you expect strong spillovers between encouraged and non‑encouraged groups, or if the encouragement has a negligible effect on uptake (weak instrument). In those cases alternative designs or stronger interventions are preferable.