Write With 'People Like Me'

American science has become much more diverse in recent decades as people born outside the United States (or their children) have advanced in numerous disciplines. But do the scientists work together across ethnic and national backgrounds?

A new research paper finds that scientists disproportionately write papers with those of similar backgrounds -- and that those who break out of that mold may be producing the most influential science.

The study (abstract available here) was released Friday by the National Bureau of Economic Research. The authors are Richard Freeman, the Herbert Ascherman Professor of Economics at Harvard University, and Wei Huang, a Ph.D. student in economics at Harvard.

They used a name analysis system to categorize the authors of papers (acknowledging that such a system is not perfect, but asserting that it is generally accurate). And they note that the number of authors of science papers in the Thompson Reuters Web of Science database shows increased diversity. In 1985, 57 percent of the authors had Anglo names, but that fell to 46 percent by 2008, as other groups (most notably those of Chinese ethnicity) saw increases.

They then looked at articles published in more than 12,000 journals, focusing on those with two to four authors (although they then checked their findings with an analysis of those with up to 10 authors and found the similar results). They looked only at authors with addresses in the United States, reasoning that not all scientists have the ability to attend international conferences where one might make connections leading to co-authorship with those outside the U.S.

Then the researchers took all the names involved and calculated the odds -- if co-authorship had been determined simply by drawing names at random -- that two-, three-, and four-person papers would be written entirely by members of the same nationality names. Then they calculated the actual number of papers co-written by those of the same background -- and across groups, such "homophily" was evident beyond what it should have been by random selection of co-authors. Hence the name of the paper, "Collaborating With People Like Me."

Here are the results for two- and three-person papers for the top three name backgrounds, Anglo, Chinese, European and Hindu/South Asian.

Two-Author Papers With Scientists of Same Ethnicity

Three-Author Papers With Scientists of Same Ethnicity

The NBER paper also finds a link between the impact of science papers (based on citations and the rankings of journals based on citations) and writing them with members of one's own ethnic group. Researchers "with weak previous publications records are especially likely to write papers with persons of the same ethnicity," the paper finds. Further, it says that "homophily is associated with publication in a lower impact factor journal and fewer of citations of papers, even holding fixed the previous publishing performance of authors."

Via email, Freeman said that the findings point to the need to encourage young scientists to work with people who are not like themselves. "We would stress getting people to meet others by presenting papers at meetings," he said. "Advisers especially of postdocs should make sure the postdocs get out to present papers." And he said this means that travel funds are important -- so people can meet people with whom they might not otherwise collaborate.