Identifying National Research Strategies

This study builds upon a stream of research where publication data is used to detect national research strategies. As such, it is important to start with the seminal studies by May and King. May [4], showed that nations with larger R&D investments had larger shares of the scientific pie in terms of both paper and citation counts. Several years later, King [5] expanded upon May’s study, including a number of R&D funding variants (e.g., HERD, etc.) and extending the analysis to seven broad scientific fields. While largely affirming May’s results and showing the relationship between economic and scientific wealth, he also showed some differences between European countries based on their publication profiles across fields.

It is important to point out that these studies did not examine national research strategies per se. Rather, their focus was on national strengths based on research outcomes (publications and citations). These outcome data were normalized so that national strengths (outcomes greater than the norm) could be determined. Alternative normalization methods have been proposed which correspondingly result in a different ordering of national strengths. For instance, Leydesdorff & Zhou [6], using King’s data, identified a new group of emerging nations with high growth that were not highlighted by King. Rousseau & Rousseau [7] investigated the efficiency of European nations with GDP, R&D expenditures and population as normalizing inputs, and showed that rankings change somewhat with changes in the definition of efficiency. Cole & Phelan [8] showed that when normalized by population, productivity was no longer fully explained by wealth, but that religion, decentralization and competitiveness were also factors. Pan et al. [9] correlated country-level data on cites per paper (CPP) with R&D expenditure per researcher, finding that the correlation was threshold-dependent. Below $100,000 USD per researcher per year there is a strong correlation between CPP and spending, while above that level there is no correlation. Cimini et al. [10], using Scopus data, found that leading nations have more diverse research systems than nations whose research systems are “under construction”.

For purposes of discussion we will refer to these national strengths as national strategies. Nations do not maintain research strengths unless there is an intention to do so. For example, King’s observation that the United Kingdom had very high impact in the medical sciences (in relation to a peer group of nations) reflects a decision to spend a greater percentage of research dollars on medical research. The decision to spend relatively more research dollars on medical research can be interpreted as a national strategy, focusing more national efforts towards this end.

The above-referenced studies were all limited in the sense that they did not attempt to determine the axes of national focus or research strategy and the reasons behind the differences. Table 1 lists five studies that did attempt to answer the strategy question rather than simply looking at strengths. Dore and colleagues [11, 12] were the first to use a more sophisticated method for detecting national research strategies from large scale publication data. Using a 12-year set of publications from the Science Citation Index compiled into 18 high level (journal-based) fields, they used correspondence factor analysis to group fields using the data from 48 countries. They identified a large number of factors, but decided to focus on the first two factorial axes (these two factors had much higher eigenvalues). They mapped the location of nations on these two axes. Their first axis highlighted the differences between the natural (labeled by Doré as ‘ancient’) and life (or ‘modern’) sciences, while the second axis differentiates between agricultural sciences and geosciences. They found that nations occupied all four quadrants of their graph, and concluded that nations were choosing between four distinct research strategies.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Research strategies as identified by studies of national publication patterns.

https://doi.org/10.1371/journal.pone.0169383.t001

Our second example comes from the Second European Report on S&T Indicators 1997 (REIST-2) [13], one of whose analyses was to determine preferred research fields for scientific collaboration by country. Fields were grouped, and four publication profile patterns were identified: 1) a ‘western model’ based around clinical medicine and biomedical research, 2) a ‘former communist model’ focused on chemistry and physics, 3) a ‘Japanese model’ centered on engineering and chemistry, and 4) a ‘bio-environmental model’ with a focus on biology, earth and space sciences. Schulz & Manganote [14] used Science Watch (Thomson Reuters) country profile data and found patterns similar to those from the REIST-2 report. One novel feature of this study was that it included the social sciences, and that this inclusion differentiated England from most other European nations. Moya-Anegón & Herrero-Solana [15] also included the social sciences, although these data did not figure strongly into the results, which featured three clusters of nations–the biomedical cluster, the basic science and engineering cluster, and the agriculture cluster. Finally, Chen & Chen [16] grouped 100 nations into 12 groups using minimum spanning trees, finding that nations within each group were similar in terms of geography, ethnicity, or economic status. Further clustering placed nine of those groups into three main clusters–a Western cluster focused on biomedicine, an Asian and East European cluster focused on the natural sciences, and a third cluster associated with developing countries.

At the level of three or four major research strategies the results from these five studies are relatively robust. All five contain two clusters focused on the natural sciences (i.e. chemistry, physics, etc.) and life sciences (i.e. biomedicine), while three contain a third cluster focused on agriculture that is associated with developing countries. The use of factor analysis, or other forms of dimensional reduction, enables identification of the different research strategies that a nation can pursue.

Defining Disciplines and Strategies

A majority of the studies mentioned above used journal-based categories or higher level groupings of such categories, ranging from seven (King) and 56 (Chen) categories, to represent strategies that could be pursued by nations. However, use of journal-based categories to represent the structure and partitioning of research is problematic. Journal-based categories are disciplinary (they represent historical institutional structures) rather than cognitive (reflecting how scientists are self-organizing around problems). Collins [17] questioned the use of disciplinary categories for policy-making over 30 years ago when he wrote: “… to develop a policy with cognitive goals in view, it is essential to start by disaggregating science according to cognitive rather than institutional boundaries–that is, to think of science as being made up of sets of research areas which involve scientists who interact, or mutually refer, across institutional boundaries, because of their common cognitive interests.” This perspective is reflected in the current emphasis on interdisciplinary and translational research. [18]

The problematic nature of journal-based categories became apparent in a recent study where the accuracy of different document partitioning strategies was compared. Using Scopus data, Klavans & Boyack [19] compared seven different journal classification systems with article-level classifications at different levels of granularity created using direct citation, bibliographic coupling, and co-citation methods (see Fig 1). Comparisons were made using a Herfindahl (or concentration) index calculated using the references in over 37,000 papers with at least 100 references each. These articles with large numbers of references were considered ‘gold standards’ because they tend to provide a comprehensive review or synthesis of a topic area. The Herfindahl index measures how closely a classification system preserves the topic structures as elucidated by the authors of the ‘gold standard’ articles. A high concentration value signifies that the classification system is structured in a way that corresponds with the authors’ perception of topical structure, while a low concentration value suggests that the classification system does not correspond with the authors’ logic. Comparing the accuracy of classification systems with different numbers of clusters can be done by comparing slopes of lines from the (1,1) point to the different Herfindahl values; the method with the least negative slope is the most accurate.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Relative accuracies of 17 document clustering solutions, adapted from reference 19.

Designations of fields, disciplines, specialties and topics in terms of numbers of categories across all of science are not meant to be exact, but rather conceptual.

https://doi.org/10.1371/journal.pone.0169383.g001

Fig 1 makes a distinction between classification systems at different levels, from disciplines (hundreds of categories, as in most journal classification systems) to topics (tens of thousands of categories, similar to Kuhnian research communities). Most relevant to this study, it shows that at the discipline level article-based classification systems (DC2 and BC2) do a much better job than any of the journal classification systems at reproducing structures defined by authors, and thus can be considered to be more accurate from this point of view. Article-based classifications also have the advantage that they reflect the cognitive structure of science, and thus satisfy Collins’ argument far better than do journal-based analyses. In addition, we have recently shown that publication profiles of most institutions are more aligned with the DC2 classification system than with journal classification systems [20]. Thus, in this study we will use the 114 DC2 discipline-level categories for our analysis. These were defined by clustering roughly 50 million documents from the Scopus database using direct citation. A description of these 114 disciplines is available in S1 Table.

In summary, there has been significant progress in the ability to identify national research strategies. King and May started with literature datasets that were somewhat limited in scope. Broadly defined disciplines were used to detect national research strategies. More sophisticated methods to normalize, and to detect the underlying dimensionality of choice, have been applied. Most importantly, a more accurate method for identifying cognitive-based structures in the literature is now available. These improvements provide the foundation from which our methodology will be built.

Motives for Research

Historically, the primary justification for investing billions in research has been based in economics. When Carl Linnaeus (1707–1778) asked the King and Queen of Sweden to support his efforts at creating plant taxonomies, he argued that, if successful, he would be able to create cold-hardy plants that could be grown in Sweden, thereby allowing Sweden to develop national wealth based on agriculture [21]. In the 19th century, the U.S. government succeeded in making agriculture the basis for national wealth via the establishment of regional agricultural colleges and agricultural extension programs. In the 19th and early 20th century, different nations invested in basic research that supported national advantages in applied research (Germany in chemistry, France with its polytechnic schools). The industrial strength of the U.S. at the turn of the 20th century was due to entrepreneurs such as Carnegie, Rockefeller and JP Morgan, who exploited the link between science and invention in a nation that had few restraints on capitalism. The 1930’s saw the rise of large industrial laboratories as the source of innovation and economic growth. The 1970’s marked the decline of these large labs, a shift to open innovation systems and the resurgence of Europe and Japan as research leaders. In the past 10 years, China’s scientific and technical publication activity has risen from 24% of the U.S. output (in numbers of articles) to 97% of the U.S. output [22], with a corresponding rise in economic power. Overall, there is an extensive literature on the relationship between science, invention, innovation and economic growth that, in essence, points to economic advantage as the primary motive for research [23–27].

Why then, in this context, would the United States spend eight billion dollars to develop and launch the James Webb Space Telescope (JWST)? Although spillover benefit is often used to justify investments in astronomy, space exploration, and other basic sciences [25], the primary reason these things are funded is to gain knowledge, to seek answers to big problems. For instance, the JWST isn’t designed to seek out economic opportunities; it’s specifically designed to answer basic questions about the birth of the universe. Likewise, why do researchers choose careers in basic sciences rather than those that directly benefit the economy? Interviews with researchers in astronomy and space exploration reveal part of the answer. Astronomers are, in large measure, men and women who, as children, would look up at the stars in awe. They maintained this sense of awe in choosing majors and career paths. Personal motives for going into astronomy and astrophysics are not based on homo-economicus; they are based on curiosity–an altruistic analog. At both the researcher and funding levels, much of research is being done for reasons that are clearly not economic in nature. A different part of the national brain is being stimulated when altruistic research is considered.

We suggest that altruism is a dominant motive for research. Discovering basic principles of life and nature and making them manifest are an integral part of the human psyche. These research activities, along with those aimed at reducing illness, protecting the weak, providing world-wide educational opportunities, sustaining the earth as our home, reflecting on the meaning of life and living in a fair and just civilized state, are rooted in altruism. It is this tension–between our current belief about the role of research in economic progress and our observation that many research activities are strongly rooted in altruistic motives–that is the focus of this background discussion. As such, we will first elaborate on the traditional views of research and economic growth before we turn our attention to the possibility that altruism plays a significant role in why nations spend tens of billions of dollars on research. The background section concludes with a summary of how one can investigate these issues using publication data.

Economic Competition and Research

Joseph Schumpeter (1883–1950) significantly influenced the economic policy in the United States and Europe via his focus on entrepreneurship, technological innovation and corresponding waves of economic destruction [28]. His position countered the Marxist’ criticism of capitalism at the turn of the 20th century. And while his beliefs about the advantages of capitalism changed when the large industrial labs were being formed in the 1930’s (i.e., due to a realization that socialism and capitalism are equally likely to benefit from investment in R&D), the underlying belief that there are long-term cycles in research and economic growth still holds merit [29]. Modern day economic theories about subsidizing R&D [30, 31] and the expected growth that will result from these investments are deeply rooted in Schumpeterian theory. Since patents are a primary mechanism for capturing the benefits from research, there is a corresponding literature on what subjects are patentable [32, 33]. Overall, there is a significantly large literature in the economic discipline that assumes that R&D is, in essence, motivated by economic considerations.

The field of strategic management also focuses on economic motives for research. Early work on corporate diversification, growth and profitability pointed out that the ‘exploitation of a scientific base’ was one of the three avenues for successful diversification [34]. This stream of research on corporate diversification was re-interpreted using the concept of absorptive capacity [35]. Absorptive capacity is the ability of a firm to identify, assimilate and apply external knowledge. This is now the dominant explanation for corporate research; firms conduct research in order to learn about what moves to make in the competitive environment and whether first mover advantages exist [36]. From the perspective of strategic management, corporations do not publish scientific articles for the altruistic reasons of contributing to world-wide knowledge and well-being.

Both economists and managers agree that entrepreneurs play a critical role in converting research into economic benefits. As such, there has been an extensive literature on the role of entrepreneurs, especially in regional growth. Starting with Roberts’ study of technical entrepreneurs from MIT that formed the Route 128 firms around Boston [37], there are now many studies that look at academic entrepreneurs [38], how the entrepreneurs at universities contribute to regional growth [39] and how incubators can be created (with corresponding research subsidies) in order to foster regional innovation [40]. There is also an extensive literature on the characteristics of these entrepreneurs and how entrepreneurs recognize opportunities [41–43].

The practical problem of taking research to the marketplace has also been addressed by academics studying the innovation process. Most notable were the early work on the characteristics of successful vs. unsuccessful new products [44] and the diffusion pattern of innovation [45]. As the source of innovation shifted from the large industrial labs, there has been corresponding work showing how innovation could be financed using crowd sourcing [46].

In summary, the majority of the academic literature that addresses the question ‘why do we do research?’ is deeply embedded in the assumption that research has economic benefits. Joseph Schumpeter is one of the most influential forefathers of this research, and there are many well developed research questions that are currently being investigated from the underlying assumption that inventors convert ideas into patents and entrepreneurs convert opportunities into product and process innovations. The assumed goal is economic gain or is strongly related to economic gain, such as is the case for research that enhances national security.

Altruism and Research

Altruism provides a strong alternative to the economic motive for research. For example, the economic motive is not the dominant reason behind funding for the JWST (as mentioned above), nor is it the driver for the large amounts of funding that are directed (primarily by wealthy nations) to detect and respond to pandemics that, while they generate fear, are relatively unlikely to significantly affect those nations. At the personal level, many medical researchers chose their professions in response to a desire to find a cure for a particular disease that affected a family member or friend. In addition, many non-profit organizations (NPOs), both large and small, fund research related to their causes and missions, all of which have a strong altruistic component. Altruism plays a significant role in research.

Despite this role, the relationship between altruism and research has not been subject to the same level of study as has the economic motive. Altruism has had no intellectual champion such as Schumpeter. While the economic motive for research has been the subject of thousands of articles in dozens of research communities, a sampling of which is mentioned above, the literature on altruism is focused on topics other than its relationship with R&D. Most altruism research is performed within the context of topics on biological fitness and behavior, cooperation (in the evolutionary dynamics and game theory sense), organ donation, professionalism (particularly in medicine), medical education, volunteerism, citizenship, and monetary donation. These are all important topics, but none of them address the relationship between altruism and R&D.

In order to better understand the diversity of altruistic motives in non-profit organizations, we recently developed a ‘map of altruism’ based on a textual analysis of the websites of over 125,000 NPOs in the U.S. [47]. The relative positions of these institutions are shown in Fig 2, where each dot represents a specific NPO. Neighboring NPOs in map space share similar mission statements as reflected by the words used on their websites. Each NPO is color coded using the National Taxonomy of Exempt Entities (NTEE) system used by the Internal Revenue Service to classify NPOs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. A map of altruistic missions for 125,000 NPOs in the United States.

Each NPO is represented by a dot, and is color coded using high-level National Taxonomy of Exempt Entities (NTEE) categories (see legend).

https://doi.org/10.1371/journal.pone.0169383.g002

Not surprisingly, one of the most populated areas on this map of altruism deals with the treatment of disease (lower right). NPOs in this area actively fund medical research without any expectation of economic gain. Their motives harken back to the same motives that were mentioned in the introduction–to extend life [1].

A comparable level of activity is associated with Civics. This area of the altruism map includes NPOs concerned about the type of society we live in and social dilemmas, as well as others that focus more on political issues. It is interesting to note that within Civics we see a large number of NPOs devoted to the issue of technological innovation. Consider, for example, the number of non-profits and regional development organizations that have been formed to encourage entrepreneurial behavior. While economists might argue that the motive of the entrepreneur (and the venture capitalist funding the entrepreneur) is personal gain, the NPOs helping entrepreneurs must (by law) be more altruistic.

NPOs interested in environmental issues are well represented in Fig 2 (top left) and have their own NTEE code. Many of these NPOs fund research on climate change and are concerned with the world’s ability to produce food for billions of people in the face of climate uncertainty. Related research on food production is part of the Agriculture strategy highlighted in Table 1.

Interestingly, many of the topics of altruism research–specifically organ donation, medical education, citizenship, volunteerism, and monetary donation–underpin the areas of altruism shown in Fig 2, and reflect the variety of ways that people and institutions give of their time and resources to altruistic causes. This diversity of mission space is reflected in the scientific literature on altruism. To the extent that NPOs fund scientific research related to these missions, there is an implicit link establishing altruism as a motive for scientific research. However, this relationship as a motive for research has not been investigated in a scientific way.

Some might be tempted to equate the conception of altruism-driven research with so-called Mode 1 research [26], which is defined as research motivated by curiosity or the desire for knowledge. However, we note that research driven by altruism can be application-oriented just as easily as it can be knowledge-oriented. In fact, the mission space of Fig 2 provides ample evidence that many, if not most, altruistic activities are aimed at making some sort of positive change rather than simply learning something new. Overall, we suggest that altruism is a strong motivator for R&D, despite the fact that the economic motive dominates the current research agenda. There are no indicators that tell us if an area of research is driven by altruistic or economic motives. There are no studies that utilize publication profiles of thousands of institutions from around the world to characterize fields of research. Finally, there are few studies that systematically explore a number of alternative reasons for a nation to focus on one field or another. The first two gaps are methodological, while the third is theoretical and is correspondingly addressed in the following sections.

Institutions

We identified a set of 4429 institutions for use in this study. Institutions were identified manually several years ago by grouping Scopus affiliation profiles into institutions for those institutions publishing at least 50 papers per year (at the time). Many institutions have only a single affiliation profile in Scopus. However, many of the larger institutions have medical (or other) schools or research institutes that, while they are clearly part of the parent institution, have separate affiliation profiles in Scopus. For example, our institutional profile for the University of California, San Diego (UCSD) contains six affiliation profiles including those for the Scripps Institution of Oceanography, UCSD School of Medicine, and San Diego Supercomputer Center. Each institution was also assigned to a sector (e.g., academia, industry, government), which was, in most cases, an obvious choice. Given that these institutional groupings are several years old, they do not reflect recent mergers or contain institutions that were below the threshold at the time, but that are now publishing larger numbers of papers. The distribution of institutions by sector is shown in Table 2. On average, academic institutions publish larger numbers of papers than other types of institutions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Institutions by sector (2010–2013).

https://doi.org/10.1371/journal.pone.0169383.t002

As mentioned above, the DC2 document classification system is used in this study. Thus, publication profiles for each institution were represented as article counts by DC2 category using publications from the 2010–2013 time period. These data also enable calculation of the fraction of publications in each DC2 that are authored by industrial institutions, which indicates whether a DC2 discipline is aligned more with economic motives (a relatively high industrial fraction) or with altruistic motives (a relatively low industrial fraction). We use this fraction as an indicator of motive later in the study.

National Characteristics

Publications in the Scopus database are assigned to 240 different countries. National publication counts range from over 1.8 million (USA) to less than a single paper (nine nations, e.g., Tuvalu) over the 2010–2013 time period using fractional counting and excluding any publication which could not be assigned to a DC2 discipline. The distribution of national publication counts in Fig 4 shows different regimes. For example, the distribution is relatively linear on a log-log scale for the first 40 nations, and is also relatively linear for the next 70 nations but with a different slope. Beyond the first 110 nations, the dropoff in publication rate is more severe. As such, we will limit our analysis to the ‘top 40’ nations (where one can expect a more intensive infrastructure for research) and the ‘next 70’ nations (with a relatively weak research infrastructure).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Distribution of national publication counts.

https://doi.org/10.1371/journal.pone.0169383.g004

As mentioned above, one of the goals of this research is to correlate national research profiles with national indicators of wealth, education, capitalism, culture, religion and language to determine which of these other factors are associated with altruistic motivation at the national level. As such, we include a number of national level indicators in this study. The specific indicators used and their sources are listed in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. National indicators and sources.

https://doi.org/10.1371/journal.pone.0169383.t003

A description of these indicators along with the correlations we expect to see are given here.

• Wealth: Per capita GDP is a commonly used indicator of national wealth. We expect that only wealthy nations will fund high levels of research in altruistic disciplines and fields.
• Education: The United Nations’ Education Index is based on expected and actual years of schooling for each nation. Given the current belief that economies are largely knowledge-based [48], we expect that altruistic research will be associated with high levels of education at the national level.
• Capitalism: The Frazer Institute’s Economic Freedom Index represents a strong libertarian focus on issues such as personal choice (instead of collective choice), voluntary markets, freedom to enter and compete in markets and protection of property rights. This index also represents Schumpeter’s initial premise that R&D is more likely to flourish in capitalistic systems. We have no working hypothesis about which form of government is more likely to support altruistic research.
• Culture: We used two indexes created by the Hofstede Center related to culture. The Power Distance Index expresses the extent to which the less powerful members of society accept and expect that power is distributed unequally, while the Individualism Index reflects the degree to which people believe that they should take care of themselves or prefer being in a tight-knit group in exchange for loyalty to that group. We have no working hypothesis about whether the cultural aspects of a nation are more or less likely to be associated with altruism.
• Religion: We include an indicator of religion simply because religion is clearly an important element in altruistic motivations, as indicated by our map in Fig 2, and because religion has played a historical role in how science has evolved over centuries. We use percent Christian as the indicator since it is the dominant religion in 73 of the 110 nations included in the study, although data on other religions (Muslim, Buddhist, Hindu, Jewish, etc.) are also available from the original source.
• Language: Degree of English-speaking is used as our indicator of language. This indicator is included because the Scopus database used to generate publication profiles focuses mostly on English publications. We note, for example, that there are extensive literatures on Chinese medicine and Ayurvedic medicine that are not well covered in Scopus (or other major databases that cater to the Western world). Many engineering conferences and other conferences of a more applied nature are published in local languages and may be under-represented. Our working hypothesis is that English-speaking nations are more likely to be associated with altruism, but that this may reflect database bias.

Indicator values for the 110 nations in our sample are provided in S1 Table. Complete data (for all 110 nations) were available for three of the indicators: wealth, religion and language (see Table 4). The education indicator had 4 missing values while the culture indicators are missing for 24 nations. Values for language are missing for 38 nations; we have used estimated values of 1% English-speaking for the 4 missing values among the top 40 nations, but have not used estimates for the other 70 nations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Characteristics of the nine research strategies identified using factor analysis.

https://doi.org/10.1371/journal.pone.0169383.t004

In summary, our hypotheses are mostly exploratory in nature. We expect that factor analysis will group disciplines into higher-level fields of research that are internally consistent and valid. We expect that the participation of industry in a field of research will create a reasonable ordering of those fields in terms of altruistic motivation. While researchers have suggested that wealth, education, capitalism, culture, religion and language will explain why nations focus on different strategies, we consider these as alternative hypotheses and will let the data tell us which explanations best explain variations in national focus along a potential altruism-economic axis.

Factor Analysis

Nine research fields were identified by grouping of DC2 disciplines using factor analysis. Inputs were the publication profiles of the 4429 institutions across the 114 disciplines. Oblique rotated factors were used to allow factors to be inter-correlated because, a priori, there is no reason to believe that fields of research are orthogonal [49, 50]. The factor analysis generated, as output, vectors that represent ‘underlying dimensions of choice’. In essence, disciplines that occur together across institutional publication profiles tend to load strongly together on the same factor, and are thus grouped in the same field.

The factor analysis produced seven factors with an eigenvalue of greater than 1.0. Two of these factors represented alternative (or opposed) fields, where there was one group of disciplines with highly positive coefficients and another group of disciplines with highly negative coefficients. For instance, one of the pairs of opposed fields is sustainability and applied physics; institutions are typically focused on one or the other, but not both. These two factors resulted in four fields. The remaining five factors each represented a single field rather than a pair of opposed fields in that positive coefficients were high and negative coefficients were near zero. In total, a set of nine fields were identified. Characteristics of the 114 DC2 disciplines and their assignments to the nine fields are given in S2 Table.

Field Characterization

The nine research fields identified using factor analysis have been characterized in a number of ways. In this section we name and describe each field, and quantify each with respect to industry participation. Industrial authors were identified as those associated with one of the 291 industrial institutions from Table 2, or whose affiliation data as indexed in Scopus contained one of the industry-related strings given in S3 Table. The major source for these strings was Wikipedia (https://en.wikipedia.org/wiki/Types_of_business_entity), with some additional strings suggested by Sugimoto et al. [51]. We note that the string-based search was essential given that the results from our 291 (large) industrial institutions only identified 1/3 of the industry authorships compared to the full search strategy. This suggests that 2/3 of industry-authored papers are coming from smaller companies, those that publish less than 50 papers per year.

Following is a discussion of each field of research, starting with the label we have used for naming. Labels for each field were assigned manually and were based on examination of the DC2 disciplines and the institutions with the top coefficients for each field. Fields are ordered by increasing fraction of industry authorship. For purposes of discussion, we have defined two industry participation thresholds. Overall, industrial authors participated on 7.07% of papers from 2010–2013. We define disciplines with more than 10.0% industry authorship to be in the ‘economic’ range (26 disciplines, 21.6% of papers), while those with less than 4.0% industry authorship are in the ‘altruism’ range (23 disciplines, 20.5% of papers). Numbers in parentheses in the descriptions below refer to DC2 discipline numbers (see S2 Table).

9—Engineering.

All of the 16 disciplines associated with engineering have industry involvement that is higher than the average for the full database (>7.07%). The most economic-oriented disciplines deal with Petroleum Engineering (105), Power & Electricity (43), Nuclear Science (94) and Geological Engineering (53).

Overall, the factor analytic approach has created a very coherent grouping of 114 disciplines into nine fields, characteristics of which are given in Table 4. Medicine is by far the largest field with nearly 21% of papers, while Infectious Disease and Basic Physics are both quite small, with around 3% of papers each. The rest of the fields are more similar in size. In addition, Fig 5 shows that not only do the nine fields reflect a continuum from less industry involvement (more altruistic) to more industry involvement (more economically motivated), but that there are few altruistic disciplines in economic fields, and few economic disciplines in altruistic fields, suggesting that some fields (and disciplines) are inherently more altruistically motivated while others are more economically motivated. Industry involvement provides a very reasonable ordering of fields and disciplines within each field (see S2 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Industry involvement in 114 DC2 disciplines assigned to nine fields of research.

Each discipline is represented by its number as listed in S2 Table. Circle sizes reflect discipline sizes while filled circles indicate average industry authorship by field.

https://doi.org/10.1371/journal.pone.0169383.g005

The field with the least involvement by industry is Civics. In addition to being the most altruistic field, Civics is the second largest and fastest growing field. However, it has a surprisingly low level of international collaboration; the average paper in this group had 1.41 national affiliations. Those fields with the highest industrial involvement are where one most feels the Schumpeterian winds of creative destruction, especially the fields of research based on Applied Physics and Engineering.

International collaboration patterns may indicate whether the field of research represents local or global issues. For example, Basic Physics, Medicine, and Sustainability exhibit relatively high international collaboration. Conversely, those fields where industry is most involved–Engineering, Computing Technology, and Applied Physics–have lower values. The lower level of international collaboration for the Civics field may indicate that this strategy is mostly dealing with local, rather than global governance issues.

National Research Strategies

Using the fields defined and characterized in the previous sections, we can now examine which nations focus on which fields of research. Fig 6 provides relative publication fractions by field for six different groups of nations, with each group having a different profile. Relative publication fraction is the fraction of documents by a nation in that field divided by the fraction of documents worldwide in that field (“Focus” in Table 4). For example, the US has 22.6% of its output in Civics, which is 1.44 times the world average of 15.7%. National publication fractions by field are given in S4 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6.

Relative publication fraction portfolios for six different groups of countries: a) civics portfolio, b) balanced portfolio, c) engineering portfolio, d) nanotech portfolio, e) disease portfolio, and f) sustainability portfolio.

https://doi.org/10.1371/journal.pone.0169383.g006

Fig 6A shows publication fractions for five countries (usa, gbr, aus, nld, hrv) whose portfolios by field are roughly opposite to the industry publication profile by field that is shown for comparison as the dashed line. These countries publish proportionally more in fields with relatively low industry participation, and proportionally less in fields with relatively high industry participation. One might say these countries have a Civics-oriented portfolio that is not consistent with overall industry (or economic) profile. Fig 6B shows a much more evenly distributed set of portfolios belonging to several European countries (deu, fra, ita, esp), with the exception of the spike in Basic Physics for Germany and Italy (and France to a lesser degree) that is due to their involvement in CERN.

Fig 6C and 6D show two different types of portfolios that both roughly follow the industry participation (or economic) profile. They differ in that Engineering predominates for some countries (Fig 6C: chn, irn, mys, rou, dza), while Applied Physics or materials and nanotechnology dominates for others (Fig 6D). There are two subtypes among the Applied Physics dominant countries, those that also focus on Basic Physics (jpn, ukr) and those that focus on Computing applications (kor, twn, sgp). The final two panels show countries whose portfolios are focused on Infectious Disease (Fig 6E: zaf, tha, nga, ken, eth) and Sustainability (Fig 6F: bra, mex, arg, chl), respectively. Chile also shows a spike in Basic Physics, which reflects its relative strength in Astronomy associated with their world-class observatories.

These six different types of portfolios show a great diversity in national research focus, and suggest that motives other than economic benefit, such as health, quality of life, and environmental concern, are drivers in many nations. Although some of the groupings in these examples seem to fall along geographical lines, there are also many cases where geographies are mixed in non-intuitive ways. For example, Iran, Romania, Algeria (irn, rou, dza) have Engineering-dominant profiles similar to that of China, and Croatia (hrv) has a Civics-dominant profile like that of the USA.

To give an overall idea of whether the research agenda of a nation is more aligned with an economic or altruistic profile, and for the purposes of correlating national research profiles with other characteristics, we have reduced motivation to a single value. We propose a motivation index based on the distribution of a nation’s publications over the nine fields. The motivation index is calculated for a nation as M = ∑ f p_f, where f is the field number (1–9) and p_f is the fraction of papers in that field for that nation. Fractions are listed for each nation in S4 Table, along with the calculated value of the motivation index. Fig 7 shows these index values as a function of number of publications for all nations with at least 10,000 papers published (summed fractional counts) from 2010–2013. It also shows that the motivation index for full set of 110 nations in this study is 4.65. The fact that this index value is less than 5.0 reflects the fact that there are more publications in the more altruistic fields (1–4) than in more economic fields (6–9).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Motivation index for the 54 nations with at least 10,000 publications from 2010–2013.

Index values for various aggregations of nations are also shown for comparison.

https://doi.org/10.1371/journal.pone.0169383.g007

The two top-producing nations provide a sharp example of the difference between a more altruistic research strategy (USA, M = 3.90) and a more economic research strategy (China, chn, M = 6.28). The USA has a majority of its publications in Civics and Medicine, while China focuses on Engineering, Computing and Applied Physics. In contrast, many European nations, including Germany and France (deu, M = 4.67; fra, M = 4.71), have a more balanced research portfolios with motivation index values very close to the world average. In general, northern European nations (e.g., nld, dnk, nor, swe) have a slightly more altruistic focus than the rest of the European nations.

Fig 7 also shows the motivation index for a number of continents and two additional aggregations of countries–Commonwealth countries and Post-Soviet States (assignments are given in S1 Table). Several of these aggregations are simple reflections of their dominant country–e.g., North America and the US, South America and Brazil, and the Post-Soviet States and Russia. The Middle East and EU (and more particularly the EU without the UK) have the most balanced strategies. The Commonwealth, consisting mostly of former British territories, is comprised of two components–one which is less economically motivated (i.e., aus, can, nzl, zaf, nga) and mirrors the UK (gbr), and one which is more economically motivated (i.e., ind, mys, sgp) and reflects the Asia-Pacific philosophy. In the latter case, geography seems to matter more than the former British influence. Possible reasons for national differences are examined in the following analyses.

Field-Indicator Correlations

The final step in our analysis was to correlate national indicators with the motivation index to determine if any of the indicators might suggest reasons for differences between altruistic and economic motivations for research. Table 5 shows that the values of these correlations for the top 40 nations and the next 70 nations (those ranked 41–110) are significantly different. For the top 40 nations, the two most explanatory national indicators are individualism and religion. Although power, language and education have higher correlations with the motivation index than does religion, they are highly correlated with individualism (>0.63 in each case), and are thus less explanatory than religion, which has much smaller correlations with these variables. The next 70 nations have much weaker correlations overall, with individualism having only the fifth highest correlation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Correlations between motivation index and national indicators.

https://doi.org/10.1371/journal.pone.0169383.t005

A common perception is that only the most developed nations can afford expensive research or research that is not directly aimed at economic gain. Indeed, we expected to find that altruistic research would be associated with high levels of wealth.–However, this was not substantiated by the data. The correlations between wealth and motivation are among the lowest for both sets of nations. Moreover, Fig 6 shows a number of poorer nations that have research portfolios that are similar to more wealthy nations, and also shows that nations with particular needs in Disease and Sustainability focus much of their research in those areas rather than pursuing research in more economically oriented fields.

We also expected that motivation would correlate strongly with education. It is more strongly correlated than is wealth, but it is also highly correlated with individualism, which is the more explanation feature. We did not expect any correlation between motivation and capitalism, and indeed there is very little.

Fig 8 visually shows the relationships between our motivation index and the individualism index for the top 40 and next 70 nations. The correlation between individualism and motivation for the top 40 nations is strong and negative (Fig 8, left); nations with a high individualism index are likely to have research portfolios that align more strongly with altruistic motives, while those with low individualism indexes are likely to be more economically motivated. This correlation flips, however, for the next 70 nations–the correlation is positive rather than negative, and there is also much more variance in the distribution (Fig 8, right). The two distributions and corresponding correlations are markedly different, suggesting that our choice to split the nations into two regimes (top 40 and next 70) was reasonable. We stress here that these are correlations only, and do not indicate causality. In addition, it is clear that none of the national indicators that correlate highly with the motivation index are independent, and thus the indicators are suggestive rather than explanatory.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Motivation index as a function of individualism for the top 40 nations and next 70 nations.

Legend as in Fig 7.

https://doi.org/10.1371/journal.pone.0169383.g008