Alberto Costantiello

The Regional Innovation Scoreboard is the principal evidence base for European regional innovation policy. This paper asks what that source can identify, using the sales of new-to-market and new-to-firm innovations across 218 European regions over four Community Innovation Survey waves as a demonstration case. Three boundaries emerge. First, the annual file replicates biennial survey values, so the nominal eight-year panel holds four independent observations per region; ignoring this understates standard errors on registry indicators by up to 40 per cent and reverses one coefficient sign. Second, the leading regressor is negatively associated with innovative sales between regions and positively within them, and leave-one-country-out validation fails in levels but not on within-region deviations, locating much of the between variation in a national component, whether it reflects survey implementation or genuine national differences. Third, flexible learners outperform the linear specification only when validation folds share information within regions and country-waves, a result that rests on 218 regions and 21 countries, since the four survey waves alone could not rank the learners. Within these boundaries, only the SME product innovation incidence, partly linked to the outcome by definition, and trademark activity are associated with innovative revenue, and all estimates are correlational. A taxonomy built on the inputs alone is nearly orthogonal to the official classification. The Scoreboard supports longitudinal inference about a region’s own trajectory, not cross-sectional benchmarking.

The Regional Innovation Scoreboard ranks European regions by a composite index but reveals little about how innovation performance is composed, how persistent it is, and whether it translates into prosperity. Using data for 245 regions in 31 countries over 2016–2023, this study addresses three research questions with panel econometrics, cluster analysis, and machine learning. First, dynamic panel estimates that treat the indicators as endogenous indicate substantial persistence in performance and show that its most robust correlates are non-R&D investment and innovation and collaboration among small- and medium-sized enterprises. Second, clustering the components of the index, rather than the index itself, identifies ten innovation profiles that are weakly aligned with the official performance groups and correspond to distinct development models, from science-intensive metropolitan cores to fast-growing catching-up regions. Third, with GDP per capita as an external outcome and region-grouped validation, the indicators account for about two-thirds of cross-regional differences in prosperity, and their composition provides additional predictive information beyond the composite index, although they carry little information on short-run growth. Read together, the results show that stable rankings conceal changing configurations, that convergence has relied largely on factors other than innovation, and that policies should address region-specific bottlenecks.

Renewable natural capital, one component of the comprehensive wealth whose maintenance per head defines sustainability in wealth accounting, is falling per capita across most of the world. This study asks what the decline records and whether environmental, social and, in particular, governance (ESG) indicators carry information about the asset base once demographic dilution is held constant. Using the World Bank Sovereign ESG Data Portal and the Changing Wealth of Nations accounts for 149 to 151 countries over 1995–2020, it decomposes per capita growth exactly into stock growth and population growth. On a fixed sample of 148 countries, the global stock grew by 5.9 per cent while population rose by 36 per cent, although fisheries, mangroves and timber declined in most countries holding them; within countries, per capita growth is mostly stock change, and across countries, mostly population growth. Joint specifications on a common sample, with pillar equations from six panel estimators as benchmarks, measure what each pillar adds. The environmental pillar carries almost all the incremental within-country information, with climatic stress acting through hydropower. Once demography is held constant, governance adds little information, no more than about two percentage points of explanatory power across countries, and ranking countries by total-stock growth reverses its association with measured sustainability. Per capita criteria should be complemented by total-stock growth and asset-level reporting, since neither measure records stewardship, and governance scores should not be read as proxies for it.

Does digital adoption raise regional prosperity in the region where it occurs, or does it work through neighbouring regions? We assemble a georeferenced panel of 234 EU NUTS-2 regions observed from 2017 to 2024, linking Eurostat data on GDP per capita in purchasing power standards, household broadband and internet use, employment and social exclusion to NUTS 2024 boundaries. Global Moran's I is positive and significant for every variable and year (0.39–0.78), and LISA maps reveal persistent high-income and low-income clusters. Non-spatial estimators give a positive digital coefficient: 0.21% higher GDP per capita per percentage point of adoption under two-way fixed effects, and 1.83% when lagged adoption instruments a noisy, survey-based regressor. However, fixed-effects residuals remain spatially autocorrelated, and a Spatial Durbin Model (ρ = 0.70) shows that the own-region (direct) effect is not significant, while the spillover (indirect) effect is positive (0.20% per point, p = 0.039) and robust across three spatial weight matrices. Of six clustering algorithms compared on eleven validity indices, Ward hierarchical clustering performs best and identifies seven spatially contiguous regional profiles. Of six machine-learning regressors, a random forest validated by region-grouped cross-validation performs best (out-of-sample R² = 0.57). Its SHAP decomposition reproduces the econometric signs: own adoption dominates in levels, while neighbours' adoption and employment dominate within regions. Digital policy in the EU should therefore be targeted at the level of functional macro-regions, not individual regions.

Nuclear power is usually treated as a determinant of emissions rather than as an outcome to be explained, and the few studies that explain it examine environmental, social and governance conditions one pillar at a time. Using an unbalanced panel of twenty-nine nuclear economies over 2004–2020, with 466 country-year observations, we analyse the nuclear share conditional on possessing capacity and estimate a single equation in which all three pillars enter jointly, comparing six estimators that separate variation between countries from variation within them. Fossil dependence is negatively associated with the nuclear share under every estimator, but the robustness is cross-sectional rather than temporal. Electoral democracy reverses sign across dimensions, positive between countries and negative within them, where it rests on a small number of closure episodes. Female parliamentary representation is negatively associated with the nuclear share across cross-sectional estimators, which provides ecological evidence consistent with the gender gap in nuclear risk perception; the mechanism itself is not observed. The methodological implication concerns validation: with 13.9 per cent of the variance being temporal, random cross-validation overstates the performance of flexible algorithms through panel leakage. The estimates are associations, rather than causal effects or evidence on adoption.

The European Innovation Scoreboard classifies medium- and high-technology exports among the impacts of innovation, implying that institutions and public demand shape what European economies export. We test that premise on thirty-nine economies, 2017 to 2024, applying panel econometrics, unsupervised clustering and machine-learning regression to one non-circular specification. All three give a negative answer. Institutional quality explains 1.5 per cent of the cross-country variation and carries a standardised effect of −0.27; procurement carries nothing. The employment share in high and medium-high technology carries 0.725 and raises the fit to 0.617. An unsupervised partition reproduces the Scoreboard's ranking at a Cramér's V of 0.929 while leaving the outcome unseparated. Validation designs that do not separate countries return out-of-sample R² above 0.94, which unit-separated folds drive below zero.

European packaging policy is built almost entirely around treatment, while the accompanying prevention obligation lacks an equivalent apparatus of measurable objectives. This article addresses this upstream question directly: what determines the quantity of plastic packaging waste that European economies place on the market? Using Eurostat data, we regress per capita plastic packaging waste on the three pillars of the World Bank Sovereign Environmental, Social and Governance framework for 29 countries between 2014 and 2023. Each pillar is examined through three modules. The first is a panel comparison of pooled, between, fixed-effects, random-effects, weighted, and dynamic estimators, with inference reassessed using a wild cluster bootstrap. The second is an unsupervised classification comparing six algorithms across eleven validity indices, with the dependent variable withheld for external validation. The third is a supervised comparison of six learners under leave-one-out cross-validation, used to validate functional form rather than to forecast. Three findings emerge. Energy intensity and carbon intensity per capita are robustly associated with generation and survive correction for multiple testing, yet environmental country types do not significantly explain differences in packaging burden. Social configurations separate generation along a binary post-transition boundary that does not survive controlling for income, and the sign reversal in life expectancy between the cross-sectional and time dimensions is confirmed independently by the learners. Internet penetration is the strongest association in the study, robust to common time effects and to income but extending to every packaging fraction and therefore not specific to delivery formats, while composite institutional quality is insignificant in every linear specification and suggests, under unrestricted functional form, a threshold rather than a gradient, a shape that a parametric test on the same cross-section cannot locate and that is therefore reported as a hypothesis. Prevention policy must therefore engage determinants beyond waste management. Among these determinants, the diffusion of online retail is currently the most informative.

Europe measures the innovative performance of its regions largely by counting patents, and allocates public resources accordingly. This paper asks what regional patenting reflects, using Regional Innovation Scoreboard data for 245 European regions observed annually from 2016 to 2023, and applies three methods. A correlated random-effects specification decomposes each association into within- and between-region components. K-Means clustering, constructed without reference to the dependent variable, yields four exploratory innovation profiles within which coefficient equality is formally tested. Predictive validation compares seven algorithms under designs withholding random observations, entire regions, entire countries, and later years. Business research effort is the dominant correlate; science–industry collaboration is weaker and does not survive country effects; trademark activity is positive and does. Regional capacity proves remarkably immobile: for the four variables modelled here, differences between regions account for roughly 95 per cent of the variation, so short-run movements carry little information. No difference in the research–patenting association across the four profiles is detected, at a precision the paper reports, while levels of research differ sharply. The ranking of predictive methods reverses as the validation design tightens, flexible algorithms falling behind ordinary least squares once regions and countries are withheld.

Comprehensive wealth accounting defines net investment in natural capital as the change in the stock, yet country-level ESG research has concentrated on flow measures of environmental pressure and left this outcome unexamined. Drawing on the World Bank Sovereign ESG Data Portal, this paper takes the annual change in renewable natural capital per capita, from the Changing Wealth of Nations accounts, as its dependent variable, for 149 to 151 countries between 1995 and 2020. It is explained three times, through parallel equations differing only in the ESG pillar of origin of their regressors, so that the three sets of determinants are estimated as distinct and comparable effects rather than compressed into a composite index. Each equation is developed through panel econometrics across six estimators, a comparison of six clustering algorithms ranked on eleven internal validity indices, and six supervised learners used to validate the specification rather than to predict. Median renewable natural capital per capita falls by 29.7 per cent and its share of comprehensive wealth halves, yet the aggregate stock is roughly constant: the coefficient on population growth is −0.913, indistinguishable from the value implied by the accounting identity, and becomes insignificant on the total stock. The decline is therefore demographic arithmetic. Climatic stress is the most robust environmental determinant, and its coefficient more than doubles when agricultural land is excluded from the dependent variable. Institutional trajectory dominates institutional level, a distinction recovered independently by the clustering. Relationships differ systematically between the two panel dimensions, and the renewable energy share reverses sign across them.

European packaging policy is built almost entirely around treatment, while the prevention obligation that accompanies it lacks an equivalent apparatus of measurable objectives. This article addresses the upstream question directly, asking what determines the quantity of plastic packaging waste that European economies place on the market. Plastic packaging waste generated per capita, from Eurostat, is regressed on the environmental, social and governance pillars of the World Bank Sovereign ESG framework for twenty-nine countries between 2014 and 2023. Each pillar is examined through three modules: a panel comparison of pooled, between, fixed effects, random effects, weighted and dynamic estimators, with inference re-assessed by wild cluster bootstrap; an unsupervised classification comparing six algorithms across eleven validity indices, with the dependent variable withheld for external validation; and a supervised comparison of six learners under leave-country-out cross-validation, used to validate functional form rather than to forecast. Three findings emerge. Energy intensity and carbon intensity per capita are robustly associated with generation and survive correction for multiple testing, yet environmental country types carry no information about packaging burden. Social configurations do separate generation, along a binary post-transition boundary, and the sign reversal on life expectancy between the cross-sectional and time dimensions is confirmed independently by the learners. Internet penetration is the strongest association in the study, while composite institutional quality is insignificant in every linear specification and emerges under unrestricted functional form as a threshold rather than a gradient. Prevention policy must therefore engage determinants that lie outside waste management, of which online retail is currently the most important.