pith. sign in

arxiv: 2404.16835 · v3 · submitted 2024-01-19 · 💻 cs.DL

Quantifying Lifetime Productivity Changes: A Longitudinal Study of 320,000 Late-Career Scientists

Marek Kwiek , Lukasz Szymula This is my paper

Pith reviewed 2026-05-24 04:55 UTC · model grok-4.3

classification 💻 cs.DL
keywords research productivitycareer mobilitylongitudinal studyproductivity persistencepath dependencebibliometricsscientist careersScopus data
0
0 comments X

The pith

Half of global top-performing scientists remain top performers across their entire careers.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper follows 320,000 scientists with 25 or more years of publishing in 16 disciplines across 38 countries to map how their output ranks change from early to mid to late career. It reports that scientists largely remain in the same productivity decile over decades, with half of those in the top 10 percent and one-third in the bottom 10 percent staying put. Moves from bottom to top or top to bottom occur in 1 percent or fewer cases. Regression results single out earlier top-decile status as the strongest predictor of later top status. The pattern points to strong persistence in individual research output over long periods.

Core claim

The global science system is highly immobile: half of global top performers continue their careers as top performers and one-third of global bottom performers as bottom performers. Jumpers-Up and Droppers-Down are extremely rare in science. The chances of moving radically up or down in productivity classes are marginal (1% or less). Our regression analyses show that productivity classes are highly path dependent: there is a single most important predictor of being a top performer, which is being a top performer at an earlier career stage.

What carries the argument

Decile-based productivity classes built from raw Scopus publication counts, followed longitudinally across early-, mid-, and late-career stages for scientists with at least 25 years of output.

If this is right

  • Prior membership in the top productivity decile is the dominant predictor of remaining in that decile decades later.
  • Bottom-decile scientists show comparable persistence, with roughly one-third staying in the lowest class.
  • Radical rank changes between productivity classes occur at rates of 1 percent or lower across the sample.
  • Productivity class at any stage strongly forecasts class at later stages in regression models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Sustained high output appears tied to early establishment rather than later acceleration.
  • Policies that allocate resources based on early output may compound existing rank differences over time.
  • Examining what enables the small fraction of movers to change classes could reveal mechanisms that the aggregate persistence masks.

Load-bearing premise

Decile classes from raw publication counts give a stable, comparable measure of output across disciplines, countries, and career stages without major coverage or normalization biases.

What would settle it

A follow-up analysis of the same scientists using field-normalized impact measures or an independent database that finds more than a few percent moving from bottom to top decile or top to bottom decile between career stages.

Figures

Figures reproduced from arXiv: 2404.16835 by Lukasz Szymula, Marek Kwiek.

Figure 1
Figure 1. Figure 1: Flowchart and major steps in data preprocessing: from all scientists in the Scopus database to late-career scientists in our sample. At each career stage, current late-career scientists showed their annual individual productivity. Consequently, their productivity was calculated for the recent five-year period and for two earlier periods: when they were early-career scientists and mid-career scientists. Our… view at source ↗
Figure 2
Figure 2. Figure 2: Distribution of academic age: Kernel density plots. Late-career scientists, all academic disciplines combined (top panel) by gender. Late-career scientists by academic discipline (bottom panel) and gender (N = 320,564) [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (Left panel) How were the top-performing (N=32,063, productivity decile 10) mid-career scientists distributed by productivity percentiles (range: 0–100) when they were in their early-career stage? (Right panel) How were the bottom-performing (N=32,063, productivity decile 1) mid￾career scientists distributed by productivity percentiles (range: 0–100) when they were in their early-career stage? Kernel densi… view at source ↗
Figure 4
Figure 4. Figure 4: (Left panel) How were the top-performing (N=32,063, productivity decile 10) late-career scientists distributed by the productivity percentiles (range: 0–100) when they were in their mid-career stage? (Right panel) How were the bottom-performing (N=32,075, productivity decile 1) late-career scientists distributed by productivity percentiles (range: 0–100) when they were in their mid-career stage? Kernel den… view at source ↗
Figure 5
Figure 5. Figure 5: Scientists’ mobility between productivity classes in the three stages of a scientific career. All academic disciplines combined (TOTAL), current late-career scientists. All observations ranked and clustered into productivity deciles, top- (upper 10%, productivity decile 10) and bottom- (bottom 10%, productivity decile 1) productivity classes only (N = 320,564) (percentages, top class, and bottom class 100%… view at source ↗
Figure 6
Figure 6. Figure 6: Scientists’ mobility between productivity classes in the three stages of a scientific career. All social science disciplines combined (SOCIAL), current late-career scientists. All observations ranked and clustered into productivity deciles, top- (upper 10%, productivity decile 10) and bottom- (bottom [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Scientists’ mobility between productivity classes in the three stages of a scientific career. All STEMM academic disciplines combined, current late-career scientists. All observations ranked and clustered into productivity deciles, top- (upper 10%, productivity decile 10) and bottom- (bottom 10%, productivity decile 1) productivity classes only (N = 307,979) (percentages, top class and bottom class, 100% e… view at source ↗
read the original abstract

The present study focuses on persistence in research productivity over the course of an individual's entire scientific career. We track 'late-career' scientists - scientists with at least 25 years of publishing experience (N=320,564) - in 16 STEMM (science, technology, engineering, mathematics, and medicine) and social science disciplines from 38 OECD countries for up to five decades. Our OECD sample includes 79.42% of late-career scientists globally. We examine the details of their mobility patterns as early-career, mid-career, and late-career scientists between decile-based productivity classes, from the bottom 10% to top 10% of the productivity distribution. Methodologically, we turn a large-scale bibliometric dataset (Scopus raw data) into a comprehensive, longitudinal data source for research on careers in science. The global science system is highly immobile: half of global top performers continue their careers as top performers and one-third of global bottom performers as bottom performers. Jumpers-Up and Droppers-Down are extremely rare in science. The chances of moving radically up or down in productivity classes are marginal (1% or less). Our regression analyses show that productivity classes are highly path dependent: there is a single most important predictor of being a top performer, which is being a top performer at an earlier career stage.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript analyzes persistence in research productivity over entire careers using a sample of 320,564 late-career scientists with at least 25 years of experience from Scopus data across 16 disciplines and 38 countries. It reports high immobility in decile-based productivity classes, with approximately 50% of top performers remaining in the top decile and 33% of bottom performers staying in the bottom, and very low rates of radical mobility (1% or less). Regression analyses indicate strong path dependence, with prior top performer status being the strongest predictor of later top performance.

Significance. If the mobility estimates prove robust, the large-scale longitudinal evidence would substantially advance understanding of cumulative advantage and path dependence in scientific careers, with direct relevance to evaluation policies and funding allocation.

major comments (2)
  1. [Abstract; data construction] Abstract and data/methods sections: Productivity classes are constructed from decile bins on raw (unnormalized) publication counts pooled across 16 disciplines. Without field-specific or career-stage normalization (e.g., per-discipline deciles or fractional counting), disciplines with higher average output rates will disproportionately occupy top deciles, so the reported transition probabilities (50% top stay top; <1% radical jumps) may partly capture field norms rather than individual changes. This directly affects the central immobility and path-dependence claims.
  2. [Regression analyses] Regression results section: The models treat the same unnormalized decile classes as both predictors and outcomes. Absent robustness checks using alternative metrics (field-year normalized counts, h-index, or citation-based measures), the finding that 'being a top performer at an earlier career stage' is the single most important predictor risks being an artifact of the measurement choice.
minor comments (2)
  1. [Methods] Clarify the exact productivity metric (total publications? fractional? including or excluding certain document types) and any handling of co-authorship in the methods section.
  2. [Results tables] Report cell counts, standard errors, and any sensitivity to the 25-year threshold in the transition matrices and regression tables.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which raise important methodological considerations regarding our measurement of productivity. We address each major comment in turn and indicate the revisions we will undertake.

read point-by-point responses

  1. Referee: [Abstract; data construction] Abstract and data/methods sections: Productivity classes are constructed from decile bins on raw (unnormalized) publication counts pooled across 16 disciplines. Without field-specific or career-stage normalization (e.g., per-discipline deciles or fractional counting), disciplines with higher average output rates will disproportionately occupy top deciles, so the reported transition probabilities (50% top stay top; <1% radical jumps) may partly capture field norms rather than individual changes. This directly affects the central immobility and path-dependence claims.

    Authors: We acknowledge that the use of raw, unnormalized publication counts pooled across disciplines means that field-specific differences in average output rates could influence decile membership. Our primary aim was to document absolute productivity persistence at the level of the global science system rather than relative standing within fields. To address the concern directly, we will add a set of robustness analyses that recompute deciles within each discipline-year and report the resulting transition matrices and regression results. We will also include a discussion of how the choice of raw versus normalized measures affects interpretation of immobility. revision: yes

  2. Referee: [Regression analyses] Regression results section: The models treat the same unnormalized decile classes as both predictors and outcomes. Absent robustness checks using alternative metrics (field-year normalized counts, h-index, or citation-based measures), the finding that 'being a top performer at an earlier career stage' is the single most important predictor risks being an artifact of the measurement choice.

    Authors: The regressions rely on the same raw decile classification for both the lagged predictor and the outcome. We will expand the regression section with supplementary models that substitute field-year normalized counts, as well as available citation-based and h-index metrics, to test whether the dominance of prior top-performer status persists under alternative operationalizations. These checks will be reported in a new appendix with a summary in the main text. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical transitions and regressions on external bibliographic data

full rationale

The paper defines decile productivity classes directly from raw Scopus publication counts for the sampled scientists, then reports observed transition frequencies across career stages and regression coefficients with prior class as a predictor. These quantities are computed outputs from the external dataset rather than quantities defined in terms of themselves or forced by author-fitted parameters. No self-citation chain is invoked to establish uniqueness or to substitute for the mobility statistics. The analysis remains self-contained against the Scopus source without reducing any central claim to a definitional loop.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central empirical claims rest on the unstated assumption that raw publication counts (or an equivalent metric) can be ranked into stable deciles without field-specific normalization and that the Scopus coverage for the selected disciplines and countries is sufficiently complete and unbiased to support the reported immobility rates.

free parameters (2)
  • productivity metric
    The abstract does not specify whether productivity is measured by publication count, fractional count, citations, or another quantity; this choice determines the decile boundaries and mobility statistics.
  • 25-year threshold
    The cutoff defining 'late-career' is chosen by the authors and directly shapes the sample and the observed path dependence.
axioms (2)
  • domain assumption Scopus provides sufficiently complete and unbiased coverage of publications by late-career scientists in the 16 disciplines across 38 OECD countries.
    The 79.42% global coverage claim and all mobility percentages depend on this.
  • domain assumption Decile ranks constructed at different career stages are comparable without additional normalization for discipline, collaboration size, or career length.
    This assumption is required for the reported 1% jump/drop rates and the regression predictor ranking.

pith-pipeline@v0.9.0 · 5775 in / 1613 out tokens · 21559 ms · 2026-05-24T04:55:15.082807+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

9 extracted references · 9 canonical work pages · 1 internal anchor

  1. [1]

    A., & Caprasecca, A

    Abramo, G., D’Angelo, C. A., & Caprasecca, A. (2009). The contribution of star scientists to overall sex differences in research productivity. Scientometrics, 81(1), 137–156. Abramo, G., D’Angelo, C. A., & Soldatenkova, A. (2017). An investigation on the skewness patterns and fractal nature of research productivity distributions at field and discipline le...

    work page doi:10.1162/qss_a_00019 2009

  2. [2]

    Dusdal, J., & Powell, J. J. W. (2021). Benefits, motivations, and challenges of international collaborative research: A sociology of science case study. Science and Public Policy, 48(1), 235–245. Fox, M. F. (1983). Publication productivity among scientists: A critical review. Social Studies of Science, 13(2), 285–305. Fox, M. F., & Mohapatra, S. (2007). S...

    work page 2021

  3. [3]

    J., & Moktan, S

    Heckman, J. J., & Moktan, S. (2018). Publishing and promotion in economics. The tyranny of the Top Five. NBER Working Paper 25093. Hermanowicz, J. (2012). The sociology of academic careers: Problems and prospects. In J. C. Smart & M. B. Paulsen (Eds.), Higher education: Handbook of theory and research (pp. 207–248). Springer. Horta, H., & Santos, J. M. (2...

    work page 2018

  4. [4]

    Kwiek, M. (2016). The European research elite: A cross-national study of highly productive academics across 11 European systems. Higher Education, 71(3), 379–397 Kwiek, M. (2018). High research productivity in vertically undifferentiated higher education systems: Who are the top performers? Scientometrics, 115(1), 415–462. Kwiek, M. (2021). What large-sca...

    work page 2016

  5. [5]

    Kwiek, M., & Roszka, W. (2024b). Are scientists changing their research productivity classes when they move up the academic ladder? Innovative Higher Education, Online first, 1–40. https://doi.org/10.1007/s10755-024-09735-3. Kwiek, M., & Szymula, Ł. (2024). Quantifying attrition in science: A cohort-based, longitudinal study of scientists in 38 OECD count...

    work page doi:10.1007/s10755-024-09735-3 2024

  6. [6]

    35 Lindahl, J. (2018). Predicting research excellence at the individual level: The importance of publication rate, top journal publications, and top 10% publications in the case of early career mathematicians. Journal of Informetrics, 12(2), 518–533. Liu, L., Jones, B.F., Uzzi, B., & Wang, D.. (2023). Data, measurement and empirical methods in the science...

    work page doi:10.1080/03075079.2021.1942822 2018

  7. [7]

    Preston, A. E. (2004). Leaving science. Occupational exit from scientific careers. Russell Sage Foundation. Priem, J., Piwowar, H., & Orr, R. (2022). OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts. ArXiv. https://arxiv.org/abs/2205.01833 Rørstad, K., & Aksnes, D. W. (2015). Publication rate expressed by age, g...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1371/journal.pone.0294562 2004

  8. [8]

    F., Morgan, A

    Way, S. F., Morgan, A. C., Clauset, A., & Larremore, D. B. (2017). The misleading narrative of the canonical faculty productivity trajectory. Proceedings of the National Academy of Sciences, 114(44), E9216–E9223. 10.1073/pnas.1702121114 Weingart, P. (2004). Impact of bibliometrics upon the science system: inadvertent consequences? In H. F. Moed, W. Glänze...

    work page doi:10.1073/pnas.1702121114 2017

  9. [9]

    37 5.54 6.83 8.58 11.11 15.72 89.39 STEMM 0.00 1.06 2.34 3.43 4.54 5.78 7.36 9.20 12.18 17.81 221.27 TOTAL 0.00 1.06 2.34 3.29 4.37 5.54 6.83 8.58 11.11 15.72 89.39 40 Journal prestige–normalized (exponential) approach to individual research productivity If a linear function was used, the value of an article that is published in the journal in the 99th jo...

    work page 2023