China’s technological prowess is frequently invoked by U.S. policymakers hoping to get votes, attention, or enough bipartisan support to pass a bill. Competition with China was a central motivating factor in federal legislation like the CHIPS and Science Act and the Inflation Reduction Act, not to mention the work of the House Select Committee on the Strategic Competition Between the United States and the Chinese Communist Party. “Beating China”—specifically in science and technology development—is a key driver of U.S. governance and, ahead of the 2024 presidential race, elections.
George Washington University Assistant Professor of Political Science Jeffrey Ding published a recent article that explains why measuring a state’s scientific and technological power should include not only how many innovations a country can check off, but also the degree to which new technologies are integrated into the economy and society. Ding spoke with Johanna Costigan about his new paper. The following is an edited transcript of their conversation. —Johanna Costigan
Johanna Costigan: Can you explain the distinction between innovation capacity and diffusion capacity?
Jeffrey Ding: Innovation capacity refers to a state’s ability to pioneer new-to-the-world breakthroughs in science and technology. Diffusion capacity refers to a state’s ability to spread those innovations across a wide range of productive processes in an economy. I argue in my paper that we tend to gravitate toward measures of innovation and pay insufficient attention to diffusion.
How did diffusion occur in the second Industrial Revolution in the U.S., and why does that example show diffusion is an essential metric?
In the late 19th century, the U.S. was not at the forefront of science and technology—our best and brightest went to Germany to study. Even though other nations were leading the way in fields like chemical engineering, the U.S. was much stronger in terms of diffusion capacity; it was better at applying those uses across a wide range of processes. If you only looked at which countries were winning the most Nobel Prizes or had the most advanced research institutes, you would be underestimating the scientific and technological prowess of the U.S.
Assessments that are more oriented around diffusion capacity would have predicted the U.S. to sustain its economic rise and become the preeminent economic power. Back then, there wasn’t a global innovation index, but the case study was to go back and see if we rank different countries based on this, only using the innovation indicators provides an incorrect view. Diffusion is central to a state’s ability to convert science and technology into economic strength.
You write that academic research in the U.S. was more closely tied to commerce than it was in Europe, which allowed American breakthroughs to spread more rapidly. Given the close relationship between government and industry in the PRC, why might that not be the case in contemporary China?
The government acts as a bottleneck within China’s scitech ecosystem, stifling organic industry-university collaborations. A lot of the research happens at government institutions as opposed to corporate-sponsored R&D [research and development]. The channels between universities and industry are not as robust, so some of that has to do with surrounding legal regimes and whether university research can be translated into a startup company. Unlike China, the U.S. has a really good set of legal rules that enable that to happen.
But Chinese companies are trying to address this; more AI companies, for example, are trying to set up labs in universities to foster that. But if you look at co-authorship rates in publications on AI (papers that have at least one author from a university and one from industry), those rates are very low in China compared to countries in Europe and the U.S. A lot of the story here is just the vestiges of central planning that are still in China’s scitech ecosystem and that just don’t foster the organic fast-acting processing that is required for diffusion.
How would you characterize American assessments of China’s science and technology power? What’s missing?
The consensus significantly overstates China’s scitech capability. One recent example from the Australian Strategic Policy Institute (ASPI), which while it is not American has been used to justify more extreme China policy, says China is leading the U.S. in 37 of 44 considered technologies, all of which were focused on innovation. (I don’t even think that conclusion is accurate based just on innovation.) It reflects the United States’ obsession with innovation capacity when it comes to assessing scitech power. Even if you go back and review Biden’s first remarks to Congress, he argued that China is closing in quickly and that it’s a race to see who can dominate new innovations in these technologies.
The contribution of my paper is partly to argue that assumptions about innovation shape the perception that China is improving rapidly and has already taken over the U.S. in some critical technologies. Reorienting ourselves to a diffusion-centric framework is an important first step to having a more balanced understanding of what’s happening in China’s scitech ecosystem.
Members of the Biden administration like Jake Sullivan have described moves like export controls aimed at China as “narrowly focused on technology that could tilt the military balance.” What do you think of that depiction, and how do judgments of relative military strength fit into the diffusion/innovation breakdown?
When it comes to competition with China over military AI applications, I think the Biden administration’s approach is overly preoccupied with maintaining a lead in innovation capacity. Their theory of victory, in my view, seems to be all about preventing China from building the biggest and baddest autonomous weapon, trained on the largest amount of data using the most amount of computing power. If the history of military electrification is a useful guide for how AI will affect the military balance of power, as I argue in a recent article, then AI’s most substantial impact on military power will take decades as advances diffuse across a broad range of applications in logistics, decryption, targeting, and intelligence. Ensuring that the military is able to tap into a broad base of AI engineering talent is a more effective route to ensuring that AI’s effect on the military balance will favor the U.S.
You write: “A rebalanced evaluation of China’s potential for S&T leadership requires looking beyond multinational corporations like Huawei, first-tier cities like Beijing, and flashy R&D numbers to the humble undertaking of diffusion.” Why do you think such limited analyses are tempting? Why are they dangerous?
Firstly, it’s just a lot harder to get diffusion capacity indicators. Many metrics of innovation, like government R&D funding indicators, patent rates, and publication numbers are all tracked, whereas it takes more work to get indicators of diffusion capacity. Finding systematic and reliable numbers of diffusion capacity is crucial.
And misleading assessments carry a few dangers. Having an accurate sense of where you stand provides a solid foundation for science and technology policy. There is something to be said about having a true—or truer—understanding of the world. Additionally, overestimating China’s scitech capabilities may lead the U.S. to engage in more reckless policies and provide more momentum for containment-type measures that backfire on both sides. It could lead to the mentality we saw in the Cold War where the U.S. was concerned about the missile gap with the Soviet Union that turned out to be illusory and resulted in wasted expenditures, spiraling fears, and an arms race that put the two superpowers on a path toward conflict.
There are people in the U.S. government who probably agree with me in that a lot of assessments on China’s tech development are overhyped, but they think it’s necessary to pump up China’s prowess in order to motivate certain policies. But that could easily backfire and is a dangerous precedent to set. While that justification might in some cases be used to make sound decisions, in other cases it could be used to enact dangerous policies.