=Paper=
{{Paper
|id=Vol-1229/dynak2014_paper1
|storemode=property
|title=Exploring Dynamic Embeddedness: a Network Analysis of the Global Pharmaceutical Industry 1991-2012
|pdfUrl=https://ceur-ws.org/Vol-1229/dynak2014_paper1.pdf
|volume=Vol-1229
|dblpUrl=https://dblp.org/rec/conf/pkdd/ShijakuLU14
}}
==Exploring Dynamic Embeddedness: a Network Analysis of the Global Pharmaceutical Industry 1991-2012==
https://ceur-ws.org/Vol-1229/dynak2014_paper1.pdf
Exploring dynamic embeddedness: a network analysis of
the global pharmaceutical industry 1991-2012
Elio Shijaku1, Martin Larraza-Kintana2 and Ainhoa Urtasun-Alonso2
1
Dept. de Empresa, Universitat Autònoma de Barcelona, Catalonia, Spain
elio.shijaku@uab.cat
2
Dept. de Gestión de Empresas, Universidad Pública de Navarra, Pamplona, Navarre, Spain
(martin.larraza,ainhoa.urtasun)@unavarra.es
Abstract. We analyze the global pharmaceutical industry network using a
unique database that covers strategic transactions (i.e. alliance, financing and
acquisition collaborations) for the top 90 global pharmaceutical firms and their
ego-network partnerships totaling 4735 members during 1991-2012. The net-
work evolution is traced via a novel method based on the concept of dynamici-
ty that quantifies individual network members (i.e. actors) contribution to the
longitudinal period. Specifically, we observe dynamic embeddedness defined
for key network centrality measures, and capture the impact of the 2007-2008
global financial crises and the subsequent global and Eurozone recession ef-
fects on the strategic transaction flows between the industry’s key players as
well as their partners. Results suggest the feasibility of dynamicity as a dynam-
ic network indicator as well as the importance of constellation strategic trans-
actions in the study of large network perturbations.
Keywords: longitudinal social network, strategic transaction, dynamicity, dy-
namic embeddedness
1 Introduction
Organizations are inherently embedded actors of social networks, whose
structures evolve dynamically, and as a result of each actor’s involvement
offer important clues on organizational strategic behavior. Inside a dynamic
network, organizations exist as highly mobile entities with their relationships
and positional structures continuously changing in time. As such, under-
standing organizational behavior involves first and foremost capturing organ-
izational dynamics often done by analyzing the longitudinal context where
network dynamics is observed. While most literature on longitudinal net-
works focuses on a more holistic evolution of their structure [2, 4, 11], more
recent studies have highlighted the contribution of each actor to the overall
network dynamics [1, 6]. This actor-level approach embodied by the concept
of dynamicity, relies on the assumption that capturing organization’s dynam-
ic behavior in a given network should be based on a combined analysis of
both static and dynamic network topologies [12]. Additionally, dynamicity
enables researchers to study the effect of specific critical events (i.e. pertur-
bations) that greatly alter the structure of the panel network.
However, quantifying actor involvement and contribution in longitu-
dinal networks, and modeling its behavior against specific perturbations has
been limited, with research confined to the effect that organizational crisis
has had on organizational communication networks [6, 13]. These few avail-
able longitudinal network studies on actor contribution have analyzed net-
work dynamic evolution relying on embeddedness, a well-known concept in
social network analysis, long considered a highly strategic resource with im-
portant impacts on firm’s performance [5, 8]. On this matter, network litera-
ture has often relied on centrality-based embeddedness to provide a dynam-
ic image of social network evolution [9, 14].
Furthermore, the majority of research on embeddedness has ap-
proached the concept from a dyadic (i.e. a group consisting of only two ac-
tors) perspective, bypassing multiple types of firm interaction. Even those
studies that focus on the so-called constellation (i.e. interactions between
more than two actors) perspective [3, 7], miss out at the relevance of actors
engaging in constellation ties of multiple kind, by considering only a single
type of collaboration. Additionally, embeddedness’ studies have focused
heavily on strategic alliance collaborations, a choice well-grounded by the
interorganizational collaborations in any given industry, but that often fails
to embrace the full picture of strategic interactions’ multitude.
We fill these shortcomings by focusing on longitudinal networks gen-
erated from strategic transactions, a conceptualization of interorganizational
collaborations engaged by a firm with its network partners including strate-
gic alliances, acquisitions and financing collaborations, analyzed under both a
dyadic and constellation lens. By doing so, we contribute not only to the lit-
erature of alliance collaborations but enhance the currently undernourished
network literature on acquisition and financing collaborations as well which
play an important role in the dynamics of strategic organizational behavior.
Our study addresses the above gaps by developing and testing a the-
oretical framework that links the concepts of dynamicity, embeddedness and
strategic transactions. By doing so, we uncover the dynamic evolution of the
global pharmaceutical industry chosen for its intensive collaboration envi-
ronment. Given the novel nature of dynamicity as a concept, we attempt to
examine some fundamental questions that develop a theory-based under-
standing of dynamicity and its relationship with embeddedness, as well as
analyze the impact that strategic transactions have on such structure: How
does dynamicity of centrality measures evolve in a longitudinal network?
What is the role of strategic transactions in the evolution of such dynamicity?
How does actor´s dynamicity behave in the presence of exogenous events
that critically alter the network structure?
Specifically, we expand the actor-level approach on dynamic networks
by introducing the concept of dynamic embeddedness, defined as the indi-
vidual actor’s central position variability in a longitudinal network setting
compared to its central position variability in a aggregated network. For the
purpose of this paper, our focus is exclusively on the dynamicity of structural
embeddedness and particularly on the dynamicity of key network centrality
measures such as degree, betweenness and closeness. Specifically, we build
on a dynamicity model [12] by exploring the critical impact that large exoge-
nous perturbations, such as the 2007-2008 financial crisis, the subsequent
2008-2009 global recession and the more local Eurozone recession of 2011-
2013 have on longitudinal networks between top-level actors and their ties
in the global pharmaceutical industry.
2 Data and measures
We conduct our analysis on a longitudinal dataset (t = 22 years, 1991-2012)
comprising the strategic transactions of 90 leading firms from the pharma-
ceutical industry in Western Europe, United States, Asia, Africa and Australia.
The sample is selected by identifying those firms that appear at least once in
the top 50 of the Pharmaceutical Executive Magazine yearly editions for the
period 2002-2013. We then use the Pharma and Medtech Business Intelli-
gence database to collect all the strategic transactions that involved the
firms in question from 1991 to 2012. During this period, the 90 firms of the
sample engaged in alliance, financing and acquisition collaborations with
4645 other firms and institutions creating a total of 12055 strategic transac-
tions.
Due to our selection process, we consider two types of firms, the core firms
comprised of the top 90 pharmaceuticals and the periphery firms including
the rest of the actors, with a total population of 4735 firms whose full list is
available from the authors. The obtained longitudinal data for both core and
periphery firms presents missing actors, since some firms are acquired by
others, or simply are not active for any particular year. Our analysis also in-
cludes financial data obtained from COMPUSTAT and DATASTREAM data-
bases, supplying missing data when possible using company annual reports.
Since the financial data concerns firms from different countries, we convert
all currencies to USD with an exchange rate based on the particular year the
data is retrieved.
We model each year over the sample period as a separate social
network and analyze each network based on a similar approach for the glob-
al banking network analysis [10]: (i) the core network, referring to the ties
between the top 90 actors; and (ii) the full network comprising all available
data from a total of 4735 actors. In our analysis, we consider a weighted un-
directed tie approach, defined as an N x N “weight” matrix, whose generic
entry wij = wji > 0 measures the interaction intensity between any two actors
(zero if no link exists between actor i and j). Following this framework and
using the software R, we build 22 symmetric 90 x 90 matrices to track the
evolution of the core network and 22 symmetric 4735 x 4735 matrices to
track the evolution of the full network for the period 1991-2012. Additional-
ly, for dynamicity calculation purposes, we build two matrices which include
the aggregated strategic transactions of the entire 22 years period for both
types of network.
Network indicators. The network measures of our analysis include
three centrality variables (degree, betweenness and closeness centrality) and
the dynamicity variable representing the variability of the structural positions
of an actor in all short-interval networks compared to its structural position
in the aggregated network [12] as shown in equation 1:
m
i
¦D
t
t ,t 1 u OVAN OVt
DDA (1)
m
where DDAi is the degree of dynamicity shown by ith actor, OVAN is the
observed value (i.e. degree centrality) for the aggregated network, OVt is
the observed value (i.e. degree centrality) for tth yearly network for the ith
actor, m is the number of yearly networks considered in the analysis, and
D t ,t 1 is a constant valued according to whether the actor is present or miss-
ing in the current and previous short-interval network. The presence of this
constant is of crucial important to properly count for actors that disappear
from the network due to simple inactivity or possible lack of presence due to
network dynamics. The possible combination values that D t ,t 1 can take are
given in Table 1.
Table 1. Possible combination of presence and absence of an actor in two consecutive short-
interval networks (Source: Uddin et al. 2013)
Current SIN Previous SIN
ɲt,t-1
(Present/Absent) (Present/Absent)
Present Present ɲp,p= 1.0
Present Absent ɲp,a= 0.5
Absent Present ɲa,p= 0.0
Absent Absent ɲa,a= 0.0
For the first short-interval network (i.e. D i ,0 for t = 0) of our analysis, the
value of the constant depends on the presence or absence of each actor (i.e.
either 0 or 1) at that particular period. The dynamicity model [9] differenti-
ates between two types of dynamicity measures, the dynamicity of an actor
represented by equation 1 and the average dynamicity shown by an actor of
the tth short-interval network represented by equation 2:
wt
t
¦Dt
t ,t 1 u OVAN OVt
DDN (2)
wt
where DDN t is the average degree of dynamicity shown by an actor of the
tth short-interval network meaning the contribution of each actor to the
short-interval network´s dynamicity, and wt is the total number of actors in
the tth short-interval (i.e. yearly) network. Therefore, our analytical approach
is based on three variables: degree dynamicity, betweenness dynamicity and
closeness dynamicity constructed by substituting each obtained centrality
value to equations 1 and 2.
Industry indicators. In order to analyze the effect of exogenous criti-
cal events such as financial crises and recessions on the global pharmaceuti-
cal industry, we construct two main effect variables: (i) global crisis repre-
senting the combined effect of the 2007-2008 financial crisis and the global
recession of 2008-2009 that followed as a direct consequence, and con-
structed as a dummy variable that takes the value of 1 for the years 2007-
2009 and zero for the rest, and (ii) local crisis representing the exogenous
effect of the Eurozone recession during 2011-2013, and constructed as a
dummy variable that takes the value of 1 for the years 2011-2012 and zero
for the rest.
Control indicators. We use several actor-specific measures such as
strategic transaction frequency, R&D intensity, profitability, headquarters
(HQ) location and financial leverage age and size. Strategic transaction fre-
quency represents the relative frequency in percentage with which firms
engage in strategic transactions. In the analysis, we differentiate between
the frequency in percentage of firms engaging in alliance, financing and ac-
quisition collaborations. R&D intensity represents the firm’s R&D expendi-
ture scaled by total sales while profitability is measured for each firm by
computing the ratio of net income to total assets (ROA). We define financial
leverage as the debt-to-total assets ratio including both short- and long-term
debt and control for the age of the firms, operationalized as the foundation
year minus the year considered in the 2002-2012 panel analysis, and size
operationalized as the natural logarithm of company’s employees. Finally,
since our data consists of multinational firms and knowing that the majority
of the top 90 firms are US- or EU-based, we control for headquarters (HQ)
location based on two separate dummy variables representing whether firms
are U.S. or EU-based.
Model approach. By using a two-step approach to our analysis, first
we assess the stability of dynamicity distributions in selected years to cap-
ture statistical differences throughout our data using Kolmogorov-Smirnov
(henceforth, KS) tests for both core and full networks, second by controlling
for firm-specific effects, we investigate the effect that the global crisis (in-
cluding the 2007-2008 financial crisis and the great 2008-2009 recession),
and the local crisis referring to the Eurozone recession, observed for 2011-
2012, have on degree, betweenness and closeness dynamicity. For our sec-
ond step, we run a panel regression model based on random effects (hence-
forth, RE) with robust estimations based on the model seen below:
Yit E k X it D i uit H it , i 1,...,90, t 1,...,10,
where
Yit is firm´s dynamic embeddedness considered as a dependent varia-
X
ble, it is a vector of firm and industry-specific independent variables includ-
ing global and local crises, age, size, profitability, financial leverage, R&D in-
tensity, transaction frequency and firm location,
D i is the unknown intercept
u H
for each firm, it is the between-firm error, it is the within-firm error, k is
E
the coefficient for each k independent variable, i is the number of firms (90
in total) and t is period of time considered (10 years in total or +/- 5 years
window before and after the offset of the 2007-2008 financial crisis).
3 Results
We describe the dynamics of the global pharmaceutical industry using four
key estimates: (i) tracking dynamic embeddedness evolution based on aver-
age dynamicity estimate plots, (ii) monitoring the stability variation of actors’
dynamic embeddedness based on KS-tests, (iii) constructing the top five firm
rankings based on yearly network average dynamicity estimates, and (iv)
understanding the global and local crises causative effect on dynamic em-
beddedness based on panel regression estimates. Results (i) – (iii) concern
the total panel period 1991-2012 while results (iv) concern the panel period
2002-2012.
We track dynamic embeddedness evolution by plotting the cross-
sectional averages of dynamic indicators during 1991-2012 as seen in Figure
1. Both panels show that dynamicity values present relative stability before
2007 for degree centrality but vary substantially for betweenness and close-
ness centrality throughout the study period. Specifically, for the core net-
work, degree and betweenness dynamicity drop respectively 20 percent and
17 percent while closeness dynamicity is almost halved by 40 percent during
the global crisis. The more local Eurozone crisis of 2011-2013 (of which we
analyze only one year due to sample structure) shows a similar pattern with
both networks’ dynamicity severely reduced. An exception is closeness cen-
trality, whose dynamicity shows an upward trend for the core network, with
signs of a more clustering-oriented tendency.
Fig. 1. Core network (left) and full network (right) dynamic embeddedness evolution
We monitor the stability of both core and full networks by comparing the
dynamicity distribution in the first year of each decade including last availa-
ble year’s data (1991, 2001 and 2012) with subsequent years in the same
decades, a procedure seen in global banking network analysis [10] and
whose results are given in Table 2.
Table 2. Empirical distribution stability for dynamic embeddedness
Table 2 shows the proportion of years when the dynamicity distribution is
statistically different (at 5 percent level of significance) in each decade com-
pared to 1991, 2001 and 2012. Values of zero mean that the distribution of a
particular year compared to a particular decade are statistically close, as is
the case for degree and betweenness dynamicity for the years 1991 and
2001 when compared with the 1991-2001 period. This means that in both
core and full networks, the firms have kept a similar centrality structure. On
the other hand, the distribution for the decade 2002-2012 is statistically dif-
ferent for almost all dynamicity variables in both core and full networks,
meaning that the actors’ dynamicity has been highly unstable for the second
decade. An exception concerns betweenness dynamicity for the full network,
whose results show a relatively unaffected actors’ brokerage tendency, with
only 18 percent of significant distribution change.
Looking at Table 3, we observe that the top five ranking for both de-
gree and betweenness dynamicity includes the biggest pharmaceutical firms
(based on their average total sales) which are not underlined, meaning that
these firms score high in their centrality position during the core network
evolution. Interestingly, closeness dynamicity shows only two big pharma-
ceuticals in the top five, with a clear tendency of smaller firms reducing their
mutual proximities. However, big pharmaceutical firms’ hegemony is rein-
stated in the core network with big pharmaceuticals scoring high in all cen-
trality measures.
Table 3. Firm rankings (1991-2012) for both core and full networks
Table 4. Dynamic embeddedness during exogenous perturbations: regression estimates
Looking at the main effects of the regression analysis, we observe the nega-
tive effect of the global crisis on dynamicity indicators except degree dy-
namicity, for which the effect is not significant, meaning that the combined
effect of the 2007-2008 crisis and the subsequent global recession of 2008-
2009 have not significantly affected the number of strategic transactions
originating from each of the core network members. Moreover, we find
strong statistical significance for the negative effect that the local Eurozone
crisis has had on firms’ dynamic embeddedness. Additionally, the type of
strategic transaction is found to influence dynamic embeddedness. This ef-
fect is understandable considering the relatively high distribution of alliance
transactions in the sample (about 75 percent). However, the positive and
significant effect of acquisition transactions on degree dynamicity is interest-
ing considering that both acquisition and financing transactions show similar
distributions in the sample (about 12.5 percent each). Finally, the observed
low R-squared is not necessarily a drawback for the chosen model particular-
ly if we consider that the results present statistically significant predictors
and the regressors are used in a panel setting.
4 Discussion and concluding remarks
With respect to the analyses’ objective, the results on firm´s dynamic em-
beddedness suggest that prior to the global crises the global pharmaceutical
industry has been relatively stable, with firms’ centrality reflecting their mar-
ket position. Specifically, the top pharmaceutical firms that rank high in
terms of sales have a noticeable central position in both core and full net-
works as observed in the firm rankings. Dynamically speaking, the global
pharmaceutical industry has reduced its activity to even lower levels than the
beginning of our sampling data, year 1991. While the reduction varies for
specific centrality measures, its effect is more prominent after 2007, which
coincides with the offset of the 2007-2008 financial crises. The regression
results confirm this by showing significant dynamicity reduction during both
crises. Furthermore, the regression results indicate that the Eurozone reces-
sion has had a far deeper negative effect on global pharmaceutical industry
than the global recession.
This study also highlights the importance of acquisition transactions in
the expansion of the firms’ importance as central hubs. Specifically, the sig-
nificant effect of acquisitions on degree dynamicity demonstrates the impact
that different strategic transactions have on centrality indicators and further
reinforces the reasoning behind our choice to study the centrality measures
evolution via the dynamicity concept. However, this also raises questions as
to why comparable effects of strategic transaction types (i.e. acquisitions and
financings) respond differently to centrality-based dynamicity.
Our study’s limitations could potentially provide interesting areas of
future research. First, we should be careful when generalizing our results
about the global pharmaceutical industry, knowing that not all firms in both
core and periphery networks are dedicated to pharmaceuticals but come
from other adjacent industries such as biotechnology and chemicals. Second,
dynamicity measure calculation is based on a novel design which takes into
account missing actors during network evolution using a specific constant
which should be subject to further research for proper values’ assignment.
Finally, the dynamicity measure could be used for other centrality measures
(i.e Eigenvector, Bonacich Power) or be included in the analysis of network
measures such as actor’s structural similarity, structural holes and brokerage
elasticity.
Acknowledgments. The authors wish to acknowledge the financial support
provided by the Spanish Ministry of Science and Education under projects
ECO2010-21393-C04-01 and ECO2013-48496-C4-4-R.
References
1. Braha, D., Bar-Yam, Y.: From Centrality to Temporary Fame: Dynamic Centrality in Com-
plex Networks. Complexity, 12, 2, 59-63 (2006)
2. Brandes, U., Lerner, J., Snijders, T.A.B.: Networks evolving step by step: Statistical analysis
of dyadic event data. In: 2009 International Conference Advances in Social Network Anal-
ysis and Mining (ASONAM 2009), IEEE Computer Society, pp. 200-205 (2009)
3. Das, T.K., Teng, B-S.: Alliance Constellations: A Social Exchange Perspective, The Academy
of Management Review, 27, 3, 445-456 (2002)
4. Gull, K.C., Angadi, A.B., Malagi, K.B.: Framework for Analysis of Dynamic Social Networks
Evolution. The International Journal of Engineering and Science, 1, 2, 260-268 (2012)
5. Hoffmann, W.H.: Strategies for managing a portfolio of alliances. Strategic Management
Journal, 28, 8, 827-856 (2007)
6. Hossain, L., Murshed, S.T., Uddin, S.: Communication network dynamics during organiza-
tional crisis. Journal of Informetrics, 7, 16-35 (2013)
7. Inkpen, A.C., Tsang, E.W.K.: Social Capital, Networks, and Knowledge Transfer. The Acad-
emy of Management Review, 30, 1, 146-165 (2005)
8. Karamanos, A.G.: Leveraging micro- and macro-structures of embeddedness in alliance
networks for exploratory innovation in biotechnology. R&D Management 42, 1, 71-89
(2012)
9. Lin, Z., Peng, M. W., Yang, H., and Sun, S. L.: How do networks and learning drive Man-
dAs? An institutional comparison between China and the United States. Strategic Man-
agement Journal, 30: 1113–1132 (2009)
10. Minoiu, C., Reyes, J.A.: A network analysis of global banking: 1978-2010. Journal of Finan-
cial Stability, 9, 168-184 (2013)
11. Snijders, T.A.B.: Statistical models for social networks. Annual Review of Sociology, 37,
129-151 (2011)
12. Uddin, S., Piraveenan, M., Khan, A., Amiri, B.: Conceptual quantification of the dynamicity
of longitudinal social networks. SocialCom conference. arXiv:1311.0090v1 (2013)
13. Uddin, S., Chung, K.S.K., Piraveenan, M.: Capturing Actor-level Dynamics of Longitudinal
Networks. ASONAM '12 Proceedings of the 2012 International Conference on Advances in
Social Networks Analysis and Mining (ASONAM 2012), 1006-1011 (2012)
14. Yang, H., Lin, Z(J)., Peng, M.W.: Behind acquisitions of alliance partners: exploratory learn-
ing and network embeddedness. Academy of Management Journal, 54, 5, 1069-1080
(2011)