R&D Expenditures and Economic Performance in Europe: Evidence from NUTS Regions

Scott W. Hegerty, Ph.D.

Distinguished Professor of Economics, NEIU

World Economy Research Institute

October 22, 2026

​

Main ideas:

• R&D Spending in Europe, by region
• Spatial patterns and spatial autocorrelation
• Categories of expenditure
• Correlations with GDP and Gross Value Added

​

Variation and differences:

• Means and standard deviations of NUTS-2 regions within countries
• Distributions of NUTS-2 GVA within NUTS-2 regions
• Convergence in GERD expenditure?

​

​

​

Previous Studies:

• Petrakos et al. (2011): NUTS-2 data, 1990-2003�→ Regional divergence (growth) getting stronger
• Rios et al. (2015): “Lisbon Index” spillovers� (Educ + Emp + R&D) 2000-2010
• Kijek et al. (2020): Technological convergence,� Clear East/West Split
• Blažek and Kadlec (2019): Advanced regions have more private R&D, less-developed have more public R&D

​

This Study:

• Eurostat data: Gross Expenditure on R&D �(GERD), 2015-2024

179 NUTS-2 regions used

​

• 4 Groups: �All, Business, Government, Higher Ed
• Also GDP, Gross Value Added (NUTS-2)
• GVA and GDP (NUTS-3)

​

Methods (#1/3):

1. GERD: Euros per capita�Growth rates (1-year, vs, 10-year-max) [key variables]

​

• Map GERD (NUTS-2) growth rates across Europe

Note concentrations of each

Private R&D as a proxy for development

• Examine patterns: Visual and spatial autocorrelation (Moran’s I)

​

• Correlation analysis (NUTS 2)
• Country by country: Calculate means and standard deviations for NUTS-2 regions in the country�Compare values and rankings�

Methods (#2/3):

• GVA (Real Indices):

For all NUTS-2 regions with four or more NUTS-3 subregions (21 total)

​

Plot distributions (as boxplots)

Look for “wide” spreads

​

→ Which NUTS-2 regions have the largest variation?

�

Methods (#3/3):

• Beta Convergence for GERD at the NUTS-2 level:

​

Yt/Y0 = 𝞪 + 𝞫Y0 + 𝜺

​

​

• If a significantly negative coefficient, large initial values related to lower growth rates
• Differences in Total expenditure and among the three categories

​

�

GERD: Spending at the NUTS-2 level

• 1) Ratio of Business GERD to Total GERD

​

• 2) 1-year growth rates (2022-2023)
• Maps for ALL, Business, Government, and Higher Ed R&D
• Note clusters, high and low values
• Note private vs. public (Business vs. Government)

​

​

​

% Private R&D/Total, 2023

(Note E. Germany, Spain, �Baltics, Balkans)

Proxy for development?

ALL: Clusters (high) in RO, PL

​

​

Business: Clusters (high) in RO (also negative in some areas)

Higher Ed: Losses (neg.) in RO, PL

Elsewhere too

​

Gov’t: Clusters (high) in RO, PL

Negative in PL regions

​

(Lack of) Spatial Autocorrelation

Moran’s I (p-value):

All -0.015 (0.670)

Bus 0.010 (0.381)

Gov 0.023 (0.188)

Higher Ed 0.035 (0.043)

​

→ Just use traditional correlations

​

(Pearson) Correlations (growth rates)

Note highest correlations w/Business GERD, negative for Gov’t

Means and Standard Deviations by Country

Note Hungary (negative)

​

Poland’s means are fairly high

Rankings: Means of Measures By Country

1-year GDP growth lowest, then 1-year GVA growth

Business GERD growth often highest (3 countries), Higher Ed in 3 countries

Rankings: Means By Category Across Countries

Hungary lowest for 1-year GVA and GDP growth

Poland #2 in 10-year GVA growth

Portugal #2 for some categories and #9 for others

Analysis: NUTS-3 regions Within NUTS-2

21 NUTS-2 regions have >= 4 NUTS-3 regions

1-yr Growth Rates: Distributions w/in regions

10-yr Growth Rates

Hungary lowest for 1-year GVA and GDP growth

Industry and Agriculture Ind., Manuf., Ag.

Range vs. Means: Levels (x) and Spread (y)

Only real outliers in Turkey

Note elevated ranges (above regression lines)

Convergence Regressions: Yt/Y0 = a + bY0 + e

1. GERD (all and Business)

​

Evidence of convergence (negative sign) for all

​

Largest coefficient (abs. val.) for Business

= fastest convergence

Convergence Regressions: Yt/Y0 = a + bY0 + e

2. GERD (Gov’t and Higher Ed))

​

Evidence of convergence (negative sign) for all

​

Larger coefficient (abs. val.) for Higher Ed

= fastest convergence

Main findings

• Geographic disparities: More Government R&D, less development in Balkans
• Little spatial autocorrelation; regular correlations w/GDP and GVA highest for Business R&D and negative for Government R&D
• Growth disparities—Hungary vs. Poland, for example
• 21 NUTS-2 regions have varying levels of internal variation
• Evidence of convergence in all four categories

Conclusions

R&D spatially clustered, but links are weak

Differs by type of spending

Differs by country

Differs within regions

​

​

→ Government R&D:Negatively correlated with growth

→ Business R&D: High share in “West”� Growing in Romania

Shows signs of rapid convergence

​

​

​

Thank you!

• S-Hegerty@neiu.edu