The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation: The upstream petroleum industry

Abstract

This article analyzes important changes in technological innovation in the upstream petroleum industry. It provides evidence that shifts in sectoral patterns of innovation over the petroleum industry's lifecycle from the 1970s up to 2005 were dependent on the dynamics of knowledge base complexity (KBC), a key dimension of an industry's technological regime. Accordingly, observed shifts in innovation patterns are understood to be the aggregated strategic response of industry innovators to changes in the technological regime. The article proposes a quantitative method for exploring KBC and Schumpeterian patterns of innovation, and interactions between the two at the industry level. As the industry evolved, its knowledge base moved to higher orders of complexity creating a shift in the Schumpeterian pattern of innovation. Increased KBC was found to alter Schumpeterian patterns from Mark I toward a 'modified' Mark II. Instead of coming predominantly from 'traditional' established oil operators, technological innovation was increasingly triggered by a new class of emergent integrated service companies - 'second tier' systems integrators of the upstream sector able to cope with increased KBC.

Full text

The effect of the dynamics of
knowledge base complexity on
Schumpeterian patterns of
innovation: the upstream
petroleum industry
Ali Maleki
1
, Alessandro Rosiello
2
and
David Wield
3
1
Research Institute for Science, Technology and Industrial Policy, Sharif University of Technology,
Teheran, Iran. a.maleki@sharif.edu
2
University of Edinburgh Business School and Innogen Institute, Edinburgh, United Kingdom.
alessandro.rosiello@ed.ac.uk
3
Development Policy and Practice and Innogen Institute, Open University, Milton Keynes,
United Kingdom. david.wield@open.ac.uk
This article analyzes important changes in technological innovation in the upstream petro-
leum industry. It provides evidence that shifts in sectoral patterns of innovation over the
petroleum industry’s lifecycle from the 1970s up to 2005 were dependent on the dynamics
of knowledge base complexity (KBC), a key dimension of an industry’s technological
regime. Accordingly, observed shifts in innovation patterns are understood to be the
aggregated strategic response of industry innovators to changes in the technological
regime. The article proposes a quantitative method for exploring KBC and Schumpeterian
patterns of innovation, and interactions between the two at the industry level. As the
industry evolved, its knowledge base moved to higher orders of complexity creating a shift
in the Schumpeterian pattern of innovation. Increased KBC was found to alter Schumpe-
terian patterns from Mark I toward a ‘modified’ Mark II. Instead of coming
predominantly from ‘traditional’ established oil operators, technological innovation was
increasingly triggered by a new class of emergent integrated service companies – ‘second
tier’ systems integrators of the upstream sector able to cope with increased KBC.
1. Introduction
This article addresses important changes in inno-
vation patterns in the upstream petroleum
industry from the 1970s to 2005, arguing that they can
be explained by the dynamics of knowledge base
complexity (KBC). A knowledge base is defined as
complex if it involves integration and combination of
1
V
C2016 RADMA and John Wiley & Sons Ltd

different scientific and technological disciplines and
requires a variety of competencies (such as R&D,
design, engineering, and production). We develop a
quantitative method to explore KBC and show that its
increase has shifted innovation patterns from a
broadly Schumpeter Mark I to a ‘modified’ form of
Schumpeter Mark II. The modification is that a higher
concentration of technological innovation was not
driven by ‘traditional’ established oil operators, work-
ing as ‘first tier’ systems integrators. Rather, it was led
by a new class of emergent integrated service compa-
nies – ‘second tier’ systems integrators within the
upstream segment of the industry able to cope with
increasing KBC.
The petroleum industry has a relatively long and
complex value chain, beginning with exploration and
production (E&P) of crude oil (upstream), to transport
and refining (midstream), and ending with refining and
retail (downstream). The upstream industry comprises
a set of complementary activities: oil and gas explora-
tion, together with heavy oil, condensates and tar
sands; developing reserves for extraction, production
over an extended lifetime; and finally decommission-
ing. It includes the business activities supporting and
supplying these main activities. It is important to study
KBC dynamics in the upstream segment because of
changes that have occurred in both market environ-
ment and the nature of technological knowledge
(Grant and Cibin, 1996; Helfat, 1997; Acha, 2002).
While the notion of technological regimes has
proved useful in explaining intersectoral differences
in sectoral patterns of innovation, the analysis of the
relationship between technological regimes and pat-
terns of innovation at different stages of development
of a given industry remains rather unexplored (Krafft
et al., 2014). Our research aims to show how the
changing nature of sectoral patterns of innovation is
intrinsically related to the dynamics of technological
regimes. We provide a threefold contribution: first,
we propose a dynamic reading of the concept of tech-
nological regimes and analyze structural transforma-
tion within the upstream petroleum sector over time.
Second, we put the notion of KBC at the centre of our
analytical framework. Third, we propose a quantita-
tive method using patent data to capture the dynamics
of KBC and its relationship with Schumpeterian pat-
terns of innovation. We empirically examine these
ideas in the upstream petroleum industry, focusing on
changes in technological opportunities and KBC.
Our study is unique in that it focuses on the
dynamic relationship between KBC and the evolution
of sectoral patterns of innovation. Most other works
on technological regimes and sectoral patterns of
innovation adopt a static cross-sectoral mode which
ignores the important role of change in the nature of
knowledge for industrial dynamics. Only the recent
study by Krafft et al. (2014) provides evidence regard-
ing the relationship between change in knowledge
base characteristics and industry structure in the phar-
maceutical industry.
We also deal with temporal variation of innovation
patterns at the sectoral level. Upstream petroleum
comprises a multitude of players, some of whom are
vertically integrated, and others who concentrate on
particular subsegments along the value chain. It also
comprises different types of operators and a range of
supply and service companies. Thus, as observed in
other sectors by authors such as Corrocher et al.
(2007), at each point in time the industry and/or some
of its segments may feature particular combinations
of Schumpeterian patterns. We focus our analysis on
exploring shifts in overall patterns over time, and ask
why they took place.
The article is organized as follows. Section 2
presents a dynamic reading of the concept of techno-
logical regimes and explores knowledge gaps in liter-
ature. In section 3 we introduce our method, showing
how we measure technological opportunities, KBC,
and dynamics of sectoral patterns of innovation. Sec-
tion 4 presents our results, focusing on dynamics of
technological opportunities, knowledge base com-
plexities, and Schumpeterian patterns of innovation.
We then discuss the relationship between Schumpe-
terian patterns of innovation and the dynamics of
KBC. Section 5 concludes.
2. Literature survey
We review three related bodies of relevant literature:
Schumpeterian patterns of innovation; dynamics of
technological regimes; and complexity.
2.1. Schumpeterian patterns
In the Schumpeterian tradition, the distinction between
Mark I and II has proved a useful analytical tool to dis-
tinguish varying sectoral patterns of innovation among
different industrial sectors. In this article, we ask
whether the Schumpeterian dichotomy is helpful in
understanding how and why patterns of innovation dif-
fer in the same industry over time. Our intuition is that
innovation patterns may gradually change as a result of
technical change and the associated division of labor,
creating a shift in the Schumpeterian pattern.
Schumpeter Mark I is characterized by creative
destruction, where new firms play a major role in
innovative activities and entry barriers are low. In
contrast, creative accumulation is the main character-
istic of Schumpeter Mark II, meaning that established
Ali Maleki, Alessandro Rosiello and David Wield
2R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

firms play a major role in technological activities
whilst it is challenging for new small innovators to
enter (Schumpeter, 1934, 1942). Authors such as Mal-
erba and Orsenigo (1996) and Malerba (2007) found
empirical support for the existence of these patterns.
Notwithstanding those key findings, these studies
suffer from two limitations. First, the methodology
employed does not allow for the observation of varia-
tions within a technological class or an industry,
because the analysis relies on aggregated data. Sec-
ond, it does not allow for the observation of temporal
variation in sectoral patterns of innovation within
industries, as the time dimension is removed. So,
while it is widely accepted that patterns of innovation
change over time, observations are based on average
behavior over time for specific technology fields or
industries (Malerba and Orsenigo, 1996).
These limitations have been partly addressed in
more recent studies. For example, Corrocher et al.
(2007) observed the coexistence of both Schumpeter-
ian patterns of innovation in the ICT industry.
Grebel et al. (2007) provided similar evidence, high-
lighting the coexistence of large diversified and new
technology firms within innovation networks in
knowledge intensive industries like biotechnology
and telecommunications.
Regarding the second limitation, Malerba and
Orsenigo (1996) explicitly acknowledged the possi-
bility of change in the nature of technological regimes
over the course of time:
‘Some of these features of knowledge may
change during the evolution of a specific sec-
tor or technology (degree of codification,
independence, and complexity)’ (p. 97).
Malerba (2004) also argued that analysis of the
knowledge base is key to developing an in-depth
understanding of the innovative dynamics within sec-
tors. Malerba (2006) added that ‘change in knowledge
and knowledge base [...] goes to the heart of the evo-
lution of the industries and of the factors affecting the
change in industrial structure’ (p. 14–15). However,
such change was conceived as very difficult to iden-
tify over significant periods of time even within single
sectors, let alone the identification of regularities
across a range of industrial sectors.
2.2. Technological regimes in a dynamic
perspective
The concept of technological regimes was introduced
by Nelson and Winter (1982), referring to the knowl-
edge environment in which firms operate, or in which
their problem-solving activities take place. More
recently, four building blocks were identified:
technological opportunity, the appropriability of inno-
vations, their cumulativeness, and knowledge base
properties (Breschi and Malerba, 2000). Technological
opportunities refer to the likelihood of innovation in a
particular sector resulting from a given investment in
search processes. Over the industry life cycle (ILC),
technological opportunities may significantly change,
suggesting their dynamic nature. The standard ILC
model assumes that opportunity conditions decrease as
industries mature (Klepper, 1996). However, statistical
analysis (McGahan and Silverman, 2001), case studies
of mature industries, and research on innovation in
low-tech industries (von Tunzelmann and Acha, 2005;
Hirsch-Kreinsen et al., 2006; Robertson et al., 2009)
show that this is not necessarily the case.
The properties of the knowledge base which shape
innovative activities constitute a synthetic framework
encompassing the degree of specificity, tacitness,
complexity, and independence. Specificity refers to
the scope of applications within a particular knowl-
edge domain. Tacitness refers to the extent to which
knowledge is not articulated in standard formats such
as blue prints. Degree of independence refers to the
extent to which knowledge that is relevant to innova-
tive activities can be separated. Like other dimensions
of technological regimes, these properties of the
knowledge base can change over time as a result of
new application, interindustry knowledge flow, codifi-
cation practices (Steinmueller, 2000), and new instru-
mentation or computational capabilities (Arora and
Gambardella, 1994).
Changes in one or more of these dimensions of
technological regime are likely to have important
implications for sectoral patterns of innovation (Mal-
erba and Orsenigo, 1996; Malerba, 2007). We test this
theory in upstream petroleum with particular empha-
sis on the dynamics of KBC. According to literature
on technological regimes (Malerba and Orsenigo,
1996; Breschi and Malerba, 2000), a knowledge base
is defined as complex if (1) it involves integration and
combination of different scientific and technological
disciplines and (2) requires a variety of competencies
(such as R&D, design and engineering, and produc-
tion) for innovative activities. So far, the role of KBC
has been addressed only by a few studies, including
Vale and Caldeira (2008) investigation of the foot-
wear industry and Iizuka (2009) account of structural
change in the Chilean salmon farming industry.
2.3. Complexity
Complexity has been defined in several ways. A key
concern of scholars writing on complexity (Wang and
von Tunzelmann, 2000) is the volume of
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 3
V
C2016 RADMA and John Wiley & Sons Ltd

interdependencies and degree of interaction between
the elements of a system. This specific notion of
knowledge complexity matters when:
‘the opportunities to generate new knowledge
are conditional on the identification and inte-
gration of the diverse bits of complementary
knowledge that are inputs into the knowledge
production process’ (Antonelli, 2003, p.507).
This kind of complexity shapes industrial dynam-
ics, since the recombination of both pre-existing and
new information is key to generating new knowledge
and introducing systemic innovations (Chesbrough
and Teece, 1996).
Knowledge indivisibility is the outcome of a pro-
cess whereby systemic knowledge serves new func-
tions which are not achievable by individual bits of
knowledge. In sectors with high levels of such com-
plexity, successful innovation is not possible without
a full understanding of the compatibilities of diverse
technologies. Because the source of this complexity is
often systemic innovation (Chesbrough and Teece,
1996), it is labeled systemic complexity.
3. Method
We combine three related methods: patent analysis to
measure the dynamics of technological opportunities;
measurements of KBC; and measurements of sectoral
patterns of innovation.
We analyze the transformation of sectoral innova-
tion systems in upstream petroleum using the Derwent
Innovating Index – the patent database which classi-
fies all upstream petroleum industry patents in class
H01. This class covers exploration, drilling, well serv-
ices and stimulations, production, and their subseg-
ments of the upstream petroleum industry. We rely on
the records of Derwent International Patent Families
(IPFs), which group similar inventions registered in
different territories, in order to avoid multiple count-
ing of the same invention registered in different coun-
tries. Patent counts are used as a proxy to capture the
dynamics of innovative performance.
Patent data is the only rigorously classified infor-
mation on technological innovation covering both
long time periods and a wide range of countries. The
advantages and limitations of patent data for the anal-
ysis of innovative activities is a widely discussed issue
within literature. It is particularly important to con-
sider its limitations, such as systematic biases in data
which may produce distorted results if not treated
properly. The main disadvantages include (Pavitt,
1985; Griliches, 1990): (1) Not all inventions are
legally patentable everywhere. The classic example is
software which in many countries is protected by
copyright. Moreover, the patenting scope may differ
from one country to another depending on their partic-
ular patent law; (2) Due to differing institutional struc-
tures in various countries which affect the length,
time, and level of protection, an inventor’s incentive
to file for patents vis-
a-vis use other forms of protec-
tion varies substantially; and (3) Propensity to patent
varies across industries. Propensity to patent can also
vary over time because of changing knowledge
dynamics (such as growth of software based technolo-
gies), the changing dynamics of the sector toward
service companies which are more reliant on their
tools and techniques, and changing competitive pres-
sures in the sector.
While patents are only imperfect measures of inno-
vation, our results are less affected because the con-
clusions in this study are based on the analysis of
trends rather than absolute levels of the variables.
Therefore, we do not expect imperfections to signifi-
cantly impact trend analysis.
3.1. Measurement of technological
opportunities
Following previous studies, such as Andersen (2005)
and Fai (2007), we use patenting growth rate to cap-
ture the dynamics of technological opportunities in
upstream petroleum. We employ variation in patent-
ing rate to examine how technological opportunities
change over time.
3.2. Measurement of knowledge base
complexity
We aim to understand how the level of KBC evolved
over the ILC in different periods and how major inno-
vators coped with its dynamics.
According to the definition introduced in section
2.3, proxies to measure complexity should consider
the links and interactions between different elements
of the knowledge base and capture the recombinant
nature of knowledge. In order to measure systemic
complexity, network representation of the knowledge
base is very relevant. According to this view (Krafft
and Quatraro, 2011; Saviotti, 2011), the knowledge
base has a co-relational structure comprised of nodes
and links between these nodes. Nodes are technology
classes and links represent relationships between tech-
nologies connecting nodes together. The measure of
systemic complexity should consider the structure of
relationships between different knowledge domains.
The dynamics of complexity are understood from
Ali Maleki, Alessandro Rosiello and David Wield
4R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

changes in the pattern, strength of linkages and inter-
actions between the nodes.
Network analysis indicators treat knowledge as an
integrated system in which both the building blocks of
the system (nodes) and their interactions (ties) are
investigated at the same time. This enables us to mon-
itor how the knowledge structure changes over time
when new technologies emerge, diffuse, and are inte-
grated into the system or the old ones expire, are aban-
doned or disconnected from the knowledge base
(Krafft and Quatraro, 2011).
Proxies to measure complexity consider the network
of links and interactions between different elements of
the knowledge base and capture the recombinant
nature of knowledge and its endogenous complexity.
The knowledge base has a co-relational structure com-
prising nodes and links between these nodes (Saviotti,
2011). Nodes are seen as technology classes, and links
represent relationships between technologies connect-
ing nodes together. The dynamics of complexity are
understood by changes in the pattern and strength of
linkages and interactions between the nodes. Social
Network Analysis (SNA) is employed to examine the
dynamics of systemic KBC in upstream petroleum.
Following Krafft et al. (2011), weighted average
degree of network centrality (WADC) was used to
measure the systemic complexity of the industry’s
knowledge base (see Appendix A).
When the speed of formation of new nodes out-
weighs the formation of links, the network becomes
less connected and systemic complexity (WADC)
decreases. In contrast, when the formation of new
links is swifter than the appearance of new nodes in
the knowledge network, network connectivity
increases (Saviotti, 2011), signaling the rise of sys-
temic complexity (WADC).
3.3. Measurement of sectoral patterns of
innovation
The indicators selected for the analysis of the dynam-
ics of sectoral patterns of innovation are based on pre-
vious studies (Malerba and Orsenigo, 1996; Breschi
and Malerba, 2000). They are: concentration of inno-
vative activities (C); the number of innovative firms
(F); share of new entries (NE) to the innovation sys-
tem in terms of the proportion of patents registered by
new innovators.
Although the variables of this intertemporal
research are similar to previous cross-sectoral studies,
their operational correspondence with archetypical
Schumpeterian patterns of innovation is interpreted
differently. Due to the dynamic nature of the analysis,
we are more interested in the variables’ trends than in
their values in cross-sectional designs. In other words,
our interpretation is based on relative change of varia-
bles over time, indicating whether at different points
in time upstream petroleum was moving closer to a
typical Mark I or Mark II type.
4. Results
Our results are compiled in five subsections: our peri-
odization of trends of technological opportunities; the
dynamics of KBC over those periods; the consequen-
ces of that complexity; the dynamics of Sectoral pat-
terns of innovation; and the resulting changes in
Schumpeterian patterns of innovation.
4.1. The trend of technological
opportunities
In Figure 1, we present the innovation trend in
upstream petroleum according to the number of patent
applications in the US Patent Office (solid line) which
reflects the trend of technological opportunities. The
dash-line shows the trend of total patenting in USPTO
at 1% scale to control for changes in the overall level
of patenting. That is, to examine whether observed
dynamics of innovation is a reflection of technology
push from other sectors, or the result of internal
dynamics within the upstream petroleum industry.
From Figure 1, we can identify three distinct peri-
ods of technological innovation over the last four dec-
ades. From the early 1970s until mid-1980s, we
observe a growing trend where the number of US pat-
ent applications almost doubled (p1). The second
period runs from 1984 to 1994, with a negative trend
in innovation (p2). The third period begins after 1994
where we see a surge in innovation (p3).
The first period, (early 1970s to mid-1980s), was
characterized by a growth and diversification strategy
Figure 1. The number of US patent applications over time.
[Colour figure can be viewed at wileyonlinelibrary.com]
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 5
V
C2016 RADMA and John Wiley & Sons Ltd

driven by investment in advanced technology to
expand supply sources in more challenging reservoirs.
The gap between supply and demand was widening.
Major oil operators lost their monopoly on global
reserves leading to the first and second oil shocks
(Grant and Cibin, 1996). The rise in patents can be
explained by upstream factors, including high oil pri-
ces. Such technological efforts were enormously suc-
cessful, bringing down E&P costs and increasing
reserve replacement ratios (Fagan, 1997). The stable
trend of total patenting (dash line) in this period con-
firms that the rise of innovation is not an overall
global innovation trend but was driven by upstream
industry-specific factors.
The second period (mid-1980s to the mid-1990s),
was characterized by oil price cuts and excessive supply
combined with stiff competition. Excessive supply was
partly due to new technologies able to access offshore
fields (e.g. in North Sea and West Africa) which trig-
gered strategic change in terms of greater static and
dynamic efficiency among operators. Cost cutting, spe-
cialization, and change of strategic focus emerged as
established trends (Grant and Cibin, 1996) so that tech-
nological innovation was no longer a top priority. The
main focus was on increasing flexibility and responsive-
ness to change (Weston and Johnson, 1999). Major oil
operators restructured while new smaller specialized
supply and service companies followed horizontal as
well as vertical integration strategies (Barreau, 2002;
Babusiaux et al., 2004). Excessive supply and low oil
prices acted as a disincentive for innovation. Thus, com-
pared to a 15% decline of patents in upstream petro-
leum, over the period we observe an increase in total
patent awards of more than 70% (Figure 1).
The third period has more complex dynamics. The
sharp upward trend in innovation from 1994 was
driven neither by higher oil prices (they did not
increase for a further six years) nor by external tech-
nology (see the dash line in Figure 1). On the contrary,
difficult access to oil fields particularly in OPEC
countries pushed operators to seek alternative sources.
As suggested by the rise of exploration and develop-
ment costs (U.S. EIA, 2011) and low reserve replace-
ment ratios (Bagheri and Di Minin, 2015), this
situation created an unprecedented demand for tech-
nological innovation, especially concerning E&P in
places such as ultra-deep waters in the Gulf of Mex-
ico. Technological advances are also reflected in the
rise of well completions (WRTG, 2008)
Harnessing new technological opportunities, how-
ever, was not a priority for international operators
who continued their quest for efficiency under pro-
longed low oil price conditions. In response to low
price and volatile environmental pressures, they cut
R&D investments, conducted mega mergers and
acquisitions for scale efficiencies, and outsourced a
wider range of activities to service companies.
Meanwhile, service companies increased their
R&D investments (Bagheri and Di Minin, 2015) to
meet new market conditions, as also reflected in their
increasing share of patents compared with declining
share of operators (Maleki, 2013). Growing supply
and service companies (such as Schlumberger, Halli-
burton, Baker Hughes, and Weatherford) gradually
began to provide a broad range of packaged services
to meet their client’s expanding needs for more ambi-
tious exploration and development projects (Barreau,
2002). As their ‘integrated solutions’ gained momen-
tum, supply and service companies cultivated project
management and integration capabilities, previously
the territory of major oil operators.
These trends nudged service operators toward
unprecedented consolidation (Barreau, 2002), an
organizational industry-wide response to new techno-
logical imperatives (Teece, 1976) that spurred a wave
of innovative solutions. Overall, the intensity of E&P
activities and their knowledge content significantly
increased over time, which led Rajan (2011) to
observe that ‘if all technological innovations pro-
duced by the oil and gas industry were added up, they
would probably rival NASA’s space program or the
Industrial Revolution’. (p. 11).
4.2. The dynamics of knowledge base
complexity
The dynamics of KBC in the upstream petroleum
industry are presented in Figure 2 using the WADC
measure. Systemic complexity shows a downward
trend over most of the first period (p1), which indi-
cates decreasing connectivity within the knowledge
Figure 2. Dynamics of knowledge base complexity in the
upstream petroleum industry. [Colour figure can be viewed at
wileyonlinelibrary.com]
Ali Maleki, Alessandro Rosiello and David Wield
6R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

network. This process is driven by a higher rate of cre-
ation of new nodes (or new technological classes),
compared to new links between new and existing
nodes (Saviotti, 2011). In this phase, the sector is
mostly in its random search period and exploration
strategy is dominant. Because the structure of the
knowledge base is changing and is not yet established,
both cognitive barriers to entry and the degree of
knowledge cumulativeness are relatively low.
Historically, this was the period of rapid technolog-
ical progress, when technologies like 3-D seismic and
horizontal drilling were first introduced. When prom-
ising new technological fields are explored, it takes
time for innovators to understand the relationships
between new and existing knowledge domains. The
introduction of new technologies may be expected to
create new but poorly connected nodes, and temporar-
ily reduce the systemic connectivity of the knowledge
network (Saviotti, 2011). The first period in upstream
petroleum reflects this hypothesis.
The situation began to change when the direction
of systemic complexity reversed in the beginning of
the second period (p2) around 1986, as connectivity
within the knowledge network increased. This trend
continued almost up to the end of p3. The diffusion
and establishment of new technological fields
explored in p1 helps to explain the changing overall
pattern in p3, when the rate of creation of new links
overtakes the rate of emergence of new nodes. It does
not imply that the emergence of new technological
domains stopped, rather that it became lower com-
pared to the established technological fields.
By the end of p1 and during early p2, the most
promising fields had become known to the industry’s
incumbents. Historically, this is when integrated serv-
ice companies began to emerge. As it was difficult for
established operators to manage the increasing range
of specialized subcontractors in different technical
domains and coordinate technological interfaces, inte-
grated service companies took on this role. They
introduced total and integrated solutions combining
related technologies in unified packages (Chafcouloff
et al., 1995; Barreau, 2002). Following Krafft and
Quatraro (2011) and Krafft et al. (2014), we argue
that search strategies of industry players gradually
became organized rather than being random. Explora-
tive behavior was gradually replaced by exploitative
strategies applied in the most productive technologi-
cal areas. Innovation increasingly happened within
technological classes which proved promising and
fruitful, with a lower dispersion of R&D investment
across fields. As a result of emergent complementar-
ities, the knowledge base of the sector is not easily
divisible or decomposable. The rise of knowledge
network connectivity over most of p2 reflects these
dynamics.
The post 2002 decline in WADC seems odd, but
still compatible with our theoretical argument. It
resembles a period of technological discontinuity
whereby the speed of new links in knowledge net-
works falls behind new nodes, so knowledge connec-
tivity declines. We suggest such change a trend is a
consequence of the fact that information about knowl-
edge links is delayed (in the data set) compared to
information about the nodes. In other words, systemic
innovations which result from combinations (links) of
previous innovations (nodes) appear later.
4.3. The expected consequences of
knowledge base complexity
During the early phases (p1 and early p2) of the ILC,
the complementarities between new and old knowl-
edge domains were not fully explored and knowledge
linkages were not fully operational. Access to a wide
range of complementary knowledge was not neces-
sary for the innovation process. Therefore, we expect
to observe new entrants taking a greater role relative
to big and established companies when the sector
moves toward a Schumpeter Mark I pattern.
When systemic complexity increased in periods 2
and 3, the sector moved toward a more organized
search period and exploitative strategies became more
pervasive (Krafft et al., 2011). Core technological
domains were defined, technological trajectories were
relatively clear and most productive complementar-
ities and technical interdependencies were explored
by industry participants. Innovative companies which
connect and integrate different bits of knowledge
were able to benefit from economies of scale and
scope in both knowledge generation and exploitation
processes.
High systemic complexity presents strategic advan-
tages for technologically diversified actors who occupy
key positions within knowledge networks, compared
to marginal players (Antonelli, 2003). As a result,
more knowledgeable incumbents are expected to be
better placed to benefit from cross-fertilization between
different knowledge domains and their wide range of
applications. The entry barriers for new companies
tend to be higher and growth opportunities for small
ones limited, pushing the sector toward a Schumpeter
Mark II pattern. These propositions are examined next.
4.4. Dynamics of sectoral patterns of
innovation in upstream petroleum
In this section, we analyze the direction of change in
the sectoral pattern of innovation of upstream
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 7
V
C2016 RADMA and John Wiley & Sons Ltd

petroleum industry. As shown by Corrocher et al.
(2007), the same industry can comprise a combination
of two Schumpeterian patterns of innovation. In the
very early days, upstream petroleum resembled a
Schumpeter Mark I dominated by individual entrepre-
neurs until a monopolistic structure similar to Mark II
materialized because of the emergence of Standard
Oil. Although the dismantling of Standard Oil in 1911
decreased the level of concentration, the fundamental
oligopolistic structure dominated by the ‘Seven Sis-
ters’ remained until the 1960s when increased compe-
tition shifted the industry in the direction of Mark I
again (Inkpen and Moffett, 2011).
In this article, a more detailed account of the
dynamics of the sector is provided for the period
1970–2005. Following the extant literature (Malerba
and Orsenigo, 1996), we use a set of variables to
examine how the sector evolved over the three periods
(see section 3.3).
Table 1 summarizes the archetypical Schumpeter-
ian patterns of innovation and the direction of the vari-
ables over time that we expect to observe in each
typical mode (see 3.2.2). A Schumpeter Mark I sector
is relatively open to the entrance of new or small
firms. Therefore, we expect that new firm entry and
the number of innovating firms will increase over
time, resulting in a decrease of the concentration of
innovative activities. Malerba and Orsenigo (1996)
term this process ‘widening’.
In contrast, a typical Schumpeter Mark II sector is
relatively closed to new or small innovators and works
in favor of large innovators. Therefore, we expect to
observe a decreasing trend in the contribution of new
firms. The number of firms may be relatively stable
(as shown in Table 1) or even decrease over time,
depending on the size of existing firms. This implies a
rise in concentration of innovative activities which
Malerba and Orsenigo (1996) term ‘deepening’.
Comparing the observed trends with the evolution
of technological opportunities (see Table 1) helps to
reveal the dominant pattern. We stick to the three
main periods defined in section 3. In order to smooth
the trends and ignore short term fluctuations, we col-
lapse the data, as shown in Figure 3. The length of the
first period is 14 years, but the length of both the sec-
ond and third periods is 10 years. Therefore, we divide
p1 into one introductory subperiod (p1-0) and two
other subperiods (p1-1, p1-2). This means that all
three main periods cover 10 years with two 4-year
subperiods at both sides and a two-year gap in the
middle, leaving out the introductory subperiod of p1-
0. Using this periodization helps to control for the
impact of change in technological opportunities on
the selected variables, and therefore helps to
unravel the role of KBC in the dynamics of sectoral
patterns of.
4.4.1. Concentration and number of innovators
The top part of Figure 4 shows the trend of concentra-
tion over time for different size groups using a cor-
rected version of Herfindahl index of concentration.
This measure is used to explore how the relative share
of big vs. small innovators in the sector changes over
time. The advantage of this corrected version is that it
controls for small sample bias (Corrocher et al.,
2007). We repeated the indicator for different subsets
of companies defined by innovation size (for N<40,
N<100, N>40, N>100 and all companies: Nis the
number of patents each company holds) to check the
robustness of the results in different size groups. The
top left side of the Figure 4a displays concentration
(C) for large innovation size group and top right side
of the Figure 4b shows it for smaller sizes. Regardless
of the size categories, all of the indicators present an
overall U shape pattern reaching their lowest points in
Table 1. Expected Schumpeterian patterns of innovation in a dynamic perspective
Schumpeterian patterns of innovation Schumpeter Mark I
Widening
Schumpeter Mark II
Deepening
Concentration (C) #"
Number of firms (F) "#-
Entry of new firms (NE) "#
Stability of ranking (STR) L H
Figure 3. Periodization of the analysis.
Ali Maleki, Alessandro Rosiello and David Wield
8R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

p1-2 or p2-1. The two lowest figures show the number
of innovative firms over time, by innovation size.
According to these figures, concentration (C)
decreases in p1 (and even up to p2-1 for larger
groups). In parallel, firm numbers (F) increase in
almost all size categories. High technological oppor-
tunities driven by high oil prices seem to have worked
as a powerful incentive for smaller firms to catch up
with major innovators. The increasing number of
innovative companies in all groups also confirms the
key role of new innovators in p1.
When oil prices collapsed in p2-1, innovative efforts
were no longer rewarding within the industry. Over
p2, F slightly decreased and C took a clear upward
trend. One reasonable explanation is the higher vulner-
ability of some smaller firms, when continued low
opportunity dries up innovative efforts. Due to the
high risk and uncertainty involved in innovative activ-
ities, firms cut R&D investments in low profit condi-
tions, though with delays (Wintersteller, 1993; Acha,
2002). As discussed in section 4.1, the trend of patents
is negative in p2. Increasing concentration of innova-
tive activities, combined with reduction in the number
of innovative firms, suggests vulnerability of smaller
firms leaving the system of innovation. Indeed, p2 is
the only period with negative net entry.
The beginning of the third period (p3) presents an
interesting and puzzling pattern. By the end of p2 and
the beginning of p3, a new wave of innovative entry
can be noted, resulting in a sharp rise of F (Figure 4d)
in all size categories, excepting super big innovators
(N>100) (Figure 4c). This was driven by the jump in
technological opportunities observed after p2-1.
Although F transforms from a negative trend in p2 to
a sharp positive trend in p3,there is no expected corre-
sponding drop in C. In contrast, C continues its
upward trend which is reinforced over p3.
This pattern reflects the relative low and weakening
share of new entrants in p3, compared to big incum-
bents (Figures 4a and 4b). In addition, the short-term
jump of F before p3-1 (Figure 4b) turned into a rela-
tively stable trend in p3, whilst concentration gained
momentum.
These patterns suggest a fundamental difference
between p1 and p3. On the one hand, flourishing
opportunity environments in both periods encourage
new innovators to enter the sector – reflected in the
rise of F. On the other hand, C presents an opposite
trend – decreasing in p1, but increasing in p3. These
different behaviors were driven by the changing
nature of technological regimes, especially the rise of
KBC. In particular, we observe that the increasing
Figure 4. Concentration of innovative activities (a & b) and number of innovative firms (c & d): by innovation size. [Colour figure
can be viewed at wileyonlinelibrary.com]
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 9
V
C2016 RADMA and John Wiley & Sons Ltd

systemic complexity in p3 was associated with a
higher concentration of innovative activities.
To summarize, our results show that during p1, small
innovators benefited from abundant opportunities
because of low systemic complexity, which was no lon-
ger the case in p3. Systemic complexity in p3 increased
the cognitive barriers to entry for small and newcomer
companies. Although high technological opportunities
emerged and were driven by knowledge recombination
processes (see section 4.2), they were mostly exploited
by knowledgeable and technologically diversified com-
panies, with integrative and combinational capabilities.
Small and new firms continued to innovate in special-
ized niche technical areas, but became less relevant.
4.4.2. Share of new entry to the system of
innovation
This section analyzes the ability of new innovators to
contribute to the development of the knowledge base
of the industry in comparison to that of incumbents.
Table 2 shows the number of patents (by international
patent family IPF) of existing and new firms in each
subperiod; and also the new innovators’ share of pat-
ents (NE) in each subperiod. This is measured for
three different innovation sizes of firms (with mini-
mum patent size of 1, 5 and 10), in order to gain
insight into the role of size for successful entry.
According to Table 2, the 1% increase in the share
of new entry during period 1 (p1) reflects the rise of
success rate of new innovators. Growth of new entries
seems higher for bigger innovators (about 2% and 4%
for 5 and 10 IPFs minimum size), suggesting the
increasing possibility of advancement among larger
firms. Overall, the new entry indicators confirm
increasing opportunities over p1 for new innovators
of all sizes.
The transition from p1 to p2 is accompanied by a
10% reduction of new entrants for all size ranges. The
arrival of low opportunity conditions in p2 works
against new entry, as expected returns on R&D are
reduced. Over p2, when low opportunity conditions
established themselves and companies adjusted to the
external shock, some new entrants’ losses were recov-
ered. This is reflected in the rise of new entrants’ share
of innovators. Such a result is rather counter intuitive,
because low opportunity conditions are not normally
conducive to new entries.
One reasonable explanation, supported by data from
Weston and Johnson (1999), is that new innovative
companies emerged as a consequence of accelerated
outsourcing by operators whereby part of the innova-
tion process was transferred to a new class of agents
(service companies). Consequently, we attribute the
Table 2. New entries to the innovation system by different innovation size.
Subperiods 1 IPFs min size 5 IPFs min size 10 IPFs min size
IFPs
existing
innovators
IPFs
new
innovators
Share of
new entry
(NEP)
IFPs
existing
innovators
IPFs
new
innovators
Share of
new entry
(NEP5)
IFPs
existing
innovators
IPFs
new
innovators
Share of
new entry
(NEP10)
p1-1 1297 693 34.82 1272 222 14.86 1222 108 8.12
p1-2 2205 1232 35.85 2166 442 16.95 2062 298 12.63
p2-1 2802 1033 26.94 2714 195 6.70 2588 67 2.52
p2-2 2528 1112 30.55 2459 223 8.31 2348 97 3.97
p3-1 4226 1957 31.65 4127 595 12.60 3952 337 7.86
p3-2 5291 1732 24.66 5089 203 3.84 4853 45 0.92
Ali Maleki, Alessandro Rosiello and David Wield
10 R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

rise of new entries over p2 to the emergence of a new
division of innovative labor in the industry.
The distinction between short-term and long-term
responses to low opportunity conditions within the
sector is an interesting finding. The short-term
response of industry to low opportunities was to
reduce new entries. However, the long-term response
entailed the formation of a new division of innovative
labor, or more precisely a new ‘industry architecture’
(Brusoni et al., 2009). This favored new entrants and
triggered new knowledge dynamics. Transition from
the low opportunity conditions of p2 to high opportu-
nity conditions in p3 amplified the number of entries,
as reflected in the continued rise of NE for all size
ranges from p2 to p3.
Over p3, we observe a relative reduction of new
entrants in all groups, to their lowest levels over the
whole 1970–2005 period. In contrast to the high
opportunity conditions of p1 over which new entries
experienced their maximum level, the possibility of
new entries over p3 is most limited. Ceteris paribus,
the standard theory of patterns of innovation predicts
a positive relationship between opportunities and new
entries. These predictions however are conditional on
the nature of technological regimes. For example,
high new entry is expected under low cumulativeness
conditions when potential innovators are not at major
disadvantage with respect to incumbent firms (Breschi
and Malerba, 2000). Our analysis in sections 4.2 and
4.3 suggests that the difference between p1 and p3 in
terms of new entries can to a significant extent be
attributed to the dynamics of KBC. New entrants are
at a high disadvantage in p3 compared to p1 because
of the change in underlying technological regimes.
The rise of systemic complexity over p3 involves
higher cumulativeness, implying higher cognitive bar-
riers to entry, which hindered the exploitation of exist-
ing technological opportunities by new and small
companies in this period.
4.5. Schumpeterian patterns of innovation
and KBC
So far, the dynamics of the sectoral pattern of innova-
tion in the upstream petroleum industry have been
analyzed using three indicators over the main periods.
Table 3 summarizes the changing pattern of these
indicators over each period. The arrows in Table 3
specify the magnitude of changes in the indicators
over that period. Accordingly, p1 is characterized as
strong Mark I, because of a considerable reduction in
the degree of concentration (C), a large increase in the
number of firms (F) and the rise of new entrants (NE).
The second period presents a pattern which is simi-
lar to Mark II, although its intensity seems weak. C
began a slight upward trend and F reduced to some
extent, as technological opportunities were relatively
low. Although NE shows an upward trend over p2,
this can be explained by the increased reliance of oil
operators on outsourced services, a trend driven by
low oil prices (Weston and Johnson, 1999). In the
absence of this structural change, higher concentra-
tion and a lower number of innovative firms and new
entries would probably have been observed. Hence,
this period could be labeled as Mark II, but with some
effect of structural change on new entries. These
results suggest that KBC contributes to explain
change in innovation pattern as the industry entered
p2, because increased connectivity within the knowl-
edge network meant larger incumbents were in a bet-
ter position to exploit technological independencies.
The signs of Schumpeter Mark II are considerably
stronger when technological opportunities increase
over p3 (as the patterns of indicators show in Table 3).
Although technological opportunities are high, new
entries are reduced and the number of firms stays rela-
tively stable. Most importantly, the upward trend of
concentration accelerates. When the three indicators
are combined, comparing Tables 1 and 3 shows the
emergence of a progressively stronger Mark II, in
which the relative advantage of big innovators coin-
cides with higher KBC. The rise of technological
complexity in the sector was driven by more sophisti-
cated upstream exploration and complex production
projects. As a result, only big technologically
advanced companies had access to the required range
of sophisticated technologies required for complex
projects.
Our results also suggest that change in technologi-
cal opportunities tends to affect the pace of change in
Table 3. Observed Schumpeterian patterns of innovation
Periods 1st period 2nd period 3rd period
Schumpeterian pattern of innovation Strong I Weak II Strong II
Concentration (C) ## " ""
Number of firms (F) "" # -
Entry of new firms (NE) "" #
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 11
V
C2016 RADMA and John Wiley & Sons Ltd

existing patterns of innovation. The existing pattern
of innovation is weakened when changing from high
to low opportunity (as observed over the transition
from p1 to p2) and is reinforced when changing from
low to high (as observed over the transition from p2 to
p3). However, this evidence by itself is unable to
explain the shift from Mark I to Mark II. This is best
understood by looking at the two extremes of p1 and
p3, when two different patterns of innovation are
observable with high technological opportunities. If
the concept of technological regimes is convincingly
to explain the shifts in the mode of Schumpeterian
pattern, other factors should be taken into account. In
the case of upstream petroleum, we stress the role
played by systemic KBC. Reduction of systemic com-
plexity over p1 is consistent with Schumpeter Mark I.
When systemic complexity of the knowledge base
increases in early p2, the features of Mark II emerge
in the sector. Then, higher opportunities in p3 rein-
force this pattern.
These findings fit well with the propositions outlined
in section 4.3. As predicted, the upstream sector seems
to move toward Mark I over p1 and shift toward Mark
II over p3. This leads us to posit a novel analytical
framework which explains how the upstream petroleum
industry evolved through different patterns of innova-
tion in parallel with technological opportunities and
KBC. The impact of the combination of these two
dimensions of technological regimes on change of pace
and mode of Schumpeterian pattern of innovation over
the three periods is demonstrated in a 2 x 2 matrix in
Figure 5. The vertical axis specifies high vs. low techno-
logical opportunities and the horizontal axis represents
the pace of systemic complexity. As argued in section
3, the dynamics of systemic complexity could favor the
dominance of two different modes of Schumpeterian
pattern of innovation (Mark I on the left and Mark II on
the right of the matrix differentiate these two types).
Increasing (decreasing) technological opportunities
tends to reinforce (weaken) the pace of existing pattern,
whether it is Mark I or II, but do not alter its mode.
5. Conclusions
This article argues that shifts in sectoral patterns of
innovation are intrinsically related to the dynamics of
technological regimes. We provide a threefold contri-
bution. First, we propose a dynamic reading of the
concept of technological regimes to understand struc-
tural transformations of an industry over time. Sec-
ond, we conceptualize, placing KBC at the center of
our analytical framework. Third, we propose a quanti-
tative method using patent data in order to capture the
dynamics of KBC and their relationship with sectoral
patterns of innovation.
Our study highlights three distinct periods starting
from 1970. The first period corresponds to high oil
prices when operators dominate and actively invest in
technology. The second period is characterized by a
collapse in oil prices and a reduction in R&D invest-
ments, with a negative effect on innovation and an
expansion of specialized supply and service compa-
nies. The third period saw the gradual emergence of
new large integrated service companies which
increased R&D investments and offered innovative
solutions to complex upstream projects.
We focus on the coevolution among KBC, techno-
logical opportunities and sectoral patterns of innova-
tion. Our evidence suggests that decreasing systemic
complexity tends to be associated with a transition
toward Schumpeter Mark I, while the rise of systemic
complexity implies a shift toward Mark II. Nonethe-
less, it is also evident from our findings that the Schum-
peterian dichotomy is not completely adequate to
capture the dynamics of complex sectoral innovation
systems. We observed that the third and last period of
the study in upstream petroleum is not a typical Mark
II, but rather a modified type in which a new class of
innovators emerges to cope with increasing technologi-
cal complexity. These innovators were integrated serv-
ice companies playing the role of ‘integrators of
technological knowledge’ (Maleki, 2013, p. 98).
Other empirical analyzes show that specialization
strategies of operators and service companies can
entail varying innovation trajectories due to their dif-
ferent ‘technology frames’ (Acha, 2002). The pattern
of division of labor in p3 allowed service companies
to take an important position as technological innova-
tors. Following the model proposed by Jacobides
(2006), oil operators remained ‘first tier’ system
Figure 5. Technological regimes and innovation patterns.
Ali Maleki, Alessandro Rosiello and David Wield
12 R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

integrators as they provided a ‘new product architec-
ture’ and/or new ‘field development models’ (Acha,
2002, p. 82). However, as systemic complexity
increased, a significant proportion of innovative activ-
ities were undertaken by service companies, acting as
‘second tier’ system integrators. As observed in other
complex domains such as the aircraft industry, coexis-
tence of multiple systems integrators along the same
value chain is possible.
We also argue that the dynamics of technological
opportunities alone cannot explain the observed shift
of mode in innovation pattern, although they can
help to reveal changes in the pace or strength of
existing Schumpeterian patterns. In the period under
consideration, while the nature of knowledge com-
ponents underlying the sector may have not changed
considerably, the intensity of interactions between
knowledge components was progressively increas-
ing, leading to higher systemic complexity and
knowledge cumulativeness of the sector. Only when
the dynamics of technological opportunities are ana-
lyzed in combination with those of KBC, can they
convincingly explain the dynamics of Schumpeter-
ian patterns both in terms of pace and mode. Change
in systemic complexity could alter the Schumpeter-
ian mode, while a rise (or decline) of technological
opportunities tends to weaken (or strengthen) the
existing mode without altering it. This situation
characterizes a (modified) Schumpeter Mark II
mode.
This article offers novel insights into our under-
standing of Schumpeterian patterns of innovation,
while also providing a wider contribution to a strand
of literature in which business history interacts with
economics and management. In this sense, the dynam-
ics of technological complexity constitute an impor-
tant, albeit less understood, factor behind firm-level
strategy and the evolution of an industry structure (see
Mowery, 2015).
Our analysis offers promising opportunities for fur-
ther research. In particular, since it is confined to
upstream petroleum, it would be interesting to exam-
ine the transferability of our findings to other indus-
trial sectors.
References
Acha, V.L. (2002) Framing the past and future. PhD disser-
tation, SPRU, Brighton, Sussex.
Andersen, B. (2005) Paradigms and trajectories of techno-
logical opportunities 1890-1990. In: Moed, H., Glanzel,
W., and Schmoch, U. (eds), Handbook of Quantitative
Science and Technological Research. NL: Springer.
133–161.
Antonelli, C. (2003) Knowledge complementarity and fun-
geability: implications for regional strategy. Regional
Studies,37, 6-7, 595–606.
Arora, A. and Gambardella, A. (1994) The changing tech-
nology of technological change: general and abstract
knowledge and the division of innovative labour.
Research Policy,23, 5, 523–532.
Babusiaux, D., Favenne, J., Bauquis P.R., Bret-Rouzaut N.,
and Guirauden, D. (2004) Oil and Gas Exploration and
Production: Resources, Costs, Contracts.Paris:Editions
Technip.
Bagheri, S.K. and Di Minin, A. (2015) The changing com-
petitive landscape of the global upstream petroleum
industry. The Journal of World Energy Law & Business,
8,1,1–19.
Barreau, S. (2002) Innovations and external growth strat-
egy: the case of oil and gas supply and service compa-
nies. Oil & Gas Science and Technology - Rev. IFP,57,
2, 193–203.
Breschi, S. and Malerba, F. (2000) Sectoral innovation sys-
tems: technological regimes, Schumpeterian dynamics,
and spatial boundaries. In: Edquist C. and McKelvey
M.D. (eds), Systems of Innovation: Growth, Competitive-
ness and Employment. Cheltenham, Edward Elgar. 249–
261.
Brusoni, S., Jacobides, M., and Prencipe, A. (2009) Strate-
gic dynamics in industry architectures and the challenges
of knowledge integration. European Management
Review,6, 4, 209–216.
Chafcouloff, S., Michel, G., Triice, M., Clark, G.,
Cosad, C., and Forbes K. (1995) “Integrated services.”
Available at http://www.slb.com/~/media/Files/resources/
oilfield_review/ors95/sum95/06951125.ashx accessed on
01/09/2015.
Chesbrough, H.W. and Teece, D.J. (1996) Organizing for
innovation (cover story). Harvard Business Review,74,
1, 65–73.
Corrocher, N., Malerba, F., and Montobbio, F. (2007)
Schumpeterian patterns of innovative activity in the ICT
field. Research Policy,36, 3, 418–432.
Fai, F. (2007) A structural decomposition analysis of
technological opportunity, corporate survival, and lead-
ership. Industrial and Corporate Change,16,6,1069–
1103.
Fagan, M.N. (1997) Resource depletion and technical
change: effects on U.S. Crude oil finding costs from
1977 to 1994. The Energy Journal,18, 4, 91–106.
Grant, R.M. and Cibin, R. (1996) Strategy, structure and mar-
ket turbulence: the international oil majors, 1970–1991.
Scandinavian Journal of Management,12, 2, 165–188.
Grebel, T., Krafft, J., and Saviotti, P.-P. (2007) On the life
cycle of knowledge intensive sectors. Revue de l’OFCE,
5,63–86.
Griliches, Z. (1990) Patent statistics as economic indica-
tors: a survey. Journal of Economic Literature,28,4,
1661–1707.
Helfat, C.E. (1997) Know-how and asset complementarity
and dynamic capability accumulation: the case of R&D.
Strategic Management Journal,18, 5, 339–360.
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 13
V
C2016 RADMA and John Wiley & Sons Ltd

Hirsch-Kreinsen, H., Jacobson, D., and Robertson, P.L.
(2006) Low-tech’ industries: innovativeness and devel-
opment perspectives - a summary of a European research
project. Prometheus,24,1,3–21.
Iizuka, M. (2009) Standards as a platform for innovation
and learning in the global economy: a case study of the
Chilean salmon farming industry. International Journal
of Technological Learning, Innovation and Develop-
ment,2, 4, 274–293.
Inkpen, A.C. and Moffett, M.H. (2011) The Global Oil &
Gas Industry: Management, Strategy & Finance.South
Sheridan Tulsa, OK: PennWell Books.
Jacobides, M.G., Knudsen, T., and Augier, M. (2006) Ben-
efiting from innovation: Value creation, value appropria-
tion and the role of industry architectures. Research
Policy,35, 8, 1200–1221.
Klepper, S. (1996) Entry, exit, growth, and innovation over
the product life cycle. The American Economic Review,
86, 3, 562–583.
Krafft, J. and Quatraro, F. (2011) The dynamics of techno-
logical knowledge: from linearity to recombination. In:
Antonelli C. (ed.), Handbook on the Economic Complex-
ity of Technological Change. Cheltenham: Edward Elgar.
pp. 181-200.
Krafft, J., Quatraro, F., and Saviotti, P.P. (2011) The
knowledge-base evolution in biotechnology: a social net-
work analysis. Economics of Innovation and New Tech-
nology,20, 5, 445–475.
Krafft, J., Quatraro, F., and Saviotti, P.P. (2014) Knowl-
edge characteristics and the dynamics of technological
alliances in pharmaceuticals: empirical evidence from
Europe, US and Japan. Journal of Evolutionary Econom-
ics,24, 587–622.
Maleki, A. (2013) Can a mature resource-based indus-
try become high-tech? Transformations of sectoral
innovation systems in upstream petroleum industry.
Journal of Science & Technology Policy,5, 4, 83–
105.
Malerba, F. (2006) Innovation and the evolution of indus-
tries. Journal of Evolutionary Economics,16,1,3–23.
Malerba, F. (2007) Schumpeterian patterns of innovation
and technological regimes. In: Hanusch H. and Pyka A.
(eds), Elgar Companion to Neo-Schumpeterian Econom-
ics. Cheltenham: Edward Elgar.
Malerba, F. and Orsenigo, L. (1996) Schumpeterian pat-
terns of innovation are technology-specific. Research
Policy,25, 3, 451–478.
Malerba, F. (ed.) (2004) Sectoral Systems of Innova-
tion Concepts, Issues and Analyses of Six Major
Sectors in Europe. Cambridge, UK: Cambridge Uni-
versity Press.
McGahan, A.M. and Silverman, B.S. (2001) How does inno-
vative activity change as industries mature? International
Journal of Industrial Organization,19, 7, 1141–1160.
Mowery, D.C. (2015) Steven Klepper and business history.
Industrial and Corporate Change,24, 4, 755–768.
Nelson, R.R. and Winter, S.G. (1982) An Evolutionary
Theory of Economic Change. Cambridge, MA: Harvard
University Press.
Pavitt, K. (1985) Patent statistics as indicators of innovative
activities: possibilities and problems. Scientometrics,7,
1, 77–99.
Rajan, S. (2011) Future technologies of the oil field. The
Way Ahead,7, 3, 11–14.
Robertson, P., Smith, K.H., and von Tunzelmann, N.
(2009) Innovation in low-and medium-technology indus-
tries. Research Policy,38, 3, 441–446.
Saviotti, P.P. (2011) Knowledge, complexity and networks.
In: Antonelli, C. (ed.), Handbook on the Economic Com-
plexity of Technological Change. Cheltenham: Edward
Elgar. pp. 141–180.
Schumpeter, J.A. (1934) The Theory of Economic Develop-
ment. Cambridge, MA: Harvard Economic Studies.
Schumpeter, J.A. (1942) Capitalism, Socialism and Democ-
racy. New York: Harper & Row
Steinmueller, W.E. (2000) Will new information and com-
munication technologies improve the ’codification’ of
knowledge? Industrial and Corporate Change,9,2,
361–376.
Teece, D. (1976) Vertical Integration and Vertical Divesti-
ture in the U.S. Oil Industry. Stanford, CA: Institute for
Energy Studies.
U.S. EIA. (2011) Performance Profiles of Major Energy
Producers 2009. Washington, DC: Energy Information
Administration.
Vale, M.R. and Caldeira, J. (2008) Fashion and the gover-
nance of knowledge in a traditional industry: the case of
the footwear sectoral innovation system in the northern
region of Portugal. Economics of Innovation & New
Technology,17,1/2,61–78.
Wang, Q. and von Tunzelmann, N. (2000) Complexity and
the functions of the firm: breadth and depth. Research
Policy,29, 7–8, 805–818.
Weston, J.F., Johnson, B.A., and Siu, J.A. (1999) Mergers
and restructuring in the world oil industry. Journal of
Energy Finance & Development,4, 2, 149–183.
Wintersteller, W. (1993) The role of technology in the
upstream petroleum industry. Dissertation, Mining Uni-
versity Leoben, Leoben Austria.
WRTG. (2008) Oil Price History and Analysis. Available at
http://www.wtrg.com/prices.htm accessed on 05/04/2016.
Ali Maleki is an Assistant Professor of Science and
Technology Policy at Research Institute for Science,
Technology and Industrial Policy, Sharif University
of Technology, Tehran, Iran. He received an MBA
from Sharif University of Technology and a Doctoral
Degree in Science and Technology Studies from Uni-
versity of Edinburgh. Ali has also been a visiting
researcher at SPRU, University of Sussex. His
research focuses on national and sectoral innovation
systems and technology policy in resource-based
developing economies. He authored several policy
reports and presented articles in international
conferences.
Alessandro Rosiello is a Senior Lecturer in Entre-
preneurship and Innovation at the University of
Ali Maleki, Alessandro Rosiello and David Wield
14 R&D Management 00, 00, 2016 V
C2016 RADMA and John Wiley & Sons Ltd

Edinburgh Business School, and an associate mem-
ber of the Innogen Institute based at the Open Uni-
versity and the University of Edinburgh. He a
substantial track record of high-quality publications
with leading international journals. In particular, he
acted as guest-editor for special issues of the Interna-
tional Journal of Technology Management & Sus-
tainable Development, European Planning Studies
and Technology Analysis & Strategic Management.
Alessandro combines innovation theories, Schumpe-
terian economics and direct experience of entrepre-
neurial venturing to study innovation processes in a
variety of industrial settings; entrepreneurship and
small business finance; commercialization of new
technologies; and the processes of industrial cluster-
ing at regional level.
Dave Wield is Professor of Innovation and Develop-
ment at the Open University and a Co-Director the
Innogen Institute based at the Open University and
the University of Edinburgh. From 2007 to 2014 he
directed the ESRC Centre for Social and Economic
Research on Innovation in Genomics (Innogen), Uni-
versity of Edinburgh and Open University. He has
worked at Imperial College, University of Dar Es
Salaam, Tanzania, Eduardo Mondlane University,
Mozambique and Aston University. He has also been
a Senior Fulbright Fellow at Stanford University/Uni-
versity of California, Berkeley. Dave’s research inter-
ests focus on the policy and management of
technology; and on development policy and practice
with emphasis on industrialization and technologies.
Recent research includes projects on: innovation in
life science companies; knowledge management and
development, public-private collaboration and inter-
national development.
APPENDIX A: Measures of knowledge
base complexity
We employed Social Network Analysis (SNA) and
its powerful toolbox to characterize the
connectivity of the network as measure for com-
plexity. A matrix of co-occurrence of technologi-
cal classes is formed to represent the knowledge
network where the value of each cell is the num-
ber of inventions for which two technological
classes appeared jointed together (Krafft et al.,
2011).
The degree of centrality of a node is used as
one of the centrality measures, describing how
strong is the level of connectivity of a node
(Krafft et al., 2011). Formally, the following
equation expresses the measure of degree of cen-
trality (DC):
DCn5X
i2N6¼n
lni (2)
Where nrepresents the nodes and lrepresent the
links.
The degree of centrality is defined as the
number of links of one node with other nodes of
the network. Because this measure is affected by
the network size, it is often divided by its maxi-
mum value to provide a normalized proxy
(Krafft et al, 2011), as shown in the following
equation:
NDC5DCn=ðN21Þ(3)
In order to create a measure of connectivity at
the level of a network, we rely on the average of
the degree of centrality of all nodes in the net-
work. Following (Krafft et al, 2011), we used the
average measure of degree of centrality,
weighted by relative frequency. This takes into
account the highly unequal strength of the nodes,
giving higher weights to important technological
classes. Accordingly, the measure of complexity
of the knowledge is the weighted average degree
of centrality (WACD) as follows:
WADC5X
n
½NDCnðPn=X
n
PnÞ (4)
The effect of the dynamics of knowledge base complexity on Schumpeterian patterns of innovation
R&D Management 00, 00, 2016 15
V
C2016 RADMA and John Wiley & Sons Ltd