Abstract
A key decision in scientific work is whether to build on novel or well-established ideas. Because exploiting new ideas is often harder than more conventional science, novel work can be especially dependent on interactions with colleagues, the training environment, and ready access to potential collaborators. Location may thus influence the tendency to pursue work that is close to the edge of the scientific frontier in the sense that it builds on recent ideas. We calculate for each nation its position relative to the edge of the scientific frontier by measuring its propensity to build on relatively new ideas in biomedical research. Text analysis of 20+ million publications shows that the United States and South Korea have the highest tendencies for novel science. China has become a leader in favoring newer ideas when working with basic science ideas and research tools, but is still slow to adopt new clinical ideas. Many locations remain far behind the leaders in terms of their tendency to work with novel ideas, indicating that the world is far from flat in this regard.
One-Sentence Summary A new measure of scientific production that measures the tendency to work with new ideas indicates that in biomedicine the United States is closest to the edge of the scientific frontier and that South Korea and China have caught up with the leader on some dimensions, which stands in stark contrast with results from comparisons that focus instead on scientific impact.
Main text
Knowledge production is an increasingly global endeavor. In spite of robust increases in scientific production by the traditional leaders (including the United States, the United Kingdom, and Japan), their relative share has decreased in recent decades because the pace of growth in science by other nations (including China, South Korea, India, and Brazil) has been even more rapid (1,2). The share of international collaborations has also increased, as has the share of citations to papers with foreign authors (1,2). This spread of knowledge production has not been unexpected. It was anticipated long ago that improved communication technologies would make it easier to learn about new discoveries regardless of location and that this would lead to the pursuit of creative work in more diverse places (3).
While this perspective suggests a diminishing influence for location in scientific work, location may in fact continue to have considerable import in science. This is because learning about which new ideas exist may not have been an important benefit of location for quite some time and because location likely still impacts the fertility of creative work in other important ways (4).
One potential remaining influence of location stems from the fact that when new ideas are first discovered, they are often raw and poorly understood. The ideas only gradually mature into useful advances after a community of scientists tries them out and develops them. But such work is hard, harder than work that builds on well-established ideas. Thus, when a scientist seeks to build on a recent advance, it is beneficial to be surrounded by a community of scholars with whom to debate about which new ideas to try out and how (3–8). Daily interactions with colleagues, the training environment, and ready access to potential collaborators are therefore especially important in scientific work that is closer to the edge of the scientific frontier in the sense that it builds on recent advances. Because such local factors influence the fertility of the debates that seek to unlock the mysteries of new ideas, the tendency to work with new ideas can be expected to vary by location. This mechanism – and thus the import of location – may even be increasingly influential. For increases in training times, specialization, and teamwork indicate that reaching the edge of the frontier now involves even more work than before (10).
Therefore, even as the pursuit of science spreads to more diverse places, location may well continue to have an important influence on what kind of science is pursued – through the impact that location may have on the ability to work with novel ideas. Identifying where barriers to knowledge adoption still exist is thus crucial for understanding the role of location in knowledge production and for designing policies that can help eliminate the remaining barriers.
We calculate each nation’s propensity to publish biomedical work that is close to the edge of the scientific frontier in the sense that it builds on relatively recent ideas. The results reveal each nation’s position on the scientific frontier: what share of its contributions to biomedical science build on relatively new ideas vs. well-established ideas. We refer to the constructed measure as the edge factor. Whereas the familiar impact factor measures scientific influence (11,12), the edge factor measures an aspect of novelty of scientific work – the tendency to build on ideas close to the edge of the scientific frontier. These measures capture distinct aspects of science and are complementary tools in policy evaluation and design (13). A feature shared by them is that for each entity both measures quantify the average of a characteristic.
Our empirical analysis is focused on biomedicine because it is an important area of science and because of the availability of the Pubmed/MEDLINE database on over 24 million biomedical research papers. We use text analysis to determine the ideas that each paper built upon and also the vintage of those ideas (see Methods and Materials). Location of each contribution is assigned based on the affiliation of the first author of the paper. We select countries as the unit of analysis because borders continue to influence scientist interactions and because many important science policy decisions are set at the national level.
Figure 1 shows the edge factor for each nation based on papers published during 2015-2016. The edge factor is normalized so that the average edge factor across all contributions is 100. An edge factor of 110 for a nation indicates that on average the nation’s contributions build on relatively new ideas 10% more often compared to the average contribution in the same research area. Markers drawn in red (blue) indicate edge factors that are well above (well below) average. Markers drawn in gray indicate edge factors that are approximately average.
The results show that the United States and South Korea have the most advanced positions on the scientific frontier: scientists working in these nations build on cutting-edge ideas more often than do scientists in other locations. The propensity for novel science is well above average also in Singapore and Taiwan. Countries that come after these four countries have approximately average propensity for novel science. Such countries include China, Canada, most western European countries (including the United Kingdom and Germany), Australia, and South Africa. Other countries (including Turkey, India, Brazil, and Iran) come further behind – scientists in these countries have clearly below average propensities for novel work. Confidence intervals and results for alternative specifications (shown in Table S4) indicate that in most cases these results are robust (the one exception is Saudi Arabia, for which results from alternative specifications suggest a below average tendency for novel work). Countries examined here thus have quite different propensities for work with newer ideas in biomedicine. This indicates that location continues to exert considerable influence on what kind of science is pursued. Furthermore, even developed nations are not on an equal footing in the pursuit of novel scientific work: in some developed nations scientists take advantage of opportunities created by the arrival of new ideas much more often than do scientists in other nations.
Figure 2 shows the change in the edge factor for each nation from the 1990s to present. South Korea, Taiwan, and China have leapfrogged most developed nations. Whereas the United States is still among the leaders, the relative positions of Switzerland and the United Kingdom are less advanced now than they were in the 1990s. Overall some convergence appears to have taken place as the lagging nations are no longer as far behind the leaders, suggesting that the world of ideas may have become somewhat flatter. Analysis of the edge factor by 5-year time periods (shown in Table S5) indicates that most changes that occur are persistent. The changes thus reflect systematic changes in capabilities rather than merely year-to-year random variations.
In our approach, we compare each contribution only to other contributions that use ideas from the same idea category and are linked to the same research area (the 127 idea categories include “Amino Acid, Peptide, or Protein” and “Pharmacologic Substance”; the 125 research areas include “Biochemistry” and “Neoplasms”; see Tables S1 and S2 for the full lists). Table 1 shows the edge factor separately for four groupings of idea categories: “Clinical and Anatomy”, “Drugs and Chemicals”, “Basic Science and Research Tools”, and “Miscellaneous”, and for three groupings of research areas: “Applied”, “Basic Science”, and “Other (Both Applied and Basic Science)”. For most nations the edge factor is similar across these groupings, suggesting that the pursuit of novel work is generally dependent on capabilities that some countries possess but others lack. One important exception is China. China’s contributions linked to the idea category grouping “Basic Science and Research Tools” now have the second highest propensity for novel work (after Singapore), but its contributions linked to idea category groupings “Clinical and Anatomy” and “Drugs and Chemicals” are well below average in terms of their novelty. This result serves to highlight an important feature of our approach: it can be used to reveal not just whether a nation is facing barriers in new idea adoption but where in the idea space those barriers lie.
While our results show that differences persist even among developed nations in their propensity to work with new ideas, the results do not reveal the specific mechanisms driving these differences. One potential driver of these cross-locational differences stems from the difficulty of working with new ideas. Because novel science is harder than conventional science, novel science is more dependent on interactions with colleagues. The fertility of these scientist interactions depends on factors such as the extent of complementary tacit knowledge that is embedded in people and is transferred to others in meetings (5,16). Cross-national variation in the extent and depth of human capital investments can thus lead to cross-national variation in the tendency to adopt new ideas (17).
Willingness to try out new ideas can vary by location also due to differences in scientist demographics. For example, given that early-career scientists are the most likely to work with new ideas (9), and given that the increase in the extent of science in China is so recent and thus many of its scientists are early on their careers, the novelty of science in China may be driven in part by the youth of its scientists. Cross-national differences in new idea adoption and China’s remarkably ability to leapfrog in this regard may also be driven in part by differences in incentives to pursue novel work: it has long been understood that nations without vested interests in existing technologies have an elevated incentive to explore new ideas (18,19). Some of the variation in new idea adoption can also be driven by variation in where the ideas are first born, and by remaining delays in the spread of awareness about which new ideas exist.
Our results are consistent with findings from recent related work that measured the complexity of each country’s production structure based on its exports and found large differences in the capabilities of nations (20). Their analysis was motivated by the idea that a nation’s capabilities determine the input varieties that can be fruitfully used in production. Our work, by contrast, is motivated by the idea that capabilities determine whether a nation’s scientists can take advantage of the opportunities created by the arrival of new ideas. Moreover, whereas in this related work the complexity of goods production is measured indirectly based on exports, the edge factor is calculated directly based on the measured idea inputs. Common to these analyses is the belief that the capabilities of a nation affect which inputs it uses and both analyses are aimed at constructing new measures that reflect those capabilities.
Our finding that nations continue to differ in their ability to pursue novel science is in line with cross-country comparisons of scientific impact as measured by citations (1,2). The ability to take advantage of scientific opportunities continues to vary across locations in spite of the “death of distance” phenomenon, because locational differences in capabilities persist (21–24). But some aspects of our results also differ from the results obtained through traditional analyses of scientific productivity. Data on the tendency to produce highly cited papers point to the United States as a leader that remains far ahead of most western European nations and even further ahead of South Korea, Taiwan and China (1,2,25–27). Our analysis on the use of new ideas, by contrast, suggests that South Korea, Taiwan and China have caught up with western Europe and are now close to the United States in terms of their tendency to work with cutting-edge ideas. Moreover, we find that China is now a leader in favoring newer ideas when working with new basic science ideas and research tools. The finding that some countries are among the leaders in terms of their edge factor but lag in terms of their impact is not surprising (28). For work on an idea early – when the idea is still raw – may well have less impact than work that builds on more established ideas which properties are better understood. The early work on the idea is still crucial: it helps the idea develop and thus makes more significant advances possible. Moreover, countries investing heavily in novel science can reap significant benefits also for themselves from their focus: early work on an idea can help the country develop capabilities that enable it to take advantage of the later, more fertile, opportunities linked to the same idea.
Because the edge factor captures an aspect of science that is distinct from impact, it has potential applications also beyond cross-national comparisons. This is important because the obsession with impact – decried even by an editor of Science (34) – may have led to less healthy science: the rise of citation metrics coincided with a decline in the novelty of biomedicine (31). A singular focus on citation counts can lead to stagnant science because impact factors under-reward scientists who try out new ideas, thereby stifling work that helps ideas mature and makes more meaningful advances possible (14,15). By using measures like the edge factor in conjunction with impact-based metrics, university administrators and funding agencies can strike a better balance between rewarding innovative but risky work that develops ideas early on and rewarding work that takes advantage of the ideas in their more mature stages.
Acknowledgements
I thank Jay Bhattacharya, Bruce Weinberg, Partha Bhattacharyya, Richard Freeman, Horatiu Rus, Joel Blit, David Autor, Larry Smith and Peter Tu for discussions. I acknowledge financial support from the National Institute on Aging grant P01-AG039347.