Assessing China as a technological innovator

The dramatic rise of domestic patent publications by Chinese inventors has raised questions about the extent of Chinese innovations. This column uses a newly assembled dataset of Chinese patents to understand the Chinese innovative ecosystem. The data reveal an innovation system that is more broad-based, more university-centred, and less dominated by state-owned enterprises, large national champions, or US-trained returnees than many standard narratives suggest. The findings challenge preconceptions about where Chinese innovation is occurring and suggest the importance of better understanding the mechanisms behind central coordination and its impact.

Domestic patent publications by Chinese inventors have risen dramatically from an extremely low base, far exceeding in recent years the levels of patents issued by the US Patent and Trademark Office (USPTO) to US inventors. These effects are particularly striking in critical technology areas such as artificial intelligence (Fang et al. 2026). 

The interpretation of these trends, however, has been hotly debated. A number of analysts have claimed that the commonly used innovation indicators overstate Chinese progress. In the context of patents, critics have pointed to the powerful incentives that exist for Chinese entities to undertake patent filings and the very low share of these filings that are also filed overseas (Li 2012, Helmers et al. 2011, Eberhardt et al. 2017). More generally, Congressmen and other US leaders have questioned the extent to which innovations have truly originated in China, as opposed to being results of intellectual property misappropriation and/or immigration of skilled personnel from Western nations. 

Others point to instances of the dramatic Chinese technological progress. The most discussed example in the last two years has been the rise of Chinese large language models (LLMs). In January 2025, for instance, the Chinese firm DeepSeek released its R1 large language model, which was then believed to approach the performance of leading US models at substantially lower cost and compute intensity. The release was widely described as a ‘Sputnik moment’ for American technological leadership (e.g. Acemoglu 2025, Cassidy 2025, Ruwitch 2025). By mid-2026, Chinese large language models represented up to 46% of token use by US firms (Lena 2026). This and other episodes have supported positive assessments of China’s innovative capacity.

Our paper (Lerner et al. 2026) seeks to use a newly assembled dataset of Chinese patents — a detailed if imperfect measure of innovation — to understand the Chinese innovative ecosystem. In the 14 critical technology areas designated by the US Department of Defense (DoD), annual Chinese patent publications rose from fewer than 10,000 in the early 2000s to more than 400,000 by 2022 (Figure 1). The data reveal an innovation system that is more broad-based, more university-centred, and less dominated by state-owned enterprises, large national champions, or US-trained returnees than many standard narratives suggest. The results underscore the credibility of anecdotal accounts of Chinese innovative progress, such as the DeepSeek case above.

Figure 1 Patenting in the US and China

a) All domestic patents, China vs US

b) Critical technology patents, China vs US

Note: Figure 1a reports the annual counts of all domestic patent publications in China (red) and the US (blue) patent graphs between 1985 and 2023. Figure 1b reports the account counts in 14 Critical Technology Areas designated by the US Department of Defense.

Our paper examines five frequently articulated perceptions about Chinese innovation.

We first address the observation that China’s patent explosion coincided with weak productivity growth. Wei et al. (2017) have highlighted that Chinese total factor productivity (TFP) growth slowed substantially during the 2010s, even as the patent measures delineated above accelerated. This seeming contradiction raises concerns about the meaningfulness of Chinese patents as an economic indicator. Perhaps they only reflect incremental innovation, not meaningful breakthroughs. 

To address this concern at least partially, we conduct two exercises. First, we compare Chinese patenting with its scientific output in international journals. We quality-weight both measures. We find that the adjusted Chinese patenting is strongly correlated with independent measures of scientific output within technology areas and over time. This correlation suggests that patents have meaningful information about underlying innovative activity. We also explore whether the Chinese rate of patenting is anomalous. While Figure 1 reveals that aggregate Chinese patent counts are large, they are less unusual when scaled by the size of China’s science, technology, engineering, and mathematics workforce. In fact, by this measure they closely resemble the domestic rates seen in the US.

A second perception relates to the role of large firms. Many US policy discussions and actions have focused on the very largest Chinese firms such as Huawei, which are often seen as enjoying favoured treatment by the state and Chinese Communist Party.

Chinese patents, even in frontier technologies, are not dominated by a handful of giants. In fact, domestic patenting in China is considerably less concentrated among large firms than in the US. Moreover, this concentration is declining over time. The share in Chinese critical technology patents assigned to the ten largest patentees declined from roughly 10% in the late 1990s to 4% by 2022. This ratio in the US has been nearly constant at roughly 20%.

The third perception concerns the involvement of the Chinese government in the direct production of innovation. We look at the filing of patents by state-owned enterprises and the government. Note that we do not consider here other efforts of state direction, such as subsidies and explicit coordination, topics we are exploring in follow-on work. We find that the directly linked patents account for less than 4% of Chinese critical technology patents. Innovation in critical technologies is concentrated among private firms (58% of these patents) and universities (27%). The university share is particularly striking, as it is roughly eight times larger than the US share of 3.3%.

We also address a fourth belief: that returnees who benefited from working and advanced training in the US have driven Chinese innovation. We find little evidence in support of this claim. Fewer than 10% of Chinese critical technology patent publications involve an inventor with prior US work experience (a share that has declined since 2019). The fraction involving inventors with US PhDs remains below 1% throughout the sample. Indeed, there is a striking pattern pointing in the other direction. The share of US patents involving inventors with non-US Western work experience has climbed to 22%, more than twice the corresponding Chinese peak.

Our last analysis explores whether the rapid expansion of Chinese patenting was associated with a deterioration of patent quality. As traditional measures such as citations and market reactions are less revealing in this setting, we evaluate patent quality using four complementary text-based measures. These capture in turn textual novelty, textual influence within and across countries, and proximity to the scientific frontier.

We examine the quality of domestic Chinese innovation over time, comparing it to domestic USPTO patents. The four metrics — while certainly individually imperfect – tell consistent stories: we see no evidence of decline in quality of Chinese patents relative to the US, despite the dramatic growth highlighted in Figure 1. For instance, Chinese patents in virtually all critical technologies gained textual importance relative to US patents between 2006 and 2021. The share of Chinese critical technology patents that mention economically salient bigrams, as extracted from US patents and earnings calls (Kalyani et al. 2025), rose from under 2% to 16% between 1995 and 2022. And Chinese patents are closer to the scientific frontier, as measured by the set of most highly cited academic articles within each technology, than their US counterparts in several technologies.

By constructing a large-scale dataset linking Chinese patents, organisations, inventors, and scientific activity across critical technologies, we provide new evidence about the nature of the Chinese innovation ecosystem. We also respond directly to works – by both academics (for instance, Boeing et al. 2024 and Kang et al. 2025) and practitioners (e.g. He 2021) – arguing that the explosion in Chinese patenting volume has been driven by incremental activity in response to government incentives. The findings also challenge preconceptions about where Chinese innovation is occurring and suggest the importance of better understanding the mechanisms behind central coordination and its impact.

Source : VOXeu