Analyses using aggregated data may bias inference. In this work we show how to avoid or at least reduce this bias when estimating quantile regressions using aggregated information. This is possible by considering the unconditional quantile regression recently introduced by Firpo et al (2009) and...