Data and R code for a Data Envelopment Analysis of national innovation system efficiency in the EU-27, 2022–2024

This record contains the compiled dataset and R code for a Data Envelopment Analysis of national innovation system efficiency across the 27 EU member states. Each country is treated as a unit converting research inputs into scientific and technological outputs, and efficiency is estimated relative to the observed best-practice frontier. All indicators are three-year means for 2022–2024, expressed in natural units. Raw extracts are drawn from Eurostat, the OECD Bibliometric Indicators database, the European Innovation Scoreboard 2025, and the WIPO Global Innovation Index 2025. The data/ folder holds these source extracts, the intermediate workbooks used to assemble them, and two analysis-ready files: NIS_DEA_data.csv, containing the six candidate inputs and three outputs used in the specification, and NIS_SII_2025.csv, containing the Summary Innovation Index scores used for external comparison. All nine candidate variables that reached the specification stage are retained in NIS_DEA_data.csv, including the two inputs subsequently dropped by the correlation screen, so the screen itself can be reproduced. A raw extract for employment in knowledge-intensive activities, excluded earlier on conceptual grounds, is also included. The R scripts/ folder contains nine numbered scripts covering the main DEA estimation, the Simar–Wilson bootstrap, the rank comparison with the Summary Innovation Index, returns-to-scale classification, collinearity screening, an IP-revenue sensitivity check, external validity against the Global Innovation Index, and a stochastic frontier identifiability test. Run them in numerical order with the working directory set to the pack root. Outputs are written to results/, which already contains the files from a complete run, including a session log. Estimation used R 4.6.1 with the Benchmarking package version 0.33; the bootstrap seed is fixed at 123. Source extracts in data/ are redistributed from Eurostat, the OECD, the European Innovation Scoreboard and WIPO, and remain subject to the terms of their respective publishers. The CC BY 4.0 licence applies to the compiled dataset, derived files, R code and results produced by the authors.