Thanks for this example Sam, though I need more guidance to understand it fully. How do you define "nationalism" and are you conceptualizing this as a country-level variable or an individual level variable? If a country level variable, it is indeed a challenging exposure to evaluate. Maybe it is worth breaking down into the components for which you can conceptualize an intervention to modify. For example, if you specifically hypothesize that communities with high in-migration rates from diverse places of origin experience health benefits from that in-migration, that is easier to evaluate. If you hypothesize that people residing in places with enforcement of immigration rules that is particularly aggressive experience worth health, that is also potentially testable (though not easy...one of your classmates proposed this).
There is some controversy in the causal inference literature about how broadly you can frame the question with a counterfactual perspective (see Krieger and Davey Smith's recent commentary responding to Vandenbroucke and Broadbent). My view is that if you can tie the exposure to a specific policy, it is amenable to evaluation of the health effects. Sometimes we don't have the data we need though!
Maria