New paper on causal inference and neighbourhood effects on health
Imagine you are interested in understanding whether living in an area with a high poverty rate impacts on your health irrespective of who you are. Typically, you might collect some data on individuals from a variety of neighbourhoods and then examine whether health status varies by the level of poverty.
Such observational data may be biased through unobservable confounding factors or residential sorting patterns. If we wanted to minimise these issues we would take our individuals, randomise their locations and then see how their health responds to their new environment (akin to how new medicines are tested).
The paper proposes using migration as an event in which individuals themselves alter their own environment. Through using matching methods we can minimise confounding factors and more accurately understand how individuals respond to neighbourhood features.
Dr Green, who led the study, said “We hope that our new approach forms an important method for causal inference in neighbourhood effects research.”
“While our study explored the impact of neighbourhood poverty, it has many applications for varying features of neighbourhoods. Imagine being interested in whether living in an area with lots of fast food outlets influences your diet. Comparing the experiences of individuals who moved to areas with lots of fast food outlets against those who did not would help us to separate out whether they are important or not”.