Policy makers frequently use education as a welfare policy instrument. We examine one such case, where students from large and financially constrained families, were given the opportunity to be transferred to university departments in their hometown as part of the social policy of the Ministry of Education in Greece. Multiple law changes meant that there was a large and quasi-random variability in the number of transferred students over time, which was orthogonal to the quality of receiving students. We construct a novel dataset by linking students' characteristics and pre-university academic performance with their university academic record until graduation for the top economics department. We present consistent evidence showing how a social policy that is meant to help poor families and to alleviate inequalities has gone bad educationally, by lowering the academic performance of receiving students. [This paper was produced as part of the Centre's Education & Skills Programme. Financial support was provided by Drasi II (2019/2020) AUEB.]