High Technology Performance, Innovation, and Networks

1
High Technology Performance, Innovation, and
Networks

• - an Applied Econometric Analysis of Firms in
  Scottish High Technology Clusters

Dr. Vandana Ujjual
University of St. Andrews
SPRU, University of Sussex,
Sponsor Scot Econ Net
Feb.6th, 2009
2
Objectives

• Dynamic Innovation Process in Firms

• How Firm effectively utilised various
  Internal Resources,
• and

• External Demand Technological
  opportunities

• Further, it estimates the importance of
  these factors in
• impacting firm Performance

3
Specific Objectives

• To establish empirically the
• Link between the Innovation Input and the
  Innovation Output, Its two-way relation with the
  firm Performance,
• By solving for Selectivity and Simultaneity
  biases
• (Lööf Heshmati 2003 Janz et al 2003 Klomp
  Van Leeuwen 2002)

• To check the robustness of the applied
  econometric model to the ways of data handling
  and estimation methods

4

• Content
• The Scottish Context
• Policy Approaches- the case of Scotland
• Evolution of the High Technology Cluster
• Current High Technology Structure in Scotland

• Innovation Indicators and Measurement
  Issues
• Innovation and Performance- Literature
• Model and Hypothesis
• Estimation Results
• Conclusion

5
Rationale Key Issues

• Several Factors, Intrinsic to the High Technology
  
• - Have well-positioned several economies
  world-wide
• - Competing on innovativeness, Strong
  Performance
• - Start-ups, Potential to create jobs,
  Revive stagnant industries
• (Audretsch 1998 Brusco 1998 Becattini 1989
  Camagni 1991 Steiner 1998 Storper 1993 Best
  1990 Piore Sabel 1984).

Characteristics Features of Hi-tech
Kodama (1991) - Technological
diversification for survival - The
targeting of RD effort towards demand-side
initiatives - Simultaneous investment in
different stages of technological trajectories
- High degree of product obsolescence due to
rapid innovation cycle - Invisible
competitors - The ratio of RD to capital
investment is usually greater than one -
Technological fusion, the combining of existing
technologies in new ways.
6
Rationale Key Issues
UK - Innovation Review Commission, DTI (2004)
Highlighted Key issues on Innovation in Firms
(Pittaway et al. 2004 Leseure et al. 2004
Edwards et al. 2004)

• Firms Relationship with External
  Organizations, affects
• its ability to innovate and perform

• The evidence on the Adoption of promising
  practices

• Strategies for Value Creation, innovating
  to produce
• products or services generating more
  revenue

• Strategies to move higher up the value or
  supply chain,
• where the products services generate
  more value.

7
High Technology in Scotland

• Driven by a conscious design implementation
  of public policy
• (SE, SDA, Locate in Scotland)

• 275 foreign owned plants, 70 US owned,
  one-fifth of total
• employment, by 1980
• By 1990, Electronics accounted for 42 of
  manufactured
• exports, 20 of gross manufacturing output.
  Half of all UK
• exports of computers/peripheral

• Scottish Electronics less Embedded, (Turok
  1993 Young et al. 1988).
• - Exogenous knowledge generation
• - Without endogenous knowledge networking
• - Low local product demand
• - Few backward value-adding Linkages,
  (Molina Kinder 2001).

8
Scottish Enterprise - Key Initiatives
Smart, Successful, Scotland - A qualitative
shift from Low-cost advantages, Low
value-added Labour intensive to
Productivity, Competitiveness, Innovation (SE
2001)

• Linking foreign investors to local supply
  chains by deepening
• links with Local Research providers
  Supplier Development
• programs (Raines 2001 Brown Raines
  2000)

• Development of Technology Clusters,
• Capitalizing on Scotland's Research Base

• Market-driven Demand-led funding for
  commercially
• focused RD
• Alba Centre, Proof of Concept Fund,
• 3 Intermediary Technology Institutes,
  (ITI Techmedia, ITI Life Sciences ITI Energy)

9
Scottish Enterprises Cluster
Development Approach A Smart, Successful
Scotland Ambitions for the Enterprise Networks,
SE (2001)
10
General Strata Used (836 Firms)
SIC at 4-digit level (Butchart 1987)
Non-SIC based (DTI 2000, 2001)
Life Sciences 150 Biotech
85 Medical Devices
44 Support Services 21
Optoelectronics 90 Service
11 RD
10 Academic 8 Creative
Digital Media 189 Software
230
Microelectronics 187 Silicon mfrs.
7 Design
75 Product Development 26
Embedded Software 79

• Multi-stage sampling technique
• Cluster sampling, Stratified sampling
• (Scottish Enterprise 1998 -The Cluster Approach,
  Scottish Technology Industry Monitor 2002,
  Biotech Scotland- Framework for Action 2002)

11
PopulationSample, Sector-wise Distribution
Sample representative of the Hi-tech Population
(?2 tests significant at 5 ). All sectors
equally represented in the sample (?2 tests
significant at 5 ).
12

Innovative Non-innovative Sample

• 92 have introduced new products in the
  past 5 years
• or will introduce new products in the
  next 5 years
• Only 77 of the firms have innovation
  expenditures.

• Overlap of firms with innovation input and
  output
• 76 of the full sample,
• 82 of the sample with product
  innovation,
• 95 of the sample with innovation input.

• Not all innovative firms can attribute their
  innovation output
• to its internal innovation expenditures

• 17.27 involved in product innovation have no
  expenditures

13
Innovative Non-innovative Sample

• Innovative firms are significantly larger

• Export Oriented
• Greater Patent Intensity
• RD organised as a department, continuous
  RD

• Greater Customer Network links, followed by
  Research
• Greater emphasis on Product Innovation
  Strategies,
• followed by Demand Expansion Strategies
• Firms Internal Sources is important
• Significance of the obstacles, economic
  firm specific
• Performance not significantly different

14
However, these simple mean comparisons cannot be
taken as evidence of the impact of
innovation input and output on
performance, As this requires controlling for
certain other variables like cooperation,
strategy etc., and the joint impact of the other
variables in a multivariate analysis.
15
Dynamic Innovation Process

• Broaden the collective understanding ,
  measurement
• Knowledge generating, Transmitting,
  Innovation process

• Beyond traditional innovation indicators-
  RD, Patent
• Oslo Manual - (OECD Eurostat 1992, 1997)

• Community Innovation Surveys (CIS 2001, 2004)
  
• New, internationally harmonised survey
  data, in 1993
• (Klevorick et al, 1995 Pavitt et al, 1987
  Klomp Van Leeuwen 2001)
• Innovation Input Stage, Throughput Stage,
  Output Stage
• (Klomp 2001 Acs Audretsch 1988, Geroski
  1990 Griliches 1995)

• Links between Different Stages of Innovation
  
• (Love Roper 2001 Oerlerman Meeus
  2002 Oerlermans et al
• 2001 Fischer Varga 2002 Belderbos et
  al 2004)

16
RD, Innovation Productivity

• Cobb-Douglas Model, common empirical approach
  (Griliches 1986,
• 1995 Jaffe 1986 Sassenou 1988 Hall
  Mairesse , 1995 Wakelin, 2000)

• Knowledge Production Function, (Pakes
  Griliches 1984)
• Transformation of Innovation Input to
  Innovation Output.
• It relates RD to Patents, (Griliches 1990)
  or
• RD to Innovations, (Klinknecht Mohnen
  2002 )

• Added a Spatial Aspect (Acs et al., 1994
  Oerlermans et al., 2001
• Fischer Varga, 2002 Belderbos et al.,
  2004)

• CDM (Crepon, Duguet and Mairesse, 1998)
• First attempts to explain productivity by
  innovation output,
• and innovation output by research
  investment in a
• structural model data for French
  manufacturing firms

17
Simultaneity Bias

• Knowledge production function has to be
  estimated
• not as a Single equation but as a System of
  equations

As the process in transforming new ideas to
innovation output or performance is complex
(Pakes Griliches 1984)

• When several links are considered in a
  Simultaneous
• equation framework some Explanatory
  variables very
• often not be safely assumed to be
  exogenously given

• Endogenous Variable Variables that are
  determined
• within the System. Correlated with the
  error term

18

• Innovation input is an explanatory variable
  in the
• Innovation Output equation, and
• Innovation Output is one of the explanatory
  variables in
• the Performance equation.

• Because of the endogeneity of these
  variables, it cannot be
• assumed that the explanatory variables and
  the disturbances
• are uncorrelated

• As a result, an Ordinary Least Square
  regression applied
• will be biased and inconsistent

19

• CDM Model accounts for both Selectivity and
  Simultaneity Bias

• Klomp van Leeuwen (1999, 2001)
• Extended the model by using alternative
  specifications, estimation procedures, and
  innovation and business performance measures
  simultaneous equation system data for Netherlands

• Loof Heshmati, (2001, 2002)
• Different model specifications and estimation
  techniques splitting the model in 2 parts and
  applying the IV approach only to the second part
  data for Sweden

20
Sample Selection Bias

• Innovative Firms
• 80.13 with Innovation input, innovation
  expenditures
• 92 with innovation output (new products in
  the past 5 years)

• 17.3 of these firms with innovation output
  have no
• Innovation Expenditures

• Not all innovative firms can attribute their
  innovation
• output to its internal innovation
  expenditures. This model
• takes into account that not all firms are
  innovation firms

Selection Bias - When only the innovation sample
is used in some parts of the model, the firms are
not randomly drawn from the population
21
Sample Selection Bias

• Key variable innovation expenditures zero for
  19.87
• Missing variables problem encountered
• Performance variable missing for 19.87 of the
  firms
• Computation of the performance variable,
• Performance Ratio of turnover to
  innovation expenditures

Introduces a Selection equation in the model-
Determines the Factors influencing a firms
propensity to Innovate
22
Applied Econometric Model
Two-Stage, Four-equations Knowledge Production
Function Model
First Stage- Selection Equation (1) To Innovate
or Not?
Second Stage - Three established relationships

• Innovation Investment, Equation (2)
• Links Research to its Determinants

• Knowledge Production Function, Equation (3)
• Relates innovation output to innovation input

• Performance, Equation (4)
• Relates innovation output to performance

Each has its idiosynchratic common determinants
Non-innovative Firms retained used in the
Selection equation to estimate the selection
corrected variable, IMR (Heckman 1979)
23
Conceptual Framework of the Model
24
Applied Econometric Model
Two-Stage, Four-equations Knowledge Production
Function Model

• First stage -Selection equation (1)
• Determines the factors influencing a
  firms propensity to
• innovate (innovation expenditures), is
  estimated separately.

• Second stage involves the last three
  equations,
• Estimated in a simultaneous equations
  framework.

• Heckmantype sample selection bias corrected
• Iterative 3 stage least squares (I3SLS)
  Estimation

• Allowing the endogeneity of Innovation Input,
• Innovation Output and Performance.

• Solves the problem of sample selection and
  simultaneity

25
Model Specification
1 if
gi x1ib1 u1i gt 0 with innovation
exp. 1. Selection Eq. gi
0 if gi
x1ib1 u1i lt 0 no innovation exp. 2.
Innovation Input ki a2qqi
x2ib2 IMR u2i 3. Innovation Output
ti a3kki a3qqi x3ib3 IMR
u3i 4. Performance qi
a4tti x4ib4 IMR u4i
ki ,ti and qi Endogenous, x1i , x2i ,
x3i and x4i Exogeneous
26
Contributions

• One of the few studies to Estimate the
• Causal effect of Innovation Investment on
  Innovation Output,
• Causal effect of Innovation Output on firm
  Performance in both Manufacturing Service
  industries, using a Simultaneous equations
  framework
• (Ebersberger Lööf 2005 Benavente 2006 Loof
  2004 Loof Heshmati 2002)

• Traditional analysis of the link between RD
  Performance extended and developed using

• New, firm-level data from primary sources
• A novel Structural Economic Framework that
  Endogenize
• Innovation in explaining performance
  heterogeneity

27
Contributions
3. The results provide insights into the

• Dynamic innovation process in High Technology
  firms
• The effect of firms External Networks
• Demand-pull Technology-push factors.

4. It captures the direct impact of these
factors on

• Each of the different stages of the
  innovation process
• (innovation decision, innovation input,
  innovation output),
• How these factors indirectly impact
  performance?

28
Contributions
Measurement of all stages, from RD, Patenting,
New Product Sales (Hagedoorn Cloodt 2003)

• Propensity to Innovate - Innovation Input
  Decision

• Innovation Intensity - Innovation
  Expenditure per Employee

• Innovation Output - Ratio of New
  Product Sales in the
• 
  past five years, to Total Sales

• Patent Intensity - Patents per
  Employee (granted filed in
• 
  the last five years)

• RD Organisation - RD Department

• Performance - Returns from
  innovation investments

29
KEY QUALITATIVE VARIABLES
- Increased productivity (cost-push) -
Improved products (technology-push) - Increased
or retained market share (demand-pull)

• Strategies

• - Internal (RD staff, marketing staff)
• Market (customer, supplier, competitor)
• Educational Public (govt., universities)

• Sources of
• Knowledge Spillover

• Customer alliances per employee
• Research alliances per employee
• Government alliances per employee

• Network Intensity

• Obstacles to
• Innovation

- Economic (cost, finance, pay-off uncertainty)
- Firm specific (lack of skilled personnel)
30
Causality Structure of the Model
X
31
Demand-pull, Cost-push Technology-push
(Schmookler 1966 Rosenberg 1974 Mairesse
Sassenou 2003 Benavente 2006 )

• Emphasize on Demand-side variables
• (Griliches1957 Schmookler 1962,1966
  Scherer 1982)

• Technology-push to offer radical product
  innovation in contrast
• to focussing on an existing Market-pull
  (Prajago Ahmed, 2006
• Scherer 1965 Phillips 1966 Parker
  1972 Rosenberg 1974)

• Both the demand-pull differences in
  technological opportunity,
• must be taken into account for an adequate
  conception of how
• technological change occurs (Scherer 1982)
  
• Firms need to maintain a balanced focus on
  both strategies
• (Betz 1998 Cohen Levin 1989 ).

32
Innovation Networks

• Impact of Networks on Innovation Input
  separately
• from the Impact on Innovation Output
  stage

• Differences in impact of these same network
  variables on the
• Innovation Input Output

• Firms objectives factors driving firms to
  collaborate
• (Criscuolo Haskell 2003 Janz et al.
  2003 Vanhaverbeke et al. 2004)

• Direct Impact on the innovation Stages,
  Indirect Impact
• on Performance

• By controlling for Impacts of different
  Knowledge Spillovers, (Zucker et al. 1998a, b
  Feldman 2001), Internal Innovation Effort (Acs
  et al 1984 Audretsch Feldman 1996b Fischer
  Varga 2002 Oerlermans et al 2001 Oerlerman
  Meeus 2002 Belderbos et al 2004).

33
Embeddeddness

• Knowledge flows are highly localized
  (Gertler Levitte 2005)
• New product development linked to the
  University Star Scientists
• in their geographical area (Zucker et
  al. 1998a).
• Local links critical for developing basic
  research (Niosi 2000b),
• research capability (Cooke 2002a)

• Local VC- Commercial Success in Biotech
  firms, Key inputs
• (Zucker et al. 1998b Gertler Levitte
  2005 Cooke 2001b Powell et al. 2002)

• Early stages of development more
  science-focused, unlike firms in
• advanced commercial stages (Powell et
  al. 2002 Cooke 2002a).
• Though research activity is not very
  highly concentrated spatially,
• commercialisation efforts are (Cortright
  Mayer 2002)

34
Estimation Methods

• Comparison of the Estimation results of the
  Second Stage
• Simultaneous equations model (I3SLS) ,
  accounting for
• Endogeneity, and correcting for sample
  selection bias,
• Single Equation Estimation, (2SLS), which
  estimates one
• equation at time, but corrects for
  Endogeniety
• Tobit Esimation which ignores the inherent
  simultaneous
• bias and interdependence of innovation
  input, innovation
• output and performance

All corrects for Sample Selection Bias
35
Key Findings

• Results highlight the Sample Selectivity
  issue

• The (relative) importance of variables
  referring to the external
• environment firm-specific innovation
  characteristics
• diverges from
• The estimated impacts of the
  Single-equation approach
• when taking into account the Simultaneous
  nature of the
• variables, (Klomp van Leeuwen 2001).

36
(No Transcript)
37
(No Transcript)
38
Innovation Input Equation- 2
Research network links greatly increase
the internal innovation
capabilities (Zucker et al. 1998 Belderbos et
al. 2004 Tether 2002 Monjon
Waelbroeck 2003).

• Supports the Absorptive Capacity
  argument, that, in order
• to assimilate the scientific
  and technological information
• received from universities and
  research institutes, it is
• necessary to have increased
  innovation capability in firms,
• (Zahra George 1999 Feldman
  2001 Prevezer 2001)

It conjectures that cooperation with
science requires higher internal RD
skills, thus higher innovation expenditures
(Klomp Leeuwan 2001 Zucker Darby
1996).
39
Innovation Input Equation 2
A Complementary mix of Internal and External
resources, (Freel 2003, 2002
Zahra George 1999 Cohen Levinthal 1990
Dodgson 1993).
Internal strengths - Continuous,
routinised, long term, pooling resources,
organising as RD department (Klomp van
Leeuwen 1999), - Greater patent intensity,
rather than market share strategies
(Hadjimanolis 2000 Nelson 2000 Oerlemans et al
1998).
External opportunities - Research
networks, - Taking advantage of the
spillover benefit from market sources like
customer, supplier and competitors
40
Innovation Input Equation 2
Technology-push factors are very important.
It is the firms radical innovation
attempts in order to develop innovative
product and patents that determine the innovation
input intensity of hi-tech firms (Zucker et al.
1998). Continuous RD, organised as a
department, and patenting are the factors that
increase the innovation input. Other attempts
such as increasing market share or expanding
internationally does not increase it.
41
Innovation Output Equation 3
The innovation sales of hi-tech firms are
determined by the Demand-pull factors
Specifically, Market extent of the firm,
Customer collaboration and Knowledge spillovers
from Market (von Hippel 1988 Tether 2002)
The innovation output significantly greater for
exporting firms and firms with intense customer
networks (Kristensson et al 2002 Prahalad
Ramaswamy 2000 von Hippel et al 1999Wikstrom
1995 )
Hi-tech firms that allocate almost all of
their innovation expenditures on RD has
less resources dedicated to Near Market
Non-RD activities
This explains why higher innovation input
intensity does not result in commercialisation of
innovation and innovation sales Benavente (2006)
42
Innovation Output Equation 3
Only Firms that overcome obstacles are successful
in achieving commercial success
Positive impact of Research alliance on
innovation output is not over and above
its impact on innovation input.
Negative impact of government alliance on
innovation output is not over and above
its effect on innovation input.
43
Conclusions

• Hi-tech firms with aggressive innovation
  strategies and
• international markets, still find it
  very important to be
• embedded in local networks, which in
  turn increase their
• performance (Gertler Levitte 2005
  Zucker et al. 1998a).
• In this way this research supports the
  fundamental ideas of
• the Innovative Milleu concept (Aydalot
  1986 Keeble1998 Lawson
• Lorenz 1999 Camagni 1995 Malliat
  1995)
• Spatial concentration enables face-to-face
  networking,
• Common labour markets and the Diffusion
  of knowledge,
• in particular the tacit knowledge that
  are difficult to
• codify (Maskell et al 1998 Nooteboom
  19 Lawson Lorenz 1999).

44
Conclusions

• The firms doing innovation on a continuous,
  routinised organised
• as a department, not only has greater
  probability of having
• innovation investments but also has
  significantly greater
• innovation input intensity.
• (Klomp van Leeuwen 1999 Meeus
  Oerlemans 2000 Ebersberger
• Lööf 2005).

• Firms with greater patent intensity has less
  incentive to invest in
• innovation expenditures, but once they
  decide to invest, the
• innovation expenditure intensity rises
  with patent intensity

45
Conclusions

• The likelihood of innovating increases more
  than
• proportionally with firm size, but after
  controlling for
• networks, Innovation input intensity is not
  influenced by
• firm size for firms involved in RD (Loof
  Heshmati 2002)
• Larger hi-tech firms have higher propensity
  to invest in
• innovation, and larger firms have a higher
  performance,
• in terms of returns to innovation
  investment.

• Findings confirm the Schumpeterian view of
  innovation,
• as an activity undertaken by larger
  monopolistic firms.
• However, size does not increase the
  innovation intensity,
• neither Innovation Input Intensity, nor
  Innovation Sales

46
Policy Implications

• The hi-tech firms unable to achieve
  commercial success from
• their innovation
• Importance of the Near Market activities of
  Innovation

• Challenge- Lack of Firm specific resources,
  Internal staff,
• Commercial capability Necessary funds
  for Non-RD
• activities, for successful completion and
  launch

• Greater role by Government Org. in creating
  an environment
• for Commercialisation of research
• Industry-government links, bridging, other
  Local
• Intermediary agencies to bring together
  organisations with
• Complementary Capabilities to increase New
  product sales,
• Legal uncertainties involved in IP when
  firms collaborate on
• innovation

47
Thank You
Acknowledgements
ScotEcon.Net for Funding the
Project
University of St. Andrews, where the
project was Undertaken