A cautionary note on using industry affiliation to predict income

Many literatures investigate the causal impact of income on economic outcomes, for example in the context of intergenerational transmission or well-being and health. Some studies have proposed to use employer wage differentials and in particular industry affiliation as an instrument for income. We demonstrate that industry affiliation is correlated with fixed individual characteristics, specifically parents’ education and own height, conditional on the covariates typically controlled for in these studies. These results suggest that there is selection into industries based on unobservables. As a result the exclusion restriction in many IV studies of this type is likely violated.