This paper uses a gravity framework to investigate the effects of distance as well as subnational and national borders in knowledge spillovers. Drawing on the NBER Patent Citations Database, we examine patent citations data at metropolitan level within the U.S. and the 38 largest patent-cited countries outside the U.S. We present 3 key findings: First, we find strong subnational localization effects at the Metropolitan Statistical Area and state levels: more than 90% of intranational border effects stem from the metropolitan level rather than state. Second, border and distance effects decrease with the age of cited patent, which implies that new knowledge faces the largest barriers to diffusion. However, over time, border and distance effects are interestingly increasing. Finally, we find that (assignee) self-citations and aggregation bias are two sources of overestimated aggregate border effects of knowledge spillovers. While self-citations are only 11% of total citations, they account for approximately 50% of MSA and national border effects.