Can government programs be improved by applying insights from the behavioral sciences? How large are the effects we can realistically expect when we build behaviorally informed interventions into programs and scale them up?

In 2020, the Office of Evaluation Sciences (OES) was featured in a very large review that found behavioral interventions to be generally effective in improving government programs. Researchers Elizabeth Linos and Stefano DellaVigna, of the University of California, Berkeley, analyzed results of interventions designed and tested by OES and by the Behavioral Insights Team North America. They found these interventions resulted in an overall improvement of 1.4 percentage points across a wide variety of program and policy outcomes, such as the percentage of people contributing to a retirement savings plan and the percentage of people making repairs in order to meet city housing code regulations. This translates to an average improvement of 8.1% on key policy outcomes.

This was the largest review of its kind and gives us a more complete and accurate picture of the effectiveness of behaviorally informed interventions than we have ever had before. The authors analyzed nearly 350 interventions, across a wide range of policy areas, that collectively reached over 23 million people.

The reason this review was able to provide an accurate picture of behaviorally informed interventions is OES’ consistent commitment to report results from all completed evaluations.. All of our results can be found on our website. OES’ project process includes posting analysis plans before we work with data, so that users of our results can be confident that we didn’t cherry-pick results that appear to be significant but in fact reflect statistical noise. Transparency as an evaluation principle helps ensure that, over time, results across studies yield an unbiased picture of what works to make government better.

Additional OES analysis found that 87 percent of the OES interventions analyzed were of no or low marginal cost, meaning that applying behavioral insights can be very cost effective. A second blog on OES’ additional costing analysis can be found here.