Source-linked AI summary
Growth rates of modern science: A latent piecewise growth curve approach to model publication numbers from established and new literature databases
Lutz Bornmann, Robin Haunschild, Ruediger Mutz
TL;DR
The study examines how scientific publication output grows over time and addresses challenges in modeling growth processes. Using publication data from four literature databases and regression approaches, it finds exponential growth and identifies a five-segment model as the best fit.
Problem
Characterizing scientific growth requires investigating growth processes and addressing different possible growth patterns.
Method
The study analyzes publication data from four literature databases and estimates regression models, including piecewise regression approaches.
Results
4.10% overall growth rate corresponds to a 17.3-year doubling time, while the five-segment model fits the publication data best.
Takeaways & Limitations
Exponential growth explains the data quite well, and segmented growth models provide a useful way to represent differing growth phases.
Takeaways & Limitations
Comparisons between scientific and economic growth rates should be interpreted with great care, and previously noted limitations remain valid.
Abstract
from arXiv · showhide
Growth of science is a prevalent issue in science of science studies. In recent years, two new bibliographic databases have been introduced which can be used to study growth processes in science from centuries back: Dimensions from Digital Science and Microsoft Academic. In this study, we used publication data from these new databases and added publication data from two established databases (Web of Science from Clarivate Analytics and Scopus from Elsevier) to investigate scientific growth processes from the beginning of the modern science system until today. We estimated regression models that included simultaneously the publication counts from the four databases. The results of the unrestricted growth of science calculations show that the overall growth rate amounts to 4.10% with a doubling time of 17.3 years. As the comparison of various segmented regression models in the current study revealed, the model with five segments fits the publication data best. We demonstrated that these segments with different growth rates can be interpreted very well, since they are related to either phases of economic (e.g., industrialization) and / or political developments (e.g., Second World War). In this study, we additionally analyzed scientific growth in two broad fields (Physical and Technical Sciences as well as Life Sciences) and the relationship of scientific and economic growth in UK. The comparison between the two fields revealed only slight differences. The comparison of the British economic and scientific growth rates showed that the economic growth rate is slightly lower than the scientific growth rate.
1 Introduction
The study addresses the lack of precise growth-rate estimates based on reliable publication data by comparing four bibliographic databases across the history of modern science. It also examines field-specific growth and the relationship between scientific and economic growth in the UK.
- Research gap: Precise growth-rate estimation from reliable publication data remains an unresolved need despite earlier evidence for exponential scientific growth.Bibliometric databases provide large-scale, multidisciplinary coverage, while publications are a central means of communicating scientific results.
- Research design: New databases Dimensions and Microsoft Academic enable investigation of scientific growth over centuries and can be compared with Web of Science and Scopus.The comparison uses four major multidisciplinary literature databases to assess whether growth results are robust across data sources.
- Research design: The study analyzes publication growth for all records and for Physical and Technical Sciences and Life Sciences, excluding social sciences and humanities from the field comparison.The authors state that publication data can be used as a valid proxy for research activity only for the two selected broad fields.
- Research design: A further analysis compares annual scientific growth with UK economic growth, using GDP because long historical economic time series are available for the UK.Worldwide comparison was not possible because long-term publication and economic-growth series are unavailable at that level.
2 Methods
The methods combine bibliometric and economic data from multiple databases with growth-function, segmented-regression, and latent-growth-curve analyses. The study defines broad subject categories, processes publication counts by year, and addresses differing database coverage and missing information.
- Data processing: Publication counts are obtained from five bibliographic databases and converted into annual statistics, with cumulative publication numbers used for growth analysis.The retrieved data represent publications published in each year, while cumulative counts form the basis of the growth analysis.
- Subject classification: Physical and Technical Sciences and Life Sciences are defined using subject classifications from Web of Science, Scopus, and Microsoft Academic.The study uses broad categories rather than additional fields and excludes patents from the relevant Microsoft Academic publication analysis.
- Data limitations: The databases differ in historical coverage, indexing strategy, document types, and completeness of document-type assignments.Microsoft Academic may favor publications with a digital footprint, and 77,227,143 indexed items lack an assigned document type.
- Statistical analyses: The analysis combines unrestricted and restricted exponential growth functions, segmented regression, and latent growth curve models.These approaches are used to address different possible growth patterns and the statistical analysis of publication time series.
3 Results
The analyses compare exponential-growth models across four bibliographic databases and identify historically interpretable periods with distinct publication-growth rates. They also examine broad scientific fields and UK science–economic growth relationships.
- Model comparison: Discarding the first five years of each database time series left data spanning 1670–2018, with database-specific starting years.The actual starting years were 1670 for Dimensions, 1805 for Microsoft Academic, 1905 for Web of Science, and 1866 for Scopus.
- Model comparison: Model M9 with five segments and first-segment intercept–slope covariance fit Physical and Technical Sciences best, while M8 with four segments fit all publications and Life Sciences best.Adding a fifth segment for the Second World War produced only negligible improvement for all publications.
- Growth rates of science: 4.10% was the unrestricted overall annual growth rate, corresponding to a 17.3-year doubling time.The unrestricted growth estimate summarizes publication-count growth across the considered databases.
- Scientific fields: 5.51% was the post-1945 growth rate for Physical and Technical Sciences, versus 4.79% for Life Sciences.The corresponding doubling times were 12.9 and 14.8 years, respectively.
4 Discussion
The study finds that publication growth is broadly exponential but varies across historical periods, with segmented rates associated with economic and political developments. It also reports similar growth across two broad fields and a slightly higher scientific than economic growth rate in the UK, while emphasizing limits of publication counts and historical GDP comparisons.
- Limitations: Publication counts remain an imperfect measure because they may not reflect actionable knowledge, and annual historical indicators without missing values are scarce.The paper also notes that the study uses multidisciplinary databases and recommends future work with mono-disciplinary databases.
- Overall growth: 4.10% is the overall annual growth rate, corresponding to a doubling time of 17.3 years across the publication databases.This unrestricted-growth estimate differs from the 2.96% Web of Science rate reported for 1980–2012 because this study covers 1900–2018 and multiple databases.
- Segmented growth: Five segments fit the publication data best, with differing growth rates associated with economic phases such as industrialization and political developments such as World Wars.The study interprets the segmented pattern historically but does not empirically establish why growth speeds differ across periods.
- Historical interpretation: War efforts produced a visible decline in publication output, although research continued and wartime results may have contributed to post-war discoveries.The passage links reduced openly available research to security restrictions and researchers’ reassignment to war-related work.
- Field comparisons: 5.07% annual growth and a 14.0-year doubling time characterize Life Sciences, versus 5.51% and 12.9 years for Physical and Technical Sciences.The study describes these differences as slight rather than fundamental.
Supplementary Information
The supplementary information identifies the paper and presents supplementary tables and a figure covering segmented regression models and unrestricted growth.
- The paper is titled “Growth rates of modern science: A latent piecewise growth curve approach to model publication numbers from established and new literature databases.”
- Lutz Bornmann, Robin Haunschild, and Rüdiger Mutz are listed as the authors.
- Table S1 reports mixed effects segmented regression models with four or five segments and missing imputation.
- Table S2 reports fixed effects segmented regression models for UK GDP and publication counts.
- Figure S6 plots unrestricted and segmented unrestricted growth using publication numbers from four bibliographic databases.