University admission is organized very diff erently throughout the world. Prospective students in the U.S. as well as graduate students in the UK need to apply directly at their preferred universities. All undergraduate students in the UK, however, need to apply with the central authority "UCAS" (Universities & Colleges Admission Service) and merely indicate a preference list of universities. In Germany, the central clearinghouse ('ZVS') allocates students to universities in those subjects which exhibit a shortage of university places. As university admission procedures diff er substantially between countries and subjects, it is straightforward to ask: Is the centralized or decentralized procedure better suited to match prospective students to universities?

To answer this question, I exploit the decentralization of university admission in the field of German law studies as a natural experiment. In this subject the centralized admission procedure via the central clearinghouse was replaced by a decentralized procedure in 2002. Using administrative data on all students at German universities and applying a diff erences-in-di fferences strategy, the efficiency and quality of the student-university matching is compared between the centralized and the decentralized procedure. My outcome variables measuring matching efficiency and matching quality are (i) the number of first-year students, (ii) the number of unassigned university places, and (iii) the dropout rate.

During the reform process, admission to all law schools except the ones in the largest German federal state of North Rhine-Westphalia was decentralized. Therefore, I use the regional variation over time between law schools in and outside of North Rhine-Westphalia for identi fication. As a second control group, I employ medical schools because admission to medical schools was centrally administered over the entire observation period. Important to note is that both changes induced by the decentralization process - first, the abolishment of the central clearinghouse and, second, the abolishment of admission restrictions - are considered by my analyses.

The results of my study show that the number of fi rst-year students has increased and the number of unassigned places has decreased after the decentralization. This increase in the matching efficiency is mainly driven by enabling law schools to abolish all admission restrictions. My estimates with respect to the drop-out rates are not signi ficantly affected by the decentralization. Nevertheless, a comparison between the eff ects for all treated law schools and for the subgroup of law schools which kept admission restrictions suggests that abolishing admission restrictions could be associated with increasing drop-out rates.