Do EU regions benefit from Smart Specialisation principles?

Figures & data

Figure 1. The Smart Specialisation framework.

Source: After Balland et al. (2019).

Table 1. Descriptive statistics for sample cities, 2011–15.

Key variables	Minimum	Median	Maximum
Patent Sum	101	469	11,198
Population	221.6	760.8	13,557.0
Popln Density	1.76	18.05	295.14
GDP Level	6.24	22.41	690.65
Mfg Share	0.036	0.158	0.403
Herfindahl	0.01	0.02	0.22
External Links	5	10	58

Note: The population, density, gross domestic product (GDP) level and manufacturing share values are reported for 2015. The values of the other variables are given for the period 2011–15. The population is reported in thousands. Population density is measured as population divided by land area. GDP is measured in billions of 2005 euros. The technology Herfindahl is bounded by the values 0–1, with 0 representing an even distribution of patents across technology classes, and 1 indicating that all patents are found in a single class. The number of external links for each city is measured as the number of K-cores (Seidman, 1983) using inventor collaboration data.

Figure 2. Technological relatedness in the European Union knowledge space: (a) 1981–85 and (b) 2011–15.

Note: The eight aggregate Cooperative Patent Classification (CPC) technology classes are: Electronics, Instruments, Chemicals, Biotech, Industrial process, Machinery and transport, Consumer goods and Climate change technologies.

Table 2. Top technology fields by complexity, 2011–15.

CPC Patent Class	Technology field	Complexity (indexed)
H04L	Transmission of digital information	100
H03M	Code conversion	96.7
H03K	Control of electronic oscillations	95.3
G06F	Electronic digital data processing	95.2
H01L	Semi-conductor devices	89.1
H03K	Pulse techniques	88.6
H04N	Pictorial communication	88.3
H01S	Devices using stimulated emissions	88.0
B81C	Manufacture of microstructural devices	87.4
G05F	Systems for regulating electric/magnetic variables	84.4

Note: CPC, Cooperative Patent Classification.

Table 3. Complexity scores in European Union cities.

Rank and city	Complexity score		Rank and city	Complexity score
	1981–85	2011–15		1981–85	2011–15
1. Rennes, France	59.02	84.11	136. Bologna, Italy	44.95	44.61
2. Dublin, Ireland	49.23	72.11	137. Ingolstadt, Germany	48.34	44.37
3. Stockholm, Sweden	52.82	70.28	138. Bielefeld, Germany	49.38	44.15
4. Antwerp, Belgium	55.70	70.25	139. Coventry, UK	41.63	44.12
5. Grenoble, France	59.81	69.29	140. Osnabruck, Germany	44.72	43.61
6. Tampere, Finland	46.25	69.21	141. Rouen, France	48.98	41.60
7. Nice, France	59.38	68.08	142. Reggio nell’Emilia, Italy	41.32	41.29
8. Caen, France	50.70	67.45	143. Aberdeen, UK	45.58	41.28
9. Helsinki, Finland	51.84	67.11	144. Amiens, France	43.22	39.67
10. Eindhoven, the Netherlands	73.49	67.03	145. Reims, France	45.15	39.15

Figure 3. Smart Specialisation in European Union cities by entry and exit, 1981–2015.

Table 4. Relationship between Smart Specialisation and economic growth.

Smart Specialisation quintiles (normalized)	Employment: rate of growth	GDP: rate of growth
1	0.30184	0.68426
2	0.29769	0.66836
3	0.25388	0.58820
4	0.21746	0.55201
5	0.19738	0.56507

Note: Gross domestic product (GDP) data are in constant 2005 euros. Simple growth rates for employment and GDP are measured between consecutive five-year periods and then summed.

Table 5. European Union city gross domestic product (GDP) growth and Smart Specialisation (dependent variable: GDP growth).

Independent variables	Model 1	Model 2	Model 3	Model 4	Model 5: (IV)GMM	Model 6: (IV)GMM
Smart Spec Index	0.0308* (0.0172)		0.0411** (0.0174)		0.0706** (0.0353)
EE Relatedness		0.0084 (0.0255)		0.0277 (0.0256)		0.1200** (0.0535)
EE Complexity		0.0512** (0.0243)		0.0534** (0.0244)		0.0917** (0.0405)
Lag GDP Level	0.0035* (0.0016)	0.0032* (0.0015)	0.0044** (0.0022)	0.0043** (0.0022)	0.0042 (0.0051)	0.0041 (0.0052)
Lag Pop Density			−0.0422*** (0.0126)	−0.0421*** (0.0126)	−0.1103** (0.0472)	−0.1116** (0.0475)
Lag Mfg Share			2.7288** (1.2829)	2.6721** (1.2858)	6.5087* (3.6640)	6.1779** (3.7068)
Lag Herfindahl			−0.1419 (2.0620)	−0.2687 (2.0703)	−0.4019 (4.4369)	0.0552 (4.4707)
Lag External Links			0.0271*** (0.0083)	0.0273*** (0.0083)	0.0209 (0.0203)	0.0209 (0.0205)
Lag Patent Sum			0.0000 (0.0001)	0.0000 (0.0001)	0.0003** (0.0001)	0.0004*** (0.0001)
Lag WoS per Capita			0.0434 (0.0428)	0.0421 (0.0428)	−0.0955 (0.0921)	−0.1118 (0.0926)
Constant	0.6134*** (0.0793)	0.6071*** (0.0795)	0.8789* (0.4579)	0.8901* (0.4583)
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations	870	870	867	867	867	867
R^2 within	0.39	0.40	0.43	0.43
Anderson LM-statistic					43.270***	43.251***
Cragg–Donald F-statistic					5.650***	5.013***
Sargan statistic					0.664	0.569

Note: Smart Specialisation measures (relatedness and complexity) are normalized. EE, entry and exit. All standard errors are robust and reported in parentheses. *Significant at 0.1, **at 0.05 and ***at 0.01. In the instrumental variables (IV) estimation of model 5 and 6, the Anderson LM-statistic tests for under-identification and the Cragg–Donald F-statistic tests for weak identification. These test statistics assume heteroscedastic robust standard errors rather than errors that are i.i.d. All models are city-level fixed effects panel regressions that include period fixed effects.

Table 6. European Union city employment growth and Smart Specialisation (dependent variable: employment growth).

Independent variables	Model 1	Model 2	Model 3	Model 4	Model 5: (IV) GMM	Model 6: (IV) GMM
Smart Spec Index	0.0319 (0.0201)		0.0373* (0.0208)		0.0467 (0.0420)
EE Relatedness		0.0120 (0.0260)		0.0221 (0.0303)		0.1115* (0.0620)
EE Complexity		0.0504* (0.0285)		0.0514* (0.0291)		0.0410 (0.0479)
Lag Employment Level	0.0004 (0.0003)	0.0003 (0.0004)	0.0007** (0.0004)	0.0007* (0.0004)	0.0004 (0.0009)	0.1115* (0.0620)
Lag Pop Density			−0.0490*** (0.0151)	−0.0489*** (0.0151)	−0.1080* (0.0569)	−0.1104* (0.0571)
Lag Mfg Share			2.4795 (1.5217)	2.4163 (1.5250)	5.9684 (4.1524)	5.8948 (4.1725)
Lag Herfindahl			2.8065 (2.4463)	2.6595 (2.4566)	6.6494 (5.1936)	7.1718 (5.2142)
Lag External Links			−0.0003 (0.0098)	−0.0001 (0.0098)	−0.0236 (0.0232)	−0.0236 (0.0233)
Lag Patent Sum			0.0001 (0.0001)	0.0000 (0.0001)	0.0003* (0.0001)	0.0003** (0.0001)
Lag WoS per Capita			−0.0085 (0.0508)	−0.0100 (0.0508)	−0.0430 (0.1083)	−0.0630 (0.1087)
Time fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
Observations R^2 within Anderson LM-statistic Cragg–Donald F-statistic Sargan statistic	870 0.12	870 0.12	867 0.14	867 0.14	867 43.265*** 5.649*** 0.244	867 43.428*** 5.035*** 0.259

Note: See Table 5.

Balland, P. A., Boschma, R., Crespo, J., & Rigby, D. (2019). Smart specialization policy in the European Union: Relatedness, knowledge complexity and regional diversification. Regional Studies, 53(9), 1252–1268. https://doi.org/10.1080/00343404.2018.1437900

 Web of Science ®Google Scholar

Seidman, S. (1983). Network structure and minimum degree. Social Networks, 5(3), 269–287. https://doi.org/10.1016/0378-8733(83)90028-X

 Web of Science ®Google Scholar