Many universities in the US offer on-campus housing opportunities to incoming as well as already enrolled students. Recent research has theoretically as well as experimentally shown that the most common student assignment mechanism used in the US is subject to serious efficiency losses. In this paper we first show that a particular mechanism which is currently in use at the MIT for about two decades is in fact equivalent to a natural adaptation of the well-known Gale-Shapley mechanism of two-sided matching theory. Motivated from the increasing popularity and success of the Gale-Shapley mechanism in a number of markets, we next experimentally compare the performances of the MIT mechanism with that of the leading theory mechanism Top Trading Cycles. Contrary to theory, the MIT mechanism
performs better in terms of efficiency and participation rates, while we observe no significant difference between the two mechanisms in terms of truth-telling rates.