We model a higher education system that admits students according to their admission signal (e.g., matriculation GPA, SAT), which is, in turn, affected by their cognitive ability and socioeconomic background. We show that subsidizing education loans increases neither human capital stock nor aggregate consumption, but only yields income redistribution mainly among the upper class. We show that the policies aimed at compensating for poor socioeconomic background result in a higher aggregate consumption, as well as income redistribution from top to bottom. We test the model using a unique dataset that includes proxies of socioeconomic background and cognitive ability. Results show that the high school matriculation GPA is a weak predictor of academic achievements. We demonstrate that, while the high school matriculation GPA is explained by proxies of cognitive ability and socioeconomic background, academic GPA is solely explained by cognitive ability proxies. Finally, the lack of a matriculation certificate is associated with a poor socioeconomic background.