Can commitment devices improve usage of online weight loss tools? A field experiment

Background A commitment device is some voluntary, deliberate arrangement an individual makes in order to bind their future choices and influence their future behaviour, in pursuit of some long run benefit. This may take the form of a public pledge (reputational commitment) or some monetary cost (financial commitment) that encourages individuals to stay on track with their behaviour change goals. Commitment devices are widely applied in the health sector to bring about desired behaviour change such as weight loss, but their effects are largely unproven. Aims This paper asks whether commitment devices can increase the usage of an online weight loss tool to bring about more consistent self-monitoring of behaviours. Method An online field experiment tests the causal effects of commitment devices on weight loss. 364 participants were recruited from the client base of an online weight loss service over July – November 2013, and randomly assigned to one of three groups: limited commitment, financial commitment (monthly subscription fee) and reputational plus financial commitment (monthly fee plus naming one person who would verify their weight loss). The outcome variable is usage of the online account, measured by frequency of food journal and self-reported weight entries over a one-month period. Results The ‘limited commitment’ group use their accounts more frequently than the financial and reputational commitment groups. Being offered the reputational commitment reduces usage. These results run counter to the hypothesised relationship between a commitment strategy and behaviour change. Conclusion Commitment devices may be better characterised as a signal of some innate motivation at the time of take up, rather than a substantive instrument for behaviour change, on average. However, heterogeneity of treatment effects may play an important part in explaining how and when commitment devices are most effective. If commitment devices can work well for identifiable sub-groups, there may yet be a targeted role for them in the public policy toolkit.