Innovation and Growth in the Lagging Regions of Europe. Neil Lee London School of Economics

Transcription

1 Innovation and Growth in the Lagging Regions of Europe Neil Lee London School of Economics

2 Introduction Innovation seen as vital for growth in Europe (Europa 2020) Economic growth Narrowing disparities Innovation highly uneven geographically Lagging regions may find it harder to translate inputs into innovation Does the impact of innovation on economic growth differ in different regions? Is innovation a viable strategy for lagging regions?

3 Growth theories and innovation Neo-classical model Endogenous growth theory New Economic Geography Evolutionary economics

4 Theoretical work on innovation Camagni (1995) suggested that lagging regions lacked the innovative milieu which helps create innovation Rodriguez-Pose (various) innovation prone and averse societies Social filter conditions innovation / growth relationship

5 Empirical evidence (1) Crescenzi (2005) European regions, Innovation (index) impact different according to regional conditions (peripherality / education) Sterlacchini (2008) European regions, Innovation (R&D) only impacts on growth beyond threshold GDP Impact differs Northern / Southern Europe Antonelli et al. (2011) Total factor productivity (TFP) in European regions Show an inverted u-shaped

6 Empirical evidence (2) Firm level evidence Frenkel (2000) local environments important and access to innovative inputs important in helping firms succeed Beugelsdijk (2007) firm level factors more important than local context Lee and Rodriguez-Pose (under review) context can matter - creative employment related to innovation activity in local firms

7 Gaps Evidence often dated e.g. Sterlacchini (2008) considers Much more evidence on drivers of innovation (e.g. Crescenzi and Rodríguez-Pose, 2013) than link innovation / growth Mostly focused on cross-sectional evidence

8 The data NUTS2 regions Eurostat Regio database Countries only included if 7 + regions (to allow calculation of lagging ) Sample: unbalanced panel of 191 regions, 12 countries

9 Defining lagging regions Lagging regions Less than 75% of national average GDP in the based year (1998) 52 lagging regions, from 191 (27%) Alternative measures Structural funds Peripherality

10 Sample of regions by country Country Not-lagging Lagging Total Belgium 8 (73) 3 (27) 11 Czech Republic 6 (75) 2 (25) 8 Germany 25 (73) 9 (27) 34 Greece 9 (69) 4 (31) 13 Spain 14 (73) 5 (27) 19 France 16 (73) 6 (27) 22 Netherlands 9 (75) 3 (25) 12 Poland 11 (69) 5 (31) 16 Portugal 5 (71) 2 (29) 7 Romania 6 (75) 2 (25) 8 Sweden 6 (75) 2 (25) 8 UK 24 (73) 9 (27) 33 Total 139 (73) 52 (27) 191 Parenthesis give share of regions by country.

11 Defining innovation Patenting Accounts for technological frontier But biased by sector Only 6 percent of innovation active firms patent (Hall et al. 2013) Variable: Total patents (linear interpolation) Research and development. An input rather than an output Biased by sector Also can be used to assimilate innovations from elsewhere Variable: Total R&D expenditure (log) [not per GDP/ capita] Missing values filled with linear interpolation

12 The model Y it = α + β 1 Innovation it + β2 Employment it + β3 HighSkill it + β4 Manufacturing it + β5 Youth it + ε it GPD pc GDP per capita, natural log Employment - Total employment (1,000s), natural log Manufacturing - Manufacturing employment as % of total Skilled workers - Share of those in employment with first and second stage of tertiary education ( Young workers - Share of labour force aged 15-24, natural log Estimated as fixed effects panel regression.

13 Does innovation have different effects in lagging regions?

14 Panel estimation: Patenting & regional growth (1) (2) (3) (4) (5) Dependent variable: GDP per capita, natural log Patents per capita (ln) *** *** *** *** (0.0158) (0.0122) (0.0134) (0.0182) Patents per capita (ln) * Non-lagging *** (0.0225) Patents per capita (ln) * Lagging *** (0.0183) Total employment (ln) (0.129) (0.317) (0.124) Manufacturing (%) * ** ( ) ( ) ( ) Skilled workers (%) 0.242*** 0.322** 0.224** (0.0728) (0.149) (0.0919) Young workers (%) *** ** *** (0.0593) (0.118) (0.0728) Constant 9.295*** 9.280*** 9.775*** 10.49*** 9.097*** (0.0655) (0.0740) (0.837) (2.066) (0.815) Sample Full Full Full Lagging Non-lagging Observations 1,451 1,451 1, ,053 R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

15 Panel estimation: R&D & regional growth (1) (2) (3) (4) (5) Dependent variable: GDP per capita, natural log R&D (ln) 0.229*** 0.173*** (0.0473) (0.0423) R&D (ln) * Non-lagging 0.248*** 0.248*** 0.188*** (0.0535) (0.0535) (0.0481) R&D (ln) * Lagging 0.195*** 0.195*** 0.147** (0.0652) (0.0652) (0.0599) Total employment (ln) (0.118) (0.118) Manufacturing (%) ** ** ( ) ( ) Skilled workers (%) 0.171** 0.176** (0.0724) (0.0713) Young workers (%) *** *** (0.0573) (0.0568) Constant 8.493*** 8.427*** 8.808*** 8.427*** 8.603*** (0.241) (0.237) (0.694) (0.237) (0.692) Sample Full Full Full Full Observations 1,013 1,013 1,013 1,013 1,013 R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

16 Panel estimation: R&D & regional growth (1) (2) (3) (4) (5) Dependent variable: GDP per capita, natural log R&D (ln) 0.229*** 0.173*** (0.0473) (0.0423) R&D (ln) * Non-lagging 0.248*** 0.248*** 0.188*** (0.0535) (0.0535) (0.0481) R&D (ln) * Lagging 0.195*** 0.195*** 0.147** (0.0652) (0.0652) (0.0599) Total employment (ln) (0.118) (0.118) Manufacturing (%) ** ** ( ) ( ) Skilled workers (%) 0.171** 0.176** (0.0724) (0.0713) Young workers (%) *** *** (0.0573) (0.0568) Constant 8.493*** 8.427*** 8.808*** 8.427*** 8.603*** (0.241) (0.237) (0.694) (0.237) (0.692) Sample Full Full Full Full Observations 1,013 1,013 1,013 1,013 1,013 R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

17 Removing convergence countries (1) (2) GDPPC GDPPC Patents per capita (ln) * Non-lagging * ( ) Patents per capita (ln) * Lagging (0.0113) R&D (ln) * Non-lagging * (0.0148) R&D (ln) * Lagging (0.0276) Constant 7.269*** 7.085** * (0.323) (0.365) Controls Yes Yes Observations 1, R-squared Estimated Number as of fixed NUTS2 effects panel regressions. All regressions 144 include 138 year dummies and full controls. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

18 Removing convergence countries (1) (2) GDPPC GDPPC Patents per capita (ln) * Non-lagging * ( ) Patents per capita (ln) * Lagging (0.0113) R&D (ln) * Non-lagging * (0.0148) R&D (ln) * Lagging Impact larger in non-lagging regions Impact in lagging regions larger, but higher standard errors (0.0276) Constant 7.269*** 7.085** * (0.323) (0.365) Controls Yes Yes Observations 1, R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies and full controls. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

19 Diminishing returns to innovation?

20 Diminishing returns to innovation (1) (2) (3) (4) GDP per capita, natural log Patents per capita (ln) *** *** (0.0193) (0.0144) Patents per capita (ln) ( ) ( ) R&D 0.610*** 0.517*** (0.0656) (0.0661) R&D *** *** ( ) ( ) Controls No Yes No Yes Constant 9.640*** 9.733*** 8.097*** 7.289*** (0.0689) (0.818) (0.208) (0.577) Observations 1,451 1, R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

21 Diminishing returns to innovation (1) (2) (3) (4) GDP per capita, natural log Patents per capita (ln) *** *** (0.0193) (0.0144) Patents per capita (ln) ( ) ( ) Patenting is associated with growth, but little evidence of non-linearities R&D 0.610*** 0.517*** (0.0656) (0.0661) R&D *** *** ( ) ( ) Controls No Yes No Yes Constant 9.640*** 9.733*** 8.097*** 7.289*** (0.0689) (0.818) (0.208) (0.577) Observations 1,451 1, R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

22 Diminishing returns to innovation (1) (2) (3) (4) GDP per capita, natural log Strong evidence of diminishing returns to R&D Patents per capita (ln) *** *** (0.0193) (0.0144) Patents per capita (ln) ( ) ( ) R&D 0.610*** 0.517*** (0.0656) (0.0661) R&D *** *** ( ) ( ) Controls No Yes No Yes Constant 9.640*** 9.733*** 8.097*** 7.289*** (0.0689) (0.818) (0.208) (0.577) Observations 1,451 1, R-squared Number of NUTS Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

23 Link between R&D and patents

24 Impact of R&D on patenting R&D (ln) Not lagging (1) Patents (ln) 0.312* (0.167) R&D (ln) Lagging 0.412*** (0.121) Constant 2.040*** (0.717) Obs 995 NUTS2 170 R Estimated as fixed effects panel regressions. All regressions include year dummies. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

25 Some conclusions (1) Innovation has less of an impact on growth in lagging regions This applies for both an input (R&D) and an output (patents) However, difference is marginal - impact is still positive in (most) lagging regions But impact may be more variable

26 Some conclusions (2) Evidence of diminishing returns of R&D Input into innovation, rather than output But would be expected to help absorb innovation from elsewhere Less evidence of patenting Output measure of innovation Patents indicate position on frontier Quantile regression evidence (not shown) suggests even impact Tentative evidence R&D may lead to innovations in lagging regions (but no controls for social filter )

27 Some conclusions (3) Does not imply innovation has no impact on economic growth in lagging regions Actually, if value growth more in lagging regions trade-offs may be finer Highlights the equity versus efficiency trade-off

28 Some limitations Lagging regions variable endogeneous? Lack of controls for convergence GMM-SYS / removing eastern europe Interpolated values but checks show unimportant