This paper has two objectives. The first is to analyze the effectiveness of the scarcity management strategy used for the Fall 2004 U.S. flu vaccine shortage, based on defining priority groups and calling on self-restraint to favor these groups. The second is to reveal differentiated behavioral responses across categories of individuals and the apparent motivations behind these responses. To do this, we observed the responses of the members of a campus population to two distinct randomized treatments in a designed field experiment during the flu vaccine shortage. Corresponding to the management strategy followed by the Center for Disease Control, one treatment provided information about a sharply reduced number of vaccination clinics (scarcity) and their schedule (deadlines); and the other provided the same information plus an appeal to self-restraint to favor priority groups. Regarding effectiveness of the scarcity management strategy, we find that information about scarcity and deadlines induced a sharp 110% increase in demand, while calls on self-restraint only helped reduce this demand by 38%, resulting in a 31% increase in demand. The distribution strategy chosen, enforced by only soft screening, was not effective. Indeed there was a 17% net increase in vaccines distributed compared to no strategy. An analysis of confidential survey responses from candidates as to being members or not of a priority group suggests that, perversely, the net increase in demand and in vaccines distributed originated entirely in non-priority individuals. Regarding behavioral responses to information about scarcity and calls on self-restraint, we find that it is the non-priority individuals who had used vaccination services the previous year that increased their demand sharply in response to scarcity and deadlines, demonstrated modest self-restraint, and contributed most to cheating. The surprising finding is that the priority population exercised uncalled for self-restraint, thereby canceling among them the increase in demand due to scarcity and deadlines. We use the stated reasons to seek vaccination among non-priority individuals to infer that the strong increase in demand by previous users was more likely due to loss aversion than to greater risk aversion. For the priority-group members who responded to calls on self-restraint, we find that greater community identification and concerns with reputation effects may have been the drivers of their behavior.