Beamtimes and knowledge production times: how big-science research infrastructures shape nations’ domestic and international science production

Frontier scientific advancements increasingly rely on big-science research infrastructures (BSRIs) like supercolliders and synchrotron light sources. These facilities often involve significant international collaboration, creating a balance between national interests and global scientific progress. Developed nations have long leveraged BSRIs to foster scientific breakthroughs and enhance national standing. Emerging economies are increasingly investing in such facilities, aiming for similar benefits and a shift from knowledge absorption to knowledge creation. This study focuses on the Shanghai Synchrotron Radiation Facility (SSRF), China's first 3rd generation synchrotron light source, operational since 2009. The researchers aim to assess SSRF's impact on both national (domestic researchers) and international scientific production, acknowledging the challenge of endogeneity in evaluating BSRIs. Unlike previous bibliometric analyses, this study employs a difference-in-differences (DID) approach at the discipline level to better capture the facility's indirect influence on scientific advancement. The researchers hypothesize that disciplines relying on synchrotron radiation (treatment group) will show greater improvements in scientific output post-SSRF compared to disciplines not relying on it (control group).

Existing literature primarily uses bibliometric methods to assess the direct outputs of BSRIs, such as publications and patents. However, this approach has limitations in capturing indirect impacts. The authors draw from the theory of technology catch-up, differentiating between indigenous and collaborative innovation. For latecomer nations, international collaboration offers access to knowledge and resources, but the ultimate aim is to foster indigenous innovation. Big science infrastructures are highlighted as playing a crucial role in this transition. A previous study using DID design examined the impact of China's National Supercomputing Center on regional innovation, but this study uniquely focuses on discipline-level differences between national and international science production.

The study uses a difference-in-differences (DID) econometric method to estimate the effect of the SSRF on discipline-level scientific knowledge production. Data include 1.75 million Web of Science indexed publications (1998-2015) with at least one Chinese author, excluding publications from multidisciplinary journals. The opening of SSRF in 2009 serves as a quasi-natural experiment. The treatment group consists of 117 narrowly defined scientific disciplines whose research directly benefits from SSRF, identified through literature review and expert input from SSRF staff. The control group comprises 96 disciplines not reliant on synchrotron radiation. The DID model includes measures of research output (high impact factor (IF) publication percentage and mean IF), a treatment dummy, a post-2009 dummy, control variables (NSFC funding, postgraduate students, overseas returnees), discipline fixed effects, and year fixed effects. The researchers investigate both national publications (by Chinese researchers) and international publications (cross-border collaborations) separately. To validate the ‘parallel trend’ assumption, they estimate dynamic effects across time points pre- and post-SSRF opening. Robustness checks include a falsification test with randomized treatment/control groups and placebo tests with pre-2009 treatments. Heterogeneous effects were explored across six broader scientific fields (physics, chemistry, materials science, engineering, biomedical science, and environmental science). A bibliometric analysis of publications directly stemming from SSRF research was also conducted.

The DID estimations reveal that SSRF significantly increased the percentage of high-impact national publications within treatment disciplines by 12.4%, compared to the control group. The effect on high-impact international publications was not statistically significant. However, SSRF led to a notable increase in the average IF of both national and international publications from the treatment group. Dynamic effect estimations and robustness tests confirmed the pre-2009 parallel trends and the causal relationship between SSRF and enhanced scientific output. The falsification test, using randomized treatment groups, showed insignificant effects, supporting the causality of the SSRF's impact. Placebo tests with pre-2009 treatments further underscored that the positive effects materialized only after SSRF's launch. Analysis across six broader scientific fields indicated that while SSRF generally had a positive effect on international publications, its impact on national publications varied by field, with some fields exhibiting less significance. This is potentially explained by the high level of international collaboration in certain disciplines.

The findings strongly suggest that investment in BSRIs like SSRF can significantly enhance the production of high-impact scientific publications, especially for national science. This aligns with the concept of latecomer nations leveraging international collaboration to initially gain knowledge and eventually transition to indigenous innovation. The increased average IF of both national and international publications highlights the overall positive impact on research quality. The more pronounced effect on national publications underscores the facility's contribution to domestic scientific development, balancing national interests with international collaboration. The heterogeneous effects across fields suggest that the impact of BSRIs may vary depending on the nature of research and the degree of international collaboration in a specific discipline.

This study provides strong evidence for the positive impact of major BSRIs on scientific output, particularly enhancing high-impact national publications. The findings emphasize the potential for developing nations to use such facilities to boost their scientific standing. Future research should address limitations like unobserved shocks and heterogeneous effects across narrower disciplines, and should also explore impacts beyond publications, such as industrial R&D and scientific networks. The study also prompts discussions on the optimal balance between national and international collaboration in science and the strategic allocation of R&D investments.

The DID estimations may be affected by unobserved shocks disproportionately impacting treatment and control groups post-2009. The study focuses on the average treatment effect on the treated (ATT), and cannot fully capture heterogeneity across narrower disciplines. The analysis focuses primarily on scientific publications and does not fully capture other potential impacts of SSRF, such as those on industrial R&D and scientific networks. Generalizability of the findings to other types of BSRIs requires caution.