“…To deal with these issues, existing empirical applications make use of the exogenous variation in the supply of PC induced by a policy reform over time and across geographical areas. Then, they measure the substitution effect at the level of the GP practice or small geographical area by using a linear panel data model (Dolton & Pathania, 2016; Dusheiko et al, 2011; Fortney et al, 2005; Pinchbeck, 2019; Whittaker et al, 2016) or a Poisson model (Iezzi et al, 2014; Lippi Bruni et al, 2016) and a difference in differences approach, which is corroborated by a propensity score matching or an instrumental variable approach in some applications. These studies find evidence of a substitution effect at the GP practice level, although the patient level substitution effect might be different due to ecological fallacy, and the effect in the total population might be different from the effect in areas targeted by the policy.…”