nach oben

Empirical Economics

Erschienen in:

07.04.2022

Productivity spillovers in endogenous coauthor networks

verfasst von: Wei Cheng

Erschienen in: Empirical Economics | Ausgabe 6/2022

Einloggen

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config

KI-gestützte Suche

Aus

Abstract

This paper estimates the spillover effects of collaboration networks by quantifying the impact of collaborator quantity and quality on research productivity in cardiovascular medicine. Given the endogeneity of coauthor networks, we propose a two-stage approach for identifying spillover effects. The first stage analyzes the influencing factors in a scientist’s choice of collaborators. The second stage utilizes characteristics of the predicted collaborators from the first stage to construct instruments for key features of a scientist’s actual collaborator networks. Our results reveal that, after controlling for researcher fixed effects, the productivity of prior collaborators is no longer a significant determinant of a researcher’s own productivity, while the quantity of previous collaborators remains significant. Moreover, the magnitude of the spillover effects of possessing more collaborators varies by researcher characteristics.

Vorheriger Artikel The global propagation of the US–China trade war

Nächster Artikel Estimating empirical marginal adjustment cost function: a power series approach

Sie haben noch keine Lizenz? Dann Informieren Sie sich jetzt über unsere Produkte:

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

über 67.000 Bücher
über 340 Zeitschriften

aus folgenden Fachgebieten:

Bauwesen + Immobilien
Business IT + Informatik
Finance + Banking
Management + Führung
Marketing + Vertrieb
Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

Jetzt informieren

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

über 102.000 Bücher
über 537 Zeitschriften

aus folgenden Fachgebieten:

Automobil + Motoren
Bauwesen + Immobilien
Business IT + Informatik
Elektrotechnik + Elektronik
Energie + Nachhaltigkeit
Finance + Banking
Management + Führung
Marketing + Vertrieb
Maschinenbau + Werkstoffe
Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

Jetzt informieren

Nur mit Berechtigung zugänglich

Cardiovascular diseases (CVD) cause more than half of all deaths across the European Region. CVD caused 31 percent of all global deaths. Data from World Health Organization.

More than 90% of all cardiovascular journal articles have multiple authors.

In this study, we examine spillovers from previous coauthors. This is different from Waldinger (2012), which investigates spillovers from colleagues in the same department/institution and finds no evidence for spillover effects. Although overlapped, coauthors and colleagues are two different groups of peers a scientist interacts with. Cheng and Weber (2021) provides an analysis considering spillovers from both groups of peers.

Some examples include Horizon Europe, which is European Union’s key funding program for research and innovation, and National Institutes of Health, which is the primary agency of the United States responsible for biomedical and public health research.

There is a large and emerging literature on network formation such as Jackson and Wolinsky (1996); Fafchamps and Gubert (2007); Calvó-Armengol and Ílkılıç (2009); Hsieh and Lee (2016); Goldsmith-Pinkham and Imbens (2013) and Leung (2015).

A zero element on the diagonal of \(\Omega \) would suggest that the effect of a certain attribute of the potential collaborator on the probability of collaboration are homogeneous.

General complementarities, especially the potential complementarity between features of collaborators and other research inputs, are not easily accommodated in our analysis and would be interesting future research questions.

\(\Theta ^{(t)}\), is a column vector of individual fixed effects for researchers at time t. As we have an unbalanced panel, the dimension of \(\Theta ^{(t)}\) varies with t. Gender and ethnicity are dropped from the regressions when individual fixed effects are controlled for.

\(\Psi ^{(t)}\) is a column vector of field fixed effects for researchers at time t. Its dimension of \(\Psi ^{(t)}\) varies with t.

If anything, the mechanical issue would make our estimates on the effect of collaborator quality susceptible to overestimation. But our findings in later sections reveal that the quantity matters more than the quality of collaborators.

Wooldridge (2002) shows that the usual 2SLS standard errors and test statistics are asymptotically valid if the parameters are consistently estimated in constructing the instruments (pp. 116-117).

In other words, the exclusion variable is whether a pair of scientists are of distance 2 at \(t-2\). The network distance at \(t-2\) affects their probability of collaboration at \(t-1\) and then partly determines the predicted prior collaborators at t, which form our main regressors.

MEDLINE contains over 30 million citations dating back to the ninetieth century. Over 5,600 titles are currently indexed in MEDLINE. Journals are selected based on their scientific merits. Both English journals and journals in other languages are included. More details are found at https://www.nlm.nih.gov/pubs/factsheets/jsel.html

The Medical Subject Headings Section staff are subject specialists responsible for areas of the health sciences wherein they have knowledge and expertise. Professionals in various disciplines are also consulted regarding broad organizational changes. Close coordination is maintained with various specialized vocabularies. Another useful feature of MeSH terms is that a paper’s major topics are tagged to be distinguished from the less substantive terms. This information is particularly valuable when measuring a scientist’s research interests. More details can be found at http://www.nlm.nih.gov/pubs/factsheets/mesh.html.

For example, “James” as “Jim,” “Joseph” as “Joe,” with and without middle name initial, etc.

Author-ity disambiguates author names by pairwise comparisons on an extensive set of features including shared language, MeSH terms, affiliations, etc.

Data available at https://jcr.incites.thomsonreuters.com/JCRJournalHomeAction.action.

A full list contains Nature, Science, The New England Journal of Medicine, The Lancet, Cell, Proceedings of the National Academy of Sciences, Journal of Clinical Oncology, JAMA, Nature Genetics, PLoS ONE, Nature Medicine, Blood, The Lancet Oncology, British Medical Journal, The Cochrane Database of Systematic Reviews, and The Journal of Clinical Investigation. An anonymous reviewer kindly pointed out that another strategy of identifying cardiovascular research articles would be extracting articles with relevant MeSH terms. However, that procedure would require a well-defined set of MeSH terms related to cardiovascular research.

We limit the publications to research papers, excluding book review, letters, etc.

Considering that there could be lumping errors in the author disambiguation algorithm, we delete outliers in the top 0.5% in the distribution of number of publications. In the online appendix, we also repeat the estimation using the full sample of researchers and the sample excluding top 1% researchers. The results are rarely changed.

There is now an updated version of Author-ity that covers more recent years.

The calculation of first publishing year and last publishing year is based on data of all available years instead of the time window we focus on.

Ethnea maps people into 26 ethnicity groups. They are English, Hispanic, Chinese, German, Japanese, French, Italian, Slav, Indian, Arab, Korean, Vietnamese, Nordic, Dutch, Turkish, Israel, Greek, African, Hungarian, Thai, Romanian, Baltic, Indonesian, Caribbean, Mongolian, and Polynesian.

It is worthwhile to note that, in biomedical fields, there is a convention of giving more credit to the first and last authors. However, it is hard to determine the exact shares of credit allocated to each of the authors on a paper. Our productivity measure equally assigns the credit; therefore, for researchers who are disproportionately first or last authors, this would underestimate their true productivity.

For example, if two of an individual’s collaborators co-publish a paper, this paper will be counted twice if using normal counts to represent the productivity of each of his or her two coauthors.

We add one to all journal impact factors to differentiate non-publishing from publishing in a journal with zero impact factor. This is equivalent to giving each article an additional citation. It will not alter the ranking of impact factors of articles, nor the ranking of researcher productivities. However, as kindly pointed out by the anonymous reviewer, it would affect the number in terms of percentage changes. For example, after adding one, a journal with an actual score of 1 is worth twice a journal with a score of 0, while a journal with a score of 2 is worth 50% more than a journal with a score of 1. In order to investigate how this affects our estimation, we tried to construct our productivity measure with and without adding one. It turned out that the estimated effects in percentage terms are quite close.

For collaborative research, some of the authors on an article might focus on work relating to some of the MeSH terms and the others work on the other terms.

Note that this is not implemented on list of unique terms but rather all occurrences of any terms. Therefore, one term can show up multiple times if a researcher publishes more than one papers on it.

The average number of terms a researcher publishes on is usually much higher in real data. For illustration purposes, I formulate this simple example.

We conducted several interviews with renowned cardiovascular scientists and inquire their opinions on the measure of research interest. We obtained very favorable feedback.

It is noteworthy that breadth of a scientist’s research area is not mechanically correlated with productivity. Scientists with more publications do not necessarily have broader research interests. It depends on the relative frequencies. For example, in Table 1, though Scientist B works in more subfields of cardiovascular science than Scientist A, Scientist A may still publish more papers than Scientist B. For example, based on Table 1, Scientist A might publish 9 papers, and Scientist B might publish 5 papers. Note that one paper has more than one Mesh terms.

Note that researchers who have previously co-published do not necessarily have close intellectual proximity. The MeSH terms from the co-published paper are not necessarily their common research interest. Whether the terms from the coauthored articles show up in both or either scientist’s set of research interests depends on the relative frequencies of these terms compared to other terms that the scientists have published on. If one or both scientists publish more papers on other keywords, the MeSH terms on this coauthored paper will not be their common research interests.

Note that this number refers to the number of cardiovascular scientist collaborators, which is only a subset of all of an individual’s collaborators. Cardiovascular scientists frequently work with researchers outside cardiovascular fields such as computer scientists and statisticians.

The organization of Mesh terms is not an exact tree structure as one child term may have multiple parent terms.

Unfortunately, to the best of our knowledge, there is no established test of weak instrument suitable for our analysis. To assess the power of our instrument, we compare the predictions of the two models with and without the distance-2 dummy variable. If this variable is omitted, only less than 1.5% distance-2 pairs yield correct predictions. However, the rate of correct predictions rises to 14% if the distance-2 dummy variable is included.

The estimates in Column (1) of Table 4 imply that the maximum of the inverted-U shape is realized at 26, which is within the support of the variable Common research interests (parent) according to Table 3.

Anderson and Richards-Shubik (2021) estimates, at article level, how the number of authors affect the paper’s impact factor. Ductor (2015) uses percent of coauthored papers as the measure of coauthorship. Konig et al. (2020)’s estimate of the spillover parameter is on the interaction between participants’ efforts allocated in the project.

This may be because of the smaller sample size of female cardiovascular scientists so that the estimates are not precisely estimated.

Agrawal A, Goldfarb A (2008) Restructuring research: communication costs and the democratization of university innovation. Am Econ Rev 98:1578–90CrossRef

Agrawal A, Goldfarb A, Teodoridis F (2016) Understanding the changing structure of scientific inquiry. Am Econ J Appl Econ 8:100–128CrossRef

Anderson KA, Richards-Shubik S (2021) collaborative production in science: an empirical analysis of coauthorships in economics. Rev Econ Stat, 1–45

Azoulay P, Graff Zivin J, Wang J (2010) Superstar extinction. Quarterly J Econ 25:549–589CrossRef

Azoulay P, Graff Zivin JS, Li D, Sampat BN (2018) Public R&D Investments and Private-sector Patenting: Evidence from NIH Funding Rules. Rev Econ Studies 86:117–152CrossRef

Beaver D, Rosen R (1979) Studies in scientific collaboration: Part II. Scientific co-authorship, research productivity and visibility in the French scientific elite, 1799–1830. Scientometrics 1:133–149CrossRef

Becker GS, Murphy KM (1992) The division of labor, coordination costs, and knowledge. Quarterly J Econ 107:1137–1160CrossRef

Bikard M, Murray F, Gans JS (2015) Exploring trade-offs in the organization of scientific work: collaboration and scientific reward. Manag Sci 61:1473–1495CrossRef

Borjas GJ, Doran KB (2012) The collapse of the soviet union and the productivity of American mathematicians. Quarterly J Econ 127:1143–1203CrossRef

Bramoullé Y, Djebbari H, Fortin B (2009) Identification of peer effects through social networks. J Econ 150:41–55CrossRef

Calvó-Armengol A, Ílkılıç R (2009) Pairwise-stability and Nash equilibria in network formation. Int J Game Theory 38:51–79CrossRef

Card D, DellaVigna S, Funk P, Iriberri N (2019) Are referees and editors in economics gender neutral? Quarterly J Econ 135:269–327CrossRef

Cheng W, Weber G (2021) Productivity spillovers in two overlapping networks. Spatial Econ Anal, 1–27

Chung KH, Cox RAK, Kim KA (2009) On the relation between intellectual collaboration and intellectual output: evidence from the finance academe. Quarterly Rev Econ Finance 49:893–916CrossRef

Cosslett SR (1981) Efficient Estimation of Discrete-Choice Models. In: Manski CF, and Daniel McFadden (ed.) Structural Analysis of Discrete Data with Econometric Applications. MIT Press, pp. 51–111

Dessein W, Santos T (2006) Adaptive Organizations. J Political Econ 114:956–995CrossRef

Ductor L (2015) Does Co-authorship Lead to Higher Academic Productivity? Oxford Bull Econ Stat 77:385–407CrossRef

Ductor L, Fafchamps M, Goyal S, van der Leij MJ (2014) Social networks and research output. Rev Econ Stat 96:936–948CrossRef

Eisenberg L, Pellmar TC (2000) Bridging disciplines in the brain, behavioral, and clinical sciences. National Academies Press (US). ISBN 0309183677

Fafchamps M, Gubert F (2007) Risk sharing and network formation. Am Econ Rev, 75–79

Fafchamps M, Leij MJ, Goyal S (2010) Matching and network effects. J Euro Econ Assoc 8:203–231CrossRef

Freeman RB, Ganguli I, Murciano-Goroff R (2014) Why and Wherefore of Increased Scientific Collaboration. In: The Changing Frontier: Rethinking Science and Innovation Policy. University of Chicago Press, pp. 17–48

Freeman RB, Huang W (2015) Collaborating with People Like Me: Ethnic Co-authorship within the US. J Labor Econ 33(3):S289–S318CrossRef

Goldsmith-Pinkham P, Imbens G (2013) Social networks and the identification of peer effects. J Bus Econ Stat 31(3):253–264CrossRef

Gonzalez-Brambila C, Veloso FM (2007) The determinants of research output and impact: a study of Mexican researchers. Res Policy 36:1035–1051CrossRef

Haeussler C, Sauermann H (2020) Division of labor in collaborative knowledge production: the role of team size and interdisciplinarity. Res Policy 49:103987CrossRef

He Z-L, Geng X-S, Campbell-Hunt C (2009) Research collaboration and research output: a longitudinal study of 65 biomedical scientists in a New Zealand university. Res Policy 38:306–317CrossRef

Hollis A (2001) Co-authorship and the output of academic economists. Labour Econ 8:503–530CrossRef

Hsieh C-S, Lee L-F (2014) A social interactions model with endogenous friendship formation and selectivity. J Appl Econ 31:301–319CrossRef

Jackson MO, Wolinsky A (1996) A strategic model of social and economic networks. J Econ Theory 71:44–74CrossRef

Jones BF (2009) the burden of knowledge and the death of the Renaissance Man: Is Innovation Getting Harder? Rev Econ Stud 76:283–317CrossRef

Kelejian HH, Piras G (2014) Estimation of spatial models with endogenous weighting matrices, and an application to a demand model for cigarettes. Reg Scid Urban Econ 46:140–149CrossRef

Kelejian HH, Prucha IR (1998) A generalized spatial two-stage least squares procedure for estimating a spatial autoregressive model with autoregressive disturbances. J Real Estate Finance Econ 17:99–121CrossRef

Konig M, Liu X, Hsieh C-S, Zimmermann C (2020) Collaboration in Bipartite Networks, with an Application to Coauthorship Networks. Centre for Economic Policy Research (CEPR) Discussion Paper

Leahey E, Beckman CM, Stanko TL (2016) Prominent but Less Productive: the Impact of Interdisciplinarity on Scientists? Research*. Adm Sci Quarterly 62:105–139CrossRef

Lee LF (2003) Best spatial two? Stage least squares estimators for a spatial autoregressive model with autoregressive disturbances. Econ Rev 22:307–335CrossRef

Lee LF, Liu X, Lin X (2010) Specification and estimation of social interaction models with network structures. Econ J 13:145–176

Lee S, Bozeman B (2005) The impact of research collaboration on scientific productivity. Social Stud Sci 35:673–702CrossRef

Leung MP (2015) Two-step estimation of network-formation models with incomplete information. J Econ 188:182–195CrossRef

Lin X (2010) Identifying peer effects in student academic achievement by spatial autoregressive models with group unobservables. J Labor Econ 28:825–860CrossRef

Lyubarova R, Itagaki BK, Itagaki MW (2009) The impact of National Institutes of Health funding on US cardiovascular disease research. PLoS One 4:e6425CrossRef

Manski CF (1993) Identification of endogenous social effects: the reflection problem. Rev Econ Stud 60:531–542CrossRef

Mohnen M (2021) Stars and Brokers: Knowledge Spillovers Among Medical Scientists. Manag Sci

Oettl A (2012) Reconceptualizing Stars: Scientist Helpfulness and Peer Performance. Manag Sci 58:1122–1140CrossRef

Rhoten D, Parker A (2004) Risks and rewards of an interdisciplinary research path. Science 306:2046CrossRef

Romer PM (1990) Endogenous technological change. J Political Econ 98:S71–S102CrossRef

Sarsons H, Gërxhani K, Reuben E, Schram A (2021) Gender differences in recognition for group work. J Political Econ 129:101–147CrossRef

Singh J, Fleming L (2010) Lone inventors as sources of breakthroughs: Myth or reality? Manag Sci 56:41–56CrossRef

Smith BN, Singh M, Torvik VI (2013) A search engine approach to estimating temporal changes in gender orientation of first names. In: Proceedings of the 13th ACM/IEEE-CS joint conference on Digital libraries, pp. 199–208. ACM

Torvik VI, Agarwal S (2016) Ethnea-an instance-based ethnicity classifier based on geo-coded author names in a large-scale bibliographic database. International Symposium on Science of Science, Library of Congress, Washington DC, USA

Torvik VI, Smalheiser NR (2009) Author name disambiguation in MEDLINE. ACM Trans Knowledge Discovery from Data (TKDD) 3:11CrossRef

Waldinger F (2012) Peer effects in science: evidence from the dismissal of scientists in Nazi Germany. Rev Econ Stud 79:838–861CrossRef

Wooldridge JM (2002) Econometric analysis of cross section and panel data. MIT Press, Cambridge

Wuchty S, Jones BF, Uzzi B (2007) The increasing dominance of teams in production of knowledge. Science 316:1036–1039CrossRef

Titel: Productivity spillovers in endogenous coauthor networks
verfasst von: Wei Cheng
Publikationsdatum: 07.04.2022
Verlag: Springer Berlin Heidelberg
Erschienen in: Empirical Economics / Ausgabe 6/2022
Print ISSN: 0377-7332
Elektronische ISSN: 1435-8921
DOI: https://doi.org/10.1007/s00181-022-02235-3

Springer Professional

Abstract

Bitte loggen Sie sich ein, um Zugang zu Ihrer Lizenz zu erhalten.

Sie haben noch keine Lizenz? Dann Informieren Sie sich jetzt über unsere Produkte:

Springer Professional "Wirtschaft"

Springer Professional "Wirtschaft+Technik"

Weitere Artikel der Ausgabe 6/2022

Unconventional monetary policies and expectations on economic variables

Correction to: Heterogeneous decision-making and market power: an application to Eurozone banks

GDP effects of pandemics: a historical perspective

The effect of health on economic growth: a meta-regression analysis

Heterogeneous decision-making and market power: an application to Eurozone banks

Monetary policy and the racial wage gap

Premium Partner