How Robust Are National Rankings of Highly Cited Scientists? Research Capacity Adjustment and Data Availability

Author: Andy A. Acosta-MonterrosaRepository: https://github.com/AndyPaaai/global-scientific-inequality-research-capacity Overview This repository contains the raw source datasets, reproducible R analytical workflow (src/code.R), and supplementary tables (Supplementary Material/) accompanying the manuscript: How Robust Are National Rankings of Highly Cited Scientists? Research Capacity Adjustment and Data AvailabilityGlobal inequalities in highly cited scientist representation: The role of research capacity and economic developmentAuthor: Andy A. Acosta-Monterrosa The study conducts a cross-sectional, country/economy-level bibliometric analysis of the August 2026 update (Version 9) of the Ioannidis science-wide author databases (2025 single-year citation impact), integrated with research and development (R&D) indicators from the UNESCO Institute for Statistics (UIS), macroeconomic and demographic data from the World Bank World Development Indicators (WDI), World Bank FY2026 income classifications, and United Nations M49 geographical regions. 1. Repository Structure global-scientific-inequality-research-capacity/ ├── LICENSE ├── README.md ├── src/ │ └── code.R # Main reproducible R workflow ├── Supplementary Material/ │ ├── Supplementary_Table_S1.csv # Data availability by income group & UN region (12 rows) │ ├── Supplementary_Table_S2.csv # National counts, FTE rates, O/E ratios & rankings (98 rows) │ ├── Supplementary_Table_S3.csv # Bootstrap rank stability (500 resamples; 98 rows) │ └── Supplementary_Table_S4.csv # Sensitivity analyses across windows, FTE cutoffs & LOO (40 rows) └── data/ ├── classifications/ │ └── OGHIST_2026_10_08.xlsx # World Bank historical income classifications (FY2026) ├── ioannidis_v9_2026/ │ ├── FREQUENTLY ASKED QUESTIONS 2026.docx │ ├── Table_1_Authors_career_2025_pubs_since_1788_wopp_extracted_202608.xlsx │ ├── Table_1_Authors_singleyr_2025_pubs_since_1788_wopp_extracted_202608.xlsx │ ├── Table_2_field_subfield_thresholds_career_2025_pubs_since_1788_wopp_extracted_202608.xlsx │ ├── Table_2_field_subfield_thresholds_singleyr_2025_pubs_since_1788_wopp_extracted_202608.xlsx │ ├── Table_3_maxlog_career_2025_pubs_since_1788_wopp_extracted_202608.xlsx │ └── Table_3_maxlog_singleyr_2025_pubs_since_1788_wopp_extracted_202608.xlsx ├── unesco_uis/ │ ├── data.csv # UNESCO UIS SDG 9.5.1 & 9.5.2 extract │ ├── disaggregations.csv │ ├── footnotes.csv │ ├── glossary-terms.csv │ ├── indicators.csv │ └── README.md └── world_bank/ ├── GDP_per_capita_PPA/ │ ├── API_NY.GDP.PCAP.PP.KD_DS2_en_csv_v2_1629.csv # WDI GDP per capita, PPP (constant 2021 international $) │ ├── Metadata_Country_API_NY.GDP.PCAP.PP.KD_DS2_en_csv_v2_1629.csv │ └── Metadata_Indicator_API_NY.GDP.PCAP.PP.KD_DS2_en_csv_v2_1629.csv └── total_population/ ├── API_SP.POP.TOTL_DS2_en_csv_v2_5254.csv # WDI total population (matched to FTE reference year) ├── Metadata_Country_API_SP.POP.TOTL_DS2_en_csv_v2_5254.csv └── Metadata_Indicator_API_SP.POP.TOTL_DS2_en_csv_v2_5254.csv When src/code.R is executed, it automatically creates (or updates): results/ — Intermediate audit tables, sample eligibility files, Table_1_model_IRR.csv, Table_1_model_fit.csv, Table_2_rank_agreement.csv, reproducibility_report.txt, and sessionInfo.txt. Supplementary Material/ — Regenerates Supplementary_Table_S1.csv through Supplementary_Table_S4.csv. figures/ — Exports high-resolution publication panels (Figure_1A_income_coverage, Figure_1B_region_coverage, Figure_2A_absolute_vs_FTE, Figure_2B_FTE_vs_adjusted, and Figure_3_bootstrap_stability) in .svg, .png (600 DPI), and .pdf formats. 2. Original Data Sources All raw input snapshots used in the manuscript are staged under data/: Ioannidis Science-Wide Author Databases (Version 9, August 2026):Ioannidis JPA. August 2026 data-update for "Updated science-wide author databases of standardized citation indicators", Version 9. Elsevier Data Repository. DOI: https://doi.org/10.17632/btchxktzyw.9. Primary outcome: single-year 2025 citation database (Table_1_Authors_singleyr_2025_pubs_since_1788_wopp_extracted_202608.xlsx). Validation & master universe construction: career-long database (Table_1_Authors_career_2025_pubs_since_1788_wopp_extracted_202608.xlsx). Additional companion files from Version 9 (Table_2_*, Table_3_*, and FAQ documentation) are included in data/ioannidis_v9_2026/ for completeness. UNESCO Institute for Statistics (UIS):SDG Indicator 9.5.1 — Research and development expenditure as a proportion of GDP (EXPGDP.TOT) and SDG Indicator 9.5.2 — Researchers (in full-time equivalent) per million inhabitants (RESDEN.INHAB.TFTE), located in data/unesco_uis/data.csv. Most recent available observations within 2019–2023 were selected independently per economy and indicator. Portal: https://uis.unesco.org/ World Bank — World Development Indicators (WDI): NY.GDP.PCAP.PP.KD: GDP per capita, PPP (constant 2021 international dollars) for reference year 2023 (data/world_bank/GDP_per_capita_PPA/API_NY.GDP.PCAP.PP.KD_DS2_en_csv_v2_1629.csv). SP.POP.TOTL: Total population matched to each economy's researcher FTE reference year to compute total national researchers FTE (data/world_bank/total_population/API_SP.POP.TOTL_DS2_en_csv_v2_5254.csv).Portal: https://databank.worldbank.org/source/world-development-indicators World Bank Country and Lending Groups (FY2026):Historical income classification workbook (data/classifications/OGHIST_2026_10_08.xlsx), using fiscal year 2026 (FY26) income categories (Low income, Lower middle income, Upper middle income, High income). Portal: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups United Nations M49 Geographical Regions:Standard country/area codes and regional classifications obtained via the countrycode R package (v1.6.1). Portal: https://unstats.un.org/unsd/methodology/m49/ 3. Running the Analysis System Requirements R version: R 4.4.2 (tested on Windows 11) Required R packages: install.packages(c( "readxl", "readr", "dplyr", "tidyr", "countrycode", "ggplot2", "MASS", "sandwich", "svglite" )) Execution From the repository root, run: Rscript src/code.R Or inside an R / RStudio session with the working directory set to the repository root: source("src/code.R") (Optional) You can also explicitly set the PAPER1_ROOT environment variable to the repository root path prior to running the script. Workflow Summary src/code.R performs the following sequential steps without modifying any raw files in data/: Verifies that each required data source pattern matches a single file under data/ and records file sizes and MD5 checksums (results/input_manifest.csv). Aggregates country-level author counts from the 2025 single-year and career-long Ioannidis databases and audits unassigned ISO3 records. Integrates UNESCO UIS indicators, World Bank WDI indicators, FY2026 income classifications, and UN M49 regions into a 224-economy master dataset and country-year panel. Constructs the analytical samples (Sample A: $n = 101$ economies; primary Sample B: $n = 98$ economies) and computes data coverage stratified by income group and UN region (Supplementary_Table_S1.csv). Fits Poisson dispersion diagnostics and two nested negative binomial models with $\log(\text{FTE})$ as an offset and heteroskedasticity-consistent HC3 standard errors (Table_1_model_IRR.csv, Table_1_model_fit.csv). Computes national rankings across three representations — (i) absolute researcher counts, (ii) listed researchers per 10,000 national researchers FTE, and (iii) model-adjusted observed-to-expected (O/E) ratios (Supplementary_Table_S2.csv) — and evaluates pairwise rank concordance using Spearman's $\rho$ and Kendall's $\tau$ (Table_2_rank_agreement.csv). Evaluates model-adjusted ranking stability across 500 nonparametric bootstrap resamples (BOOT_SEED = 20261106L; Supplementary_Table_S3.csv). Runs sensitivity analyses restricting to economies with $\ge 1{,}000$ FTE ($n = 85$), testing alternative R&D reference windows (2015–2023, 2019–2023, 2019–2024, 2021–2023), and conducting leave-one-country-out analyses (Supplementary_Table_S4.csv). Generates publication figure panels in figures/ and logs session metadata (results/reproducibility_report.txt, results/sessionInfo.txt). 4. Expected Analytical Checks & Key Results The script enforces strict sample-size assertions to guarantee exact reproduction of the manuscript snapshot: | Metric | Expected Study Output | |---|---:| | Annual database rows (2025 single-year) | 249,310 | | Annual rows with valid ISO3 country code | 249,141 | | Master dataset economies | 224 | | Economies with R&D expenditure and FTE in 2019–2023 (Sample A) | 101 | | Economies with additional 2023 PPP GDP per capita (Sample B) | 98 | | Annual author records captured in Sample B | 215,130 (86.3%) | | Economies with zero annual author records in Sample B | 5 | | Valid bootstrap replicates (BOOT_SEED = 20261106L) | 472 of 500 (94.4%) | Summary of Principal Findings Data availability (Supplementary_Table_S1.csv): Complete indicators were available for 43.8% (98/224) of economies, with marked disparities across income groups (56.3% high-income vs. 16.0% low-income) and regions (82.6% Europe vs. 25.9% Africa and 5.0% Oceania). Negative binomial regression (Table 1): Model 1 (R&D % GDP + FTE offset): R&D expenditure $\text{IRR} = 1.204$ (95% CI: $0.996\text{--}1.457$, $p = 0.055$), $\text{AIC} = 1298.03$. Mode