Universal productivity patterns in research careers

Editors' Suggestion
Open Access

Universal productivity patterns in research careers

Andre S. Sunahara, Matjaž Perc, and Haroldo V. Ribeiro

Phys. Rev. Research 5, 043203 – Published 5 December 2023

Abstract

A common expectation is that career productivity peaks rather early and then gradually declines with seniority. But whether this holds true is still an open question. Here we investigate the productivity trajectories of almost 8500 scientists from over 50 disciplines using methods from time-series analysis, dimensionality reduction, and network science, showing that there exist six universal productivity patterns in research. Based on clusters of productivity trajectories and network representations where researchers with similar productivity patterns are connected, we identify constant, u-shaped, decreasing, periodic-like, increasing, and canonical productivity patterns, with the latter two describing almost three-fourths of researchers. In fact, we find that canonical curves are the most prevalent, but contrary to expectations, productivity peaks occur much more frequently around midcareer rather than early. These results outline the boundaries of possible career paths in science and caution against the adoption of stereotypes in tenure and funding decisions.

Received 27 June 2023
Accepted 5 November 2023

DOI:https://doi.org/10.1103/PhysRevResearch.5.043203

Published by the American Physical Society under the terms of the Creative Commons Attribution 4.0 International license. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.

Published by the American Physical Society

Physics Subject Headings (PhySH)

Complex systems Network structure Physics & society

Interdisciplinary PhysicsNetworks

Authors & Affiliations

Andre S. Sunahara ¹, Matjaž Perc ^2,3,4,5,6, and Haroldo V. Ribeiro ^1,*

¹Departamento de Física, Universidade Estadual de Maringá–Maringá, PR 87020-900, Brazil
²Faculty of Natural Sciences and Mathematics, University of Maribor, Koroška cesta 160, 2000 Maribor, Slovenia
³Department of Medical Research, China Medical University Hospital, China Medical University, Taichung, Taiwan
⁴Alma Mater Europaea, Slovenska ulica 17, 2000 Maribor, Slovenia
⁵Complexity Science Hub Vienna, Josefstädterstraße 39, 1080 Vienna, Austria
⁶Department of Physics, Kyung Hee University, 26 Kyungheedae-ro, Dongdaemun-gu, Seoul, Republic of Korea

^*hvr@dfi.uem.br

Article Text

Click to Expand

Supplemental Material

Click to Expand

References

Click to Expand

Issue

Vol. 5, Iss. 4 — December - December 2023

Subject Areas

Reuse & Permissions

Author publication services for translation and copyediting assistance advertisement

Images

Figure 1
Clustering patterns of researchers' productivity curves. The central panel displays a network representation, where each node represents a researcher and weighted edges connect those with similar productivity trajectories. Ten distinct communities, represented by different colors and labeled 1 to 10, are identified and correspond to groups of researchers with similar productivity patterns. The surrounding panels display the productivity curves of researchers in each community, with the black curves representing the average behavior of each cluster. The lengths of researchers' careers in each group are scaled to the unit interval and the numbers and fractions of researchers in each group are shown within each panel. The ten clusters are further grouped into six categories: constant (cluster 1), u-shaped (cluster 2), decreasing (cluster 3), periodic-like (cluster 4), increasing (clusters 5 and 6), and canonical-like (clusters 7 to 10) curves. Increasing and canonical-like patterns describe almost three-fourths of the researchers in our study, while periodic-like curves are the least common. Clusters and nodes that are close together share similar productivity patterns (see Ref. [43] for an interactive visualization).
Reuse & Permissions
Figure 2
Career length and cohort effects on the prevalence of productivity patterns. (a) Probability distributions of career lengths for each of the ten clusters of productivity trajectories, as determined by kernel density estimation. All clusters encompass a broad range of career lengths, but these distributions are more localized in distinct positions (Table S1 in the Supplemental Material [32]). (b) Prevalence of productivity patterns across four categories of career length: 10–14 years, 15–19 years, 20–24 years, and greater than 24 years. The dominant pattern among researchers with shorter careers, which also correspond to younger scholars, is the increasing productivity curve. This pattern becomes less prevalent among researchers with longer careers, which corresponds to more experienced scholars. Canonical-like trajectories exhibit the opposite behavior and are significantly more prevalent among senior researchers. Periodic-like curves are also more common among researchers with long careers, while constant, u-shaped and decreasing trajectories occur more among young researchers. (c) Comparison of the prevalence of productivity patterns in the initial career years of senior researchers with those exhibited in later career stages. The left bars show the fractions of each productivity pattern obtained when considering the initial 14 career years of researchers with careers longer than 24 years, and the right ones show the prevalence of patterns when considering the full range of their careers. The connections between the left and right bars indicate the migration flow among the productivity patterns. Almost half of canonical senior careers are classified as increasing curves in their beginnings; however, only 9% of senior researchers who exhibit early career increasing productivity sustain this pattern with career progression.
Reuse & Permissions

Physical Review Research