We know that instrumental variable (IV) estimates a causal effect if the instrument satisfies a monotonicity condition. When this condition is not satisfied, we only know that IV estimates the difference between the effect of the treatment in two groups. This difference could be a very...