Can scalable behavioral science strategies create lasting exercise habits? We investigated this question by conducting a meta-analysis of a massive randomized controlled trial with 20 separate interventions (and 57 different treatment arms) as part of a 28-day digital program designed to promote physical activity. The sample was drawn from a large national gym chain and included at least 600 participants per treatment arm. The 20 interventions tested different behavioral science insights to identify ways to create lasting gym habits. The simplest treatment included elements that have been previously shown to effectively create healthy habits: planning (Milkman et al., 2011; Duckworth et al., 2011), receiving reminders (Karlan et al., 2016), and receiving repeated incentives for exercising (Beshears et al., 2018; Charness & Gneezy, 2009; Royer et al., 2015). In this treatment group, participants registered online and were prompted to create a workout schedule for the next week. During the 28-day intervention period, these participants received text message reminders before each scheduled gym visit, weekly emails with their workout schedules, and points redeemable for roughly $0.21 in Amazon Cash each time they went to the gym. Other interventions built on this baseline treatment by adding additional behavioral science ingredients. For example, some interventions varied the size and types of incentives that participants could earn (e.g., incentivizing returning to the gym after missed workouts, incentivizing planning workouts), while others added additional behavioral science ingredients to the treatment (e.g., describing social norms about exercise). The key dependent measure was participants’ gym attendance (tracked by our gym partner) during the intervention period and up to 52-weeks post-intervention.
This presentation will focus on one of the unprecedented features of this research: It allows us to test and compare the efficacy of twenty different behavioral science interventions, each tested simultaneously on a comparable population. We will present results from a meta-analysis of the interventions, pinpointing which interventions proved most effective at producing enduring behavior change (i.e., sustained increased exercise over the post-intervention follow-up period). We will identify which strategies worked best compared to the baseline condition and a holdout control condition offered no treatment, but also which work best compared to one another. Additionally, by combining the results into a meta-analysis, we will be able to identify whether the effects of our various behavior change interventions differed across sub-populations or groups (e.g., men, women, frequent exercisers, infrequent exercisers, those under 30, etc.). These findings will help inform future interventions using incentives and other behavioral science insights to build lasting habits.
