China's Papers Are Up, So Why Aren't Citations?

Research output and overseas citations are different metrics

China’s research output has grown enormously. In the Nature Index¹ 2026 rankings, which tally 2025 papers, 9 of the top 10 institutions were Chinese. By overall institutional ranking, the Chinese Academy of Sciences (CAS) came in first, Zhejiang University second, and Harvard University third. This ranking tallies each institution’s contribution to papers in a selected set of journals — it isn’t a measure of the quality of all research or the value of any individual paper.

But the rise in research output hasn’t translated into a proportional rise in overseas citations. That’s what recent large-scale citation analyses show. That said, not being cited doesn’t mean not being read. Citation data alone can’t tell us whether someone actually read a paper, or what a researcher’s intentions were.

To gauge how much China’s research is actually being used internationally, we need to separate paper counts from citation counts — and look at how the gap varies by field.

If research isn’t sufficiently transmitted to other countries, there’s a real risk of missing the chance to build on findings that already exist. I found myself connecting this problem to something closer to home: how Korea evaluates foreign technology and industry.

The Citation Gap Two Studies Revealed

Two studies published this year have put data behind this phenomenon.

The first is an NBER working paper released in January 2026 by Abhishek Nagaraj of UC Berkeley and Randall Yao of MIT. The researchers analyzed 44 million papers published between 1980 and 2022. According to NBER’s summary, 58–68% of citations received by Chinese papers came from within China itself, and this domestic concentration held even for high-impact papers. Citations from abroad amounted to roughly one-third to two-fifths of the total.

The second is an analysis posted to arXiv² in April 2026 by a joint Chinese-Dutch research team. Covering 39.35 million papers from 2000 to 2022, the study compared collaboration and citation patterns after adjusting for each country’s research output. China was increasingly chosen as a partner for international collaboration, but it received fewer citations than the benchmark the researchers had set. In other words, collaboration ties and citation ties don’t necessarily move in the same direction.

When interpreting these figures, we also need to account for home bias³. Xu, Steinwender, and Azoulay compared actual citations against a benchmark that accounts for each country’s size. Once adjusted this way, China’s citation ranking dropped from 2nd place—behind only the US—to 4th, behind the UK and Germany. This shows that rankings shift depending on the adjustment method used; it isn’t evidence that individual citations were unfair or manipulated.

In this study, home bias appeared not just in China but in other countries as well. The fact that China has a high rate of domestic citation and the finding that it’s cited less than expected abroad can be examined together. It’s difficult to judge an entire research ecosystem from either figure alone.

There are also field-specific differences. In Nagaraj and Yao’s analysis, China was strong in physical sciences like engineering and chemistry, while the US was strong in biomedical and health sciences. When researchers in a particular field are concentrated in one country, citations within that country can naturally increase. This is exactly why a single domestic-citation ratio shouldn’t be equated with research isolation or misconduct.

The English draft matches the Korean source accurately in meaning, structure, numbers, and terminology. No corrections needed.

Factors That May Affect How Research Crosses Borders

The citation gap is often attributed to trust in the research itself, information about institutions, and the environment for international collaboration. How much each of these factors actually explains the gap still needs separate verification.

The first factor is trust in the research. According to a study analyzing the Retraction Watch database, 30,977 papers with at least one China-affiliated author had been retracted as of February 2025. That figure represented 55.3% of all retracted papers recorded in the database at the time. This is the share among retracted papers — it does not mean 55.3% of Chinese papers were retracted. It’s also part of the backdrop for China’s crackdown on paper mills⁴ and its overhaul of research evaluation systems. Any individual paper still needs to be judged on its own methods and data.

The second factor is information about research institutions and authors. Oxford’s Dorothy Bishop has pointed out that overseas researchers may struggle to distinguish between Chinese research institutions. In other words, unfamiliar names or institutions could plausibly influence how a paper is evaluated. But being affiliated with a well-known institution doesn’t, by itself, guarantee a paper’s credibility. Checking the methodology, the available data, and whether results can be reproduced is a far more direct basis for judgment.

The third factor is the environment for international collaboration. US-China relations and research-security investigations can raise the cost and burden of joint research. A study examining NIH investigations reported that US researchers with Chinese collaborators saw their productivity decline relative to other internationally collaborating researchers. This doesn’t mean collaborating with Chinese researchers inherently lowers productivity — rather, the study looked at the effects of the investigative and regulatory environment itself.

These factors can reinforce one another. As joint research and conference exchanges decrease, so too can the opportunities to encounter and evaluate each other’s work. Rather than pinning the citation gap solely on individual researchers’ bias, we should also examine the conditions under which research can be verified and shared.

Reflecting on How Korea Views Chinese Industry

Reading this research, I kept thinking about how domestic coverage handles Chinese industry.

Something I’ve pointed out for a long time is the tendency to worry intensely about China’s semiconductor catch-up while underrating its battery competitiveness. This isn’t a conclusion from a quantitative analysis of all domestic reporting—it’s a judgment I’ve formed from following the relevant news.

In semiconductors, even when news breaks about SMIC’s advanced process nodes, you need to check yield and production scale before you can compare it to Korean firms. Nor should the achievements of memory maker CXMT and foundry SMIC be lumped together as if they represent the same process story. In batteries too, understanding competitiveness requires looking beyond CATL and BYD’s market share to price, technology, and overseas production capacity.

I’m wary of the tendency to exaggerate whatever field feels threatening while assuming that familiar advantages will persist indefinitely. That said, these studies don’t prove that the same underlying cause drives both domestic industry perception and international academic citation patterns—connecting the two is my own interpretation.

DeepSeek is another case that forced a reexamination of existing assumptions. Its open model and technical report made it possible to verify performance and cost efficiency directly. But the fact that Nvidia’s stock dropped sharply after the release doesn’t by itself prove the technology’s value. Stock market reactions and technical verification need to be considered separately.

Missing research from other countries can mean unknowingly repeating work that’s already been done. But redoing the same experiment isn’t always wasted effort—independent replication is an important part of building research credibility. There’s a meaningful difference between verifying known prior work and starting from scratch without knowing it exists.

Korea’s own achievements also deserve a closer look at which metrics are being used. In Nature Index 2025, which tallies 2024 papers, Korea ranked 7th, with its adjusted Share up 4.1% from 2023. Here, Share is the sum of a country’s contributions to papers in a selected set of journals—it doesn’t mean Korea’s share of all global research rose by 4.1 percentage points.

What I find significant is that Korea has access to both American and Chinese research and can collaborate with both. To actually use this as an advantage, we need to look at the substance and verifiability of research before asking which country it comes from. No single citation metric can capture the full value of that kind of relationship.

Oswarld’s Lens

I look at this issue less through the lens of “science” and more through the lens of “GTM (go-to-market strategy).”

In building GTM strategies, I’ve seen cases where even a good product takes time to earn trust. That’s why the gap between China’s research output and its international uptake caught my attention. A rise in research output doesn’t automatically translate into a shift in reputation or collaborative relationships. That said, academic citation isn’t the same as product sales, and trust alone can’t account for the entire gap.

When international exchange shrinks, so do the opportunities to directly verify research quality. I think rigorously scrutinizing the retracted-paper problem and maintaining normal research collaboration need to happen together, not one at the expense of the other.

I think Korea, too, needs to examine how it evaluates foreign research and competitors. Whether we’re worried about China catching up in semiconductors or assessing its progress in batteries, the standard for judgment should be concrete data.

I’d like Korean researchers and companies to keep the channels open for comparing and verifying achievements on both sides. My concern isn’t about picking which side to root for — it’s about making sure we don’t miss out on research we actually need to use.

Closing

China’s rising paper output alongside its lagging citation rates is worth examining. But if we treat low citation ratios as proof of dismissiveness or bias, we end up repeating the very snap judgments we set out to guard against.

Citation counts and research rankings are just a starting point. We still need to check whether a paper’s methods and data hold up, whether its results replicate in other settings, and what it actually contributes to our field.

If there’s Chinese research or technology in your field that you’ve recently had to reassess, let me know in the comments.

References & Further Reading

Primary sources

• Abhishek Nagaraj & Randol Yao, “The Geography of Science”, NBER Working Paper #34694, 2026. : This is the core study, analyzing citation data from tens of millions of papers published between 1980 and 2022. It shows that research output and international citations don’t grow at the same rate.
• Leyan Wu et al., Persistent geographical biases in global scientific collaboration and citations, April 2026. : A preprint analyzing the gap between cross-country collaboration and citation patterns.
• Shumin Qiu, Claudia Steinwender & Pierre Azoulay, “Paper Tiger? Chinese Science and Home Bias in Citations”, Journal of International Economics 157, 2025. : Shows how, after correcting for home bias, China’s global citation ranking drops from 2nd to 4th. The comparison between home bias in trade and home bias in scientific citation is especially interesting.
• Reckoning with Retractions in Research Funding Reviews: The Case of China, Publications, 2025. : A study examining the distribution of retracted papers in the Retraction Watch database as of February 2025.
• “More than half of all retracted papers are from China, analysis finds”, Chemistry World, April 2026. : The latest report summarizing China’s retraction figures based on Retraction Watch data.
• Nature Portfolio, Nature Index 2026 Research Leaders announcement, June 2026. : Explains the 2025 paper-based institutional rankings and the change in scope of coverage.
• “Nature Index 2025 Research Leaders”, Nature Index, June 2025. : The original country-level research performance data, including Korea’s rise to 7th place, China’s expanding lead at 1st, and the decline of major Western countries.

Background

• “China-U.S. Science Collaborations Are Declining, Slowing Key Research”, Scientific American, July 2024. : Tracks the chilling effect of NIH investigations and the real impact of declining U.S.-China co-authorship on research productivity.
• “Home bias and China’s global standing in science”, CEPR VoxEU, February 2025. : A column that lays out the core findings of the Qiu, Steinwender & Azoulay paper in accessible terms for non-specialists.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

Footnotes

1. Nature Index: A metric that tallies the number of papers in selected journals along with each author’s or institution’s fractional contribution (Share). The 2025 edition covered 145 journals in the natural and health sciences; the 2026 edition expanded to 177 journals plus one academic society, adding applied and social sciences. It is not a ranking of overall research quality. ↩

2. arXiv: A preprint server where academic papers are posted before peer review. It’s widely used in physics, mathematics, and computer science, and is one of the fastest channels for accessing the latest research. ↩

3. Home bias: The tendency of researchers to cite papers from their own country more than expected. It works on the same logic as the preference for domestic goods in trade, and the same pattern shows up in scientific citation. The 57.2% figure refers to the share of self-citations among all citations received by Chinese papers in that study — it differs from bias figures adjusted for country size. ↩

4. Paper mill: An operation that mass-produces and sells fake or extremely low-quality academic papers. They get papers published in journals through a combination of fabricated data, plagiarized structure, and bribed peer review. ↩