This paper examines the effect of relative productivity levels on international knowledge flows. Using the Coe and Helpman (1995) data on research, I show how separating the sample into distinct groups allows the relationship between relative productivity and research spillovers to be estimated. Since the data are nonstationary, the panel cointegration techniques recently developed by Kao and Chiang (1999) are used. I find that the observed flow of knowledge is consistent with a simple model of catch-up to the technological frontier. Countries benefit more from the research of more-advanced nations than from the research of nations with similar productivity levels.