The measurement boundary and the mismeasurement of national innovation: is the R&D-intensity target a biased proxy for knowledge investment?

Author:
Preprint
Publisher website
Google Scholar
Abstract
Innovation policy in Europe is steered by a single headline number: gross expenditure on research and development as a share of gross domestic product. The 3%-of-GDP target has anchored European Union strategy for two decades, and R&D intensity remains the backbone of the scoreboards that benchmark national performance and help justify the allocation of funds. Yet R&D sits inside a narrow accounting boundary that leaves out most modern intangible investment: organizational capital, design, brand, and training. We ask whether the indicator that drives policy is a faithful proxy for national knowledge investment and, if it is not, in which directions it misjudges which economies. We use the Global INTAN-Invest database, whose asset detail separates intangibles inside the national-accounts boundary (R&D, software, other intellectual property) from those outside it, together with Eurostat R&D data, and we compare each European economy’s rank on the R&D target with its rank on full Corrado–Hulten–Sichel intangible intensity, year by year from 2010 to 2024. R&D intensity is positively but imperfectly related to knowledge investment (Spearman’s ρ = 0.53 across the EU in 2019, rising to 0.70 once the Irish and Luxembourg multinational hubs are removed). The residual is not noise. The same economies are over-credited and under-credited year after year, and a stochastic acceptability analysis over the entire space of admissible asset weights shows that the reordering is weight-invariant and survives the addition of 15% estimation noise to the modelled assets. Austria is the limiting case: third in the Union on R&D intensity, sixteenth of twenty-four on total intangibles, over-credited in every year of the panel and under 99% of weightings. The pattern is not unique to Europe: extended to a global pool of economies using UNESCO R&D data, the same biased-but-informative relationship holds (ρ ≈ 0.71), with Japan the clearest over-credited case (over-credited under 99% of asset weightings) and the United States under-credited. We read the R&D-intensity target as biased-but-informative: informative about effort, biased in directions that can be named. We then draw out what that means for indicator-based innovation policy.Anonymized for review; author identity, affiliations, acknowledgements, funding, and the declaration of generative-AI use are provided on a separate title page.

This publication has 30 references indexed in Scilit: