J. Sylvan Katz

What is Research Collaboration

Although there have been many previous studies of research collaboration, comparatively little attention has been given to the concept of ‘collaboration’ or to the adequacy of attempting to measure it through co-authorship. In this paper, we distinguish between collaboration at different levels and show that inter-institutional and international collaboration need not necessarily involve inter-individual collaboration. We also show that co-authorship is no more than a partial indicator of collaboration. Lastly, we argue for a more symmetrical approach in comparing the costs of collaboration with the undoubted benefits when considering policies towards research collaboration.

Where Is Science Going

Do researchers produce scientific and technical knowledge differently than they did ten years ago? What will scientific research look like ten years from now? Addressing such questions means looking at science from a dynamic systems perspective. Two recent books about the social system of science, by Ziman and by Gibbons, Limoges, Nowotny, Schwartzman, Scott, and Trow, accept this challenge and argue that the research enterprise is changing. This article uses bibliometric data to examine the extent and nature of changes identified by these authors, taking as an example British research. We use their theoretical frameworks to investigate five characteristics of research said to be increasingly pervasive—namely, application, interdisciplinarity, networking, internationalization, and concentration of resources. Results indicate that research may be becoming more interdisciplinary and that research is increasingly conducted more in networks, both domestic and international; but the data are more ambiguous regarding application and concentration.

The self-similar science system1

Abstract A system with a self-similar property is scale-independent and statistically exhibits that property at all levels of observation. In addition, a power law describes the distribution of a scale-independent property. Many investigators have observed social activities and structures, particularly in the science system, that are best described by a power-law distribution. However, unlike classical physical power laws that are used in the design of complex technical systems, social power laws are not used to develop social policy. Using the science system as a model social system and peer-reviewed publications and citations to these papers as the data source we will demonstrate the existence of two power law distributions that are then used to predict the existence of two additional power laws. In fact, it will be shown that in four UK sectoral, six OECD national, a regional and the world science systems the Matthew effect can be described by a power-law relationship between publishing size (papers) and recognition (citations). The exponent of this power law is 1.27±0.03, it is constant over time and relatively independent of system size and nationality. The policy implications of these robust self-similar social properties as well as the need to develop scale-independent policy are discussed.

Scale-independent indicators and research evaluation

This paper demonstrates that some conventional indicators used in research evaluation may fail to account for the non-linearity between size of institution and performance. This can result in an over- or under-estimation of the research performance of both large and small institutions and nations. This paper shows that a power law relationship exists between recognition or impact and (a) publishing size of scientific communities within an OECD science system and (b) publishing size of a research community across OECD science systems or institutions in a science system. Also, a power law relationship exists between the amount of various types of collaboration and the publishing size of institutions. A new class of scale-independent indicators is developed to overcome the inequity produced by some non-linear characteristics commonly measured when evaluating research performance. Copyright , Beech Tree Publishing.

Scale-independent bibliometric indicators

Van Raan (this issue) makes an excellent case for using bibliometric data to measure some central aspects of scientific research and to construct indicators of groups: research groups, university departments, and institutes. He claims that, next to peer review, these indicators are indispensable for evaluating research and can be used in parallel with peer review processes. By way of an example, van Raan provides a table containing nine indicators for a German medical research institute. Two of these indicators—articles (P) and citations (C)—are established proxy measures for the size of a group and the impact of its published research (Katz & Hicks, 1997). The ratio between citations and articles (CPP) and the ratios between CPP and the mean Journal Citation Score and between the field-based world average and the Germany-specific world average—which are uniquely defined CPP reference values—are used to construct a set of indicators that van Raan suggests can be used to assess international research performance. This commentary focuses solely on the use of bibliometric indicators to compare international research performance. It addresses the fundamental question of whether CPP or measures like CPP can be used to accurately compare the performance of groups of different sizes.

Power laws and athletic performance.

In a previous study, we showed that the 1992 mens world record running times in the 100 m to 200 km could be represented accurately by the equation T = cDn, where T is the calculated record time for distance D, and c and n are positive constants. Here, we extend that to cover the years 1925-65 at 10-year intervals and 1970-95 in 5-year intervals for distances of 100 m to 10 km. Values of n for all years lie along a straight line with a small negative slope. A regression analysis yields an equation for values of n covering the period 1925-95. Values of c from 1925 to 1995 were fitted by a quadratic equation. These two equations define a surface in three-dimensional space (log(T), log(D), data) for all mens world record runs over the 70-year period for distances of 100 m to 10 km. We also demonstrated previously that event times, t, do not scatter randomly with respect to the values of T but form a consistent pattern about the straight lines in log(T) versus log(D) plots. In this study, we show that the pattern of (t-T)/t as a function of date has remained constant for the past 70 years.

Web indicators for complex innovation systems

The Web and innovation systems are interacting complex systems that impact each other. Web indicators used to inform decision-makers about the impact they are having on each must be reproducible and relevant. Web metrics research is young and best-practice methodologies for producing robust indicators are evolving. This study describes a methodology for producing robust indicators of the web presence of the European and Canadian innovation systems. It demonstrates that the emergent properties and scaling characteristics expected of complex systems are captured by these indicators. It illustrates how these indicators can be used to measure the amount of recognition a nation or provinces web presence receives from other nations and provinces in their innovation systems. Copyright , Beech Tree Publishing.

What Is a Complex Innovation System

Innovation systems are sometimes referred to as complex systems, something that is intuitively understood but poorly defined. A complex system dynamically evolves in non-linear ways giving it unique properties that distinguish it from other systems. In particular, a common signature of complex systems is scale-invariant emergent properties. A scale-invariant property can be identified because it is solely described by a power law function, f(x) = kxα, where the exponent, α, is a measure of scale-invariance. The focus of this paper is to describe and illustrate that innovation systems have properties of a complex adaptive system. In particular scale-invariant emergent properties indicative of their complex nature that can be quantified and used to inform public policy. The global research system is an example of an innovation system. Peer-reviewed publications containing knowledge are a characteristic output. Citations or references to these articles are an indirect measure of the impact the knowledge has on the research community. Peer-reviewed papers indexed in Scopus and in the Web of Science were used as data sources to produce measures of sizes and impact. These measures are used to illustrate how scale-invariant properties can be identified and quantified. It is demonstrated that the distribution of impact has a reasonable likelihood of being scale-invariant with scaling exponents that tended toward a value of less than 3.0 with the passage of time and decreasing group sizes. Scale-invariant correlations are shown between the evolution of impact and size with time and between field impact and sizes at points in time. The recursive or self-similar nature of scale-invariance suggests that any smaller innovation system within the global research system is likely to be complex with scale-invariant properties too.

The Scaling Relationship between Citation-Based Performance and Scientific Collaboration in Natural Sciences.

The aim of this paper is to extend our knowledge about the power‐law relationship between citation‐based performance and coauthorship patterns in papers in the natural sciences. We analyzed 829,924 articles that received 16,490,346 citations. The number of articles published through coauthorship accounts for 89%. The citation‐based performance and coauthorship patterns exhibit a power‐law correlation with a scaling exponent of 1.20 ± 0.07. Citations to a subfields research articles tended to increase 2.1.20 or 2.30 times each time it doubled the number of coauthored papers. The scaling exponent for the power‐law relationship for single‐authored papers was 0.85 ± 0.11. The citations to a subfields single‐authored research articles increased 2.0.85 or 1.89 times each time the research area doubled the number of single‐authored papers. The Matthew Effect is stronger for coauthored papers than for single‐authored. In fact, with a scaling exponent <1.0 the impact of single‐authored papers exhibits a cumulative disadvantage or inverse Matthew Effect.

The power-law relationship between citation-based performance and collaboration in articles in management journals: A scale-independent approach

The objective of this article is to determine if academic collaboration is associated with the citation‐based performance of articles that are published in management journals. We analyzed 127,812 articles published between 1988 and 2013 in 173 journals on the ISI Web of Science in the “management” category. Collaboration occurred in approximately 60% of all articles. A power–law relationship was found between citation‐based performance and journal size and collaboration patterns. The number of citations expected by collaborative articles increases 21.89 or 3.7 times when the number of collaborative articles published in a journal doubles. The number of citations expected by noncollaborative articles only increases 21.35 or 2.55 times if a journal publishes double the number of noncollaborative articles. The Matthew effect is stronger for collaborative than for noncollaborative articles. Scale‐independent indicators increase the confidence in the evaluation of the impact of the articles published in management journals.