On the relationship between economic performance and position of

businesses in supply chain networks

Andreas Seiler
Salford Business School, University of Salford, United Kingdom

Christos Papanagnou
Salford Business School, University of Salford, United Kingdom

Phil Scarf
Salford Business School, University of Salford, United Kingdom

Abstract

This paper investigates how the network position of a focal company in a scale-free supply
chain network impacts the economic performance of that company. Network flows of a
sample of focal companies are merged to create a section of the scale-free supply chain
network. Social network analysis provides data on network position of the focal companies.
Network position is studied for its impact on a variety of financial performance measures.
Several findings confirm dependencies between network position and economic
performance. The analysis of different performance measurement models recognises
Bonacich power (connectedness to connected companies) as one major influencing factor.

Keywords

Performance measurement, supply chain management, social network analysis, continuous

improvement

1
Introduction

Many organisations, to be more effective and efficient, link with other organisations and form
supply chains, and the nature and extent of the linkages in these supply chains are expected
to impact upon business dynamics and company performance. Furthermore, global
manufacturing has led to complex supply chain networks, which require a new framework for
performance analysis (Bourne et al., 2000; Melnyk et al., 2014). In addition, organisations
realise that, in the future, competition to attract more end customers will be transferred to the
supply chain level, pitting one supply chain against another. Thus, it is crucial to understand
competitive advantage by establishing a set of metrics that can measure and evaluate the
performance of the entire supply chain, and thereby guide strategic interventions.
In general, performance measurement systems are associated with reference models
containing a standard description of management processes, a framework of relationships
among the standard processes, and standard metrics to measure process performance
(Bititci et al., 1997). However, companies often base their management and analysis on
experience and intuition, neglecting to use performance measurement tools (Simchi-Levi et
al., 2004). Veen-Dirks (2010) studied performance measurement for two distinct cases,
distinguishing the set of measures that managers present within periodic evaluation and the
set of measures used to determine the managers’ variable rewards.
A wide categorisation of performance measures is obtained by sorting the different
measures into financial and non-financial measures, both often referred to as synthetic
process indicators (Merchant and Van der Stede, 2007). Many companies, when typically
monitoring their broad set of activities and processes, use financial performance measures.
When these processes are linked to external entities, the focus needs to be extended
beyond the boundary of organisations, constructing inter-organisational management and
network accounting systems (Chenhall, 2005). These systems should better support a single
organisation (called a focal organisation) working within a network. They should also better
support the management and performance of the entire network, by monitoring the ability of
the network itself to leverage network output.
Thus, it is no longer sufficient to analyse performance only from the narrow perspective of
the focal firm and only in financial terms (Bititci et al., 2006). Therefore, the research that we
present in this paper uses social network analysis to investigate the linkage between
companies in supply chain networks and their performance. In so doing, we evaluate overall
network performance by measuring key position variables that are derived from social
network theory and that reflect the entire network structure and its relationships. Our study
focuses on SMEs in the German plastics processing industry and develops a new
methodology to consolidate the supply chain networks of focal companies. This approach
identifies common business partners of focal companies and reveals important connections.
We study a single, complete supply chain network of fifteen focal companies and their
suppliers and customers. A software tool created by us merges the supply chain data of the
focal companies into the single, scale-free network. Our approach is more holistic than the
standard analysis of dyadic connections of company-specific supply chains. We collect data
on the network position of the focal companies by detecting and interpreting patterns of
bonds among all stakeholders within the network. We evaluate financial performance of
these companies using business reports drawn from the business register of the German
Federal Gazette and data available from commercial providers of digital business
information such as Bisnode. In our network, we study 448 companies in all, the fifteen focal
companies and their 433 suppliers and customers. Connections between the companies are
evaluated on the basis of value flows (cash flows for products or materials).

Literature review

Social network analysis has shown its potential in many disciplines including supply chain
networks. The framework of Bellamy and Basole (2013) distinguishes research on system
architecture, system behaviour and system strategy. They found relatively little research that

2
attempts to use social network analysis in supply chain management. As we aim to find
those network characteristics that impact performance, the present study spans system
architecture and system strategy.
Jarillo and Stevenson (1991) compare companies that became successful because they
turned competitors into allies. The focus here is the establishment of strategies that help
reap the benefits of cooperation rather than striving to compete in markets. Harland et al.
(2001) compare supply chain networks with different types of managerial challenge.
Empirically, they present a taxonomy of supply chain networks, and claim that depending on
the network type, managers face different challenges with respect to operating and network
formation. By embedding social network analysis into supply chain networks, Kim et al.
(2011) perform a quantitative study that aims to develop a conceptual framework for the
analysis of the supply chain networks of three different automotive product lines. Based on
the idea of comparing different supply chain networks, these studies give rise to the question
as to whether we can link network position to quantitative performance measures at the
node level.
The literature in supply chain management suggests that there are many performance
measures, both qualitative and quantitative (Beamon, 1998). Qualitative measures are those
with no single direct numerical measurement, although some aspects of them may be
quantified. The most important are: (i) customer satisfaction, (ii) flexibility, (iii) effective risk
management, and (iv) supplier performance. In contrast, quantitative performance measures
are directly described numerically, and may be categorised by (i) objectives that are based
directly on financial level indicators (e.g., return on assets), and (ii) objectives that are based
on non-financial indicators (e.g., customer responsiveness) (Morgan, 2007). In our research
we opt for a quantitative approach by focusing on measuring performance beyond
organisational boundaries.
One approach to study whether performance depends on different network-based
characteristics is the calculation of the so-called ego-network quality (Borgatti and Li, 2009).
Ego-network quality facilitates a comparison among focal companies by taking into account
context and relationships. To substantiate types of relationships, it is conceivable to apply
the concept of hubs (pointers to popular companies) and authorities (pointers from popular
companies). However, we want to contribute to knowledge by taking a “bird’s eye” view of a
scale-free supply chain network. In so doing, we address the lack of a study “integrating the
insights gained from conceptual, empirical, and modelling/simulation work on supply chain
system architecture” (Bellamy and Basole, 2013). As with Pavlov and Bourne (2011), the
inclusion of the network characteristics into a performance measurement model creates new
opportunities for performance improvement, strategic rethinking, and recommendations for
action.

Methodology

Our purpose is to test the strength of links, node centrality and link diversity for their
influence on economic performance. Our analysis considers a sample of businesses,
labelled the focal companies, operating in the German plastics processing industry. To
describe our methodology, we first set out our research questions. We then formulate these
as specific hypotheses. Next, we define measures of economic performance (Table 1) and
network position (Table 2 and Table 3) with which we test these hypotheses. Finally, we
describe data collection on the network of interest and how the values of our measures are
calculated.

Research questions and hypotheses

A strong relationship with partners across the network is expressed by trust and has
advantages in terms of information sharing, sharing of investments and fast, constructive
feedback. Links of various strength characterise a company’s position within the supply
chain network. Thus, our first research question asks:

3
RQ1: Does the performance of a company depend on the strength of its links within the
supply chain network?
Each relationship between a focal company and its business partners is characterised by
cash flow related to procurement or sales. To draw conclusions regarding the impact of the
strength of the links on the economic performance, social network analysis allows us to
calculate aggregated strength. Thus, we calculate aggregated strength based on the
proportion of such cash flows between focal companies and their mutually shared business
partners. Table 3 provides the definition. Our precise hypothesis is then:
H1: The higher is the aggregated strength of links of a company in the supply chain
network, the better is the performance of that company.
One might argue that larger companies will tend to have stronger links than smaller
companies. Therefore, we also consider the effect of node centrality in the supply chain
network on economic performance. Our second research question asks:
RQ2: Does the performance of a company depend on the centrality of that company within
the supply chain network?
Social network analysis provides a variety of centrality measures. Initially originating from the
flow-mechanism, centrality is reflected in the quantity of direct links to partners. A central
position in the network ought to strengthen the negotiating position with partners. Following
Robins (2015, p. 182), it is best to at least focus on degree centrality and betweenness
centrality. In our case of short network paths between focal companies and their business
partners, both measures would correlate (see Table 3). Therefore, we omit betweenness
centrality and test the following hypothesis:
H2-1: The higher is the degree centrality of a focal company in the network, the better is the
performance of that company.
Bonacich power generalises the notion of centrality to accommodate circumstances in
which being connected to well-connected nodes brings positive (β > 0, see Table 3) or
negative (β < 0) consequences (Bonacich, 1987). Since we expect effects of the immediate
environment of focal companies on economic performance, we test the following hypothesis:
H2-2: The higher is the Bonacich power (β > 0) of a focal company in the network, the
better is the performance of that company.
Finally, a network position that is characterised by diverse links ought to reduce
dependency because alternatives are available. Firstly, we study diversity based on the
product flows of focal companies. The third research question then asks:
RQ3: Does the performance of a company depend on the diversity of product flows of that
company in the supply chain network?
In order to assess diversity on the level of product flows, we apply the concept of hubs and
authorities (Borgatti and Li, 2009). By formulating the following hypotheses, this concept
allows us to evaluate the diversity both on the procurement side, as well as on the sales side
of focal companies:
H3-1: The higher is the share of product flows from a focal company to its customers in the
supply chain network (hubs), the better is the performance of that focal company.
H3-2: The higher is the share of product flows to a focal company from suppliers in the
supply chain network (authorities), the better is the performance of that focal
company.
A focal company gets a high-value of HUB for delivering to customers which have many
focal companies as suppliers. Further, a focal company gets a high-value of AUTH for being
supplied by suppliers which have many focal companies as customers. The underlying idea
is that the more a focal company is linked to hubs and authorities, the more diverse are its
relationships within the network.

4
 Additionally in relation to diversity, we anticipate that companies which are able to satisfy
the needs of different markets, e.g. via a large product variety, should be more successful.
Consequently, our final research question asks:
RQ4: Does the performance of a company depend on diversity in its affiliation to different
complementary sectors of industry besides the focal industry?
Looking for structural isomorphism, we create classes of industries that refer to aspects of
network role theory (Borgatti and Everett, 1992). Nodes that are structurally similar to each
other are reduced to classes that share an affiliation in the same industry. Assuming links to
various markets, the vulnerability to fluctuations in demand ought to be reduced and
exogenous influences may have less drastic consequences (Klibi et al., 2010). Thus, based
on the number of different complementary industries IND we test the precise hypothesis:
H4: The higher the number of complementary sectors of industry besides the focal
industry to which a company is connected, the better is the performance of that
company.

Definition of financial and network position variables

To evaluate financial performance comprehensively, we chose measures that cover aspects

of profitability, liquidity and efficiency (Table 1). The notation in Table 2 defines entities that
are used in the definitions of the network position measures or variables in Table 3.
In Table 3, we briefly indicate the nature of each measure. We make some further
detailed comments in relation to Bonacich power here. Bonacich power (BP) is a more
flexible measure of centrality than degree centrality (C) and eigenvector centrality (EC) and it
generalises these measures. If β = 0, then BP is equivalent to C. Further if β = 1/λmax, BP is
equivalent to EC. When it may be advantageous to be connected to those who are
themselves not well-connected then β < 0 is appropriate, wherein power derives from being
connected to the powerless, and to the contrary, having many powerful partners can reduce
one’s own power. Finally, we note that β acts as a weight on the centrality score of the
neighbours of a node, whereby a small absolute value of β gives more weight to local
network structure than distant network structure, and vice versa. Choosing β is a matter of
the analyst (Borgatti et al., 2013). The parameter α is a scaling parameter that is typically
chosen so that BP does not depend on the size, v, of the network. In fact, all the centrality
measures that we use are rescaled in this way.

Table 1 - Notation and definition of financial performance variables

Operating 𝑂𝑃 = 𝑝𝑟𝑜𝑓𝑖𝑡 𝑎𝑓𝑡𝑒𝑟 𝑡𝑎𝑥 OP is the difference between revenue
profit, OP +𝑡𝑎𝑥 (sales) and expenditure (costs).
+𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡
+ 𝑑𝑒𝑝𝑟𝑒𝑐𝑖𝑎𝑡𝑖𝑜𝑛
Return on 𝑅𝑂𝐴 = 𝑝𝑟𝑜𝑓𝑖𝑡 𝑏𝑒𝑓𝑜𝑟𝑒 𝑡𝑎𝑥 ROA is an indicator of profitability. This
Assets, ROA 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 financial performance measure is generally
independent of the size of a company.
Asset 𝐴𝑇 = 𝑟𝑒𝑣𝑒𝑛𝑢𝑒 OP and ROA focus on profitability. AT
turnover, AT 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 instead indicates the efficiency with which a
company is able to deploy its assets.
Dynamic debt 𝐷𝐷𝑅 = 𝑑𝑒𝑏𝑡 DDR measures financial risk and indicates
ratio, DDR 𝑐𝑎𝑠ℎ 𝑓𝑙𝑜𝑤 the time in years it would take a company to
pay off its debts.

5
 Table 2 - Notation used to define network position variables
𝐺D (𝑉, 𝐴) The directed graph that is the set 𝑉 of nodes (companies) and the set 𝐴 of arcs
(cash flows). The companies are the focal companies, their suppliers, and their
customers.
𝑣 𝑣 = |𝑉|, the number of nodes in 𝐺D (𝑉, 𝐴) (the number of companies in the
network).
𝑢 The number of focal companies, 𝑢 < 𝑣.
𝑤 The number of industries in which the focal companies trade.
𝑥𝑖𝑗 Defined on 𝐺D (𝑉, 𝐴), 𝑥𝑖𝑗 ≥ 0 for all i and j is the weight of the arc from node i to
node j (monetary value of the procurement by company i from company j). In
terms of the supply chain network we study, this is the cash flow from a company
to its supplier to pay for materials or the cash flow from a customer to a company
to pay for manufactured product. Note: in the network we consider, a company
may act as both a customer and a supplier.
X The 𝑣 × 𝑣 matrix (𝑥𝑖𝑗 ) of cash flows, called the cash flow matrix.
𝐺U (𝑉, 𝐸) The undirected graph that is the set 𝑉 of nodes (companies) and the set 𝐸 of
edges (links).
𝑦𝑖𝑗 Defined on 𝐺U (𝑉, 𝐸), for all i and j, 𝑦𝑖𝑗 = 1 if company i trades with company j and
𝑦𝑖𝑗 = 0 otherwise; 𝑦𝑖𝑗 indicates the presence or absence of an edge (link) in
𝐺U (𝑉, 𝐸).
Y The symmetric 𝑣 × 𝑣 matrix (𝑦𝑖𝑗 ), called the adjacency matrix.
𝐻D (𝑊, 𝐵) The directed graph that is the set 𝑊 of nodes (the u focal companies and w
industries in which they operate) and the set 𝐵 of arcs (from a focal company to
an industry if the focal company operates in that industry). This network
simplifies the network 𝐺D (𝑉, 𝐴) by aggregating, into industries, the trade between
focal companies and their suppliers and customers.
𝑟𝑖𝑗 Defined on 𝐻D (𝑊, 𝐵), 𝑟𝑖𝑗 = 1 if company i operates in industry j, 𝑟𝑖𝑗 = 0
otherwise.
R The 𝑢 × 𝑤 matrix (𝑟𝑖𝑗 ), called the affiliation matrix.
𝑝𝑖𝑗 Defined on 𝐺D (𝑉, 𝐴), 𝑝𝑖𝑗 is the number of different types of product procured by
company i from company j in 𝑉. The product types procured by i from j each
have a corresponding cash flow that sum to 𝑥𝑖𝑗 .
P The 𝑣 × 𝑣 matrix (𝑝𝑖𝑗 ), called the product-mix matrix
𝑔𝑖𝑗 Defined on 𝐺D (𝑉, 𝐴), 𝑔𝑖𝑗 is the number of arcs in the shortest path from node i to
node j
G The 𝑣 × 𝑣 matrix (𝑔𝑖𝑗 ), called the geodesic distance matrix.
ℎ𝑖𝑗𝑘 Defined on 𝐺D (𝑉, 𝐴), the shortest path from node i to node k that passes through
node j.

6
 Table 3 - Definition of network position variables
𝑥𝑖𝑗 𝑥𝑗𝑖
Aggregated 𝐴𝑆𝑖 = ∑𝑗 ∑ + ∑𝑗 ∑ The aggregated share of cash flows
strength, AS 𝑘 𝑥𝑘𝑗 𝑘 𝑥𝑗𝑘 from and to company i.

Degree 𝐶𝑖 = ∑𝑗 𝑦𝑖𝑗 Total number of companies with direct

centrality, C links to company i.

Betweenness 𝐵𝐶𝑗 = ∑𝑖<𝑘 ℎ𝑖𝑗𝑘 /𝑔𝑖𝑘 How often company j lies on the
centrality, BC shortest path between any two other
companies (Borgatti et al. 2013, p. 174)
Eigenvector The eigenvector centrality vector, A centrality measure in which
centrality, EC e, is the solution of the set of connections (links) to well-connected
linear equations Ye = λe for nodes score more highly, in relative
which λ, a scalar, is maximum. terms, than connections to less well-
This λ, λmax, is the largest connected nodes.
eigenvalue of Y. The ith
component 𝑒𝑖 of e is the
eigenvector centrality of node i.
Bonacich 𝐵𝑃𝑖 (𝛽) = ∑𝑗(𝛼 − 𝛽𝐶𝑗 )𝑦𝑖𝑗 A more general measure of centrality
power, BP than C and EC.
𝑝𝑖𝑗 Similar to aggregated strength, but this
Hubs, HUB 𝐻𝑈𝐵𝑖 = ∑
∑𝑘 𝑝𝑘𝑗 is not the value of product (cash flow)
𝑗
but the proportion of product types sold
that is aggregated, and the more
diverse the product types a company
provides upstream in the supply chain,
the higher its HUB score
𝑝𝑗𝑖 Essentially the complement of HUB, so
Authorities, 𝐴𝑈𝑇𝐻𝑖 = ∑
∑𝑘 𝑝𝑗𝑘 that the more product types a company
AUTH 𝑗
procures from the downstream supply
chain, the higher its AUTH score.
Industries, 𝐼𝑁𝐷𝑖 = ∑𝑗 𝑟𝑖𝑗 The number of different industries to
IND which company i is connected.

Testing the hypotheses

We test the hypotheses in five steps: (i) the processing of ego-network data, (ii) network
creation, (iii) the evaluation of business reports and the supply chain network, (iv) the
statistical analysis, and (v) the development of performance measurement models.
In the first step, the real enterprise data on each ego-network of each focal company is
processed. Each ego-network consists of a focal company, its customers and suppliers and
the relationships originating from cash flows between each focal company and its suppliers
and customers. As we create network data in an indirect way, we do not “rely on the often
inaccurate recollections of respondents” (Nooy et al., 2011, p. 26) when assessing
relationships via questionnaires. We develop a software-tool that allows us to process the
cash flows (procurement and sales) using the structured query language SQL.
In the second step, we merge the different ego-networks of the focal companies. As all
focal companies are comparable we make one large section of the scale-free supply chain
network visible. To achieve this, we inspect the individual supply chain networks for common
business partners between different focal companies and highlight the connections. The

7
output is a network GD(V,A) with v = 448 nodes. This network is our first subject of
inspection. The second subject concerns the business reports of focal companies. Network
flow data and financial data are matched for the same fiscal year.
In the third step, we collect data on network position characteristics. We use social
network analysis software (Pajek1 and UCINET2) to analyse the network position of focal
companies in the supply chain network. We evaluate for each focal company (i) strength of
the links, (ii) degree centrality, (iii) Bonacich power, (iv) hubs, (v) authorities and (vi) the
number of complementary industries. These measures, defined in table 3, are the
independent variables (IVs) of our statistical analysis.
The dependent variables (DVs) measure the economic performance of the focal
companies. We obtain the DVs by quantitative analysis of business reports. We evaluate: (i)
return on assets, (ii) asset turnover, (iii) operating profit and (iv) dynamic debt ratio.
In the fourth step, we analyse the statistical association between network position
properties and financial performance measures. We use simple linear regression 𝑌 = 𝑏0 +
𝑏1 𝑋, where Y is the dependent financial performance measure DV, X is the independent
network position variable (IV), and 𝑏0 and 𝑏1 are the regression parameters. To verify the
significance of our results, we use two-sided t-tests.
Finally, in the fifth step, using multiple linear regression, we investigate the combined
effect of the network position variables (predictors) on the financial performance measures
(response). To find evidence for network position characteristics influencing economic
performance, we test our hypotheses. The inclusion of network position variables is based
on an overall F-test for the multiple linear regression models and a t-test for the partial
coefficients.

The network and its associated measures

Following our methodology, we create a scale-free supply chain network 𝐺D (𝑉, 𝐴), using a
software tool created by us to merge the revenue (sales) and procurement data of the
sample of focal companies studied. As there is no need to identify the different companies
by name, we meet ethics requirements by not disclosing them. However, we go beyond the
analysis of each individual network and highlight common business partners of different focal
companies. Thus, before encoding the names of companies, we verify each dataset for
proper naming. Otherwise the network generation would not be able to identify common
nodes between different focal companies.
This process results in a network with 15 focal companies and their customers and
suppliers, 448 companies (nodes) in all. In our analysis, we assume an equilibrium (of
exchange of goods for cash) in the connections between companies. Thus, the relationships
up- and downstream in the supply chain are assumed to be equally important. We generally
do not distinguish connectivity up- and downstream the supply chain. Only in terms of
diversity do we differentiate between hubs (customer side) and authorities (supplier side).
In the statistical analysis, we focus on the results of the focal companies because we do
not possess financial information for the other 433 companies in the network. Thus, to test
our hypotheses, we collect characteristics on network position of our sample of 15 focal
companies. Our accidental sample embodies typical manufacturing companies from the
plastics processing industry. These companies are small and medium-sized enterprises that
operate in the automotive plastics supply chain in Germany. These 448 companies
represent approximately a 1% convenience sample of the whole population of automotive
plastic converters in Germany. The data describe the financial year, 2012. Access to the
supply chain data of these companies was facilitated through a software supplier common to
the focal companies. The inaccessibility of data, for reasons of confidentiality, would make it
very difficult to collect a random sample of a reasonable size from this industry or one
similar. Nonetheless, we anticipate that important insight will be gained using our

1
http://pajek.imfm.si/doku.php
2
https://sites.google.com/site/ucinetsoftware/home

8
methodology on our data, and we are, to our knowledge, the first to perform a quantitative
network analysis on company performance to this level of detail.
The values of the measures of interest for the 15 focal companies are shown in Table 4.
Social network analysis software (here UNICET v6.560) implements the required algorithms
to calculate the network position variables. In the calculation of Bonacich power (BP), as
discussed in the methodology, we have to pay attention to the value of β. By default,
UCINET sets β = 0.108 = 0.995/λmax. All values are normalised automatically by UNICET.
The testing of our hypotheses is invariant to these scaling factors.
We expect further insights whether companies are sales or procurement oriented. Based
on 𝐺D (𝑉, 𝐴), the product-mix matrix P allows to distinguish diversity in product flows up- and
downstream the supply chain.
Although all focal companies of our egocentric network study are manufacturing
companies in the same industry, these companies might produce for other industrial sectors.
We classify the industries of focal companies based on the WZ2008 classification
(Statistisches Bundesamt, 2008). Based on the separate affiliation network 𝐻D (𝑊, 𝐵) we
connect focal companies to their industries and create classes of structural isomorphism.
When presenting the financial measures, where appropriate, cash flows are given in millions
of monetary units (mmu). The few missing values of financial measures result from
inaccessibility or imperfect fulfilment of accounting policies.

Table 4 - Values of all dependent variables (DV) and independent variables (IV)
Dependent variables DV Independent variables IV

Comp. Empl. ROA AT OP DDR AS C BP HUB AUTH IND

(%) (mmu) β=0.108

FOC1 135 -3.45 2.43 0.53 9.96 1.41 0.085 49.18 0.00 1.29 0
FOC2 218 4.63 1.89 2.48 6.66 3.99 0.065 14.15 0.79 2.79 0
FOC3 103 20.21 3.05 1.80 2.29 1.19 0.038 19.81 0.06 0.65 2
FOC4 80 31.97 2.19 2.33 0.46 0.95 0.036 16.72 0.47 0.92 3
FOC5 380 7.69 1.34 9.52 3.65 5.10 0.161 185.05 0.20 7.22 1
FOC6 230 3.66 2.51 1.20 14.75 4.89 0.092 46.44 1.55 4.49 1
FOC7 415 6.57 2.05 5.67 5.62 3.35 0.087 36.02 0.80 1.17 2
FOC8 50 1.27 0.011 10.52 0.66 0.64 0
FOC9 270 3.15 0.081 24.72 0.47 3.60 1
FOC10 250 44.55 1.97 14.87 0.45 3.40 0.174 236.81 0.00 1.27 0
FOC11 60 -0.43 2.70 0.35 12.27 1.17 0.043 28.67 0.00 1.44 1
FOC12 126 -4.68 0.77 3.40 6.90 1.54 0.029 7.04 0.00 1.03 0
FOC13 261 10.60 1.99 7.47 2.01 6.79 0.130 35.04 3.19 1.61 0
FOC14 47 2.51 1.07 0.063 12.07 0.84 0.50 2
FOC15 86 6.26 1.52 1.61 1.08 1.74 0.043 26.20 0.97 0.38 0
Median 6.42 2.05 2.41 4.64 1.74 0.065 26.20 0.47 1.27 1
Mean 10.63 2.07 4.27 5.51 2.73 0.076 49.90 0.67 1.93 0.87
St. dev. 14.71 0.61 4.39 4.78 1.82 0.048 67.28 0.83 1.88 0.99

Analysis

Results for simple linear regressions

Based on the collected data, we study the statistical association between each DV and IV in
turn. For BP, we present the results for β = 0.108 = 0.995/λmax . Table 5 shows the Pearson
correlation coefficients of each variable of interest against every other. Table 6 presents the
results of the simple linear regressions, in which each DV is regressed on each IV, one at a
time in turn.

9
 Table 5 - Pearson correlation coefficients
AT OP DDR AS C BPβ=0.108 HUB AUTH IND
ROA 0.188 0.600** -0.667** -0.011 0.355 0.527* -0.085 -0.177 0.324
AT -0.336 0.268 -0.230 -0.151 -0.202 0.010 -0.232 0.459
OP -0.519* 0.502* 0.848*** 0.856*** 0.057 0.249 -0.195
DDR -0.003 -0.185 -0.275 -0.064 0.268 -0.133
AS 0.729*** 0.378 0.666*** 0.624** -0.284
C 0.843*** 0.198 0.551** -0.188
BPβ=0.108 -0.222 0.446* -0.191
HUB 0.032 -0.142
AUTH -0.037
(*** significant at 1%, ** 5%, * 10%)

Table 6 - Results of the simple linear regressions, showing coefficients, their standard errors
(s.e.), and the coefficient of variation, R2
AS C BPβ=0.108 HUB AUTH IND
ROA -0.011 0.355 0.527* -0.085 -0.177 0.324
s.e. 0.316 0.296 0.269 0.315 0.311 0.299
2
R 0.000 0.126 0.278 0.007 0.031 0.105
AT -0.230 -0.151 -0.202 0.010 -0.232 0.459
s.e. 0.293 0.298 0.295 0.301 0.293 0.268
2
R 0.053 0.023 0.041 0.000 0.054 0.211
OP 0.502* 0.848*** 0.856*** 0.057 0.249 -0.195
s.e. 0.274 0.168 0.163 0.316 0.306 0.310
2
R 0.252 0.718 0.733 0.003 0.062 0.038
DDR -0.003 -0.185 -0.275 -0.064 0.268 -0.133
s.e. 0.316 0.311 0.304 0.316 0.305 0.313
2
R 0.000 0.034 0.076 0.004 0.072 0.018
(*** significant at 1%, ** 5%, * 10%)

Results for multiple linear regressions

To test our hypotheses, we go one step further and analyse the influence of several network
related IVs on each financial DV. Using multiple linear regression, we develop performance
measurement models of the type in Equation 1:

DV = b0 + b1 ⋅ IV1 + b2 ⋅ IV2 + … + br ⋅ IVr (1)

Starting with the full model, we optimise the models using a backwards elimination
procedure: the number of variables is reduced as long as the adjusted coefficient of
determination barely differs. Table 7 shows the summarised results for these regressions,
with the coefficient of determination, R2, the same adjusted for the number of parameters in
the regression, and the regression coefficients, for the full model, the optimised model and a
conservative model. The conservative model has one less explanatory variable than the
optimised model.

10
 Table 7 - Standardised regression coefficients for the multiple regression models
Multiple regression R2 R2 adj. AS C BPβ=0.108 HUB AUTH IND
ROA Full 0.84 0.65 0.590 -1.215 1.891** 0.325 -0.735* 0.549**
Optimised 0.71 0.61 0.873*** -0.557** 0.479**
Conservative 0.47 0.35 0.615** 0.445
AT Full 0.33 -0.34 -0.548 1.125 -0.839 0.012 -0.070 0.407
Optimised
Conservative
OP Full 0.94 0.87 1.334** -0.577 1.165** -0.450 -0.724** 0.164
Optimised 0.89 0.85 0.454** 0.897*** -0.425**
Conservative 0.76 0.70 0.934*** -0.173
DDR Full 0.46 -0.18 -0.566 1.247 -1.617 -0.391 0.682 -0.300
Optimised
Conservative
(*** significant at 1%, ** 5%, * 10%)

Discussion of results

The results of the simple linear regression (Table 6) show one significant result for ROA and
three significant results for OP. The multiple linear regression (Table 7) confirm two
significant models, namely ROA and OP. As both models are indicative of the profitability
aspect of performance, we answer our research questions in the light of profitability.
 RQ1 is answered with yes because H1 (AS) is confirmed by the OP model.
 RQ2 is answered with yes because H2-2 (BP) is confirmed by both models. H2-1 (C)
is included by H2-2 (BP).
 RQ3 is answered with no because H3-2 (AUTH) is rejected by both models.
 RQ4 is answered with yes because H4 (IND) is confirmed by the ROA model.
Comparing the two models of ROA and OP, one would recommend OP as the best fitting
solution. The adjusted R2 indicates that 61% of the variability in ROA is explained by BP,
AUTH and IND, compared to 85% in OP explained by AS, BP and AUTH. However, looking
at Table 5 we recognise a correlation between ROA and OP. Thus, using the companies’
number of employees to control our findings, we think that the solution for OP is too
dependent on the size of a company. With respect to the number of employees, a correlation
coefficient of 0.03 for ROA compared to 0.61 for OP indicates Equation 2 as the most
reliable network oriented performance measurement model:

ROA = 3.071 + 0.176 ∙ BP0.108 - 4.155 ∙ AUTH + 6.842 ∙ IND (2)

To better understand the implications of (2), we analyse the different IVs in more detail.
The result for IND indicates that companies benefit from diversity because of connections to
complementary industries besides plastics. The result for BP (β = 0.108) indicates that it is
advantageous to be connected to well-connected companies. Therefore, to improve
performance, it is important to identify and to focus on relationships with well-connected
companies. A negative coefficient for AUTH suggests that it is disadvantageous to have
diversified relationships with suppliers in the network. One possible explanation for this
apparent effect is the prevailing market conditions: the industry of focus is very competitive,
which is why material suppliers have many opportunities for business. The focal companies
(manufacturing companies) cannot necessarily expect scale effects when relationships are
diversified over a variety of product types. Focal companies might rather benefit if they try to
play-off many suppliers against each other by increasing their number of different suppliers.
Such a strategy reduces the risk of strong dependencies. This explanation is also manifest in
the effect of Bonacich power, BP. The modification of the parameter β would allow us to look
at the risk of being outplayed by nodes having many alternatives. Following Bonacich

11
(1987), Bonacich power from the viewpoint of bargaining situations is not about benefiting
from connections to well-connected nodes, but the risk of being outplayed by nodes having
several alternatives at hand. As the centrality scores for BP (β < 0) are nearly the opposite of
the positive results in Table 4, we do not present the results for β < 0 separately.

Conclusion

To contribute to performance measurement in a supply chain management context, we

examine the link between network position properties and financial performance measures.
The transfer of social network analysis to the supply chain network allows us to collect data
on network position. The calculation of a variety of financial performance measures ensures
that we evaluate economic performance comprehensively.
In a two-stage statistical analysis, we study network position properties for their influence
on financial performance. We develop performance measurement models based on multiple
linear regression. By optimising the most significant models, ROA and OP, we identify the
strongest, most influencing factors. We confirm a strong correlation between Bonacich
power and financial performance.
In their work, Bellamy and Basole (2013) describe Bonacich power as a “well-known
measure not yet exploited in operations and SCM literature”. This then underlines the
timeliness of our work, and our results indicate that the success of a company is linked to the
degree centrality of its neighbours. Furthermore, we recognise that a network node benefits
from selective connectedness. If these connections with well-connected business partners
are relatively strong, as measured by cash flows, we demonstrate that such a network
position results in improved economic performance.
The implications of our study are manifold. For example, our results suggest to managers
of companies the type of business partners that should be their main focus. Furthermore, we
provide substantial evidence for the enrichment of performance measurement tools, so that
these tools have a clear network perspective.
Our conclusions are limited by the scope of the network of our accidental sample of 15
focal companies and network of 448 nodes. Nonetheless, we believe our methodology
generalizes and that it would provide the basis for a study of an even larger network. Also, it
would be interesting to verify our results for other supply chain networks of manufacturing
industries.

References

Beamon, B. M. (1998), “Supply chain design and analysis: models and methods”,
International Journal of Production Economics. Vol. 55, No. 3, pp. 281-294
Bellamy, M.A. and Basole, R.C. (2013), “Network analysis of supply chain systems: A
systematic review and future research”, Systems Engineering. Vol. 16, No. 2, pp. 235–
249
Bititci, U.S., Carrie, A.S., McDevitt, L.G. (1997), “Integrated performance measurement
systems: a development guide”, International Journal of Operations and Production
Management. Vol. 17, No. 5, pp. 522-535
Bititci U.S., Mendibil, M., Nudurupati, S., Turner, T., Garengo, P. (2006), “Dynamics of
performance measurement and organizational culture”, International Journal of
Operations and Production Management. Vol. 26, No. 12, pp. 1325-1350
Bonacich, P. (1987), “Power and centrality: a family of measures”, American Journal of
Sociology. Vol. 92 No. 5, pp. 1170–1182
Borgatti, S.P. and Everett, M.G. (1992), “Notions of position in social network analysis”,
Sociological Methodology. Vol. 22, No. 1, pp. 1–35
Borgatti, S.P., Everett, M.G. and Johnson, J.C. (2013), Analyzing social networks, Sage
Publications, London.
Borgatti, S.P. and Li, X.U. (2009), “On social network analysis in a supply chain context”,
Journal of Supply Chain Management. Vol. 45, No. 2, pp. 1–17

12
Bourne, M. C. S., Mills, J. F., Wilcox M., Neely A.D. and Platts, K. W. (2000), "Designing,
implementing and updating performance measurement systems", International Journal of
Production and Operations Management. Vol. 20, No. 7, pp 754-771
Chenhall, R. H. (2005), “Integrative strategic performance measurement systems, strategic
alignment of manufacturing, learning and strategic outcomes: an exploratory
study”, Accounting, Organizations and Society. Vol. 30, No. 5, pp. 395-422
Harland, C.M., Lamming, R.C., Zheng, J. and Johnsen, T.E. (2001), “A taxonomy of supply
networks”, The Journal of Supply Chain Management. Vol. 37, No. 4, pp. 21–27
Jarillo, J. and Stevenson, H.H. (1991), “Co-operative strategies—the payoffs and the
pitfalls”, Long Range Planning. Vol. 24, No. 1, pp. 64–70
Kim, Y., Choi, T.Y., Yan, T. and Dooley, K. (2011), “Structural investigation of supply
networks: A social network analysis approach”, Journal of Operations Management.
Vol. 29, No. 3, pp. 194–211
Klibi, W., Martel, A. and Guitouni, A. (2010), “The design of robust value-creating supply
chain networks: A critical review”, European Journal of Operational Research. Vol. 203,
No. 2, pp. 283-293
Melnyk, S., Bititci, U., Platts, K., Tobias, J. and Andersen, B. (2014), “Is performance
measurement and management fit for the future?”, Management Accounting Research.
Vol. 25, No. 2, pp. 173-186
Merchant, K. A., and Van der Stede, W. (2007). Management Control Systems. Pearson
Education Limited, London.
Morgan, C. (2007), “Supply network performance measurement: future challenges?”, The
International Journal of Logistics Management. Vol. 18, No. 2, pp. 255–273
Nooy, W.d., Mrvar, A. and Batagelj, V. (2011), Exploratory social network analysis with
Pajek, 2nd ed, Cambridge University Press, New York.
Pavlov A., and Bourne M. (2011), “Explaining the effects of performance measurement on
performance: An organizational routines perspective”, International Journal of Operations
and Production Management. Vol. 31, No. 1, pp. 101-122
Robins, G. (2015), Doing social network research: Network-based research design for social
scientists, SAGE, Los Angeles, California.
Simchi-Levi, D., Kaminsky, P., Simchi-Levi, E. (2004), Managing the supply chain: the
definitive guide for the business professional. McGraw-Hill, New York.
Statistisches Bundesamt (2008), “Klassifikation der Wirtschaftszweige”, available at:
https://www.destatis.de/DE/Methoden/Klassifikationen/GueterWirtschaftklassifikationen/C
ontent75/KlassifikationWZ08.html (accessed 5 June 2015).
Veen-Dirks, v. P. (2010), “Different uses of performance measures: the evaluation versus
reward of production managers”, Accounting, Organizations and Society. Vol. 35, No. 2,
pp. 141-164

13