Existing instrumental variable methods require latent variables to be unidimensional. In this research, we relax this restriction and develop a characterisation of the resulting sharp identified set. Applications include demand analysis and occupational choice models.
Centre for microdata methods and practice