Innovative Place-Based Triple Helix Approaches for Regional - - PowerPoint PPT Presentation

Enh nhancin ing Inno Innovativ ive Cap apabil ilit itie ies in in Per erip ipheral Reg egio ions: An n Ext xtra-Regio ional Col

• lla

laborativ ive Appr pproach to

• RIS

RIS3 Dr Phil Tomlinson, Deputy Director, Centre for Governance, Regulation & Industrial Strategy, School of Management, University of Bath Panel: Multi-level Governance and Smart Specialisation: A View from the Regions

Innovative Place-Based Triple Helix Approaches for Regional Development through Smart Specialisation Strategies 28-29 June, 2019 St Mary’s University, Waldegrave Suite, UK

• S3 is the major component in EU’s ‘Innovation Union’ programme and 2014-

220 Cohesion Policy

• ‘Place-Based’, ‘bottom up’ process, ‘entrepreneurial discovery’ linked to

regional competences, capabilities and knowledge bases

• S3 logic naturally favours dynamic regions; entrepreneurial & technological

dynamisms, good networks & absorptive capacities. And ‘good governance’. Opportunities for Industry 4.0.

• Lagging Regions at a distinct disadvantage; ‘hollowed out’ manufacturing

bases, weak networks, low knowledge/skills bases How might policy (RIS3) be utilised to turn the perceived potential of S3 into reality for lagging regions?

Overview

• We focus upon the potential role of extra-regional collaborations as a means to

facilitate S3 in lagging regions:

• 1. Technological Diversification and Relatedness is required for S3 - lagging

regions generally have low diversity of sectors and lack critical mass for the cross-fertilisation of ideas between adjacent technological domains and sectors, hence extra-regional links are likely to help (Boschma, 2015; Boschma and Iammarino 2009; McCann and Ortega-Argilés, 2015)

• 2. Technological Upgrading: links between lagging regions and core regions at

the technological frontier (McCann and Ortega-Argilés, 2015)

• 3. Entrepreneurial discovery: evidence that firms innovate and develop new

specialisations by connecting both within and outside the region (Bathelt et al., 2004; Boschma, 2014; Belussi et al., 2010; Boschma and Ter Wal, 2007)

• Existing studies are focused on specific sectors or specific countries / territories,

no systematic evidence → aim of this study

Research Focus

Data and variables

Variable Construction Description

PINT

log ෍

𝑗 𝐽

𝑂𝑠𝑢/𝑄

𝑠𝑢,𝑗

Patents per 100,000 pop PGROWTH

Pgrowthrt = log 𝑄𝑠𝑢 + 3 − log(𝑄

𝑠𝑢)

Patent growth over 3 years SHARE_COLL

𝑇𝐼𝐵𝑆𝐹_𝐷𝑃𝑀𝑀𝑠𝑢 =

𝐷𝑠𝑢 𝑂𝑠𝑢

Extra-regional collaboration REL_COLL

𝑆𝐹𝑀_𝐷𝑃𝑀𝑀𝑠𝑢 = log ෍

𝑗 𝐽

𝑄

𝑠𝑢,𝑗𝐷𝑝𝑚𝑠𝑢,𝑗

Similarity of Knowledge in Extra-Regional Collaboration (Boschma & Iammarino (2009), Miguelez & Moreno (2018))

K_STOCK

𝐿_𝑇𝑈𝑃𝐷𝐿𝑠𝑢 = 𝑂𝑠𝑢 + 1 − 𝜀 𝐿_𝑡𝑢𝑝𝑑𝑙𝑠𝑢 − 1

Tech Absorptive Capacity (Proxy)

TECH_DIV

𝑈𝐹𝐷𝐼_𝐸𝐽𝑊

𝑠𝑢 =

𝑂𝑠𝑢 𝑂𝑠𝑢 − 1 1 − ෍

𝐽 𝐽

𝑞𝑠𝑢,𝑗 σ𝑘

𝐾 𝑇𝑗𝑘𝑞𝑠𝑢,𝑘

𝑂𝑠𝑢

2

Inverse Herfindhal index weighted by relatedness across IPC classes (Corradini and De Propris, 2015)

Controls ∑𝑎𝑠𝑢−1

GDP per capita, Population Density, Education

PATSTAT-CRcovers1999 – 2013. It is based on patent applications for 285 NUTS2 regions, using a fractional IOS Dataset count of inventors to determine location at the NUTS2 regional level. A 5 digit IPC classification (approx. 650 IPC classes) is used. Time is identified using the priority date.

Share of extra-regional collaboration (4 quantiles) 1985-1990 Share of extra-regional collaboration (4 quantiles) 2005-2010

Model & Estimation

To assess how extra-regional collaboration and relatedness in external collaboration affect regional innovation performance (Yrt), we define: 𝑍

𝑠𝑢 = 𝛾0 + 𝛾1𝐿𝑡𝑢𝑝𝑑𝑙𝑠𝑢−1 + 𝛾2𝑇ℎ𝑏𝑠𝑓𝐷𝑝𝑚𝑚𝑠𝑢−1 + 𝛾12𝐿𝑡𝑢𝑝𝑑𝑙𝑠𝑢−1𝑇ℎ𝑏𝑠𝑓𝐷𝑝𝑚𝑚𝑢𝑠−1

+𝛾3𝑆𝑓𝑚𝐷𝑝𝑚𝑚𝑠𝑢−1 + 𝛾13𝐿𝑡𝑢𝑝𝑑𝑙𝑠𝑢−1𝑆𝑓𝑚𝐷𝑝𝑚𝑚𝑢𝑠−1 + 𝑎𝑠𝑢−1 + 𝜀𝑠 + 𝜀𝑢 + 𝜗𝑠𝑢 To test the differential effect of SHARE_COLL and REL_COLL for lagging regions, the above equation includes interaction terms for both variables with K_STOCK, allowing us to explore their impact across the distribution of knowledge capabilities of the regions in the dataset. Estimated using FE with cluster robust standard errors, ML with country dummies to exploit between-variation and System GMM.

Regional PINT Regional PGROWTH FE ML Sys-GMM FE ML Sys-GMM (1) (2) (3) (4) (5) (6) L.DEPVAR 1.146***

• 0.173**

(0.114) (0.075) K_STOCK 7.253*** 7.303*** 17.026*** 0.337*** 0.264*** 1.820*** (1.019) (0.406) (3.498) (0.124) (0.045) (0.509) SHARE_COLL 15.920*** 14.003*** 132.926*** 4.417*** 3.269*** 7.157*** (2.735) (1.931) (45.576) (0.471) (0.247) (1.810) K_STOCK X SHARE_COLL

• 4.012***
• 3.589***
• 12.016*
• 0.672***
• 0.465***
• 1.023**

(0.776) (0.455) (6.717) (0.100) (0.057) (0.520) REL_COLL

• 0.778***
• 0.954***

1.240

• 0.331***
• 0.128***
• 0.753***

(0.277) (0.190) (1.858) (0.052) (0.022) (0.245) K_STOCK X REL_COLL 0.207*** 0.270***

• 1.082***

0.040*** 0.007** 0.078* (0.060) (0.030) (0.228) (0.008) (0.003) (0.045) TECH_DIV 20.315*** 20.455***

• 23.878

1.918*** 0.968*** 4.888** (3.254) (2.326) (59.488) (0.491) (0.309) (2.199) TECHDIV X TECHDIV

• 15.355***
• 15.379***

61.744

• 1.136***
• 0.409*
• 3.123

(2.653) (2.095) (64.477) (0.439) (0.279) (2.216) ln(GDP)

• 1.318
• 4.671***

3.618 1.351*** 0.217***

• 0.752

(1.060) (0.566) (5.243) (0.203) (0.047) (0.945) EDUC

• 0.073
• 0.071**
• 0.421**
• 0.021**
• 0.015***
• 0.225***

(0.059) (0.032) (0.203) (0.009) (0.003) (0.057) ln(PDENS)

• 10.500**
• 0.767**
• 4.313*
• 2.245***
• 0.082***

0.241 (4.505) (0.344) (2.343) (0.685) (0.023) (0.604) _cons 42.356* 26.367***

• 93.698**
• 3.588
• 2.750***

2.242 (23.636) (4.675) (45.223) (3.826) (0.407) (7.103) N 3107 3107 3092 2608 2608 2584 N Groups 264 264 263 260 260 258 Hansen test Χ2, Prob > Χ2 119.23 (0.14) 86.91 (0.12) AR1 test, Prob > z

• 3.39 (0.00)
• 4.39 (0.00)

AR2 test, Prob > z 0.84 (0.39)

• 0.52 (0.61)

Average Marginal Effects for SHARE_COLL (a) and REL_COLL (b) across quantiles of K_STOCK The positive impact of extra-regional collaboration diminishes in stronger regions Stronger Regions seem to benefit more from extra-regional collaborations based upon technological relatedness (Maybe weaker regions should not overly focus

• n related technological collaborations)

Results II: Marginal Effects

Results III

• Extra-regional collaboration, is positively associated with patent per capita &

patent growth.

• Extra-regional collaboration appears more important for lagging regions (and

actually diminishes in stronger regions)

• Building Extra-Regional collaborations based upon technological relatedness

may be less effective for lagging regions.

• The benefits for extra-regional collaboration are asymmetric!

Policy needs to recognise this in terms of incentive structures

Some Policy Implications I

1.Developing extra-regional collaborations to foster and support the entrepreneurial discovery across a more diverse set of opportunities within lagging regions. 2. Formal International Collaborative networks (e.g. H2020/Interreg) are essential for lagging regions, & regions involved are more active in promoting concrete cross-fertilization actions (e.g. Puglia Vs Campania); this type of project is also useful to increase institutional capabilities (institutional failure being typical in lagging regions) and improving governance structures. (note; lagging regions can have ‘pockets of excellence’ e.g. Cornwall; mining technologies and expertise (MIREU, REMIX, ¡VAMOS!).

• 3. Barriers:
• weak local networks/poor governance structures in lagging regions
• unwillingness of leading regions to collaborate
• Brexit Funding concerns

These barriers may be overcome, for example, by:

• Developing transferable skills to respond to regional sector-specific shocks and

increase absorptive capacity in lagging regions (e.g. ICT skills)

• Building

a broad spectrum

• f

extra-regional collaborations involving also medium/low-tech regions

Some Policy Implications II

Thanks for Listening Dr Phil Tomlinson, School of Management University of Bath