International collaboration is one of modern science’s quiet superpowers. Increasingly, it is also a geopolitical flashpoint. This column presents new experimental evidence that shows both US policymakers and US-based scientists are systematically less likely to support otherwise identical research proposals when a key collaborator is based in China rather than in Germany. Support shifts away from unconditional approval and toward stricter, conditional backing. These penalties are strikingly consistent across scientific fields and respondent characteristics (including Asian scientists), highlighting a tension between science’s universalist ideals and today’s geopolitical realities.
Science likes to picture itself as a republic of letters — open borders, shared standards, disputes settled by evidence rather than passports. Merton (1973) coined this ethos as CUDOS: Communalism, Universalism, Disinterestedness, and Organised Scepticism. Reality is messier. Public funding ties science to national priorities: economic growth, technological leadership, and, at times, national security. When geopolitical rivalry intensifies, the ideal of universalism collides with the practical question of risk.
That collision is most visible today in the US–China relationship. Concerns over intellectual property, surveillance, human rights, and security have pushed universities and funders to scrutinise such partnerships more closely. We typically observe the downstream symptoms, shifting collaboration patterns and a ‘chilling effect’ on cross-border teams (e.g. Aghion et al. 2023, Xie et al. 2023, Jia et al. 2024, Li and Wang 2024). But one upstream mechanism is harder to observe: evaluative gatekeeping. Before collaborations happen, someone must say “yes”, “yes, but…”, or “no”. How much does geopolitics enter those judgments, especially among scientists who are expected to evaluate work on merit rather than nationality?
To isolate the effect of collaborator nationality, we conducted two large-scale randomised survey experiments in a US government funding context (Furnas et al. 2026). The key idea was simple: hold the science constant, change the country label.
In both studies, respondents evaluated a hypothetical research proposal submitted to the US National Science Foundation. The proposal was explicitly described as (i) having passed peer review, (ii) unclassified, and (iii) not subject to export controls. We randomly varied two features: collaborator nationality (Germany-based versus China-based) and field (batteries, biotechnology, robotics, or environmental science). Respondents then chose between unconditional funding, conditional funding (with additional safeguards beyond standard compliance), or rejection.
The results are striking in both samples.
Figure 1 A large and consistent ‘China penalty’
As shown in Figure 1, among policymakers, unconditional support drops from 68.2% for US–Germany collaborations to 27.6% for otherwise identical US–China collaborations. Among scientists, unconditional support falls from 47.5% (Germany) to 29.7% (China). These gaps appear across all four fields; environmental science shows a slightly smaller penalty, but the pattern remains.
If the effect were mainly about field-specific ‘security adjacency’, we would expect the China penalty to concentrate in robotics or batteries and fade elsewhere. Instead, it is broadly present – suggesting evaluators are applying a country-based lens that cuts across domain boundaries.
The most policy-relevant detail is how the penalty manifests. In both samples, it is primarily a shift from unconditional approval to conditional approval, not a shift to outright rejection.
For China-based collaborations, 59.8% of policymakers and 65.8% of scientists prefer conditional funding. For Germany-based collaborations, conditional support is much lower: 29.6% among policymakers and 50.9% among scientists. Rejection rises as well, but remains a minority outcome (12.6% vs 2.3% for policymakers; 4.5% vs 1.9% for scientists, China versus Germany).
When we ask about safeguards, respondents disproportionately impose additional conditions for China-based collaborations related to national security and surveillance – such as screening for military affiliations and prohibiting surveillance applications. Scientists also more often require that China collaborations serve a global public good (e.g. climate or health), effectively adding an extra normative hurdle. The distribution of conditions is presented in Figure 2.
These patterns clarify how geopolitics can reshape science without a single new law. At least some of the shift happens at the evaluation stage: before labs exchange data, before students visit, before papers are written. Gatekeeping is an upstream lever: when it moves, collaboration networks reconfigure downstream.
The findings also complicate a comforting story about peer review as a firewall. Even when proposals are described as peer-reviewed, unclassified, and outside export controls, scientists still penalise China-based collaboration. That does not mean scientists are acting irrationally; conditionality can be sensible risk management. The difficulty is that risk management requires criteria, and criteria require clarity. Most scientists are not trained to assess national security or human-rights risk, so absent transparent standards, evaluation can drift toward informal heuristics, common sense that is hard to justify, audit, or apply consistently.
Figure 2 Evidence of conditional support in US-China versus US-Germany collaborations
Germany is a deliberate comparison: a close US ally with strong scientific capacity and relatively low security concern. Our estimates therefore capture reluctance toward China relative to a ‘best-case’ partner, not relative to a neutral or average country.
Our design is cross-sectional. It identifies how current geopolitical relations, as reflected in a collaborator’s country label, shape judgments today. The proposals are hypothetical, and real-world panels involve deliberation and institutional rules that can amplify or dampen individual preferences. Still, because collaborator nationality is randomised while content and quality signals are held constant, the causal inference is clean: the country label itself changes evaluators’ choices in this funding context.
Could the results instead reflect racial attitudes or perceived differences in research quality? The patterns are hard to square with these alternatives. US-based Asian scientists exhibit the same support gap, and the penalty is remarkably consistent across fields, suggesting a broadly geopolitical response rather than group-based bias or discipline-specific beliefs about quality.
Science has always lived in two worlds: the world of ideas and the world of states. In calm times, the seam is easy to ignore. In tense times, it shows up right where you least want it: on the grant panel’s score sheet, in the reviewer’s margin notes, in the difference between “fund” and “fund, but…”.
Our evidence suggests that the seam is widening. The challenge for science policy is not to pretend geopolitics is absent, but to build evaluation systems that acknowledge real risks while preserving the integrity of merit-based judgment.
Source : VOXeu
Algorithms curate what users of social media see, raising concerns that they may distort attitudes…
Imagine this situation - María runs a small grocery shop, and one afternoon she receives…
The UK vote to leave the EU in 2016 led to an immediate rise in…
Connections to global markets and supplies are a precondition for trade driven development, investments, and…
The surge in inflation following the COVID-19 pandemic prompted many central banks to raise interest…
When the philosopher and poet Ralph Waldo Emerson wrote in 1841, "Every sweet hath its…