Is America losing its competitive edge in critical technologies? The House Science Committee last week held the latest in a string of congressional hearings on this question, asking where the United States stands in fields such as artificial intelligence, quantum physics and biotechnology.
A perceived exodus of these international graduates has recently prompted fears about the United States’ ability to lead in technologies such as AI. Senior computer scientist Yolanda Gil wrote on behalf of the Association for the Advancement of Artificial Intelligence that “international PhD graduates are leaving the US in larger numbers than before, in part due to immigration constraints but also due to the availability of attractive opportunities for AI overseas.”
Others in academia, the private sector and government have also sounded the alarm over the loss of U.S.-educated talent to other countries, especially China. But is this perceived brain drain real? Research that my team and I conducted actually found very little evidence of increased U.S. talent loss.
These are strong claims, but weak data
The Chinese Ministry of Education is the data source typically cited by those who fear declines — this includes an extensive recent Senate report on China’s threat to the U.S. research enterprise. Statistics published by this ministry show that 8 overseas students returned for every 10 students who went to study abroad in 2018, up from a ratio of just 1 in 10 in the early 2000s.
Yet these numbers aren’t very helpful. The ministry’s data cover all Chinese students abroad, not just those in the United States, and don’t distinguish students by degree level or area of study. We also don’t have any documentation about how Chinese officials gathered or analyzed the data. Even taken at face value, an overall increase in the number of returnees would, at best, be an indirect and weak indicator of decreasing U.S. retention rates.
We looked at better data
Our newly published study analyzes data that we collected on the career histories of 2,000 recent U.S. AI PhD graduates. We also analyzed several existing data sets. In none of these data sets did we find evidence of a decline in the number of PhDs who remain in the United States. Despite immigration difficulties and recent political tensions, the vast majority of international AI graduates stay and work in the United States.
The National Science Foundation and the Computing Research Association both run high-quality surveys of PhDs in AI-related fields, and neither has seen a recent drop in graduates’ desire to stay after completing their degree. For each year between 2005 and 2018, only about 10 percent of Chinese PhD graduates in computer science report plans to leave the country.
We also found this same trend line in the new data set we collected to examine AI PhDs’ career histories to determine actual, not intended, stay rates. In our data set, around 90 percent of those graduating from U.S. universities in 2018 remained in the United States to work immediately after graduation. This is about the same level as in 2014, as shown in the figure below.
After five years, when data coverage ends, more than 80 percent are still in the country. Long-term stay rates for Chinese graduates are even higher, with more than 90 percent staying for at least five years.
What the data don’t say
Gaps in the data mean that we don’t have all the answers. For one, good retention data exists only for PhDs, who make up just a small — though very important — portion of international students. In AI, master’s students account for 79 percent of international graduates (bachelor’s and PhDs account for 16 and 5 percent, respectively). Stay rates among bachelor’s and master’s students could very well differ from those among PhDs. We currently have no way to know for sure.
It’s also conceivable that international graduates still stay in the United States at the same rate immediately after graduation, but that recent graduating cohorts have changed their minds about remaining over the long term. We are aware of no data source that has systematically tracked how long graduates intend to stay in the United States after completing their degree.
Of course, past trends are an uncertain guide to the future. Whether student retention remains high depends in part on factors outside of graduates’ control. Immigration barriers have made it increasingly difficult to stay in the United States, especially with key visa programs such as Optional Practical Training (OPT) on the chopping block. Just because retention rates have remained high in recent years does not mean they will stay that way.
Tracking this closely is important
Without good data, policymakers can be left with an inaccurate picture of what’s happening with international talent. A prominent 2018 report by the Pentagon’s Defense Innovation Unit, for instance, incorrectly stated that “nearly all” STEM graduate students from China left the country after graduating and that this amounted to “unintentional violations of U.S. export control laws.”
There are good reasons to have security concerns about foreign talent working in fields that have military as well as civilian applications. However, if policy decisions in this area rely on incomplete or wrong data, they run the risk of creating more problems than they solve.
The U.S. government already has an infrastructure in place to help fill these data gaps. It’s not difficult to extend the annual surveys the National Science Foundation runs among PhD graduates to include master’s students as well. And collecting data about graduates’ future intentions simply requires adding a small number of questions to these surveys.
Until this data is collected, we will continue to be uncertain about exactly how many international graduates stay in the country and why. But what we can say is that almost all of the available evidence seems to contradict the general belief that the United States is retaining less top international talent than before.
Remco Zwetsloot (@r_zwetsloot) is a research fellow at Georgetown University’s Center for Security and Emerging Technology (CSET).