Matching in co-operative education programs: An experimental study

Almost all Canadian and many American universities and colleges participate in cooperative education programs whereby each year co-op students alternate between dedicated for-credit work terms and school terms. Many of these programs use a minimum sums algorithm (MS) to match students to jobs. In this first study of the algorithm, we show that it and all its variations may produce unstable outcomes. We compare experimentally the properties of this algorithm and seemingly improved variations with the deferred acceptance algorithm (DA). While the improved versions of MS sometimes lead to more truthful reporting of preferences and increase the likelihood of a stable assignment, they all fare worse than DA. Our data reveal that the superior outcomes associated with DA are the result of both the algorithm itself and the behavioral responses it elicits. The continued use of MS has large adverse consequences for human capital accumulation and lifetime earnings.