67
Marketing Management in MNC Subsidiairies: An Archetypal Analysis of Heterogeneity in Strategy and Organization _______________ David MIDGLEY Sunil VENAIK 2012/72/MKT

Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

  • Upload
    buicong

  • View
    214

  • Download
    1

Embed Size (px)

Citation preview

Page 1: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

Marketing Management in

MNC Subsidiairies:

An Archetypal Analysis of

Heterogeneity in Strategy

and Organization

_______________

David MIDGLEY

Sunil VENAIK

2012/72/MKT

Page 2: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

Marketing Management in MNC Subsidiaries:

An Archetypal Analysis of Heterogeneity in Strategy

and Organization

David Midgley*

Sunil Venaik**

August 2012

* Professor of Marketing at INSEAD, Boulevard de Constance 77305 Fontainebleau Cedex.

Email: [email protected]

** Senior Lecturer in Strategy at The University of Queensland, Brisbane St Lucia, QLD 4072,

Australia. Email: [email protected]

A Working Paper is the author‘s intellectual property. It is intended as a means to promote research to

interested readers. Its content should not be copied or hosted on any server without written permission

from [email protected]

Find more INSEAD papers at http://www.insead.edu/facultyresearch/research/search_papers.cfm

Page 3: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

2

ABSTRACT

The relative degree of control and standardization in the ways multinational corporations

operate across the globe has been a central question in international business research since

the inception of the discipline. Normatively, is it better to control from the center or should

local subsidiaries be allowed to go their own way? Or is it more appropriate to globally

standardize certain elements of operations but allow local adaptation on others? Moreover,

due to the dynamism and diversity of the business environments in which MNCs operate, they

constantly seek better strategies and decision-making structures to balance the evolving global

and local pressures on their subsidiaries. Consequently there is considerable heterogeneity in

the strategies and structures employed by individual corporations as they adapt to the specific

circumstances in which they find themselves. And paradoxically, although the literature

examines in depth the moderating influence of various pressures on strategy and structure, it

does not describe the underlying empirical heterogeneity of these strategies and structures in

any detail. In this paper, we use a relatively new statistical method—archetypal analysis—and

detailed measures of the marketing mix to describe the heterogeneity of subsidiary marketing

strategies and decision-making structures. Our preliminary results show this heterogeneity to

be more complex than recognized by the literature.

Keywords: Archetypal analysis, MNC subsidiaries, adaptation, innovation, autonomy,

networking

Page 4: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

3

INTRODUCTION

With increasing globalization of the world economy and the spread of multinational

corporations worldwide, there is continuing interest in examining the strategy and

organization of these multinational corporations (hereafter MNCs). MNCs account for a

quarter of global GDP, and their foreign affiliates ―more than one-tenth of global GDP and

one-third of world exports‖ (UNCTAD, 2011: 1). Understanding their strategy and

organization is therefore of central interest to scholars of international business.

The earlier literature on MNCs mainly focused on the strategies and decisions taken in

the headquarters of the parent company (Buzzell, 1968; Levitt, 1983). However, with growing

size and significance of MNC subsidiaries, scholarly interest shifted to understanding

strategies and decision-making at the level of the local subsidiaries. Prahalad and Doz‘s

(1987) seminal work on global integration and local responsiveness documents the diverse

pressures confronted by MNCs as they expand globally and the strategies they pursue in

response to these environmental pressures. Although Prahalad and Doz proposed a broad

framework, many of the early debates on how MNCs should respond to these pressures were

more limited in scope. For example, should firms standardize or adapt their strategies as they

expand globally? For another example, should firms centralize their decision-making in the

corporate headquarters or provide autonomy to local subsidiaries for subsidiary-level

decisions? Bartlett and Ghoshal (1989) extended these discussions by adding considerations

of subsidiary learning and innovation. In doing so, they highlight the need for internal

networking and coordinated decision-making across the MNC organization. Bartlett and

Ghoshal argue such coordination is necessary to support inter-unit learning and accelerate

innovation in MNCs. Continuing in this tradition, Devinney, Midgley and Venaik (2000)

extend the integration-responsiveness framework to include the transactional pressures on the

MNC‘s value chain and formalize the key role managerial beliefs play in the strategic choices

Page 5: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

4

made by such organizations. And, in an empirical paper, Venaik, Midgley and Devinney

(2005) show there are dual paths to better MNC performance. One path is through subsidiary

decision-making autonomy, which encourages innovation, and the other path is through

networking, which encourages inter-unit learning.

Notwithstanding the broadening of discussion in the international business literature, the

international marketing literature continues to focus on the simple standardization-adaptation

debate (Takeuchi & Porter, 1986; Lages, Jap & Griffith, 2008). While recognizing the

importance of this debate, and the deeper understanding that has grown from it, we believe

there is a need for international marketing scholars to embrace the organizational issue of

autonomy versus centralization. In addition, the critical significance of local market

innovation for MNC subsidiary competitiveness and the need for global networking to

discuss, decide, share and compare marketing best practices across the MNC also needs

recognition in the international marketing literature.

Equally, we believe the international business strategy literature could benefit by

adopting the detailed perspective seen in international marketing studies. For example, a

marketing strategy can be broken down at least to the level of the 4Ps (price, product,

promotion and place) if not to a finer level of detail. Zou and Cavusgil (2002) develop a

model of global marketing strategy that includes the four Ps plus marketing concentration and

coordination, and global market participation. Given the diversity of the global business

environment and the wide range of industries in which MNCs compete, the devil is surely in

the specific details of adaptation, innovation, autonomy and networking. Which aspects of

strategy are adapted, which are standardized? What is the country or global focus of

innovation for an MNC? Over which decisions are subsidiaries given autonomy and over

which are they not? And what are the priority topics in networking and coordination

meetings? All in all, we believe there is a need for more breadth in marketing studies and

Page 6: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

5

more detail in business strategy studies. But more importantly, our overall understanding of

MNCs would greatly benefit from more detailed data on adaptation, innovation, autonomy,

and networking.

More detailed data would require scholars to break down each of these four areas into a

number of subordinate components, develop measures of these components and collect the

necessary supporting data on MNC subsidiaries. However, having more detailed data on

these critical areas of MNC operation admittedly makes analysis of these data more

challenging. We will have a larger number of components to analyze and we believe these

may display a greater degree of heterogeneity across MNC subsidiaries than the literature

discusses. For example, if we break down marketing strategy into 4Ps this may uncover

subsidiaries where product and place are standardized but pricing and promotion are adapted.

Indeed, as we increase the number of subordinate constructs more combinations potentially

exist—be these of adaptation or standardization, autonomy or centralization, etc. Hence,

there is consequent need to identify common patterns in these data and thus represent MNC

heterogeneity in a concise, understandable and meaningful way. Existing techniques such as

cross-tabulation or cluster analysis may not meet this need. For this reason, we also believe

we need new techniques to analyze these data in ways that shed deeper insight on how MNCs

respond to pressures in their business environment.

The objectives of our paper are thus twofold. First, to demonstrate the benefits of a

more detailed view on strategy and organization. We do this by defining the appropriate

subordinate components through a ‗7Ps‘ framework. Essentially this framework builds on

marketing‘s traditional 4Ps but adds components for product positioning, MNC marketing

policies and personnel. This results in a coherent framework that can be applied across the

four areas of adaptation, innovation, autonomy, and networking. We then develop and

validate multi-item measures for these 7Ps and collect the supporting data from a sample of

Page 7: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

6

229 MNC subsidiaries. We should note that while we do this in the context of marketing our

approach could be readily extended to other areas of MNC operations. Second, we analyze

these data by archetypal analysis (hereafter AA). As will be discussed subsequently, AA

provides an advance over existing techniques because the resulting common patterns—MNC

subsidiary archetypes—have a clear definition. As a consequence these archetypes are a

powerful summary of the heterogeneity of MNC subsidiary operations. AA also uses a

mixture model. This allows insightful comparisons between MNCs that are clearly

associated with one single archetype and those with a strategy or organization that is a

mixture of two or more archetypes. A major contribution of our work is to demonstrate that

these mixed strategies and organizations are more prevalent than the literature suggests. A

second contribution is to demonstrate that the strategies and organizations the archetypes

themselves represent are more heterogeneous than those discussed in the literature. Finally,

we also believe our application of AA is novel within the international business literature.

The paper is organized as follows. The next two sections provide an overview of the

four key subsidiary level issues in international marketing, namely, the strategies of local

adaptation and local innovation, and the question of decision-making through local autonomy

and internal networking. In the section that follows these, we discuss the 7Ps framework,

introduce the basic ideas behind AA and present our research hypotheses. This is followed by

another section in which we discuss the methodology we use for our study—including

component measures, survey design, sampling and the application of AA to our data. We

conclude our paper by presenting and discussing our results and outlining the potential

contributions they make to the international business literature. The paper also includes a

technical appendix that provides deeper explanation of AA and demonstrates the convergent

and discriminant validity of our measures and archetypes.

Page 8: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

7

MNC SUBSIDIARY STRATEGY

A large body of scholarly work in international business, and especially work focusing on

international marketing, examines the impact of the twin pressures of global integration and

local responsiveness along the standardization-adaptation dimension (see Nasir & Altinbasak,

2009 for a recent review). To what extent are the MNC‘s marketing activities standardized

across its various subsidiary operations? Or to what extent do local subsidiaries adapt

elements of the 4Ps to their local market? Further, with growing global and local competition

in the local markets, MNCs are under increasing pressure to use their subsidiaries as sources

of innovations that can be deployed around the globe (Nobel & Birkinshaw, 1998). Is it

possible to leverage an innovation developed by one subsidiary across the other markets in

which the MNC operates? These two strategic imperatives---local adaptation and local

innovation—are central to thinking about MNC strategy and organization. We view local

adaptation as reflecting the degree to which price, product, promotion and place (and their

constituent elements) are modified to suit the requirements of the local market. And we view

local innovation as the degree to which the subsidiary seeks new ideas for improving elements

of its price, product, promotion and place. We now briefly review each of these two areas.

Local Adaptation

The importance of environmental and institutional forces on firm strategy and decision-

making is acknowledged widely in both the international business literature (e.g., Porter,

1990; Venaik, Midgley & Devinney, 2004) and the organization theory literature (e.g.,

Lawrence & Lorsch, 1967; Sundaram & Black, 1992). In particular, MNCs are confronted

with diverse and often conflicting environmental pressures as they expand their activities

around the globe. These pressures are often broadly referred to as the pressures of global

integration (GI) and local responsiveness (LR) (Prahalad & Doz, 1987). The GI pressures

force firms to take an integrated approach to their global activities—that is, to coordinate their

Page 9: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

8

business units and strategies to attain maximum efficiency and competitive advantage. These

pressures might lead to responses such as manufacturing product parts in a single location for

global use at efficient scale, or mandating global consistency in brand positioning.

Concurrently, firms face a countervailing set of pressures to adapt their activities to the

unique circumstances of the countries in which they operate. These pressures for LR may

prompt responses such as manufacturing parts locally to obtain tax incentives or adapting

product positioning to local market circumstances.

Within the marketing function, companies attempt to deal with these conflicting

demand and cost pressures in a host of ways (Sheth, 2011). Some firms appear to be able to

find common segments across multiple markets and develop truly global brands with

underlying production efficiency. Pringles (Pollack, 1999) and Heinz (Neff, 1999) are two

brands able to standardize without major internal tradeoffs because customer needs vary little

across the globe. Other companies give in to the pressure of sacrificing global economies of

scale for the high levels of local adaptation they believe necessary to meet widely differing

local needs or circumstances. For example, many MNCs in the insurance business extensively

adapt their marketing to fit both market and regulatory requirements across different

countries. Other corporations seek the middle ground, standardizing some elements of the 4Ps

but allowing others to be adapted to the local market. For example, Philips attempted to

appeal to a global audience through its Olympic advertising (Edy, 1999) by delivering a

common message tailored to each market by different actors taking different approaches to

the use of different Philips products. Overall, the degree of local adaptation of the 4Ps

remains an important issue in MNC subsidiary strategy, both for scholars and managers. And

the choice of where an MNC subsidiary falls on this continuum is contingent on many

environmental factors and pressures, both at the global and local level. These include

customers, competitors, regulation, marketing infrastructure (distribution channels, support

Page 10: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

9

agencies, etc.), technology, labour and the MNC‘s organisational culture and

internationalisation history (Cavusgil and Zou, 1994). The considerable diversity of these GI

and LR pressures—across nations, industries and firms—may also generate significant

heterogeneity in marketing at the subsidiary level. However, beyond discussions of the

stereotypical fully standardized or fully adapted firm, we have relatively little knowledge of

the shape and form of this heterogeneity. What are the empirical patterns of local adaptation

across MNC subsidiaries? Do most subsidiaries follow pure strategies like Pringles or hybrid

strategies like Philips? Is product standardization more prevalent than pricing standardization?

There is little evidence to answer these and many other questions, especially at the level of the

4Ps.

Local Innovation

Innovation is regarded as the fundamental basis for creating firm specific advantages that

enable a firm to achieve sustainable competitive advantage and improve its corporate

performance (Howard, 1993). With increasing global and local competition, there is a

growing imperative to tap into diverse sources of new ideas within the MNC network (Lee,

Chen, Kim & Johnson, 2008). In the context of MNCs therefore, the focus is increasingly

shifting to the MNC subsidiaries as the source of innovations (Bartlett & Ghoshal, 1986;

Gupta & Govindarajan, 1994). Although some subsidiary innovations may be created

specifically for the local subsidiary market, increasingly, subsidiaries are sources of

innovations that the MNC can leverage on a global basis. Local innovations span the entire

value chain, including customer facing product and marketing activities such as product

positioning, promotion, and sales and distribution. In this way, subsidiaries contribute to the

firm-specific advantages of the MNC, and thereby shift the generation of these advantages

―from being the sole concern of the parent company to a collective responsibility for the

corporate network‖ (Birkinshaw, Hood & Jonsson, 1998). Local innovation in MNC

Page 11: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

10

subsidiaries is therefore seen as a critical and increasingly significant determinant of MNC

competitiveness. Moreover, marketing innovation can be seen, at least in part, as deriving

from interactions between the subsidiary and its local customers, distributors and support

agencies. Whether the result is price, product, promotion or place innovation, or a

combination of these, the innovation itself stems from the opportunities and challenges the

subsidiary sees in its local environment. As for local adaptation, and corresponding to the

diversity of these local environments, we might also expect significant heterogeneity in

innovation strategies across MNC subsidiaries. Perhaps some subsidiaries see opportunities

for promotional innovation, others for product innovation? And clearly there may also be a

relationship between local adaptation and local innovation. For example, an MNC with a

globally standardized product might not primarily seek product innovation from its

subsidiaries, but rather innovation in the areas of price, promotion or place. But again as for

local adaptation, we have little evidence to answer these questions, both in terms of the

patterns of local innovation across the 4Ps and any relationship between adaptation and

innovation.

Next, we discuss the two key ways in which MNCs take decisions, namely, local

autonomy and internal networking.

MNC SUBSIDIARY DECISION-MAKING

The locus of decision-making is an important issue since the way global strategies are

implemented within the network of MNC subsidiaries impacts on the performance of

multinational firms (Kashani, 1989). The major dimensions of organization structure are

complexity, centralization and formalization (Van de Ven, 1976). However, early studies

concentrated on the issue of centralization versus autonomy since centralization was regarded

as the primary construct in organization design (Egelhoff, 1988). More recently, there has

been a realization that centralization, while remaining important, may not fully capture the

Page 12: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

11

wide range of methods and processes used by firms for taking decisions. For example, taking

decisions in multi-country teams and task forces is an important aspect of an MNC‘s

organizational structure and processes (Ghoshal, Korine & Szulanski, 1994). A network

approach to decision making is considered essential to gain deeper insights about the

complexities of the diverse markets served by large multinational firms and to respond rapidly

to changes in these markets. Following Ghoshal et al. (1994), we focus on two decision-

making constructs – local autonomy and internal networking. We view local autonomy as

reflecting the degree of decision-making freedom given to the subsidiary by the headquarters,

and internal networking as the degree to which the subsidiary uses, or is used by, other parts

of the firm for making key decisions. We now briefly review each of these two areas.

Local Autonomy

Notwithstanding extensive research in the MNC and organization literature on the issue of

locus of decision-making and its determinants and consequences, there is little research on

this issue in the international marketing literature until recently (e.g., Ozsomer & Simonin,

2004; Tong, Wong & Kwok, 2012). In the MNC literature, greater autonomy is considered to

have a strong motivating influence on the local subsidiary managers and encourages them to

take initiatives that result in marketing innovations for local and global markets. For example,

Bartlett and Ghoshal (1989) found subsidiary autonomy to have a positive relationship with

innovation in multinational firms. Birkinshaw et al. (1998) show that autonomy is associated

with the subsidiary contributing more towards firm-specific advantages at the global level, a

perspective also supported more broadly in the strategy literature (e.g., McGrath, 2001;

Zanfei, 2000). Venaik, Midgley and Devinney (2005) found local autonomy in marketing

decisions to have a significant positive effect on local marketing innovation. It might also be

argued that as subsidiaries are given greater autonomy a greater range of possibilities for

adaptation and innovation open up to them. The stereotypical centralized MNC implements

Page 13: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

12

just one strategy across the globe. The stereotypical decentralized MNC may implement

many more strategies—up to and including a unique strategy for each of the local markets in

which it operates. Thus greater autonomy could contribute, at least potentially, to the

heterogeneity of strategies at the local level we discussed earlier. However, as before we have

little knowledge of what the actual patterns of centralization/autonomy are, especially at the

level of the 4Ps. Nor do we yet fully understand how autonomy relates to adaptation and

innovation at this level of detail.

Internal Networking

In the strategic management literature, organizational networks are classified into two broad

types – external and internal. External networks are formed between firms, whereas internal

networks are formed between organizational units separated by functions, businesses or

geographic locations (Charan, 1993). Here, we are interested in internal global networks as

mechanisms for organizational decision making in MNCs. That is, the extent to which

marketing decisions in the MNC are taken in groups, such as teams, task forces and

committees, comprising managers from the corporate and regional headquarters and country

subsidiaries.

Due to rapid technological change, the knowledge base of most businesses is becoming

increasingly complex and widely dispersed. Global networking increases the intensity of

communication among organisational members, which is ‗a major determinant of

organisation‘s effectiveness in creating and diffusing innovations‘ (Gupta & Govindarajan,

1991). Working across diverse customer, competitive, and country environments, subsidiary

managers bring together multiplicity of experiences and perspectives that assist in

overcoming narrow parochial functional, departmental or geographical interests, and

evaluating problems and taking decisions in the best interest of the global corporate

organisation (Charan, 1993). By providing a range of opinions from a variety of perspectives,

Page 14: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

13

networking refines and ultimately increases the chances of success of the network-based

decisions (Powell, Koput & Smith-Doerr, 1996). Due to their flexibility, organisational

arrangements such as networks provide an effective means of quick decision-making in a

volatile environment, and are important sources of sustainable competitive advantage

(Charan, 1993; Powell et al., 1996). Cross-functional networks, such as teams and taskforces,

allow concurrent rather than sequential interaction, thus reducing the time-to-market for new

products and processes (Teece, 1996). Ghoshal et al. (1994) found significant positive

relationship between the use of networking mechanisms and inter-unit communication and

learning in MNCs. Overall, it appears MNCs increasingly use global networks for decision-

making, and to good effect, despite the increased complexity and coordination difficulties that

come with them. However, as yet we have relatively little understanding of the relative

frequency with which the 4Ps are discussed in these meetings. Nor whether there are distinct

and different patterns of internal networking across MNCs and their subsidiaries.

THE 7Ps FRAMEWORK, ARCHETYPAL ANALYSIS AND RESEARCH

HYPOTHESES

The preceding discussion demonstrates that these four areas—local adaptation, local

innovation, local autonomy and internal networking—are important topics for both scholars

and managers. Yet this discussion also raises a number of questions to which more detailed

answers are required. To begin to answer these questions, we first seek to develop a set of

reliable measures of the subordinate components that underlie local adaptation, local

innovation, local autonomy and internal networking. Such a set of measures allows us to look

at these topics at a finer level of detail. Second, using these measures and a sample of MNC

subsidiaries, we seek to identify the patterns that exist within each of the four areas. For

example, subsidiaries where products and promotions are standardized but prices and place

adapted. Or MNCs where decisions about products are discussed in networking meetings, but

Page 15: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

14

never decisions about pricing or promotion. We identify such patterns using archetypal

analysis.

The 7Ps Framework

The set of measures for subordinate components we develop around a 7Ps framework. This

framework builds on marketing‘s 4Ps of price, product, promotion and place but adds three

additional components—product positioning, MNC marketing policies and marketing

personnel decisions. Product positioning is at the heart of the marketing function around

which the other 4Ps are built to deliver customer value and competitive advantage (Kotler,

1999). Despite its importance, positioning is often ignored in the international marketing

literature, as are marketing policies and personnel decisions – the process aspects of

marketing that are required to implement the 4Ps of the marketing program (Sorenson and

Weichmann, 1975). We use five – including the 4Ps and product positioning – to study local

adaptation and innovation—that is, activities visible in the local market. And we use all

seven – including policy and personnel – to study local autonomy and internal networking—

that is, MNC organization.

By appropriately framing our survey questions, this framework can be applied across all

the four areas of adaptation, innovation, autonomy and internal networking. We do this by

framing the questions in terms of the level of adaptation and innovation seen in, for example,

local promotion activities. And by asking further questions about the level of autonomy or

internal networking when local promotional decisions are made. As discussed in the

Methodology section each of these 5 or 7Ps is measured with multiple items and for each of

the four areas, resulting in a total of 24 reliable component measures. Thus our framework

aligns the major decisions in marketing with the key areas of interest to scholars and

managers and allows us to systematically compare these within and across subsidiaries.

Indeed, the values we measure on the 7Ps provide an interesting and more detailed profile of

Page 16: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

15

each subsidiary in our sample. This also allows us to apply analytical techniques like AA to

identify patterns in these data.

However, before we discuss our hypotheses as to what we might expect to see in these

patterns, it is first necessary to outline archetypal analysis. This is because we need to justify

our choice of this technique but also because we need to show it permits the testing of more

interesting hypotheses about heterogeneity than other methods.

Archetypal Analysis

Explaining heterogeneity is an important aspect of scholarly work on international business.

Scholars do not assume their units of analysis are identical, be these countries, multinationals

or their subsidiaries. Rather scholars assume these vary in a systematic way, variation we seek

to explain by building appropriate theories. However, before we can explain heterogeneity we

must first describe it well. Such description requires the researcher (1) selects appropriate

variables with which to characterize their units of analysis and (2) conducts analyses that

illustrate and categorize the heterogeneity of these units in an insightful way. We believe we

meet the first requirement through the 7P component measures and so here we focus on the

second step. Why does AA provide better insights into these patterns of heterogeneity?

The word archetype itself means ―a very typical example of a certain person or thing‖

(Oxford Dictionary Online). This word derives from the Ancient Greek arkhe—meaning

primitive—and tupos—meaning a model. Archetypes are common in everyday language,

psychoanalysis, literature and art. For example, and in our context, the ‗centralized MNC‘ or

the ‗autonomous subsidiary.‘ Cutler and Breiman (1994) introduced archetypal analysis

(AA) as a formal statistical technique, motivating their analysis problem by a data set

containing the head measurements of a sample of Swiss soldiers. Was it possible to identify

the archetypal head shapes from these data—subject to the constraint that any individual

soldier‘s head could be represented as a mixture of these archetypes? They showed that this

Page 17: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

16

was indeed possible. The reader is referred to the Technical Appendix for an explanation of

how AA works and the various issues the researcher must be aware of when applying it to

their data. The appendix also discusses why, in the context of our study, AA may be superior

to other pattern identification methods such as cluster analysis. Here we simply summarize

the key points from the appendix.

Although AA is not a cluster analysis method its output has some similarities, namely a

small number of discrete patterns—archetypes—that summarize the data. However, the big

difference of AA to other techniques is these archetypes have a clear definition. To

understand this definition, you need to imagine data as a cloud in the hyper-dimensional space

describe by the variables of interest (here the 7Ps). Archetypes are then defined as the

influential data points that best describe the exterior surface of this cloud. And largely

because of this clear definition, AA provides a number of advantages over commonly used

techniques such as hierarchical or k-means clustering, or even more recent developments such

as fuzzy clustering. These advantage include:

Sharper and more differentiated solutions than other techniques

AA imposes no strong ‗model‘ on the data

AA is robust to noise in the data

Each archetype is associated with a real observation, facilitating interpretation

All cases in the data have scores representing the degree they are associated

with each archetype—this allows ―single‖ and ―mixed‖ cases to be separated

(where ―single‖ cases are associated with only one of the archetypes and

―mixed‖ are associated with two or more archetypes).

Overall, AA produces simple, interpretable and robust solutions where the identified

archetypes and individual case scores have defined meaning. This contrasts with the

Page 18: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

17

complexity, sensitivity to algorithmic assumptions and lack of interpretability of many other

techniques.

Research Hypotheses

Given our 5 and 7P measures and AA what might we expect to see in our data in terms of

numbers and profiles of archetypes across each of the four areas? Here we build on our

earlier discussion. Particularly the literature concerning the diverse pressures for global

integration and local responsiveness (Prahalad & Doz, 1987) and the heterogeneity of

subsidiary strategies and MNC organizations observed as a consequence (Bartlett & Ghoshal,

1989). From that literature, we expect to see the classic archetypes for fully standardized or

fully adapted subsidiaries, but we might also expect to see other archetypes that represent

composite strategies. That is, some of the 5Ps standardized and some adapted as in the

Philips example cited earlier. Unfortunately, beyond this simple conclusion, the literature is

not especially clear on how many archetypes we might expect in total, or what their profiles

across the 5Ps might look like. Similar comments can be made about the literature on local

innovation, autonomy and internal networking. So all we can say so far is we expect at least

three archetypes for each of the four areas. One archetype where all the relevant component

measures have high values, one where they have low values and one where they have a

composite of high and low values. However, we did also discuss the idea that there may be

more archetypes for strategy than for organization. This is on the basis that more autonomous

or less connected subsidiaries may develop a broader range of strategies—being freer to

choose which of the 5Ps they adapt or innovate. So here we might argue for a minimum of

four archetypes—the two classic and two composites. Based on this discussion, we propose

the following hypotheses on heterogeneity:

Page 19: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

18

H1a: Three organization archetypes—classic high, classic low and composite

organizations--will be required to describe the observed heterogeneity in local

autonomy and internal networking.

H1b: Four strategy archetypes—classic high, classic low and two composite

strategies—will be required to describe the observed heterogeneity in local adaptation

and local innovation.

Next, from arguments around economies of scale, together with the examples cited before, we

might argue that one of the adaptation strategy composites would have a standardized product

and the other a standardized message. For example, where global scale is needed to develop

and manufacture a competitive product we might see a highly standardized product but local

adaptation on the other Ps. Where marketing requires scale, or where a globally consistent

message is desired, we might see a highly standardized positioning and promotion but local

adaptation on the other Ps. Similar arguments would apply to local innovation. From this

discussion we develop the following hypotheses:

H1c: For the case of local adaptation, one of the two composite strategies will have

standardized products and the other standardized positioning and promotion. For local

innovation, one will have no local innovation on product, the other no local innovation

on positioning and promotion. With the remaining Ps being more adapted or more

innovative as appropriate.

Hypothesis H1a can be rejected if we only find the two classic archetypes in our data or we

find more than three archetypes. In terms of AA we are postulating that three exterior points

will represent the surface of the 5 or 7-dimensional cloud to a reasonable approximation. For

Hypothesis H1b the relevant number is four archetypes/exterior points. Hypothesis 1c can be

rejected if we do not find composite archetype with the hypothesized profiles.

Page 20: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

19

Thus far we have only examined heterogeneity as described by the archetypes

themselves. One advantage of AA is that, being a mixture technique; it also describes each

subsidiary as a mixture of archetypes. So we obtain a different perspective on heterogeneity

by looking at the association between subsidiaries and archetypes. Are most subsidiaries

‗pure,‘ that is, clearly associated with only one of the classic archetypes? Or are they

‗hybrid,‘ that is, associated either (a) with one of the composite archetypes or (b) with a

mixture of two or more archetypes? In terms of AA, this depends on the distribution of our

subsidiaries in the 5 or 7-dimensional space. Are the data points distributed close to the

exterior surface and therefore more likely to be clearly associated with an archetype? And if

so, is the archetype they are associated with classic or composite? Or are our data points

more towards the interior of the space and thus more likely to be mixture of two or more

archetypes? Given the scarcity of detailed data and the novelty of AA, the literature naturally

provides few answers to these questions. They might be seen more as empirical questions.

However, there is a general theme in the literature that firms often fail to make clear choices

of strategy, and are either stuck in the middle (Porter, 1980, 1985) or pursue hybrid strategies

and organization (Miller, 1992). Put together with the complexity and diversity of the global

business environment (Prahalad & Doz, 1987), we might well expect to see more subsidiaries

with hybrid strategies and organizations than those with pure ones. We therefore advance our

second, and admittedly more speculative, hypothesis on heterogeneity.

H2: The majority of MNC subsidiaries pursue hybrid strategies and organizations.

This hypothesis can be rejected if we see more pure than hybrid (mixed and composite)

cases in the data. In terms of AA, ‗pure‘ can be defined as an association score with any of

the classic strategy archetypes that is equal to or greater than 0.5. This is because the sum of

the association scores for one subsidiary across all archetypes totals to 1. A score of 0.5 or

greater therefore implies the subsidiary is more strongly associated with one archetype than

Page 21: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

20

with any other archetype or with all the other archetypes together. Figure 1 summarizes this

discussion in a two by two matrix. In the Figure the label ‗pure‘ represents those subsidiaries

that are associated with one of the classic archetypes at a score of 0.5 or greater. The label

‗hybrid‘ represents either (1) those subsidiaries associated with a composite archetype at a

score of 0.5 or greater, or (2) those associated with a mixture of two or more archetypes—that

is, having scores of less than 0.5 with every archetype. The relative sizes of the various parts

of the Figure also depict the hypothesis—we expect to find more hybrid than pure cases.

==========================

FIGURE 1 ABOUT HERE

==========================

We now present the methodology we use to test these hypotheses.

METHODOLOGY

This section is organized as follows. First, we discuss the measures we use as inputs to AA

and the various steps we took to establish their reliability and validity. Second, we discuss our

unit of analysis, our sampling procedures, and the tests we use to identity any biases or

problems with our data. Third, we outline how we apply AA to our data.

Constructs and Measures

Here we seek to identify archetypes in four major areas of MNC subsidiary operations, so we

define our four study constructs as follows. (1) Local adaptation—the degree to which MNC

subsidiaries adapt their products, services and marketing activities to the demands of the local

market place. (2) Local innovation—the degree to which MNC subsidiaries innovate in their

products, services or marketing activities at the local level. (3) Local autonomy—the degree

to which MNC subsidiaries are given the freedom to make product, service and marketing

decisions at the local level. (4) Internal networking—the degree to which MNC subsidiaries

discuss product, service and marketing decisions with other subsidiaries or the head office of

Page 22: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

21

their company. These four constructs are rich and complex, embodying multiple facets of

multinational organizations. To adequately capture their richness, the constructs are measured

with multiple questionnaire items using 7-point Likert scales. Exhibit 1 contains an extract of

the relevant sections of the questionnaire.

==========================

EXHIBIT 1 ABOUT HERE

==========================

We profile the four construct areas via a two-stage, component and item approach

(Chin, Marcolin & Newsted, 1996). The constructs of local adaptation and local innovation

are each measured with five marketing mix components of price, product, positioning, place

and promotion (5Ps). The constructs of local autonomy and internal networking are measured

with these five components plus two additional components for marketing policy and

marketing people (7Ps). In turn we measure each component with three to four questionnaire

items. For example, the price component of the local autonomy construct is measured by

asking the extent to which decisions pertaining to customer credit, price discounting, retail

pricing and wholesale pricing are made at the subsidiary or headquarters level. It is the

components themselves that become inputs to archetypal analyses; allowing us to identify

archetypes for each construct across the 5 or 7Ps. Analyses presented in the Technical

Appendix demonstrate these components are unidimensional and have adequate convergent

and discriminant validity for our purposes here.

Unit of Analysis, Sampling, and Tests of Potential Biases

Unit of analysis. MNC subsidiaries often operate in more than one area of business. To

focus our study, obtain more precise data, and reduce the time cost to the responding

manager, we chose a business unit within the subsidiary as our unit of analysis. We define a

business unit as an organizational unit that has separate and independent marketing and

Page 23: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

22

profitability objectives. Within business units, we asked respondents to answer about the

product market with the highest annual sales revenue, assuming this to be most representative

of the business unit‘s activities. The key informant here is the head of the business unit.

Sample. A stratified random sample of MNE subsidiaries was selected from the Dun

and Bradstreet WorldBase. To ensure sufficient variance, strata included manufacturing and

services, consumer and industrial products, and subsidiaries in industrialized and

industrializing countries. Questionnaires were mailed to 1128 subsidiaries with a separate

questionnaire for each of the business units in the firm. The net response rate was 20 percent,

which compares favorably with the response rates of between 6 and 16 percent reported in the

literature for international surveys (Harzing, 1997). The responses we use here represent 229

business units; with an approximate 50:50 split between those operating in consumer and

those operating in business-to-business markets. Although the subsidiaries were located in 36

countries, their parent companies were mainly large Japanese, UK, and US MNCs with a

median of 22,000 employees worldwide and 325 employees in the subsidiary. Respondents

had an average of 10 years‘ experience in their company and averaged 40 years of age.

Potential bias. Although surveys are the standard approach to research in the

international business literature, questionnaire surveys inevitably raise concerns about

potential bias. Before analyzing our data we examined three such biases, namely measure

equivalence, common method bias, and non-response bias.

Measure equivalence. One potential bias in international studies concerns the degree to

which respondents from different countries interpret measures in the same way. For example,

this is a major issue in studies of individual values (Hui & Triandis, 1985). However, our

respondents were senior managers, mostly university educated, spoke English, traveled

widely, had been exposed to the business concepts incorporated in our measures, and were

familiar with questionnaire studies. While this suggests the potential for bias is low, we did

Page 24: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

23

check the equivalence of our measures. First, for each subsidiary we computed Kogut and

Singh‘s (1988) cultural distance measure (here using the UK as reference point, the

questionnaire being in English because of its common use in MNCs and the impracticability

of translating the questionnaire into 30+ languages). Second, we ranked our subsidiaries by

cultural distance (low, medium, and high distance) and compared the means for our measures

between the high and low groups. After correcting for the known bias in multiple

comparisons, there are no significant differences between these means. Scale equivalence

problems in these data are unlikely to have biased our analyses to any significant extent.

Common method bias. Using a common 7-point scale across all measures can create a

response bias. Here this might also be exacerbated as constructs have a similar format because

of the use of common underlying components. However, factor analyses demonstrate that

there is no common factor loading on all measures (the ex post one-factor test, Podsakoff &

Organ, 1986). Further, the questionnaire itself contained intervening sections on other topics

and different phrasing of the questions for each construct. These are also steps that can reduce

common method bias (Podsakoff, Mackenzie, Lee & Podsakoff, 2003). Hence, although we

cannot rule it out, common method bias is unlikely in these data.

Non-response bias. To test for non-response bias, the original sample drawn from Dun

and Bradstreet and those subsidiaries that responded were compared on three criteria: the

number of countries, how long the subsidiary had operated, and the number of employees. We

received responses from subsidiaries in 60% (36 of 60) of the countries we sampled, so any

bias due to the countries included or excluded is likely to be small. Our data also covered all

continents. The median age (i.e., length of operation) and size of the subsidiaries responding

was 30 years and 325 employees versus 21 years and 250 employees for the non-respondents.

Overall, though these statistics suggest a slight bias to older and larger subsidiaries, we

believe our data set is more than adequate for our analyses.

Page 25: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

24

Applying AA to Our Data

Our procedure for applying AA to our data is as follows.

Preliminary setup. First, we select the robust AA algorithm available within the R

package ―archetypal analysis‖ and follow the procedures described by the authors of this

package, Eugster and Leisch (2009), as well as the various steps outlined in the Technical

Appendix. Second, of the 229 business units responding to our questionnaire, we exclude 26

with missing data, leaving 203 complete questionnaires for further analysis. Third, we

identify multivariate outliers using standard techniques. Fourth, we ran preliminary archetypal

analyses—to identify whether any of these outliers had a strong influence on the selection of

archetypes. Five cases had a strong, distorting influence and are dropped—leading to a final

database of 198 MNC subsidiaries. Fifth, we generate 100 databases of normally distributed

random numbers; each with the same number of cases and variables, and the same scale

range, as the actual data. We use these databases to test whether the results of applying AA to

the actual data could occur by chance alone.

Archetypal analysis. For each set of variables, we examine solutions from one to ten

archetypes, repeating each analysis from 100 random starting points to reduce problems of

local minima. Our variables are standardized so they have equal weight in these analyses and

we take the best fitting solution from the 100 starting points. We also examine the warnings

from the package, none of which are of concern for the solutions we report here. Further, the

degree of fit (residual sum of squares) we obtain from different starting points is both

satisfactory and similar, suggesting there is no better fit to be found in these data.

Following practice in the AA literature, we use a scree plot of fit to determine the

number of archetypes to report. We also use the fit statistics from the 100 random databases

to identify a 1% confidence level. As an example, Figure 2 shows the fit to actual and random

data for the adaptation analysis (based on the 5P components of price, product, positioning,

Page 26: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

25

place and promotion). Here the scree plot indicates six archetypes as the best solution. Also,

for the number of archetypes from two to eight, the fit we obtain from actual data is better

than the 1% confidence level by a substantial margin. In contrast, we cannot distinguish the

solutions for one, nine and ten archetypes from chance. Solutions for innovation, autonomy

and networking have similar scree plots, and for two to eight archetypes are also better than

chance. For brevity these results are not shown here but are available from the authors on

request. These scree plots indicate six archetypes for innovation (based on the 5Ps) and five

for autonomy and networking (based on the 5Ps plus policy and people). For those familiar

with cluster analysis, and the problem of slicing the database into small unstable clusters, we

should also point out this is not the case for AA. Our five and six archetype solutions are the

exterior cases/polytope vertices that best describe the complete cloud of 198 data points.

Given we can rule out local minima and chance, these are robust solutions.

==========================

FIGURE 2 ABOUT HERE

==========================

RESULTS

The results of applying AA to our data reject Hypotheses 1a and 1b. For organization, both

local autonomy and internal networking require five archetypes and not three as hypothesized.

For strategy, both local adaptation and local innovation require six archetypes and not four.

Further, examining the scree plots and the statistics on which they are based demonstrates

solutions with three or four archetypes could not be considered adequate in either case. While

there are as yet no formal tests for this conclusion, the reductions in the residual sum of

squares between three and five archetypes for organization are considerable. As are the

reductions between four and six archetypes in the case of the strategy constructs.

Page 27: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

26

Figures 3 through 6 show the resulting archetypes. These we present as simple bar

charts, one chart for each archetype, five or six charts to a figure. All charts use the same

scaling from -3 to +3 standard deviations from the mean. Within each figure we also order the

individual archetypes. This we do from highly negative values on all components (Archetype

A) at the top left of the figure to highly positive values on all components at the bottom right

(F for Figures 3 and 4 with six archetypes, and E for Figures 5 and 6 with five archetypes).

==========================

FIGURES 3 – 6 ABOUT HERE

==========================

Thus for Figure 3, the six archetypes for local adaptation, archetypes A and F present

the biggest contrast. Archetype A is the classic subsidiary that adapts very little locally except

positioning to an average amount. The other four Ps are highly standardized. In contrast, F is

the classic subsidiary that adapts everything to a high degree. The remaining four archetypes

show more focused patterns of adaptation. B is a subsidiary that adapts just price and

promotion to an average extent, the rest being highly standardized. C is a subsidiary that

adapts price, and to an average extent place and product, but has highly standardized

positioning and promotion. D is a subsidiary that adapts product, place and promotion but has

relatively standardized positioning and highly standardized pricing. E is a subsidiary that

adapts everything except product, which is relatively standardized. Setting aside the greater

number of archetypes than hypothesized, there is some support for Hypothesis 1c in Figure 3.

Archetype E represents the hypothesized composite strategy with a standardized product but

all the other Ps locally adapted. And Archetype C represents the composite strategy of

standardized positioning and promotion and relatively more locally adapted place, product

and pricing.

Page 28: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

27

Turning to Figure 4, the six archetypes for local innovation, reveals a similar but not

identical set of archetypes. The first and last charts in the figure are the classic subsidiaries

that are either not innovative at all (A) or highly innovative (F). This pattern is similar to that

for adaptation. However, the other, more focused, archetypes are not. B is a subsidiary that is

averagely innovative on promotion and place, but much less innovative on the other Ps. C is a

subsidiary somewhat similar to B except that it is noticeably more innovative on price. D is a

subsidiary that is innovative on promotion and positioning but below average on the other Ps.

And E is a subsidiary that is innovative on place, positioning and price but not on the other Ps.

There is some support for Hypothesis 1c in Figure 4, although this is not as strong as for local

adaptation. Archetype E comes closest to the hypothesized profile of no local innovation on

product, but local innovation on the other Ps. However, the promotion component with no

local innovation does not fit this hypothesis. And while archetype C is a composite strategy

with low innovation on positioning, only price is the focus of local innovation. The other Ps

having average levels of local innovation.

Turning to Figure 5, the five archetypes for local autonomy, we have two additional Ps

to consider, people and policy. Again the first and last charts are as before. Archetype A is the

classic subsidiary that is highly centralized and allowed very little autonomy on any of the

seven marketing decision areas. In contrast, E is the classic subsidiary with a high degree of

freedom to act in all seven areas. Interestingly, B is a similar subsidiary to A except it is given

more freedom in the areas of product and price. And D is a similar subsidiary to E except it is

given less freedom in the area of product. The middle archetype, C, is given less than average

autonomy to act on product, positioning and promotion, but greater than average autonomy on

people, policy and place.

Turning to Figure 6, the five archetypes for internal networking, the first and last charts

are similar to those for autonomy. We have a subsidiary, A, that is relatively unconnected

Page 29: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

28

from the rest of the organization in all the seven decision areas. And we have the opposite in

subsidiary E that is highly connected in all seven areas. However, the remaining three

archetypes do not resemble those for autonomy. B is a subsidiary that, like A, is also

relatively unconnected but somewhat more connected in the area of people. C is also

relatively unconnected except in the area of product where it does network with the rest of the

organization. And D is a subsidiary, like E, that is relatively well connected to the rest of the

organization except to a lesser extent in the areas of place and price.

Table 1 summarizes the distribution of subsidiaries by archetypes across our four

constructs. The numbers under each archetype (A to F) represent the subsidiaries which have

an association of 0.5 or more with the respective archetype. The columns ‗Total Classic‘ and

‗Total Composite‘ show the total numbers of these strongly associated subsidiaries according

to whether they belong to two types of archetype. That is, either a classic archetype with an

extreme profile or a composite archetype with a non-extreme or middle profile. The classic

subsidiaries are associated with the archetypes A and F for adaptation and innovation and A

and E for autonomy and networking. The composite cases are strongly associated with the

remaining, non-extreme archetypes. The ‗Mixtures‘ column gives the total numbers of the

remaining subsidiaries. That is, those that do not have an association of 0.5 or greater with

any archetype and thus are mixtures of two or more archetypes, And the ‗hybrid‘ column

totals are the sums of the mixed and composite totals, in other words, the subsidiaries that are

not associated with the classic archetypes. As shown in the Table, the majority of our sample

subsidiaries are hybrids across all aspects of strategy and organization. As most subsidiaries

in our sample have hybrid strategies, the results in Table 1 provide strong support to

Hypothesis 2.

Page 30: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

29

================================

TABLE 1 ABOUT HERE

================================

Figure 7 illustrates this conclusion for the adaptation measures by way of a parallel plot

of the six association scores for each subsidiary with each archetype. The density of the

plotted scores in the Figure show the archetypes A and F--the classic strategies of

standardization and adaptation respectively—are less representative of our sample than the

composite archetypes B to E. Further, many of the plot lines for individual subsidiaries are

placed towards the bottom of the figure. This indicates a large number of subsidiaries that are

mixtures of two or more archetypes.

Overall, the local adaptation strategy in MNC subsidiaries is more closely represented

by the composite archetypes (B to E) than by the classic ones (A and F). This is also the case

for local innovation, local autonomy and internal networking (Figures not shown but available

from the authors).

==========================

FIGURE 7 ABOUT HERE

==========================

DISCUSSION

First, we believe our results show AA to be a useful additional tool in the study of

multinational subsidiaries. AA provides a small number of profiles that summarize complex

data in a meaningful way. We should also not forget that, unlike cluster analysis, each of

these archetypes corresponds to one or more real subsidiaries. Moreover, there is a rationale

based in multidimensional topology for why these five or six archetype solutions are the best

summary of our data cloud. Finally, AA allows for subsidiaries to be described as mixtures of

archetypes, rather than forcing them into one profile.

Page 31: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

30

Second, these results present challenges to conventional thinking about MNC strategy

and organization, particularly as they indicate greater heterogeneity than the literature

suggests. In our terms, more archetypes are necessary to describe the data cloud than might

be hypothesized from this literature. Thus we reject Hypotheses 1a of three organizational

archetypes and Hypothesis 1b of four strategy archetypes in favor of five and six archetypes

respectively. And Hypothesis 1c on the profiles of some of these additional, or ‗composite,‘

archetypes receives only modest support. There are archetypes in our data that cannot be

predicted from the existing literature. Further, and in contrast to the simplicity of the

literature, the majority of MNC subsidiaries are ‗hybrids.‘ That is, either represented by

composite (non-classic) archetypes or a mixtures of archetypes (interior data points).

An alternative way to visualize these results is to imagine the data cloud of subsidiaries

as a spikey ball where each of the spikes represents an archetype. The existing literature

focuses on some, but not most of the spikes on this ball. Nor does this literature focus on the

many subsidiaries in the interior of the ball. This analogy is not exact as our data cloud is in

five or six dimensions, but it does demonstrate the scale of the research challenge in

formulating better theories of MNC strategy and organization.

The detailed results themselves allow two conclusions and an open question. First,

while the archetypes that are highly negative on all variables (top left), or highly positive on

all variables (bottom right), are somewhat similar across the four constructs, the other

archetypes are not. We do see subsidiaries with these stereotypical patterns—that is, highly

standardized versus highly adapted, non-innovative versus innovative, centralized versus

decentralized and isolated versus networked. But we also see many other, more focused,

patterns where there are negative values on some variables but central or positive values on

others. In fact, of 22 archetypes, 14 are like this. Second, these focused patterns are not the

same across the four constructs—suggesting a subtlety, complexity and heterogeneity of

Page 32: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

31

MNC subsidiary marketing that is not fully recognized in the literature. Third, the solutions

for the market-based constructs—adaptation and innovation—have six archetypes and those

for the organizational constructs—autonomy and networking—have five. We have examined

other numbers of archetypes for all four constructs but always reach the conclusion that six

and five are the right numbers. Essentially there is less multi-dimensional complexity in the

organizational constructs (despite them being based on two more variables) than the market-

based ones. Now it seems evident that where subsidiaries are given autonomy they will

develop more alternative strategies than where they are controlled. So we might expect more

complexity in the market-based constructs. But an interesting question for future research is

why only one more archetype is necessary to represent this added complexity? Might not

more be expected? Answering that question will of course require developing hypotheses

about the relationships between the four sets of archetypes. We may also need to incorporate

other variables; for example, descriptions of the local country environment, as these may play

an important role in the choice of adaptation or innovation strategy. Lastly, we also need to

develop explanations for those subsidiaries with hybrid strategies or organizations. What

leads decision makers to such configurations? And what are the performance consequences

of pure versus hybrid strategies or organizations?

Our study has the normal limitations, notably the cross-sectional nature of our survey

and the use of self-report data from one key informant. However, we believe it is strong

enough to suggest the international business literature needs to pay more attention to the

heterogeneity of the various marketing mix strategies and decision-making structures MNCs

employ across the globe.

Page 33: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

32

REFERENCES

Bartlett, C.A. & Ghoshal, S. 1986. Tap your Subsidiaries for Global Reach. Harvard Business

Review, November-December, 87–94.

Bartlett, C.A. & Ghoshal, S. 1989. Managing Across Borders: The Transnational Solution.

Boston: Harvard Business School Press.

Birkinshaw, J. Hood, N. & Jonsson, S. 1998. Building Firm-Specific Advantages in

Multinational Corporations: The Role of Subsidiary Initiative. Strategic Management

Journal, 19(3), 221-241.

Buzzell, R.D. 1968. Can You Standardize Multinational Marketing? Harvard Business

Review. 46(November-December), 102-113.

Cavusgil, S.T. & Zou, S. 1994. Marketing Strategy-Performance Relationship: An

Investigation of the Empirical Links in Export Market Ventures. Journal of Marketing,

58(1), 1-21.

Charan, R. 1993. How Networks Reshape Organizations – for Results. In R. Howard (Ed.),

The Learning Imperative: Managing People for Continuous Innovation: 111-132.

Boston: Harvard Business School Press.

Chin, W.W. Marcolin, B.L. & Newsted, P.R. 1996. A Partial Least Squares Latent Variable

Modeling Approach for Measuring Interaction Effects: Results from a Monte Carlo

Simulation Study and Voice Mail Emotion/Adoption Study. Proceedings of the

Seventeenth International Conference on Information Systems, 16-18 December,

Cleveland, Ohio.

Cutler, A. & Breiman, L. 1994. Archetypal Analysis. Technometrics, 36(4), 338-347.

Devinney, T.M., Midgley, D.F. and S. Venaik 2000. The Organizational Imperative and the

Optimal Performance of the Global Firm: Formalizing and Extending the Integration-

Responsiveness Framework. Organization Science, 11(6), 674-695.

Page 34: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

33

Edy, C. 1999. The Olympics of Marketing. American Demographics, 21(6), 47-48.

Egelhoff, W.G. 1988. Organizing the Multinational Enterprise: An Information-Processing

Perspective. Cambridge, MA: Ballinger.

Eugster M.J.A. & Leisch, F. 2009. From Spider-Man to Hero--Archetypal Analysis in R.

Journal of Statistical Software, 30(8): 1-23.

Ghoshal, S. Korine, H. & Szulanski, G. 1994. Interunit Communications in Multinational

Corporations. Management Science, 40(1), 96-110.

Gupta, A.K. & Govindarajan, V. 1991. Knowledge Flows and the Structure of Control within

Multinational Corporations. Academy of Management Review, 16(4), 768-792.

Gupta, A.K. & Govindarajan, V. 1994. Organizing for Knowledge within MNCs.

International Business Review, 3(4), 443-457.

Harzing, A.W. 1997. Response Rates in International Mail Surveys: Results of a 22-Country

Study. International Business Review, 6(6), 641–665.

Howard, R. (Ed.). 1993. The Learning Imperative: Managing People for Continuous

Innovation. Boston: Harvard Business School Press.

Hui, H. C. & Triandis, H. C. 1985. Measurement in cross-cultural psychology: A review and

comparison of strategies. Journal of Cross-Cultural Psychology, 16(2): 131-152.

Kashani, K. 1989. Beware the Pitfalls of Global Marketing. Harvard Business Review, 67(5),

91-98.

Kogut, B. & Singh, H. 1988. The Effect of National Culture on Choice of Entry Mode.

Journal of International Business Studies, Fall, 411-432.

Kotler, P. 1999. Kotler on Marketing: How to Create, Win, and Dominate Markets. New

York: Free Press.

Page 35: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

34

Lages, L.F. Jap, S.D. & Griffith D.A. 2008. The Role of Past Performance in Export

Ventures: A Short-term Reactive Approach. Journal of International Business Studies,

39, 304-325.

Lawrence, P.R. & Lorsch, J.W. 1967. Managing Differentiation and Integration. Boston:

Harvard University.

Lee, R.P. Chen, Q. Kim, D. & Johnson, J.L. 2008. Knowledge Transfer between

Multinational Corporations‘ Headquarters and their Subsidiaries: Influences on and

Implications for New Product Outcomes. Journal of International Marketing, 16 (2), 1-

31.

Levitt, T. 1983. The Globalisation of Markets. Harvard Business Review, 61(May-June), 92-

102.

McGrath, R.G. 2001. Exploratory Learning, Innovative Capacity, and Managerial Oversight.

Academy of Management Journal, 44(1), 118-131.

Miller, D. 1992. The Generic Strategy Trap. The Journal of Business Strategy, 13(1), 37-41.

Nasir, V.A. & Altinbasak, I. 2009. The Standardization/Adaptation Debate: Creating a

Framework for the New Millenium. Strategic Management Review, 3(1), 17-50.

Neff, J. 1999. Test it in Paris, Launch it in Paris Texas. Advertising Age, 70(23), XX.

Nobel, R. & Birkinshaw, J. 1998. Innovation in Multinational Corporations: Control and

Communication Patterns in International R&D Operations. Strategic Management

Journal, 19(5), 479-496.

Ozsomer, A. & Simonin, B.L. 2004. Marketing Program Standardization: A Cross-Country

Exploration. Intern. J. of Research in Marketing, 21, 397-419.

Podsakoff, P.M. & Organ, D.W. 1986. Self-Reports in Organizational Research: Problems

and Prospects. Journal of Management, 12(4), 531-544.

Podsakoff, P.M. MacKenzie, S.B. Lee, J.Y & Podsakoff, N.P. 2003. Common Method Biases

Page 36: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

35

in Behavioral Research: A Critical Review of the Literature and Recommended

Remedies. Journal of Applied Psychology, 88(5), 879-903.

Pollack, J. 1999. Pringles Wins Worldwide with One Message. Advertising Age, 70(2), 14+.

Porter, M. 1980. Competitive Strategy. New York: Free Press.

Porter, M. 1985. Competitive Advantage: Creating and Sustaining Superior Performance.

New York: Free Press.

Porter, M.E. 1990. The Competitive Advantage of Nations. Harvard Business Review, 68(2),

73-94.

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(1), 116-145.

Prahalad, C.K. & Doz, Y. 1987. The Multinational Mission: Balancing Local Demand and

Global Vision. New York: Free Press.

Sheth, J.N. 2011. Impact of Emerging Markets on Marketing: Rethinking Existing

Perspectives and Practices. Journal of Marketing, 75(July), 166 -182.

Sorenson, R.Z. & Wiechmann, U.E. 1975. How Multinationals View Marketing

Standardization. Harvard Business Review, 53(May-June), 38.

Sundaram, A.K. & Black, J.S. 1992. The Environment and Internal Organization of

Multinational Enterprises. Academy of Management Review, 17(4), 729-757.

Takeuchi, H. & Porter, M.E. 1986. Three Roles of International Marketing in Global Strategy.

In Porter, M.E. (Ed.). Competition in Global Industries: 111-146. Boston: Harvard

Business School Press.

Teece, D.J. 1996. Firm Organisation: Industrial Structure, and Technological Innovation.

Journal of Economic Behaviour and Organisation, 31, 193-224.

Page 37: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

36

Tong, C. Wong, A. & Kwok, E.Y. 2012. Major Determinants Affecting the Autonomy of

Multinational Corporation Subsidiaries in China. Journal of Management Research,

4(1), 1-33.

UNCTAD. 2011. World Investment Report. New York and Geneva: United Nations.

Van de Ven. 1976. A Framework for Organization Assessment. Academy of Management

Review. 1, 64-78.

Venaik, S. Midgley, D.F. & Devinney, T.M. 2004. A New Perspective on the Integration-

Responsiveness Pressures Confronting Multinational Firms. Management International

Review, 44(1), 15-48.

Venaik, S. Midgley, D.F. & Devinney, T.M. 2005. Dual paths to performance: the impact of

global pressures on MNC subsidiary conduct and performance. Journal of International

Business Studies, 36(6), 655-675.

Zanfei, A. 2000. Transnational Firms and the Changing Organization of Innovative Activities.

Columbia Journal of Economics, 24, 515-42.

Zou, S. & Cavusgil, S.T. 2002. The GMS: A Broad Conceptualization of Global Marketing

Strategy and Its Effect on Firm Performance. Journal of Marketing, 66(4), 40-56.

Page 38: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

37

Table 1: Distribution of Subsidiaries by Archetypes and Degree of Association

Associated ≥ 0.5 with

Archetypes Total

Classic

Total

Composite

Mixtures

(≤0.5) Hybrid (%) Constructs A B C D E F

Local

Adaptation 10 7 6 6 25 48 58 44 96 140 (71%)

Local

Innovation 23 12 14 5 8 49 72 39 87 126 (64%)

Local

Autonomy 24 2 8 48 62 - 86 58 54 112 (57%)

Internal

Networking 38 13 30 18 57 - 95 61 42 103 (52%)

Page 39: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

38

Figure 1: Associating Subsidiaries with Strategy Archetypes

Page 40: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

39

Figure 2: Scree Plot for Local Adaptation

Page 41: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

40

-3 0 3

Price

Product

Posi oning

Place

Promo on

F

-3 0 3

Price

Product

Posi oning

Place

Promo on

A

-3 0 3

Price

Product

Posi oning

Place

Promo on

B

-3 0 3

Price

Product

Posi oning

Place

Promo on

C

-3 0 3

Price

Product

Posi oning

Place

Promo on

D

-3 0 3

Price

Product

Posi oning

Place

Promo on

E

Figure 3: Local Adaptation

Page 42: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

41

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

A

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

B

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

C

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

D

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

E

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

F

Figure 4: Local Innovation

Page 43: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

42

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

A

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

B

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

C

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

D

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

E

Figure 5: Local Autonomy

Page 44: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

43

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

B

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

A

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

C

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

D

-3.00 0.00 3.00

Price

Product

Posi oning

Place

Promo on

Policy

People

E

Figure 6: Internal Networking

Page 45: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

44

Figure 7: Association Scores between MNC Subsidiaries and Adaptation Archetypes

Page 46: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

45

Exhibit 1: Examples of Scales for Marketing Mix Strategy and Decision-Making in

MNC Subsidiaries

MARKETING MIX STRATEGY – LOCAL ADAPTATION

In subsidiaries of multinational firms, the marketing mix elements may be standardised (i.e.,

not modified for the local subsidiary market), or adapted (i.e., completely modified for the

local subsidiary market). Please indicate the extent to which the marketing mix elements for

your local subsidiary business unit are standardised or adapted. Circle the appropriate number

for each element on a scale of 1 to 7, where 1 means standardised, and 7 means adapted.

Product Brand Name--------------------------------------- 1 2 3 4 5 6 7

MARKETING MIX STRATEGY – LOCAL INNOVATION

Marketing innovation is defined as the extent to which a business unit seeks new ideas for

conducting its marketing activities and improving its marketing mix. Please indicate the extent

to which your local subsidiary business unit is innovative, i.e., seeking new ideas for

conducting the marketing mix activities. Circle the appropriate number for each activity on a

scale of 1 to 7, where 1 means not innovative, and 7 means highly innovative.

Product Brand Name-------------------------------------- 1 2 3 4 5 6 7

MARKETING MIX DECISION-MAKING – LOCAL AUTONOMY

In subsidiaries of multinational firms, the marketing mix decisions for the local subsidiary

business unit may be centralised (i.e., the decisions are never taken in the local subsidiary),

or autonomous (i.e., the decisions are always taken in the local subsidiary). Please indicate

the extent to which the marketing mix decisions for your local subsidiary business unit are

centralised or autonomous. Circle the appropriate number for each decision on a scale of 1 to

7, where 1 means centralised, and 7 means autonomous.

Product Brand Name Decisions-------------------------- 1 2 3 4 5 6 7

MARKETING MIX DECISION-MAKING – INTERNAL NETWORKING

Networks are defined as groups, such as teams, task forces, meetings, committees, etc.,

comprised of managers from the corporate and regional headquarters, and the various

country subsidiaries of the parent company. Please indicate the extent to which the marketing

mix decisions for your local subsidiary business unit are taken in networks. Circle the

appropriate number for each decision on a scale of 1 to 7, where 1 means never taken in

networks, 7 means always taken in networks.

Product Brand Name Decisions ------------------------- 1 2 3 4 5 6 7

Page 47: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

46

TECHNICAL APPENDIX

This appendix is divided into two sections, A and B.

Section A outlines archetypal analysis (AA) and discusses issues in applying this

technique to the data typically obtained from surveys of international business. Section A

also presents arguments why we believe AA is an appropriate and strong technique for

describing heterogeneity in such data.

Section B details the tests of convergent and discriminant validity we applied to our

data, both at the level of individual construct measures and for the archetypes we develop

from these measures. Section B also presents some comparisons between standard cluster

analysis solutions and our archetypes. These comparisons also help us to demonstrate the

superiority of AA over these standard clustering approaches.

A: Archetypal Analysis

In our view, the best way to explain AA is to start from cluster analysis, the basics of which

many people know. AA is not a clustering approach, but discussing it in this manner

highlights the key conceptual difference between AA and cluster analysis, a difference that

underlies the potential superiority of AA for our purposes.

Cluster analysis. Here we will only outline two of the main techniques. In doing so, we

will briefly examine the assumptions underlying these techniques and the usefulness of the

clusters they generate. This will set the context for our discussion of AA. We should note we

have simplified this discussion to a few key points. Cluster analysis is a large and complex

topic. The reader is referred to a text such as Kaufman and Rousseeuw (1990) for details on a

broader range of techniques and more depth on the statistical foundations of clustering. The

two basic techniques we outline are (1) hierarchical clustering and (2) centroid clustering. As

for most clustering techniques, both use a measure of the ‗distances‘ between the units of

analysis to determine which units are ‗closest‘ and therefore most similar to each other. This

Page 48: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

47

distance is computed across all the variables of interest, typically using either the Euclidean or

Manhattan (‗city-block‘) metrics. For example, using the Manhattan metric the distance

between two units is simply the sum of the absolute differences between the two values of

each of the variables describing these units.

Hierarchical clustering (Ward 1963) assigns the units of analysis into a tree structure

(‗dendrogram‘). This structure is generated by merging those units closest to each other into

‗twigs,‘ those twigs closest to each other into ‗branches‘ and so forth into the main ‗trunk‘ of

the tree. This tree can be built bottom-up (‗agglomerative‘ clustering) or top down (‗divisive‘

clustering.) But whether bottom-up or top down, to arrive at the final tree we require

additional assumptions on how we link clusters to each other. For example, do we assume the

distance between two clusters is that between the two ‗nearest neighbor‘ units or do we

assume it is the average of the distances between all the units in the first cluster and all those

in the second? Different assumptions often produce quite different tree structures.

One advantage of hierarchical clustering is each cluster is derived from a real

observation and is therefore readily interpretable. One disadvantage is the tree structure itself

is often unwieldy for practical purposes as it presents a complete picture of all the data.

Choosing a simpler structure from the tree can then become more of a subjective judgment on

the part of the researcher than an outcome of the analysis itself. For that reason, systematic

methods to help with this simplification have been developed (i.e. Boudaillier & Hebrail,

1998).

The second technique is centroid clustering, typically the k-means variant (MacQueen

1967). With k-means the researcher specifies the number (k) of clusters they want and

supplies starting or ‗seed‘ units for each of these (typically randomly selected). The algorithm

then identifies which other units are closest to each seed unit and designates these as the

starting clusters. The means of each cluster on the set of variables—the centroids—are then

Page 49: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

48

computed and used as reference coordinates for a new set of distances. This process repeats

itself until the solution converges to a local minimum, with the objective of maximizing the

similarity of the units within each cluster.

One advantage of k-means is that it provides a simple clear description of heterogeneity

by way of a small number of clusters. This simple solution may be the reason k-means

appears to be more popular for business applications than hierarchical clustering. One

disadvantage is the centroid of a cluster may not resemble any real observation, blurring

interpretation. For this reason, more recent algorithm use mediods rather than centroids,

where the mediod is the observation with the smallest average distance to the other

observations in the cluster (Van der Lann, Pollard & Bryan, 2003).

From this discussion of hierarchical and centroid clustering, it can be seen we would

often prefer simple cluster solutions where each cluster can be accurately represented by a

real observation rather than a statistic. And we would prefer these solutions to be less

sensitive to the algorithmic assumptions we make. It is also notable that for both of these

techniques to work well requires (1) clusters that are clearly distinct from each other and (2)

units that clearly belong to one cluster. Unfortunately real data is often not like this. For

example, variables may not strongly discriminate between units, ―clusters‖ may overlap and

units may ―belong‖ to more than one cluster or be isolated from all clusters. The assumption

of ―lumpy clouds‖ of data in some highly dimensional space is a strong one that may not

always be valid (Elder & Pinnell, 2003). And when this assumption is not valid, both

hierarchical and centroid clustering will fail to produce useful solutions.

Moreover, unambiguously assigning each unit to a cluster is not the only way to

describe multivariate data. More recent techniques—such as fuzzy clustering—take a

different approach. Fuzzy clustering (Bezdek, 1981) does this by describing each unit as

having varying degrees of membership to all the clusters identified in the data. This

Page 50: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

49

introduces the idea of probabilistic mixtures as a viable alternative to unambiguous

assignment. Instead of saying a unit is either of cluster X or Y, we say it is 70% X and 30%

Y. This mixture approach is potentially both more flexible and more realistic given the data

we typically have available in international business.

In addition, mixtures have more in common with the nuanced ways in which human

beings think. Indeed, reasoning from universally understood prototypes and seeing other

objects as mixtures of these is common in everyday activity and language (for example,

‗…there‘s a little bit of __________ in everyone‘.). The idea of such universal prototypes—

archetypes—also has a long history in philosophy (Plato) and psychology (Jung). Archetypal

analysis builds on these ideas and is based on a mixture model. We argue it is an

improvement on all forms of clustering, including fuzzy clustering, because it has a

conceptual definition of ‗archetype‘ that builds on the topology of the data. In contrast, fuzzy

clustering (and other more recent clustering methods) continue in the mainstream clustering

tradition of defining ―clusters‖ primarily through the features of the specific algorithm used.

We will return to this point shortly, first it is necessary to outline the main features of AA.

Archetypal analysis. Cutler and Breiman (1994) introduced AA as a formal statistical

technique. They motivated the analysis problem by reference to data on the head

measurements of 200 Swiss soldiers. Was it possible to identify ―pure‖ or archetypal head

types in these data, with the constraint that each individual soldier‘s head should be

represented as a mixture of these archetypes? To achieve this goal requires an algorithm that

approximates each real head by a mixture, while maximizing the overall fit to the data. But to

find these mixtures requires we first identify the ‗archetypes‘ that generate them. And to

identify the archetypes requires we have a definition of what they are. The main insight of

Cutler and Breiman was to formally define their archetypes through the topology of the data.

Page 51: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

50

To understand this topological definition, we first need to imagine our data as a cloud of

points in the multidimensional space described by our set of variables. Next we enclose this

data cloud by a hypersurface whose vertices are exterior data points of the cloud. The

hypersurface defined by the minimum number of such data points is called the convex hull of

the data. In two dimensions we can use the analogy of snapping a rubber band around an

object—the taut band becomes the convex hull of that object. AA approximates this convex

hull by a simpler polytope (a polytope is the multidimensional extension of a polyhedron).

The vertices of this polytope are the archetypes. They are also data points and the polytope

itself encompasses a smaller hypervolume than the convex hull. Data points lying inside the

convex hull are exact mixtures of archetypes, while points lying outside the polytope are only

approximated (Cutler & Breiman, 1994). This process of approximating the convex hull by a

simpler polytope has analogies to many statistical techniques where we use a reduced form to

capture most but not all of the information in the data. For example, choosing a small number

of factors in principal components analysis or dropping non-significant coefficients from a

regression. Generally the number of archetypes is much smaller than the number of vertices in

the convex hull (Chan, Mitchell & Cram, 2003), allowing a more parsimonious

representation.

In summary then, archetypes are the specific exterior points that best account for the

shape of the data cloud. In our opinion, this topological definition of the pure types in the

mixture makes AA superior to other mixture models such as fuzzy clustering. The archetypes

in AA are defined more in terms of the overall shape of the data cloud than through the

operation of the algorithm. And largely because of this definition, AA provides a number of

advantages over the commonly used techniques such as hierarchical or k-means clustering.

First, AA can produce solutions that are sharper and more differentiated than k-means

or hierarchical clustering (Li et. al., 2003; Elder & Pinnel, 2003). Second, each archetype is

Page 52: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

51

associated with a real observation, facilitating interpretation (Elder & Pinnel, 2003). Third,

the degree of association of an individual unit is defined against the standard of the archetype

rather than a less meaningful cluster centroid (Elder & Pinnel, 2003). Fourth, AA imposes no

restrictions of orthogonality (Cutler & Breiman, 1994), nor does it impose a strong ‗model‘

on the data (Li et al., 2003). Rather the key modeling assumption is that a simple polytope is a

good representation of the cloud. Finally, simulations have shown AA is robust in the

presence of Gaussian, Poisson and systematic error noise in the data (Chan et al., 2003).

Overall, AA produces simple, interpretable and robust solutions where the key vectors—the

archetypes—have defined meaning and key assumption is the polytope. This contrasts with

the complexity, sensitivity to algorithmic assumptions and lack of interpretability of many

clustering approaches.

That said, it is necessary to clear up some lack of clarity in the literature and note some

disadvantages of this technique. First, the exterior data points defining the archetypes are not

outliers or extreme cases (here we present a different point of view to Li et al., 2003). AA is

just as vulnerable to the influence of outliers as any statistical technique (Eugster & Leisch,

2010). Extreme data points—i.e. ones far from the main cloud of data—may distort the

archetypes by extending the polytope well beyond the cloud. The researcher should either

exclude these outliers from their analysis or use robust archetypal methods (Eugster & Leisch

2010). Second, like many other techniques using numerical optimization, the alternating least

squares algorithm that Cutler and Breiman utilize to fit the polytope to data provides no

guarantee of finding a global minimum. And for AA this problem appears to get worse as the

number of archetypes increases (Cutler & Breiman, 1994; Elder & Pinnell, 2003).

Researchers need to (1) use an adequate range of random starting points and (2) demonstrate

the minimum they obtain is not the result of statistical chance. While such steps are desirable

whenever numerical optimization is used, for AA this may be a necessity. Third, AA is a

Page 53: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

52

relatively new technique with only a limited number of applications in the literature to date

and only a few published simulation studies investigating its properties. More research is

clearly needed before it becomes a standard technique.

B: Tests of Convergent and Discriminant Validity

The correlations in Tables A1 to A4 show our measures have high convergent and

discriminant validity. Individual items are considered to have convergent validity if they

correlate more than 0.7 with the component that they intend to measure. All items satisfy this

criterion. Similarly discriminant validity is evaluated by examining the cross-loadings of the

items and their components. Examining Tables A1 to A4, the correlations of the components

with their items are higher than those with the items associated with other components.

Simple factor analyses also demonstrate all components are one-dimensional, which is a basic

requirement for these component measures and statistics to be valid.

=============================

TABLES A1 – A4 ABOUT HERE

=============================

As noted before, we define a subsidiary to be associated with an archetype if its score

for that archetype is 0.5 or higher. This seems reasonable as AA normalizes the association

scores of a subsidiary with all the archetypes to sum to 1. Hence, an association of 0.5 or

more with a single archetype implies the subsidiary does not have a higher association with

any other archetype or all the other archetypes together. To examine the convergent and

discriminant validity of our AA results, we separate our sample into (1) those subsidiaries that

are clearly associated with a single archetype (score of 0.5 or more) and (2) those that are

mixtures (all scores less than 0.5). For the first sample, we test if the average association with

their respective archetypes is significantly higher than their average association with the other

archetypes. In a table of such averages we should see the same pattern as in the standard

Page 54: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

53

tables of convergent and discriminant validity shown earlier. That is, higher values in the

diagonal cells and lower values in the off-diagonal cells. This is indeed the pattern we see in

Tables A5 to A8. Moreover, for 21 of the 22 archetypes we see in these figures the on-

diagonal values are also significantly higher than the off-diagonal values. We perform this

test by computing the 95% lower confidence limit for the diagonal values and checking

whether all off-diagonal values are below this number. We use the t-distribution here because

of the small numbers of subsidiaries allocated to some cells. Thus we have some basis for

concluding there is adequate convergent and discriminant validity for 21 of the 22 archetypes

we present here. The one exception is archetype B for autonomy—there is a possibility this

composite archetype overlaps with another

============================

TABLES A5 – A8 ABOUT HERE

=============================

For the second sub-sample of subsidiaries associated with a mixture of archetypes, the

average subsidiary weights are nearly uniformly distributed across all archetypes for each of

the four constructs. For example, for the 96 subsidiaries with mixed adaptation strategies, the

average association with each of the six archetypes ranges from 0.13 to 0.25. Similarly, the

87 subsidiaries with mixed innovation strategies have average association ranging from 0.08

to 0.27. Turning to organization, the average association for the 54 mixture subsidiaries in

autonomy and the 42 mixture subsidiaries in networking ranges from 0.13 to 0.25 and 0.10 to

0.27 respectively. Thus, overall, the average association for the subsidiaries with mixed

strategies or organizations does not exceed 0.3 on any archetype. Here we cannot say

anything about convergent or discriminant validity because there is no referent. What we can

say is the interior data in our cloud are quite distinct from the exterior.

Page 55: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

54

Comparing Archetypal Analysis (AA) and Cluster Analysis (CA)

Although AA has a totally different conceptual basis to CA, it is often compared with CA (Li

et al., 2003) and the archetypes in AA considered to be similar to clusters in CA. This is not

really the case. To illustrate the differences between AA and CA, we cluster the cases based

on the component measures for each of the four constructs and compare the resulting

solutions with the archetypes reported earlier. Since these comparisons produce similar

conclusions across all four constructs, we only report those for the adaptation construct here

(the comparisons for the other constructs are available from the authors). To make the

comparisons we apply a standard clustering algorithm—k-means—to the adaptation

component measures and generate a six-cluster solution to match the six archetypes we

extracted earlier. We then compare the cluster profiles with the corresponding archetype

profiles. As shown in the last column of Table A9, the profiles of cluster 3 and archetype C

are negatively related (r=-.84) and those of cluster 4 and archetype D only weakly related (r=

.26). In contrast, the profiles of clusters 1 and 2 and the corresponding archetypes A and B

are more strongly correlated (r = .55 and r = .68 respectively) and the profiles of clusters 5

and 6 and the corresponding archetypes E and F are highly correlated (r = .95 and r = .98

respectively). From correlations between profiles, we might therefore conclude there is a

relationship between the two solutions but they are by no means identical.

=========================

TABLE A9 ABOUT HERE

==========================

However, as well as correlating profiles we also need to see how the two techniques

classify the various subsidiaries. To examine this point we cross-tabulated the cases

classified into each of the six clusters and archetypes. This cross-tabulation is shown in Table

A10 and as previously separates those subsidiaries clearly associated with an archetype from

Page 56: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

55

those that are mixtures of two or more archetypes. This separation clarifies the difference

between standard CA and AA and shows what drives the correlations in Table A9. These

correlations originate from those subsidiaries clearly associated with an archetype—74 of 102

matching in the same pattern as the correlations (namely, clusters 1, 2, 5 and 6 with

archetypes A, B, D and E). However, what is missing in CA and provided by AA is the

identification of 96 subsidiaries with mixed strategies. AA thus adds a level of insight into

the data that is missing in CA. Moreover, the CA solution can be seen to be misleading as

subsidiaries with mixed strategies are assigned to one cluster. Fuzzy clustering techniques

would ameliorate this problem but lack the clear definition the exterior points of AA provide.

It is for all these reasons we recommend AA as a technique well worth considering in

international business research.

=========================

TABLE A10 ABOUT HERE

==========================

TECHNICAL REFERENCES

Bezdek, J. 1981. Pattern Recognition with Fuzzy Objective Function. Plenum Press, New

York.

Boudaillier, E. & Hebrail, G. 1998. Interactive Interpretation of Hierarchical Clustering,

Intelligent Data Analysis, 2, 229-244.

Cutler, A. & Breiman, L. 1994. Archetypal Analysis. Technometrics, 36(4), 338-347.

Elder, A & Pinnel, J. 2003. Archetypal Analysis: An Alternative Approach to Finding and

Defining Segments. Proceedings of the Sawtooth Software Conference, Sequim, WA,

June, 113-132.

Eugster, M.J.A. & Leisch, F. 2010. Weighted and Robust Archetypal Analysis. Technical

Report Number 082. Department of Statistics University of Munich.

Page 57: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

56

Kaufman, L & Rousseeuw, P. 1990. Finding Groups in Data: An Introduction to Cluster

Analysis. New York: John Wiley & Sons.

Li, S. Wang, P. Louviere, J. & Carson R. 2003. Archetypal Analysis: A New Way to Segment

Markets Based on Extreme Individuals. ANZMAC Conference Proceedings, Adelaide,

December, 1674-1679.

MacQueen, J. B. 1967. "Some Methods for classification and Analysis of Multivariate

Observations". Proceedings of 5th Berkeley Symposium on Mathematical Statistics

and Probability. University of California Press. 281–297.

Van Der Lann, M. J. Pollard K. S. & Bryan, J. E. 2003. A New Partitioning Around Medoids

Algorithm. Journal of Statistical Computation and Simulation. 73 (8), 575–584

Ward, J. H. 1963. "Hierarchical Grouping to Optimize an Objective Function". Journal of the

American Statistical Association, 58 (301), 236–244.

Page 58: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

57

Table A1: Convergent and Discriminant Validity – Local Adaptation

Local Adaptation Component

Local Adaptation Item Product Price Place Promotion Positioning

Product brand name .77 .17 .21 .25 .21

Product design .84 .24 .28 .24 .36

Product range .81 .35 .36 .38 .38

Product packaging .82 .26 .28 .25 .21

Retail price .32 .84 .43 .19 .23

Wholesale price .34 .85 .43 .10 .19

Customer credit .17 .80 .43 .21 .22

Price discounting .25 .86 .48 .23 .25

Sales force decisions .30 .39 .82 .40 .32

Channel decisions .34 .44 .84 .35 .40

Inventory management decisions .34 .40 .79 .21 .29

Physical distribution decisions .21 .47 .85 .37 .36

Advertising theme .39 .11 .30 .84 .45

Advertising copy .33 .14 .31 .89 .34

Media mix .28 .25 .41 .84 .36

Sales promotion .23 .23 .40 .84 .41

Market segmentation .31 .28 .41 .38 .91

Target segments .32 .21 .40 .41 .94

Product positioning .36 .23 .32 .44 .91

Page 59: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

58

Table A2: Convergent and Discriminant Validity – Local Innovation

Local Innovation Component

Local Innovation Item Product Price Place Promotion Positioning

Product brand name .72 .18 .23 .22 .22

Product design .88 .12 .10 .28 .24

Product range .81 .16 .14 .29 .29

Product packaging .79 .29 .29 .31 .31

Retail price .28 .89 .43 .21 .41

Wholesale price .17 .91 .47 .15 .47

Customer credit .16 .77 .60 .34 .36

Price discounting .20 .86 .52 .31 .38

Sales force decisions .23 .46 .79 .54 .51

Channel decisions .21 .54 .86 .45 .44

Inventory management decisions .16 .48 .84 .40 .35

Physical distribution decisions .18 .50 .86 .39 .40

Advertising theme .38 .12 .37 .88 .37

Advertising copy .38 .17 .40 .92 .34

Media mix .23 .38 .54 .87 .44

Sales promotion .24 .36 .54 .79 .48

Market segmentation .24 .41 .43 .40 .93

Target segments .27 .42 .45 .40 .95

Product positioning .40 .40 .44 .44 .85

Page 60: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

59

Table A3: Convergent and Discriminant Validity – Local Autonomy

Local Autonomy Component

Local Autonomy Decision Item

Pro

du

ct

Pri

ce

Pla

ce

Pro

moti

on

Posi

tion

ing

Poli

cy

Peo

ple

Product brand name .81 .29 .27 .32 .32 .29 .17

Product design .89 .34 .31 .37 .37 .31 .16

Product range .82 .42 .43 .43 .48 .39 .27

Product packaging .83 .41 .29 .39 .37 .33 .19

Retail price .42 .82 .52 .39 .36 .42 .41

Wholesale price .42 .87 .60 .32 .45 .47 .44

Customer credit decisions .27 .76 .59 .38 .36 .54 .56

Price discounting .37 .86 .59 .38 .37 .45 .47

Sales force .27 .53 .80 .52 .42 .60 .63

Channel .36 .57 .83 .59 .61 .65 .61

Inventory .34 .53 .81 .37 .39 .51 .53

Physical distribution .35 .58 .86 .43 .40 .54 .56

Advertising theme .47 .27 .39 .84 .55 .56 .45

Advertising copy .48 .31 .46 .92 .54 .61 .50

Media mix .40 .49 .61 .90 .60 .69 .59

Sales promotion .27 .48 .64 .82 .57 .64 .61

Market segmentation .41 .45 .54 .61 .94 .70 .50

Target segments .39 .41 .51 .54 .93 .65 .50

Product positioning .48 .39 .46 .62 .90 .65 .48

Marketing policy .36 .47 .60 .63 .65 .87 .63

Market research .35 .45 .55 .65 .68 .88 .58

Marketing budget .31 .48 .63 .55 .55 .88 .71

Marketing personnel selection .22 .55 .66 .53 .47 .71 .94

Marketing personnel training .23 .54 .64 .59 .50 .70 .94

Marketing personnel performance evaluation .19 .43 .60 .49 .46 .63 .90

Page 61: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

60

Table A4: Convergent and Discriminant Validity – Internal Networking

Internal Networking Component

Internal Networking Decision Item

Pro

du

ct

Pri

ce

Pla

ce

Pro

moti

on

Posi

tion

ing

Poli

cy

Peo

ple

Product brand name .80 .18 .22 .35 .42 .37 .25

Product design .90 .28 .27 .35 .51 .42 .26

Product range .89 .50 .48 .55 .58 .56 .44

Product packaging .83 .48 .49 .63 .56 .54 .51

Retail price .41 .90 .73 .65 .55 .62 .62

Wholesale price .43 .94 .81 .68 .62 .65 .69

Customer credit .36 .91 .80 .70 .57 .64 .72

Price discounting .34 .93 .82 .70 .62 .63 .66

Sales force .36 .83 .92 .69 .65 .66 .73

Channel .42 .84 .93 .75 .67 .70 .75

Inventory .35 .71 .89 .55 .61 .60 .64

Physical distribution .40 .77 .93 .64 .60 .60 .68

Advertising theme .56 .59 .59 .88 .70 .64 .56

Advertising copy .52 .66 .64 .95 .71 .70 .65

Media mix .49 .71 .69 .94 .67 .72 .71

Sales promotion .40 .74 .76 .87 .68 .66 .69

Market segmentation .54 .59 .66 .70 .96 .73 .67

Target segments .55 .62 .68 .72 .98 .74 .68

Product positioning .60 .59 .64 .71 .95 .73 .63

Marketing policy .48 .56 .57 .58 .68 .88 .63

Market research .53 .58 .58 .69 .69 .89 .71

Marketing budget .42 .65 .66 .67 .65 .90 .73

Marketing personnel selection .38 .68 .69 .68 .64 .76 .95

Marketing personnel training .39 .69 .72 .66 .64 .75 .94

Marketing personnel performance evaluation .37 .65 .68 .63 .63 .68 .93

Page 62: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

61

Table A5: Convergent and Discriminant Validity Analysis of Case Archetypes (Local

Adaptation)

Adaptation

archetype N

Mean weight of cases in each

archetype Total

weight

SD for

max

weight

95% LCL

for max

weight

Is LCL above

mean weight

of all other

archetypes? A B C D E F

A 10 .71 .07 .08 .09 .03 .01 1.00 .21 0.56 Yes

B 7 .09 .72 .08 .02 .05 .04 1.00 .20 0.54 Yes

C 6 .08 .08 .69 .03 .09 .02 1.00 .17 0.52 Yes

D 6 .12 .05 .03 .65 .11 .03 1.00 .18 0.45 Yes

E 25 .09 .03 .07 .06 .64 .11 1.00 .12 0.59 Yes

F 48 .05 .05 .05 .04 .07 .75 1.00 .16 0.70 Yes

TOTAL 102

(Due to rounding, total weight varies slightly from the simple sum of weights)

(SD – standard deviation, LCL – lower confidence limit)

Page 63: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

62

Table A6: Convergent and Discriminant Validity Analysis of Case Archetypes (Local

Innovation)

Adaptation

archetype N

Mean weight of cases in each

archetype Total

weight

SD for

max

weight

95% LCL

for max

weight

Is LCL above

mean weight

of all other

archetypes? A B C D E F

A 23 .65 .07 .05 .05 .06 .12 1.00 .16 0.58 Yes

B 12 .13 .65 .02 .05 .04 .11 1.00 .14 0.56 Yes

C 14 .13 .02 .63 .04 .02 .15 1.00 .14 0.55 Yes

D 5 .01 .04 .14 .70 .04 .08 1.00 .20 0.45 Yes

E 8 .04 .11 .06 .02 .67 .09 1.00 .20 0.51 Yes

F 49 .12 .06 .08 .03 .05 .66 1.00 .14 0.62 Yes

TOTAL 111

(Due to rounding, total weight varies slightly from the simple sum of weights)

(SD – standard deviation, LCL – lower confidence limit)

Page 64: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

63

Table A7: Convergent and Discriminant Validity Analysis of Case Archetypes (Local

Autonomy)

Autonomy

archetype N

Mean weight of cases in

each archetype Total

weight

SD for

max

weight

95% LCL

for max

weight

Is LCL above

mean weight

of all other

archetypes? A B C D E

A 24 .69 .07 .09 .03 .12 1.00 .16 0.62 Yes

B 2 .00 .77 .01 .22 .00 1.00 .33 -2.19 No

C 8 .08 .03 .73 .05 .11 1.00 .12 0.62 Yes

D 48 .12 .05 .07 .67 .10 1.00 .12 0.63 Yes

E 62 .08 .05 .05 .10 .72 1.00 .15 0.68 Yes

TOTAL 144

(Due to rounding, total weight varies slightly from the simple sum of weights)

(SD – standard deviation, LCL – lower confidence limit)

Page 65: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

64

Table A8: Convergent and Discriminant Validity Analysis of Case Archetypes

(Internal Networking)

Networking

archetype N

Mean weight of cases in

each archetype Total

weight

SD for

max

weight

95% LCL

for max

weight

Is LCL above

mean weight

of all other

archetypes? A B C D E

A 38 .74 .02 .09 .03 .12 1.00 .17 0.68 Yes

B 13 .10 .64 .06 .12 .08 1.00 .10 0.58 Yes

C 30 .15 .05 .62 .07 .11 1.00 .10 0.58 Yes

D 18 .05 .09 .12 .63 .11 1.00 .12 0.57 Yes

E 57 .07 .05 .08 .06 .74 1.00 .17 0.69 Yes

TOTAL 156

(Due to rounding, total weight varies slightly from the simple sum of weights)

(SD – standard deviation, LCL – lower confidence limit)

Page 66: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of

65

Table A9: Comparing Cluster and Archetype Profiles

Profiles on adaptation components Correlation

CA/AA

profiles Cluster Archetype Price Product Positioning Place Promotion

1

-1.58 -0.78 -1.34 -2.08 -1.72 0.55

A -2.48 -1.36 -0.41 -2.15 -2.63

2

0.16 -0.91 -1.30 -0.59 -0.72 0.68

B 0.30 -1.46 -2.28 -2.14 0.23

3

-1.73 -0.52 0.17 -0.22 0.31 -0.84

C 0.84 -0.38 -2.35 0.19 -2.81

4

0.07 0.75 -0.28 -0.15 -0.19 0.26

D -2.11 1.04 -1.00 0.58 1.09

5

0.50 -0.49 0.44 0.49 0.26 0.95

E 0.84 -1.50 1.14 0.97 1.09

6

0.62 1.39 0.97 0.74 0.83 0.98

F 0.84 1.88 1.14 0.84 1.09

Cluster and archetype profiles generated from k-means and robust archetypal analyses of 198 MNC subsidiaries.

Table A10: Cross Tabulation of CA versus AA

Adaptation archetypes

(Scores of 0.5 or more)

Mixtures

Total

cases Clusters A B C D E F

1 8 1 1 0 0 0 5 15

2 0 6 5 0 0 0 16 27

3 0 0 0 1 0 13 23 37

4 2 0 0 5 0 0 14 21

5 0 0 0 0 25 0 37 62

6 0 0 0 0 0 35 1 36

Total 10 7 6 6 25 48 96 198

Page 67: Marketing Management in MNC Subsidiairies: An …flora.insead.edu/fichiersti_wp/inseadwp2012/2012-72.pdf · Marketing Management in MNC Subsidiairies: ... An Archetypal Analysis of