This study examines the ethnic identify of the co-authors of over 1.2 million papers with US addresses from 1985 to 2008. It finds a striking change in the ethnic composition of authors, with the proportion with English and European names falling while the proportion of names from China and other developing countries increases. The greater variety of ethnicity is associated with considerable homophily among research teams, as persons of similar ethnicity tend to work together far more frequently than can be explained by chance. The paper identifies a modest negative relation between homophily and the potential scientific contribution of the papers as measured by the impact factor of journals of publication and the number of citations, with the latter attributable to the previous publishing performance of authors. Using a Markov analysis to calculate a steady state rate of homophily, the paper finds that the rates is close to the steady state and thus likely to continue at high levels into the future. The analysis also finds that papers written by authors at different addresses and that cite larger numbers of references are more likely to get into high impact journals and to gain more citations than other papers.