Supervised machine learning algorithms fail to perform well in the presence of endogeneity in the explanatory variables. In this paper, we borrow from literature on partial identification to propose deep causal inequalities that overcomes this issue.…Read More