Treatment and Control Groups in a Dynamic Setting

The paper considers a version of the question of how to define treatment and control groups in a dynamic setting where treatments can occur at any time (but only once). The version considered pre‐supposes that treatments as well as outcomes can be conceptualised as events occurring in temporal locations of a discrete time axis. It is proposed to think of effects as being dependent on both the time when and the time since the treatment occurred. The paper develops corresponding definitions of treatment and control groups, and proposes a notion of ‘comprehensive treatment effect’ that takes into account how treatment and control groups are generated. Based on this notion, the paper discusses causal interpretations that do not pre‐suppose a potential outcomes framework.