This paper proposes an ℓ1 penalized quantile regression estimator which adapts the Hausman–Taylor instrumental variable approach in order to address the bias resulting from the shrinkage of the individual effects.