What we know about SMS reminders in public-sector programs

Across vaccination, court dates, taxes, voting, and benefits — when do text-message reminders move behavior, and when do they fail?

10 experiments synthesized · 8 positive, 1 mixed, 1 null, 0 negative

Few interventions have been tested as widely as SMS reminders. They are cheap to deploy, easy to randomize, and the marginal cost per recipient is near zero — which means agencies have run them everywhere from vaccination clinics to court systems to tax authorities. The result is one of the largest natural experiments in behavioral public policy.

The pattern that emerges is consistent: SMS reminders work when they reduce a specific friction at the moment of decision, and produce near-zero effects when they merely add information to a problem people already understand. The 'what' of the message matters less than whether there is a concrete action the recipient can take, and whether the message arrives in time to support that action.

Effect sizes typically range from null to 5 percentage points on the behavior of interest. That is small, but at the cost of a text message it remains one of the most reliably cost-effective tools available — provided the agency runs the experiment honestly and reports the null results too.

Takeaway

Run an SMS reminder pilot when the underlying friction is forgetting or missing a deadline. Expect a null when the message asks people to do something they were already not planning to do.

The underlying experiments

Positive findings

8 experiments

Mixed findings

1 experiment

Null findings

1 experiment

Open questions

  • When does personalization (using the recipient's name, condition, or balance) materially improve over a generic reminder?
  • How long does the effect persist — do recipients habituate after the third message?
  • Which populations respond more weakly, and is the gap explained by phone access, language, or trust?

If you've run a pilot that speaks to any of these, submit it to the registry.