Robust inference in instrumental variable models

Instrumental variable methods are commonly used to draw causal inferences in nonexperimental settings. A significant threat to the validity of these methods is the presence of outliers in the data. This thesis contributes by developing testing procedures that enable outlier robust inference in instrumental variable models. It demonstrates how the new robust tests can be applied in practice by revisiting several published empirical studies. In some cases, the robust tests give vastly different results compared to the original studies, highlighting the importance of the new methods.