There have been many studies of the volume-outcome relationship. In all of these, the unit of analysis is the hospital or physician. However, this level of analysis is mostly limited to the use of in-hospital mortality rates and is particularly sensitive to selective referral. Moreover, the literature on agglomeration economies highlights the importance of information spillovers within regions (Glaeser, 2010). To overcome these problems, our study is the first that examines the volume-outcome relationship on a regional (county or cancer registry) level. Using data from the National Cancer Institute's Surveillance, Epidemiology and End Results program we find that regions with relatively more of the same cancer type exhibit relatively better health outcomes.