The identification of average causal effects of a treatment in observational studies is typically based either on the unconfoundedness assumption or on the availability of an instrument. When available, instruments may also be used to test for the unconfoundedness assumption (exogeneity of the...

This paper shows that import exposure affects voting behavior because it affects local labor markets. We develop a new framework for mediation analysis where one instrumental variable is sufficient to identify three causal effects. Using German data from 1987–2009, we find that import exposure...