The evolution of a strategic alliance network –
Exploring the case of stock photography
This is an Author’s Accepted Manuscript of an article published by Edward Elgar, Cheltenham in
Boschma R and Martin R (eds) The Handbook of Evolutionary Economic Geography. 2010: 298-
315, doi: 10.4337/9781849806497.00023 available online at:
Over the recent years, increasing effort has been taken to conceptualize an evolutionary
perspective within economic geography (Boschma and Lambooy 1999; Boschma and Frenken
2006). Evolutionary approaches have theorized a variety of conceptions, among others, the
notion of path dependence in regional development (Martin and Sunley 2006), the role of
technological and sectoral variety in regional innovation (Rigby and Essletzbichler 2006;
Boschma and Iammarino 2007) and the idea of co-evolution in regional development (Schamp
2008, this book). Explicitly or implicitly, many geographical accounts of the economy rely on
the notion of networks to represent and analyze economic processes and modes of organization.
A dynamic perspective of network evolution, however, has experienced only little attention until
recently (McPherson et al. 2001; Baum et al. 2003; Kilduff and Tsai 2003). The common re-
search strategies to focus on the temporal dimension of networks have been twofold. On the one
hand, contingency models of networks have analyzed the appropriateness of certain network
forms for different environmental conditions. Empirical research demonstrates, for example,
that when firms operate under conditions of exploration, i.e. high uncertainty and high inno-
vation rates, weak ties and loose coupling are more beneficial than strong ties. In turn, when
firms operate under conditions of exploitation, i.e. stable market conditions and mature
products, strong ties are more advantageous (Rowley et al. 2000). On the other hand, sequential
models of non-linear cyclical change have been applied to the evolution of firms (Hite and
Hesterly 2001; Lavie 2004) and regional clusters (Staber 1997). Within the framework of
organizational ecology, Staber (1997) found that in the beginning of a textile cluster, dense
networks with strong ties between firms were beneficial in order to circulate knowledge and to
ally for joint lobbying. In later stages of cluster development, however, when an increasing
number of firms competed for the same resources, i.e. labour, expertise, capital etc., weak ties
became more important (Staber 2001). Though these models recognize the ephemeral nature of
network structure, they are relatively silent on the endogenous mechanisms that drive network
evolution because they always use external environmental changes to explain responses in
This paper aims at exploring and connecting network evolution with the wider project of an
evolutionary economic geography (Glückler 2007). Concretely, this paper pursues two
questions. First, are there endogenous network dynamics, i.e. mechanisms within network
structure that drive the dynamics of inter-firm networks and help to explain future network
structure? Second, is geographical proximity associated with growth patterns of these inter-firm
networks? The empirical challenge in evolutionary network research is to find appropriate
longi-tudinal structural data on sectoral or regional economic activity. This paper focuses on the
inter-firm sales partnerships in the organizational field of stock photography and uses a survey
on German stock photo agencies for the period between 1989 and 2005. Stock photography is
an interesting sectoral case since it has been undergoing major technological, institutional and
organizational change in very short time (this shift is discussed in more detail below). The
empirical analysis looks at the recent emergence of sales alliances between picture agencies as
the unit of analysis. Strategic alliances are an organizational strategy to access and share re-
sources, costs and profits with partner organizations. Especially under conditions of increased
competition and unstable market conditions, they render different advantages to individual
organizations, such as firm survival, performance, imitation and innovation (Knoke 2001; Brass
et al. 2004). A sales partnership, in particular, is a contractual licence agreement, where one
firm grants the usage right either for an intellectual property-right protected content to another
firm for a specified time and use in return for a royalty payment (Knoke 2001). Though stock
photography has a long tradition in international sales cooperation (Wilkinson 1997), the
alliance behaviour has changed dramatically over the last ten years. It has changed in
quantitative terms because more and more picture agencies seek sales partners; and it has
experienced a qualitative change because picture agencies no longer ally only internationally but
they have also started partnering with agencies in their domestic market. Section 2 discusses
some elements of an emerging theory of network evolution and develops research hypotheses.
Section 3 introduces the business context and the fundamental technological and organizational
changes occurring in the market for still images in recent history. Section 4 documents the
methodology and section 5 presents the findings of the empirical analysis before the paper
closes with some concluding remarks with respect to the emerging project of an evolutionary
2 Network evolution and geography
The essential starting point for any theory about network evolution is the question of “how do
structural dimensions of an interorganizational communication network at Time 1 affect the
interactions among member organizations – specifically, their formation of ties to other organi-
zations – at Time 2?” (Kenis and Knoke 2002, 277-278). This question pays attention to
inherent conditions of a network rather than external conditions that drive network change. The
network trajectory (Kilduff and Tsai 2003) is an appropriate concept in the analysis of network
evolution which combines the notions of evolution, network and geography: It describes a
geographically and historically specific development path of a network in which the formation
and dissolution of ties in earlier stages generates cumulative propensities for the formation and
dissolution of ties in the future and in which the mechanisms of path-disruption and variation
are endogenous (Glückler 2007). This perspective explicitly moves beyond the dyadic analysis
of single relations to the analysis of entire networks of relations. A theory of network evolution,
thus, looks at the changes that new ties produce in the existing structure and, conversely, at the
impact that the structure imposes on the formation of these next ties. In a recent, sophisticated
study Powell et al. (2005) explicitly tested a set of alternative hypotheses on the emergence of
new relationships in an expanding network. Recurring on the network literature, they focused on
a set of hypotheses that they assessed as potential mechanisms for network evolution: homo-
phily, preferential attachment and multiconnectivity. The following sections briefly review these
concepts before geography will be explored as a fourth potential condition of network evolu-
2.1 Homophily vs. diversity between firms
According to homophily theory, firms with similar attributes have a greater ease in developing
characteristic-based trust (Zucker 1986) and are therefore more likely to engage in trust-
demanding activities. The similarity between firms implies a stronger overlap of interests,
strate-gies and interpretative frameworks to deal with information on markets and products.
Hence, networks are expected to expand by a process in which new partners are chosen for
similarity or homophily (McPherson et al. 2001; Sorenson 2003; Rowley et al. 2005). The
homophily between firms may be measured as the similarity in firm size, age and form as well
as similarities in terms of strategy and products. Alternatively, however, alliances may also be
hypothesized to follow a pattern of diversity. Firms may choose to collaborate exactly because
of different, complimentary resources and competencies (Rowley et al. 2005) in order to yield
synergies from collaboration. Therefore, the dissimilarity with respect to strategy, resource
endowment etc. may be also be a driving element in the formation of strategic alliances.
2.2 Preferential attachment and multiconnectivity
Apart from firm specific characteristics that may be compared and assessed for pair-wise homo-
phily, Powell et al. (2005) also propose structural effects on future tie formation which would
produce endogenous dynamics. The preferential attachment-hypothesis expects firms with many
ties at one point in time to be more likely to receive new ties in the future than those with fewer
ties. There is obviously accumulative advantage for well positioned actors on the one side and a
liability of dis-connectedness for peripheral actors on the other, as the authors found in earlier
research (Powell et al. 1996). Alternatively, the multiconnectivity-hypothesis expects networks
to expand through a process in which firms seek diversity of relations and form multiple
independent paths. By moving in different communities and connecting to heterogeneous
groups the firm enhances the multiple reachability of partners. Empirically, however, Powell et
al. largely disprove homophily and popularity by analyzing extensively the evolution of the
strategic alliance network in US biotechnology over a 12-year period between 1988 and 1999.
Instead, they found strong support for the multiconnectivity-hypothesis because new alliances
were more likely to form between those firms that were more multiply or more diversely linked
to each other at a previous stage of network growth. This finding suggests that networks entail a
logic of multiconnectivity, where new linkages reinforce diversity and multiple cohesion
(Powell et al. 2005). Another strand of network research supports these conclusions from a
different perspective. While the former approach stresses the mechanisms of change through the
appearance of new ties in an incumbent network, Walker et al. (1997) emphasize path-
dependence and persistence in network structure through its evolution over time. In their
empirical analysis of the alliance network in biotechnology they found that early partner choice
at the beginning had a significant impact on future cooperation. A year-to-year comparison of
structural equivalence for each pair of actors showed that if two firms were connected to the
same other companies, i.e. they were structurally equivalent, they most likely remained so
throughout the subsequent expansion of the network. This finding leads Walker et al. (1997) to
argue for path dependence in the network trajectory.
2.3 Geographical proximity
Interestingly, a lot of the economic evidence on path dependence and lock-in has actually been
produced and exemplified in the context of geographic clusters (Martin and Sunley 2006). Geo-
graphical propinquity seems to be an important constraint on network formation. Sometimes
this effect is abbreviated in unfortunate terms of ‘spatial causation’. Space, however, is not a
necessary cause of economic interaction. Instead, its role is mediated through at least two
underlying social technologies: communication technology (Storper and Venables 2004) and
transport technology (Marquis 2003). Breschi and Lissoni (2008, this book) demonstrate nicely,
that the effect of geographical proximity on innovation is often mediated through inert mobility
of inventors. They tend to stay in a region over long periods of time but move between
companies in the same region to produce a pattern of locally interrelated innovation networks.
Only with respect to the actors’ communication preferences and mobility opportunities may the
contingent relation between physical space and economic interaction be established (Glückler
2007). In other words: the constraints of proximity only rule if face-to-face is the only mode of
communication and if travel is prohibitive. In any other case, especially in a world of high
mobility of people, commodities and information, proximity is contingent on the underlying
social technologies. Owen-Smith and Powell (2004) distinguish two such communication tech-
nologies to explain why knowledge intensive businesses such as biotechnology tend to favour
co-located patterns of inter-firms alliances: One is the conduit as the distribution of private
information from a sender to a number of limited receivers, and the other is the channel as the
potential spill-over of information into the entire set of connected actors in a network under
conditions of spatial proximity (Owen-Smith and Powell 2004). In fact, the more complex the
knowledge in a particular industry, the more do industries agglomerate. This is mainly because
the distribution of complex knowledge entails more problems of accurate transmission and
interpretation such that spatial proximity eases transfer and locks-out remote actors from the
knowledge flow (Fleming and Sorenson 2001; Sorenson 2005). Furthermore, Powell et al.
(2005) found unequivocal evidence for a strong geographical bias on strategic alliances in the
organizational field of biotechnology. New ties as well as repeat ties were more likely when two
firms were co-located. These studies univocally suggest the importance of geography to the
organization and dynamics of networks.
Geographical proximity is not only mediated through communication technologies at a given
moment in time but may also be a result of its historical legacy. Marquis’ (2003) work
illustrates just how geographical proximity may sustain to affect the formation of network
relations even though communication and travel technologies have changed dramatically. He
demonstrates that those networks of interlocking directorships in U.S. cities that had established
prior to the advent of air travel technology continued to be significantly more locally bound than
networks in younger cities. Despite the availability of modern travel technologies in all cities
today, even new corporate board positions were filled with local directors. This persistence of
geographical network structure in older communities supports imprinting theory (Stinchcombe
1965): organizations adopt organizational characteristics in response to the environmental
conditions during their period of foundation. Marquis’ study legitimizes the possibility that
geographical proximity endures in network practice even though the communication and travel
technologies have changed and now allow for different patterns of spatial interaction. These
arguments are challenging with respect to sales alliances in the case of stock photography
because digital technology has enabled picture agencies to operate nearly independent of physi-
cal constraints and to exchange pictures with other agencies at any place. Will geography still
matter for the creation of new alliances in the network, although the business processes are
almost fully digitized and virtualized?
3 A case from the cultural economy: the evolution of stock
Photography is an essential element of visual information and communication at all levels of
society. We encounter still images in press, television, the internet, books, exhibitions, corporate
communication, and advertisement campaigns. Wherever we find a photograph, there is a sup-
ply chain behind the placement and there are basically two ways for a photo to reach the market.
One is to assign a photographer and have photos produced in photo-shootings, the other is to
buy a licence for the use of an existing photograph from a stock picture agency. Licensing a
stock photo has a number of advantages. In contrast to a photo production, the customer avoids
the production risk, i.e. she sees what she gets, and saves the production cost which might sum
up to considerable sums for location, equipment, models, stylists and the photographer crew.
The customer only pays a royalty to the agency for using a photo which might be just a small
percentage of the production cost. Typically, it is the picture editor in a publishing house or the
art buyer in an advertising agency who takes the decision whether to assign a photographer for
production or to buy a photo “from the shelves”. In the USA, picture editors and art buyers
spend about double the amount of money on stock photography than on photo-shootings (Sachs
2003) and use about four times more stock photos than own productions (Kjemtrup 1997). To-
day, there are around 1,000 picture agencies in Europe offering photographs and illustrations
from about 21,000 photographers and generating revenues of roughly 1 billion euros (CEPIC
The historical evolution of stock photography might well be interpreted as a continuous re-
sponse to technological innovation. Ever since its invention by Louis Daguerre in Paris and W.
H. Fox Talbot in London, photography has been shaped and developed through the ongoing
advance in production and transmission technologies: Talbot was the first to develop the
principle of the negative pattern, an invention that laid ground for repeated duplication in 1840
(Frizot 1998); and Arthur Korn’s invention of the Telefotokopie in 1902, for instance, built the
technological fundament for large-distance circulation and distribution of photographs within
short time and thus made photography available as a visual medium in the established news
press at the beginning of the 20th century (Albert and Feyel 1998). The emergence of picture
agencies as intermediaries between photographers and publishing houses was a consequence of
the rapid growth of photography in the 1920s. Toward the end of the 1930s a significant number
of firms specialised on archiving and marketing photography on commission of photographers.
After this historical departure, the organizational field of stock photography has experienced
some fundamental changes. Three major organizational and technological shifts are apparent in
this context (see Glückler 2005 for an extended elaboration).
3.1 From press to publicity: expansion into commercial photography
Stock photography had long been editorial, i.e. pictures for reportage, documentation, and news,
whereas advertising agencies would usually produce their visual content through assignment
photography with commercial photographers. The 1970 brought about a major change. Given
the economic downturn and the oil crisis in 1973, the advertising industry suffered serious
budget constraints and sought for new image sources at lower cost. At the same time, the 1978
US Copyright Law, which asserted that an image was the photographer’s rather than the client’s
property, provided institutional guarantee for photographers to use stock photography as an
additional business opportunity (Frosh 2001). The Image Bank pioneered modern stock
photography by venturing from the editorial into the commercial market in 1974. In order to
compete with assignment photography, they managed to improve picture quality, attract
commercial photographers, and raise the attention of advertising agencies to buy stock at lower
cost than to assign photo productions. The company pursued an aggressive marketing strategy,
introduced picture catalogues as an instrument to gain visibility and create visual style (Frosh
2001). Today, picture catalogues are an established marketing tool. In Europe, half of all agen-
cies publish catalogues on an annual basis (CEPIC 2001). In addition, the British pioneer Tony
Stone developed a sophisticated licensing system to maximise the profit per picture such that the
agency succeeded in selling the same picture over 300 times in just one year generating
revenues of £30,000 with just one photograph (Rich 1994).
3.2 From light-box to screen: digitization and online archives
In the early 1990s the market experienced enormous growth rates of up to 25% per year and
enthusiastic expectations accompanied the advance of digital technology. However, still in 1995
no single company had launched an online archive and a survey in Britain at the time revealed
that many of the big agencies had not yet taken clear decisions when to go online. Some busi-
ness experts were even quite averse to the notion of selling pictures via the internet and saw
more threat than opportunity: “I can’t understand why any picture library can consider going on-
line“ (Sal Shuel, BAPLA, quoted in Goddard 1995). While the incumbent business was some-
how reluctant to the advent of digital technology, very powerful and resourceful investors
entered the market with high expectations. The strategic opportunity of digitization was based
on the fact that more images could be offered and distributed to more customers at essentially
less cost. British oil magnate Mark Getty and Microsoft founder Bill Gates recognized this
opportunity and made large investments to buy up basically all established and significant
players in the market. Today, Getty Images (founded 1995) and Corbis (founded 1989) are by
far the largest firms in the global market place for still images. In contrast to their incumbent
competitors, both firms drove technological change with intensive commitment. Getty was the
first big agency to launch an e-commerce platform on their online archive in 1998. While they
gained 14% of their revenues in the first year, they increased online sales to over 90% in only
3.3 From licence to property: royalty-free images
Online-archives and e-commerce platforms have not been the last innovation, though. Decreas-
ing distribution cost through digital technology has made another source of friction more obvi-
ous: the legal constraints of intellectual property right. When customers buy a reproduction li-
cense, the royalty is tailored to the intended use and depends on the size and placement of the
photography, the medium (commercial or editorial), the print run, and eventually on exclusivity.
If the customer wanted to extend the use of a photograph to e.g. brochures, homepage, flyers,
posters etc. she would have to buy a new reproduction licence and pay an additional royalty.
Managing the property and reproduction rights implies transaction cost and limits the flexibility
of the customer. As a consequence, a new licence model called ‘royalty-free’ successfully
established in stock photography. When customers buy a royalty-free (RF) photo they pay a
fixed price and acquire unlimited usage right. The royalty-free licence model thus converts
photography into a full commodity. Mark Torrence had known this licence model in the music
industry, when he founded PhotoDisc in 1991 to pioneer stock photography with RF-visual
content. Servicing graphic and web designers in the beginning, PhotoDisc launched an e-
commerce platform in 1995 and drove the company among the top ten of Americas 500 fastest
growing companies in 1997. Like many successful agencies, Getty Images acquired PhotoDisc
in order to incorporate their technology and know-how and to go online just a year later in 1998.
In 2000, the market share of RF-images was already assumed to reach 15% in the US and 7% in
Europe. Today, Getty gains about one third of their revenues from royalty-free images and at
higher margins per image than traditional rights-managed photography (Glückler 2005).
In conclusion, the market for stock photography has experienced major changes which have
only been highlighted very briefly. Nonetheless, digital technology and royalty-free photos have
dramatically reverberated the market environment, and redefined fundamental business
parameters. As a consequence, the market now looks very different from what it looked like
fifteen years ago. It has shifted from a market of medium-sized companies to an oligopolisitic
market dominated by three players that easily take a 45% share of the global revenues in this
sector; it has moved from localised enduring personal relationships to a virtual business
transaction; and it is transforming photography from a licensable content to a private
commodity (Glückler 2005). This shift has deep cutting implications for competitive strategy.
The availability of e-commerce and internet search in online picture archives disrupts existing
networks of personal relationships between agents and customers and turns a seller market into
a buyer market. The global dominance of Getty Images and Corbis forces picture agencies into
niche strategies, specialization and inter-firm alliances. Since market visibility and scale of
photo collections have become so center stage, many agencies have begun to seek sales
partnerships in order to forge alternative distribution channels or to collect additional material
and reach scale.
The aim of the empirical analysis is to analyse the growth pattern of the stock photography
field-net of inter-agency licence agreements in Germany. An organisational field-net is “a par-
ticular pattern of both present and absent links among the entire set of organizational dyads
occurring in a specified organizational field” (Kenis and Knoke 2002). The term is used to
explicitly focus on the dyadic relations and the overall structure of relations between every pair
of players in a field.
The empirical approach to test the hypotheses of homophily, geography, popularity bias, and
multiconnectivity on the alliance behaviour of picture agencies made necessary a network
survey because there are no secondary data available on alliance behaviour in the stock
photography market. The survey method, however, limits the study of evolutionary processes in
at least three ways: it cannot deliver long-term historical data; it will normally produce
incomplete response; and it produces a survival bias by missing out those firms and earlier
alliances that were abandoned in the past. The network thus represents a growth path of
surviving firms rather than the entire network evolution.
Variables. The questionnaire consisted of two parts. The first part collected attribute
information on the individual picture agencies such as firm age, product specialization, firm size
by number of employees, by image stock and by number of represented photographers as well
as international sales and membership in the national business association. These attribute items
were used to develop relational values of firm-to-firm similarity (homophily) and geographical
propinquity between all pairs of agencies (see Table 1 for variable definitions). The calculation
of relational similarities is based on measures of pair-wise differences or coincidence in the
individual scores on an item. For each variable, the similarity values are then represented in a
similarity matrix for all pairs of n x n firms, where each cell value indicates the coincidence
(e.g. association membership) or (dis)similarity (e.g. difference in firm age) of two firms on an
The second part asked firms to identify all sales partnerships with other agencies. The variable
is represented in a binary adjacency matrix, where for each pair of agencies the value one
indicates the existence of an alliance and where zero indicates its absence. Thus, the dependent
variable is the network of sales alliances trading rights managed picture licenses be-tween
picture agencies in Germany. Each relationship was further specified for three characteristics:
First, two types of licence models – rights-managed vs. royalty-free – were distinguished. Since
the alliance behaviour in royalty-free imagery is quite different from cooperation in rights
managed photography, the study focused purely on the dominant business of rights-managed
sales alliances. Second, the year of establishment of each licence agreement was assessed in
order to produce network stages on a year-to-year basis. This temporal dimension of network
formation builds the basis for the analysis of network dynamics over time and allows for testing
the effects of homophily, geographical propinquity, popularity bias and multiconnectivity on
later stages of network structure. Third, the direction of picture supply distinguishes between
senders and receivers and thus identifies reciprocal and unidirectional linkages.
Finally, four other variables were generated from the alliance data. The dyadic differences
between individual firms’ indegree and outdegree centralities were used to measure the concept
of popularity bias. The greater the differences in indegree (receiving relations) or outdegree
(sending relations) centrality between a pair of agencies, the more likely should they be to have
a relationship in the future. The concept of multiconnectivity was measured by two variables.
The point connectivity of two unconnected (non-allying) firms is the number of firms that need
to be deleted so that no indirect connection exists between them any more (Borgatti et al. 2002).
Geodesic count produces a measure that counts the number of shortest network paths between
each pair of agencies (Borgatti et al. 2002).
Survey population. There are three organizational types of picture agencies. In the dominant
case the agency is a commercial representative of the images provided by independent
photographers. The photographer holds the intellectual property rights and commissions the
agency to sell reproduction licenses for a royalty. Apart from this dominant form, there are
photographer cooperatives and archives. In contrast to agencies, the former maintain the
intellectual property rights within the cooperative and the latter can market historical public
domain images, where intellectual property rights have expired. Therefore cooperatives and
archives may offer pictures at more competitive prices since they do not have to share their
revenues with the legal originator. The population of picture agencies is defined as the number
of commercial firms that licence usage rights of still images under commission of independent
photographers. This explicitly excludes non-commercial archives, small-scale sideline
businesses and direct-marketing activities of photographers. Photographer cooperatives,
however, are included into the analysis since they often licence material also from independent
photographers. For the year 2005, 201 agencies were identified through the industry asso-
ciation’s member directory, own research and through snow-ball sampling in the course of data
collection. Altogether, 75 agencies completed the questionnaire where 70 had at least one sales
alliance in rights managed photography. The adjacency matrix of this alliance network between
the set of 70 agencies represents the independent variable. Though the response rate of 37% is
not very high, the respondents represent the majority of the actual business in the market. Many
medium and large players took part in the study and most non-respondents were micro
businesses with often just one self-employed owner, and most of them unknown to the national
business association. Therefore it seems justified to assume that the complete network would
look quite similar to the one surveyed. The only alliances missing out in the survey are
peripheral alliances between non-responding microfirms.
Network regressions. Network data violate the assumption of independence between
observations and thus require alternative methods to generate test statistics. Hence, regression
analysis uses the QAP technique to test for associations with the dependent variable. The
quadratic assignment procedure (QAP) calculates the Pearson correlation coefficient for two
observed matrices and compares it with a self-generated distribution of correlations based on
random permutations of the reference matrix (Krackhardt 1987; Kilduff and Krackhardt 1994).
All regression analyses were computed with the software package Ucinet (Borgatti et al. 2002).
Table 1: Variable labels and definitions, dyad level
Sales alliances between all pairs of agencies for each year,
Difference in years of foundation in Germany
Difference in years of launching an online picture archive
Difference in number of employees
Difference in size of picture archive
Difference in number of photographers (suppliers)
Dummy, match for two firms as specialized (=0) or universal
Dummy, match for two firms according to the kind of
specialization of picture portfolio
Difference in percentage of foreign annual sales
Dummy, match for two firms according to their membership in
the national industry association BVPA
Dummy, match for two firms according to their location in the
same metropolitan region
C. Popularity bias
Difference in indegree
Difference in outdegree
Number of different firms through which a pair of firms is
Number of shortest paths between two agencies
5.1 Visualizing network growth in German stock photography
The study of network evolution is promising especially in early stages of network growth, since
network structure has not yet become fully institutionalized (Kenis and Knoke 2002). In the
case of stock photo alliances this seems to be the case since it is a more recent reaction to
digitization and competitive shift. New agencies and new ties in the alliance network had
remained quite limited throughout the 1990s with a clear and disproportional increase since
1999. This increase is reflected in rights-managed as well as royalty-free photography and also
in the domestic rights-managed alliance network in Germany (Figure 1). 46 agencies were
founded in the considered period between 1989 and 2005, while 29 had already been established
Figure 1: Growth of alliance network by members and by licence agreements, 1989-2005
Correlation analysis of firm attribute data supports the argument that digitization and royalty-
free photography reduce transactional friction and ease inter-agency cooperation. First,
companies with RF-photos receive a higher percentage of pictures from partner agencies and
they tend to generate a higher proportion of revenues internationally. At the same time, RF-
agencies do not differ in terms of firm age or employment size. Second, there is a more general
trend that the higher the proportion of digital images, the higher the share of photo supply
through sales part-ners. Technological and institutional change thus fosters the emergence of a
new division of labour and a growth in inter-agency sales partnerships.
Figure 2 displays four stages of the domestic RM alliance network in five-year intervals
between 1990 and 2005. The visualization of this network illustrates the rapid, exponential ex-
pansion of alliances since 2000 only. The network started off from a few disconnected cores in
1990 that remained rather stable and small throughout the 1990s. Toward the end of the decade,
as the number of new ties increased, one large component emerged in which tow thirds of all
alliances are connected either directly or indirectly. One third of alliances however forms a
rather peripheral landscape of various unconnected and isolated alliances. In the next analytic
section, social network analysis will be used to search for an explanation of this emerging
1989 1991 1993 1995 1997 1999 2001 2003 2005
number of firms
RM overall firms
RM domestic firms
1989 1991 1993 1995 1997 1999 2001 2003 2005
number of alliances
RM overall alliances
RM domestic alliances
Figure 1: Stages of the domestic RM sales alliance network, 1990-2005
5.2 Explaining tie formation in the rights-managed sales alliance network
This section applies the hypotheses on network growth to the domestic RM alliance network in
Germany. Two sets of analyses are carried out. First, in a static perspective the independent
variables are tested against the network of sales alliances all in the same year 2005. The results
for bivariate QAP regressions are summarized in Table 2. This produces a number of significant
explanatory variables for further analysis. Second and in a dynamic perspective, significant
variables from the first set of regressions in the year 2005 are used for a year-to-year
longitudinal analysis between 1995 and 2005. Each independent variable from one year tn is
compared to the alliance network structure of the subsequent year tn+1 thus providing a year-to-
year estimation of the hypothesized effect on the new network structure in the subsequent year.
The regression effects for the four hypotheses are found to be quite different.
Table 2: Bivariate QAP regression analysis: all sales alliances in 2005
C. Popularity bias
Homophily. Arguments about preferential partner selection based on homophily suggest that the
more similar two partners are to each other, the more likely will they engage in a relationship.
However, this proposition receives hardly any support in the context of the alliance network
between picture agencies. Most of the similarity variables are insignificant except for employ-
ment size, specialization and business association membership. According to this finding, col-
lecting image material in the same topic specialty and being a member of the business associa-
tion increases the odds of having a licence agreement. In contrast, however, dissimilarity also
drives alliance formation. The higher the difference in employment size, the more likely are two
agencies to initiate a licence agreement. This result would argue more in favour of diversity than
for similarity. It might reflect an emerging division of labour between big agencies that source
their image material increasingly from small supplying agencies (see Glückler 2005). It has to
be emphasized that none of the significant effects explains more than one percent of overall
variance in the distribution. Hence, the impact of firm similarity may be summarized as being
Geography. A second argument about preferential partner choice holds that geographical
proximity between partners increases the likelihood for an alliance. Three quarters of all agen-
cies in the 2005 alliance network are located in one of the top five metropolitan regions,
Hamburg, Munich, Berlin, Frankfurt and Cologne (Figure 3). Regression analysis shows that if
two firms are located in the same metropolitan region, they are more likely to have a licence
agreement. Though the association is highly significant, again, location does not explain more
than 0.4% of overall variation and thus can only be an auxiliary condition for alliance
formation. The analysis of a series of year-to-year regressions offers an additional and quite
remarkable finding. Over the ten-year period since 1995, co-location has become increasingly
important in the structuration of alliances. While until 1999, the pattern of domestic sales
partnerships was independent from geography, the new ties in the network have been
increasingly constrained by geographical propinquity within the dominant metropolitan regions.
The positive effect of co-location in the same metropolitan region has grown steadily over the
last five years, yet remaining a highly contingent condition for tie formation.
Figure 3: Metropolitan clustering of stock picture agencies in Germany, 2005
Popularity bias. The argument about accumulative advantage predicts that a partner with many
previous contacts is more likely to have new ties than a partner with only few prior relations.
Evidence from the regression analysis suggests that the higher the difference in the number of
licence agreements between a pair of agencies, the more likely are they to have a relationship.
This result applies to both, licence-giving and licence-taking, and reflects the overall
centralization of alliances on a smaller number of highly connected agencies. In line with the
former hypotheses, however, popularity bias accounts for less than 1% of the variance in the
distribution of network ties.
Multiconnectivity. Finally, the argument of multiconnectivity theorizes that multiple indirect
linkages between partners enhance the formation of direct ties between these organizations. In
comparison with the other hypotheses, this argument seems best to represent the pattern of
licence agreements in the German stock photography market. Point connectivity as well as
geodesic count are highly significant and explain around 20% of the variance in the distribution
of sales alliances. Both measures particularly focus on unconnected firms in the network and
develop notions of the quality of their indirect interconnection through third parties. If there are
many different intermediate contacts between any two firms, and if there are many alternative
shortest connections between the two, they can be considered as being multiply connected
(multiconnectivity) – although there is no direct link between them. The analysis of a series of
year-to-year regressions reinforces this association. Point connectivity and geodesic count for
each pair of agencies in year tn explain roughly between 40% and 60% of the network structure
in each subsequent year tn+1. In addition, the ten-year time series since 1995 conveys that
although both measures are continuously significant, their explanatory value has suffered in the
last couple of years because more and more new ties enter the alliance network and challenge
the existing structure (Table 3). One conclusion might be that when networks face major
expansion and high rates of new tie formation, the incumbent network structure imposes less
constraint on new ties and may experience a period of path-destruction.
Table 3: R²s for year-to-year QAP regressions of the alliance network, 1995-2005
1995 to 1996
1996 to 1997
1997 to 1998
1998 to 1999
1999 to 2000
2000 to 2001
2001 to 2002
2002 to 2003
2003 to 2004
2004 to 2005
* p < 0.05, ** p < 0.01, 2,000 random permutations; R²s are adjusted. Between 1989 and 1995 there were
only eight new entries and a total of 14 alliances. Year-to-year regression would not convey interpretable
results before 1995.
The information on the location of an agency is only available for 2005 and thus used consecutively for
regressions in each year of alliance network.
6 Discussion and conclusion
This paper has sought to combine research on network evolution with an evolutionary approach
to economic geography. A number of different theoretical arguments about the nature of
network expansion were discussed and empirically applied to the context of inter-organizational
alliances in stock photography. Although stock photography is a small market in financial terms
and has been largely ignored in academic research, it is a particularly interesting organizational
field for the study of network dynamics. The market has experienced fundamental changes with
respect to various dimensions: digitization in technological terms, new licence models in
institutional terms, and new organizational practices as well as major new entrants to the
incumbent field. It has been argued that these environmental changes have given rise to new
forms of sales alliances between picture agencies. Despite the fact that agencies renounce a
significant share of revenue with a distribution partner, they increasingly now engage in
domestic partnerships. Moreover, the sheer number and rate of new alliance formation has
grown exponentially over the last couple of years. Since this phenomenon is relatively recent
and still in an early phase, the evolving alliance network represents an ideal case for the study of
The analysis revealed that organizational homophily and popularity bias are relatively weak
conditions for estimating the locus of new alliances. Only organizational diversity, i.e.
difference in firm size, and geographical co-location were highly significant, yet relatively weak
predictors for network growth. The multiconnectivity-hypothesis (Powell et al. 2005) received
stronger support. It was measured as the extent to which two agencies were multiply and indi-
rectly connected via alternative third parties. Multiconnectivity explained on average half of the
variance in the distribution of alliances across the network over a consecutive ten-year period.
In sum, two agencies were more likely to form a sales partnership if they were located in the
same metropolitan region, had different firm size, and if they were indirectly connected through
other organizations in the field. On the other hand, while over time the tendency to cooperate
locally has increased, the impact of multiple linkage on the alliance network has decreased. This
is mainly because in recent years, many new agencies were established and formed many new
alliances in the network. Given the limit of this study on stylized survey data it is difficult to
interpret the emerging effect of geographical propinquity on new tie formation appropriately.
Although the picture trading business has become mostly virtual, the new uncertainties
emerging from the dynamic market environment seem to facilitate strategic partnerships in
spatial prox-imity. The local milieu or creative fields (Scott 2000) might be argued to represent
both, an imprinted institutional mode of communication among the creative professionals in a
city or city region (Marquis 2003), and a territorialized form of culture (Maillat et al. 1995) that
conveys positive communication externalities and reduces uncertainty through more
monitorable interaction. Though the association between geography and alliance behaviour
corresponds well with other empirical evidence (Powell et al. 2005), more detailed research on
the nature of sales alliances is necessary to support the findings in the context of stock
It is an open question for further research whether and under what conditions rapid network
expansion reproduces an imprinted structure over long times to lock-in as a path of network
evolution or whether it disrupts established patterns and creates a new path to network growth
(Martin and Sunley 2006). The study presented here is only a first and exploratory analysis to-
ward a currently emerging field of evolutionary theory on networks. For economic geography,
this is a promising line of reasoning not only because geography learns from their input but also
because it may essentially contribute to the development of solutions. If geography is a
continuous constraint (or enabling condition) for networks to form, how then do networks travel
over space in the course of their development and how do regions anchor these networks to
retain their dynamics locally?
The focus of this paper has been on the mutual effects that network structure imposes on future
tie formation and, vice versa, the effect of new ties on the existing network structure. For this
purpose, the research strategy has focused on a more standardized, quantitative approach to
these patterns of variation over time. The specific regional context that reflects in a set of
interrelated institutions, the particular set of actors and a given resource endowment will also
play an important role for the understanding of network formation in a geographical perspective.
Future research should dedicate more enquiry also to institutional aspects of geographical
Albert, P., Feyel, G. (1998) Fotografie und Medien. Die Veränderungen der illustrierten Presse.
In M. Frizot (ed.) Neue Geschichte der Fotografie. Köln: Könemann, 358-69.
Baum, J. A., Shipilov, A. V., Rowley, T. J. (2003) Where do small worlds come from? Indus-
trial and Corporate Change, 12: 697-725.
Borgatti, S. P., Everett, M. G., Freeman, L. (2002) Ucinet 6 for Windows. Harvard: Analytic
Boschma, R., Iammarino, S. (2007) Related variety and regional growth in Italy. SPRU Elec-
tronic Working Paper Series 162. Brighton: University of Sussex, SPRU - Science and
Technology Policy Research.
Boschma, R. A., Frenken , K. (2006) Why is economic geography not an evolutionary science?
Towards an evolutionary economic geography. Journal of Economic Geography, 6:
Boschma, R. A., Lambooy, J. G. (1999) Evolutionary economics and economic geography.
Journal of Evolutionary Economics, 9: 411-29.
Brass, D. J., Galaskiewicz, J., Greve, H. R., Tsai, W. (2004) Taking stock of networks and
organizations: A multilevel perspective. Academy of Management Journal, 47: 795-817.
CEPIC (2001) CEPIC European Industry Survey 2001 Berlin: CEPIC.
Fleming, L., Sorenson, O. (2001) Technology as a complex adaptive system: evidence from
patent data. Research Policy, 30: 1019-39.
Frizot, M. (1998) 1838-1840. Fotografische Entdeckungen. In M. Frizot (ed.) Neue Geschichte
der Fotografie. Köln: Könemann, 22-31.
Frosh, P. (2001) Inside the image factory: stock photography and cultural production. Media,
Culture & Society, 23: 625-46.
Glückler, J. (2005) Digitalisierung und das Paradox informatorischer Reichweite in der
Agenturfotografie. Geographische Zeitschrift, 93: 100-20.
Glückler, J. (2007) Economic geography and the evolution of networks. Journal of Economic
Geography, 7: 619-34.
Goddard, L. (1995), "Focus photo libraries: Pictures at the touch of a button," in PR Week.
Hite, J. M., Hesterly, W. S. (2001) The evolution of firm networks: from emergence to early
growth of the firm. Strategic Management Journal, 22: 275-86.
Kenis, P., Knoke, D. (2002) How organizational field networks shape interorganizational tie-
formation rates. Academy of Management Review, 27: 275-93.
Kilduff, M., Krackhardt, D. (1994) Bringing the individual back in: a structural analysis of the
internal market for reputation in organizations. Academy of Management Journal, 37:
Kilduff, M., Tsai, W. (2003) Social Networks and Organizations. London: Sage.
Kjemtrup, I. (1997) ASPP Photo Editor/Researcher Survey. Part 1. The Picture Professional
(American Society of Picture Professionals) www.aspp.com.
Knoke, D. (2001) Making connections. In D. Knoke (ed.) Changing Organizations: Business
Networks in the New Political Economy. Boulder, Colo.: Westview Press, 120-63.
Krackhardt, D. (1987) QAP partialling as a test of spuriousness. Social Networks, 9: 171-86.
Lavie, D. (2004) The evolution and strategy of interconnected firms: a study of the UNISYS
alliance network. Academy of Management Proceedings E1-E6.
Maillat, D., Lecoq, B., Nemeti, F., Pfister, M. (1995) Technology District and Innovation: The
Case of the Swiss Jura. Regional Studies, 29: 251-63.
Marquis, C. (2003) The pressure of the past: Network imprinting in intercorporate communities.
Administrative Science Quarterly, 48: 655-89.
Martin, R., Sunley, P. (2006) Path dependence and regional economic evolution. Journal of
Economic Geography, 6: 395-437.
McPherson, M., Smith-Lovin, L., Cook, J. (2001) Birds of a feather: Homophily in social
networks. Annual Review of Sociology, 27: 415-44.
Owen-Smith, J., Powell, W. W. (2004) Knowledge networks as channels and conduits: The
effects of spillovers in the Boston biotechnology community. Organization Science, 15:
Powell, W. W., Koput, K. W., Smith-Doerr, L. (1996) Interorganizational collaboration and the
locus of innovation: networks of learning in biotechnology. Administrative Science
Quarterly, 41: 116-45.
Powell, W. W., White, D., Koput, K. W., Owen-Smith, J. (2005) Network dynamics and field
evolution: The growth of interorganizational collaboration in the life sciences.
American Journal of Sociology, 110: 1132-205.
Rich, M. (1994), "The Queen's Awards 1994: The picture of success - Tony Stone Images sells
photographs around the world," in The Financial Times. London.
Rigby, D. L., Essletzbichler, J. (2006) Technological variety, technological change and a ge-
ography of production techniques Journal of Economic Geography, 6: 45-70.
Rowley, T., Behrens, D., Krackhardt, D. (2000) Redundant governance structures: An analysis
of relational and structural embeddedness in the steel and semiconductor industries.
Strategic Management Journal, 21: 369-86.
Rowley, T. J., Greve, H. R., Rao, H., Baum, J. A. C., Shipilov, A. V. (2005) Time to break up:
Social and instrumental antecedents of firm exits from exchange cliques. Academy of
Management Journal, 48: 499-520.
Sachs, C. D. P. (2003) ASPP 2003 members survey. The Picture Professional 31-32.
Scott, A. J. (2000) The Cultural Economy of Cities. London: Sage.
Sorenson, O. (2003) Social networks and industrial geography. Journal of Evolutionary Eco-
nomics, 13: 513-27.
Sorenson, O. (2005) Social networks, informational complexity and industrial geography. In D.
B. Audretsch, D. Fornahl, C. Zellner (eds) The Role of Labour Mobility and Informal
Networks for Knowledge Transfer. New York: Springer, 79-95.
Staber, U. (1997) An ecological perspective on entrepreneurship in industrial districts.
Entrepreneurship and Regional Development, 9: 45-64.
Staber, U. (2001) The structure of networks in industrial districts. International Journal of Urban
and Regional Research, 25: 537-52.
Stinchcombe, A. L. (1965) Social structure and organization. In J. G. March (ed.) Handbook of
Organizations. Chicago: Rad McNelly, 142-93.
Storper, M., Venables, A. J. (2004) Buzz: face-to-face contact and the urban economy. Journal
of Economic Geography, 4: 351-70.
Walker, G., Kogut, B., Shan, W. (1997) Social capital, structural holes and the formation of an
industry network. Organization Science, 8: 109-25.
Wilkinson, H. (1997) 'The new heraldry': stock photography, visual literacy, and advertising in
1930s Britain. Journal of Design History, 10: 23-38.
Zucker, L. G. (1986) Production of trust: Institutional sources of economic structure, 1840-
1920. Research in Organizational Behavior, 8: 53-111.