Physical activity is a textbook example of time inconsistent behaviour – people continually fail to follow through on their physical activity (PA) intentions. Given this, and the importance of PA in tackling the obesity problem, PA is a popular target for behavioural interventions. However, very few such interventions reach the goal of sustainably improving PA habits. An insufficient understanding of time inconsistency in PA, and in particular its drivers, is likely an important factor in this failure. We seek to address this by designing a longitudinal survey using a theory-based approach that provides novel empirical evidence on the prevalence and drivers of time inconsistency in PA in a general population sample. This survey also allows us to formally test the appropriateness of the popular quasi-hyperbolic model of present bias in this context. The findings of this study will have key implications both for future PA intervention design as well as the development of theoretical models of time inconsistency.