Institutions in the Complexity-Relatedness Framework

Regional Economic Performance Models – Europe, NUTS‑2 (2010, 2013, 2017, 2021)Description This dataset underpins the panel regressions of regional economic performance in the accompanying manuscript. It covers 183 NUTS‑2 regions across 22 EU Member States for the benchmark years 2010, 2013, 2017, and 2021 (729 region‑year observations; 29 variables; ~0.18 MB). Outcome variables are three‑year growth rates in GDP, GVA and employment (constructed as in the paper), with initial income levels in logs. Key regressors include institutional quality (EQI) and technology‑based capability measures derived from OECD REGPAT (technological complexity and relatedness density) combined into composite indices: CD‑Mean (arithmetic mean of complexity and relatedness), CDQ‑Geo (geometric mean of CD‑Mean and EQI), and CDQ (geometric mean of complexity, relatedness, and EQI). Controls include capital stock and GFCF ratios, hours worked, population density, and R&D variables used in robustness checks. Administrative‑proximity spillovers are provided as within‑NUTS‑1 and within‑country averages of the composite indices (suffixes _W1 and _W0). Construction choices and normalisation follow the manuscript’s Methods.Filesdata_regional_performance models.csv Geographic coverageEU NUTS‑2 regions (excludes FRY overseas regions; Azores/Madeira and Spanish islands excluded for geographic/institutional consistency; countries where NUTS‑0 ≡ NUTS‑2—LV, EE, LU, CY, MT are omitted).Temporal coverageBenchmark years 2010, 2013, 2017, 2021; outcomes are 3‑year growth rates.Data sourcesPatent‑based indicators: OECD REGPAT (PCT applications, IPC 4‑digit).Institutional quality: European Quality of Government Index (EQI), QoG Institute.Controls / outcomes: European Commission ARDECO and Eurostat.Variable construction and filters (5‑year rolling averages, presence matrix, complexity algorithm, relatedness density, normalisation) follow the manuscript.manuscript_numaraliVariables (grouped overview)Identifiers: reg_nuts2, NUTS1, NUTS0, prio_yearOutcomes / initial levels: gdp_c_gr3, gva_c_gr3, totalemp_gr3 (3‑year growth rates), lngdp_c, lngva_cControls: cap_stock, gfcf, lnhwork (hours worked, log), capdens** (capital intensity variants used in robustness), population density proxy (see manuscript tables)manuscript_numaraliComposite capability indices & spatial averages:tcrd_mean (CD‑Mean),fdenseqi_geo (CDQ‑Geo),tcrd_eqi (CDQ),_W1 / _W0 suffixes = within‑NUTS‑1 / within‑country averages.R&D / structure (used in robustness): rd_pers, lnrd_pers, rd_gdp, emp_tech.Technological Entry Models – Europe, NUTS‑2 × IPC 4‑Digit (2010–2014; 2017–2021)DescriptionThis dataset underpins the logistic models of technological entry (probability of developing a new regional specialisation). It is a technology-region panel spanning 183 NUTS‑2 regions × 608 IPC 4‑digit classes, with two non‑overlapping windows (2010–2014 and 2017–2021). The dependent variable entry equals 1 if a region gains a revealed technological advantage in technology c within the window (0 otherwise). Core regressors are the institutional‑technology composites, tcrd_mean (CD‑Mean), fdenseqi_geo (CDQ‑Geo), and tcrd_eqi (CDQ), constructed from complexity and relatedness density (REGPAT) and institutional quality (EQI). Controls include R&D intensity and personnel, capital stock/GFCF, labour and density proxies. Spatial averages at NUTS‑1 (_W1) and country level (_W0) are provided. Variable construction and thresholds follow the manuscript’s Methods (augmented presence matrix; 5‑year rolling averages; normalisation).Filesdata_entry_models.csv (218,528 rows × 40 cols)Key variablesIdentifiers: reg_nuts2, NUTS1, NUTS0, ipccode, prio_year, periodDependent variable: entry (0/1)Composite capability indices: tcrd_mean (CD‑Mean), fdenseqi_geo (CDQ‑Geo), tcrd_eqi (CDQ); their spatial averages: _W1, _W0Controls: cap_stock, gfcf, hwork/lnhwork (as provided), rd_gdp, rd_pers, lnrd_pers, emp_tech.