China's 2024 R&D Spending Exceeds RMB 3.6 Trillion; Basic Research Share Lags Developed Economies

China's R&D investment has grown rapidly over the past two decades. In 2024, total national expenditure on research and experimental development exceeded RMB 3.6 trillion, firmly ranking second in the world, with investment intensity rising to 2.69%, now comparable to the OECD average. Enterprise-executed R&D accounted for 77.7% of the total. By both scale and structure, this constitutes R&D investment at a developed-economy level. However, basic research funding in the same year stood at RMB 250.09 billion, only 6.88% of total R&D spending. While a historical high, this is roughly half the average of major developed economies, which typically allocate above 15%.

Over the past decades, China has occupied a catch-up position in most manufacturing and technology sectors, and catch-up players have different basic research needs than frontier leaders. When manufacturing operates far from the technological frontier, absorbing publicly available basic knowledge from others is the more economical choice. But in the latter half of the catch-up phase, the stock of borrowable knowledge shrinks, and the source knowledge needed to break through the frontier must be produced domestically. At this stage, weak basic research constrains the ceiling of applied technology and industry, becoming a hard constraint on frontier breakthroughs. The investment system and research organization of basic research directly shape both the scale and efficiency of investment.

Basic research funding comes mainly from government, enterprises, and business income and donations of universities and research institutes. Basic research suffers from a free-rider problem: outputs are difficult to exclude and can be reused repeatedly, making it a quasi-public good. Private returns fall below social value, and the market inherently underinvests, making public finance the primary funder. China's basic research funding is relatively monolithic. From 2012 to 2021, central government finance averaged 84% of national basic research funding, including the National Natural Science Foundation, national science and technology programs, the Chinese Academy of Sciences, and central university appropriations, while local government science and technology spending concentrated mainly on applied and industrial projects.

Enterprises are not entirely absent from basic research investment. As a share of national basic research funded by enterprises, Japan and South Korea stand at around 50-60%, the UK and US between 30-40%, France about 20%, and China 6.5%, up from 3.8% in 2019. In the US in 2022, the federal government accounted for 40% of basic research investment, enterprises 37%, with the remainder from university own funds and non-profit and charitable organizations. Enterprises willing to invest typically face less financial pressure and are often large firms with competitive advantages or strong profitability in a given field, such as Bell Labs, Google, Sony, and Samsung. China has similar cases: some large central state-owned enterprises undertake national research missions, while in market-frontier competitive fields, private enterprises are the main investors, such as Tencent's New Cornerstone Science Foundation, Alibaba's DAMO Academy, Huawei's hiring of top scholars for research, and firms like High-Flyer, CATL, and BeiGene conducting relatively frontier work.

Enterprise willingness to invest in basic research depends on returns, including converting research into excludable products, cultivating absorptive capacity, gaining standards and ecosystem leadership, and reputational signaling. These return mechanisms require appropriate property rights and reputation institutions; without them, excessive free-riding discourages enterprise initiative. Public fiscal appropriations are by nature spending taxpayers' money for public interest. Administrative agencies are neither funders nor beneficiaries, so incentives for both cost-saving and delivering quality outcomes are weak.

The history of scientific breakthroughs shows that winning paths are often marginal. Deep learning remained outside the mainstream for decades; mRNA technology founder Katalin Karikó was demoted in 1995 and went years without grant funding, yet her papers later became the basis for COVID-19 vaccines. CRISPR gene editing began as a basic question with no apparent application. Basic research breakthroughs are difficult to plan in advance, knowledge is highly dispersed, and returns follow a fat-tailed distribution in which a few extreme successes carry the entire value.

Michael Polanyi, in "The Republic of Science," argued that research judgment should be returned to the scientific community. Science is a republic in which no one can grasp the whole but which can self-coordinate and self-correct; each scientist judges only the parts adjacent to their own work, countless local judgments overlap, and reputation is the currency that constrains the process. Peer review is the institutional embodiment of the scientific community, standing in for the nonexistent neutral expert administrative agency.

During the AI winter for connectionist approaches, the Canadian Institute for Advanced Research (CIFAR) funded researchers including Hinton. Founded in 1982, CIFAR specializes in funding long-cycle, unconventional basic research. The US has numerous basic research funders, including the NIH, NSF, Department of Energy, Department of Defense, NASA, and private foundations, each with its own preferences and no unified hierarchy. DARPA does not use peer review, relying instead on fully empowered project managers permitted a high failure rate to make direct decisions. The coexistence of multiple funders and multiple selection logics reduces the probability that promising but temporarily marginalized projects go unfunded.

Over the past two decades, China's basic research funding mechanisms have made considerable progress, including the National Natural Science Foundation and the National Social Science Fund. But these mechanisms still face difficulties. Yan Ning, now founding president of the Shenzhen Medical Academy of Research and Translation, shared a failed grant application on her blog in September 2014. She had applied for a key project from the National Natural Science Foundation; the review comments acknowledged the importance of the work and her research strength but rejected the application for lacking a feasibility plan. Months earlier, the core results of that research had been published in Nature to high acclaim, and she had won the Sackler International Prize in Biophysics for related work. She questioned whether key funds should support risky but important topics rather than only safe, predictable projects.

NSFC review is relatively fair; whether approved or not, reviewers provide written comments to applicants. By contrast, the National Social Science Fund's correspondence review has long only scored applications without providing written review comments, making it harder to avoid personal connections. Its review discipline repeatedly prohibits applicants from visiting or consulting discipline review experts and from inviting experts for application coaching. In 2010, Shi Yigong and Rao Yi co-wrote in Science that in China, "connections" often matter more than academic contributions in securing large research grants. NSFC's rigorous review is not itself wrong, but if everything must be predictable and guaranteed, high-risk exploration becomes impossible. The key lies in balancing prudence with fault tolerance.

Fund allocation and academic evaluation are mutually dependent. What drives the academic mechanism is not funding but recognition and priority. Merton's account of the scientific community shows that scientists publish discoveries publicly for the recognition of being "first," accompanied by norms of communal sharing and organized skepticism. Publishing first earns credit and reputation, making rapid public disclosure a rational choice, and basic research thus becomes a public good with positive externalities. Krebs' paper on the citric acid cycle was rejected by Nature in 1937 citing a backlog of manuscripts, and was only published after being submitted to a Dutch journal; the work earned him the 1953 Nobel Prize.

The lag and conservatism of peer review gave rise to the preprint model, such as arXiv in physics and biomedical preprint servers, allowing researchers to publish results immediately and free of charge while claiming priority, with formal review following later. Organized skepticism of published papers continues across disciplines; data fabrication has been uncovered in countries and institutions at all levels, including top universities such as Harvard. Academic publishing can self-renew because the evaluation system is sufficiently pluralistic and self-correcting.

To stimulate basic research investment, three approaches can be taken. On the funding side, raising the share contributed by enterprises and foundations requires letting some enterprises consistently earn profits and accumulate sufficient margins, while strengthening intellectual property protection and reputation mechanisms so that both the capacity and willingness to invest in basic research are present. On organization and evaluation, improving academic reputation and peer review mechanisms is the hardest but most necessary part of the institutional environment for the efficient conduct of basic research.