Independent Empirical Support for Porter's Generic Marketing Strategies ? A Re-analysis using correspondence analysis

Abstract

Many published studies have sought to identify distinct strategy approaches with the objective of assessing whether certain strategies yield superior performance. Empirically derived strategy clusters are sometimes contrasted to theoretically derived strategy schemas or typologies as a point of reference, for comparison and contrast, or to explain associations with dependent variables such as performance. In some cases this theory dependence of observation can be misguided if the typology used lacks validity or incorporates flawed assumptions. This paper re-analyses a published work where empirically derived strategy clusters were identified using the multivariate mapping technique of correspondence analysis. The analysis provides further insights into the relationships between the variables under study by allowing the distance between variables to be seen (visually). In this case, the technique shows how close or distant various business strategies are to one another. This is of interest because if quite similar strategies yield dissimilar performance levels, the implications are that either minor differences in strategy are extremely important; or unobserved factors are influencing the results. Conversely, if superior performance is associated with markedly different strategy, an implication for managers is to take very different approaches to strategy. The paper concludes that the use of a well known generic strategy typology (Porter's (1980) generic competitive strategies) was of little use in interpretation of the clusters that were identified. Further, it suggests that Porter's (1980) generic competitive strategy schema does not describe/fit empirical reality, and provides no support for the notion that these generic strategies are routes to superior profit.

Full text

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 36
Independent Empirical Support for
Porter’s Generic Marketing Strategies ?
A Re-analysis using correspondence
analysis.
John Dawes and Byron Sharp
Marketing Science Centre
University of South Australia
North Terrace
Adelaide, Australia.
Email: John.Dawes@unisa.edu.au, Byron.Sharp@unisa.edu.au
Abstract
Many published studies have sought to identify distinct strategy approaches with the
objective of assessing whether certain strategies yield superior performance.
Empirically derived strategy clusters are sometimes contrasted to theoretically derived
strategy schemas or typologies as a point of reference, for comparison and contrast, or
to explain associations with dependent variables such as performance. In some cases
this theory dependence of observation can be misguided if the typology used lacks
validity or incorporates flawed assumptions. This paper re-analyses a published work
where empirically derived strategy clusters were identified using the multivariate
mapping technique of correspondence analysis. The analysis provides further insights
into the relationships between the variables under study by allowing the distance
between variables to be seen (visually). In this case, the technique shows how close or
distant various business strategies are to one another. This is of interest because if quite
similar strategies yield dissimilar performance levels, the implications are that either
minor differences in strategy are extremely important; or unobserved factors are
influencing the results. Conversely, if superior performance is associated with
markedly different strategy, an implication for managers is to take very different
approaches to strategy.
The paper concludes that the use of a well known generic strategy typology (Porter’s
(1980) generic competitive strategies) was of little use in interpretation of the clusters
that were identified. Further, it suggests that Porter’s (1980) generic competitive
strategy schema does not describe/fit empirical reality, and provides no support for the
notion that these generic strategies are routes to superior profit.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 37
Introduction
In recent years several authors have undertaken empirical studies of competitive
strategy in an effort to expand our knowledge of the links between strategy and
economic performance. Some authors have approached this from an Industrial
Organisation economics viewpoint (for theoretical tenets see Caves & Porter (1977))
and have focused on a single industry, for example, Cool & Schendel (1987) and
Hatten & Schendel (1977). Such research has advanced the notion of "strategic
groups", groups of firms within a single industry which display similar conduct along
key strategic dimensions, such as scope and resource commitments. Authors such as
Douglas & Rhee (1989) have examined businesses across industries using a still
relatively restricted range of theoretically derived variables such as marketing tactics,
market scope, and business synergy and identified 'clusters' of firms with broadly
differing strategies. Other approaches have endeavoured to identify or validate a priori
strategy frameworks such as those of Porter (1980), examples being Dess & Davis
(1984) and Miller & Freisen (1986).
Other authors have taken a broader view, preferring to utilise a wide range of
strategy elements in measuring the broad strategies of firms in diverse operating
environments, for example Wong & Saunders (1993). Such endeavours are clearly
more empiricist, with their measurement of a larger number of strategy variables and
reliance upon cluster analysis rather than grouping firms according to any theoretically
based ideal/extreme types. However, the choice of which strategy variables to measure
have inevitably been theory driven or at least vaguely influenced by theory; even PIMS
which collects a vast array of data is based on an Industrial Organisation/Business
Policy industry structure - business action model. This paper analyses an important
study of this type, that of Hooley, Lynch, and Jobber (1992). These authors gathered
responses from 616 single business companies on five key marketing strategy
variables, taken from O'Shaughnessy (1984). These were:
Marketing Objectives: defensive, hold or prevent decline
Strategic Focus: expand market, win share, or focus on internal productivity.
Market Targeting: whole market, selected segments, or individual customers
Quality Positioning: quality higher, the same, or lower as competitors

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 38
Price Positioning: above, the same, or lower than competitors
Using Ward's hierarchical method of cluster analysis, Hooley et al identified five
Generic Marketing Strategy (GMS) clusters. In addition to this, the type of market the
firm operated within was examined. Variables relating to the newness or maturity of
market, fluidity of competitive structure, and speed of change in customer needs were
measured across clusters but not included in the cluster analysis. Performance was also
measured, to be analysed later as a dependent variable, in terms of sales, market share
and profits (relative to competitors and improvement over the last financial year).
Hooley et al presented the survey results in a series of tables detailing percentages of
firms corresponding to the strategy or market description across each cluster. The
tables of percentages are shown below.
Table 1
Variable Measured GMS 1 GMS 2 GMS 3 GMS 4 GMS 5
% % % % %
Marketing Objectives
Defend 1 15 13 4 89
Steady Growth 17 68 68 92 4
Aggressive Growth 82 17 19 4 7
Strategic Focus
Expand Market 39 40 25 48 17
Win Share 54 47 61 41 12
Cost reduction/productivity 6 13 14 11 71
Marketing Targeting
Whole Market 48 18 5 0 20
Selected Segments 27 59 61 67 21
Individual Customers 24 22 34 32 51
Competitive Positioning
a) Quality relative to competitors
Higher 69 79 1 100 44
The same 29 18 98 0 55
Lower 3 3 2 0 3
b) Price relative to competitors
Higher 15 88 5 0 5
The same 62 0 89 100 79
Lower 23 13 7 0 17
Market Growth
New and growing 60 40 37 56 30
Mature and stable 28 46 49 33 43
Competitive Structure

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 39
Variable Measured GMS 1 GMS 2 GMS 3 GMS 4 GMS 5
% % % % %
Fluid 35 25 23 29 18
Speed of change in customer needs
Rapid 39 30 23 35 31
Approach to new product
development
Imitate competitors 16 20 34 18 24
Lead the market 69 67 44 59 39
Role of marketing in strategic
planning
None 4 6 9 6 21
Major 50 44 34 37 25
Approach to competition
Ignores it 14 12 6 9 10
Takes on any 73 65 60 56 53
Avoid it 14 23 35 35 37
Approach to taking risks
Moderate risks 53 59 71 68 59
Performance improvement over last
financial year
Better sales 75 62 65 65 47
Better market share 57 44 38 36 20
Performance relative to major
competitors
Better profit 40 39 29 19 26
Better sales 51 39 25 22 14
Better market share 47 31 24 22 15
Further Analysis
Large tables of frequencies such as the above are always difficult to interpret. The
aim is to determine which strategy variables distinguish between the GMS clusters.
Typically this interpretation is arrived at by looking at a number of strategy variables
(rows) and comparing the clusters percentage scores across the columns. This
"eyeballing" approach to make meaning of the tables is quite normal but places
considerable demands upon the researcher and later readers (Sharp, 1995). It also has
the deficiency in that some clusters score relatively highly (lowly) on all or many
strategy variables. Looking across a row it may be seen that a cluster achieves a greater
(lesser) score than the other clusters on one particular variable but this is not to say that

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 40
this variable particularly distinguishes that cluster from the others, because that cluster
typically scores higher (lower) than the other clusters on all/most variables. A means of
avoiding this problem will be discussed later in the section on correspondence analysis.
Hooley et al's descriptions, (based on a simple eyeballing approach) of the firms
who made up each cluster (generic marketing strategy) were along the following lines:
GMS 1: aggressive growth goals, often through market share gain or total market
expansion. Aim at the whole market ...marketing of high quality products at similar
prices.
GMS 2: steady sales growth either through market share gain or market expansion.
Selected segments are targeted through higher quality products at higher prices than
competitors.
GMS 3: steady sales growth pursued ... by focusing on selected segments or
individual customers. Positioning is average quality at average prices.
GMS 4: steady growth goals with a focus on total market expansion or winning
share by targeting selected segments or individuals. High quality positioning at same
prices.
GMS 5: defensive strategy achieved through a focus on cost reduction or
productivity improvement. Very selective targeting with similar or higher quality at
similar prices.
In addition, Hooley et al endeavoured to categorise the strategy clusters in relation to
Porter’s (1980) generic strategies. Their comments on each respective strategy cluster
were as follows:
GMS 1 ...clearly this is a differentiation strategy (Porter, 1985)
GMS 2 ...This strategy resembles the focused differentiation strategy of Porter
(1985)
GMS 3 ...this most closely resembles “stuck in the middle”
GMS 4 ...again resembles the focused differentiation strategy of Porter

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 41
GMS 5 ...This strategy is similar to that of focused cost leadership
Porter’s strategy of overall cost leadership and broad market focus does not appear
to be represented but this would be expected. Porter said there can (or should be) only
one such firm in an industry or even none. In a five cluster solution, even one based on
many industries it is not unreasonable to expect that such firms, if any even existed,
might be subsumed into another cluster - most likely GMS 5.
The Philosophy of Science literature highlights the tendency of scientists to explain
or interpret phenomena using some prior theory about what sort of things the world
contains (Chalmers, 1976, Doyal and Harris, 1986). This theory dependence of
observation affects not only scientists’ choice of which things to measure but also their
interpretation of that data once collected. In this case, Hooley et al have suggested that
the clusters they identified resemble theoretical types suggested to be routes to
competitive advantage. However, is this portrayal using such hypothetical types valid
or even useful ? A small body of revisionist literature has emerged which casts doubt
on the validity of Porter’s scheme (Hendry, 1990, Sharp, 1991, Speed, 1989). One of
the criticisms mounted has been that the types are not delineated by a common
dimension (Sharp and Dawes, 1996), a prerequisite for a valid classification scheme
(Hempel, 1965). If such criticism is valid it would not be expected that endeavours to
match observed strategy clusters to Porter’s types would be possible because the
variables do not reflect parameters which truly distinguish business or marketing
strategies.
Hooley’s interpretation of the GMS clusters using the Porter dimensions can be
illustrated graphically. Such a graphical illustration utilises two of the important
dimensions upon which Porter based his strategy scheme. The first of these is the
breadth of the product market served; Porter wrote that “the {focus} strategy rests on
the premise that the firm is thus able to serve its narrow strategic target more effectively
or efficiently than competitors“. The second is the extent to which the firm either
differentiates in order to reduce monetary price sensitivity (a differentiation strategy) or
relies on achieving low costs of operations (a low cost strategy), the outcome of which
has been widely interpreted as offering low prices to customers (Sharp, 1991). While
Porter was not consistent in explaining whether low cost meant low price, the example
he provided in Porter (1980), of a crane manufacturer (Harnischfeger) was of a firm
offering a low monetary price offering. On this basis the following graph shows the
position of Hooley et al’s clusters according to the descriptions in that work.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 42
Chart 1
Breadth of
Market
Narrow
Broad
High “Differentiation/Value”
Low “Differentiation” / Low Cost
GMS 1
GMS 2
GMS 3
GMS 4
GMS 5
Differentiation
Focus - Differentiation
Stuck-in-the-Middle
Focus - Low Cost
Low Cost
The preceding illustration shows GMS 1 with a broad market scope and high degree
of differentiation. GMS 2 and 4 are situated quite close together, as both are “focused
differentiation” perhaps providing a theoretical justification of a smaller cluster solution
than five as utilised by Hooley et al. GMS 3 has a narrow market scope but is
described as being “stuck in the middle” with neither clear differentiation or low
costs/prices. Lastly, GMS 5, the “focused cost leadership” strategy is at the lower end
of the vertical axis.
This paper suggests that this interpretation of the GMS clusters according to the
Porter typology is misguided and that the relation between the generic strategies Hooley
et al revealed bear little resemblance to the explication provided. This assertion is
based, in part, upon a reanalysis using the technique of correspondence analysis which
provides a useful way of showing visually the spatial relationship between the GMS
clusters.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 43
Correspondence Analysis
Correspondence analysis provides a means of analysing tables of categorical data in
order to determine the relationships between the variables of interest and has the
advantage of making it considerably easier to see the relationship between a large
number of variables.
It is a relatively recently developed multivariate statistical technique introduced by
French statisticians (Benzecri, 1969) with later notable work being done by Greenacre
(1984), a South African statistician, and American marketing scientists Carroll, Green
and Schaffer (1987, 1986). Correspondence analysis can produce a two-dimensional
display of complex multi-variate non-metric data. It has a number of advantages over
more traditional techniques utilising principal components analysis and discriminant
analysis, particularly since these techniques were developed to deal with metric, rather
than categorical, data.
One of the major benefits of correspondence analysis over discriminant analysis is
that it can show the relationship between all variables in the analysis. Discriminant
analysis captures the relationship between independent variables and a dependent
variable but not the relationship between the independent variables. Principal
components analysis has an important assumption that the data is metric and normally
distributed, and it can not be used to display the relationship between dependent and
independent variables simultaneously. In summary, correspondence analysis provides
a multivariate representation of interdependence for non metric data which is not
possible with other methods (Bendixen, 1996, Hair, et al., 1995).
Particularly relevant to this paper is the way that correspondence analysis, due to its
multivariate nature, can reveal relationships that would not be detected in a series of
pairwise comparisons (Hoffman and Franke, 1986). As part of a series of tests for
validity, Hooley et al utilised significance tests across the rows of the data matrices with
the strength of differences between clusters being determined from visual analysis,
again across rows. Each cluster was compared to others on a particular attribute, for
example, emphasis on segments or the whole market. In comparison, correspondence
analysis simultaneously captures the relationship between, to continue the example,
emphasis on segments and all other strategy attributes and their impact on
distinguishing the clusters. As a consequence the relative differences between the
clusters, as determined by the attributes, can be visually appraised in terms of distances
between the cluster names on the two dimensional correspondence map. Likewise the
distance between attributes indicates their relationship to one another in determining the

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 44
difference between the clusters. These distances are often referred to as chi-square
distances because correspondence analysis relates the frequencies for any row/column
combination to all other row/column combinations based on marginal frequencies, a
procedure which yields a conditional expectation very similar to an expected chi-square
value (Hair, et al., 1995).
In order to use correspondence analysis on the Hooley et al data each of the
percentages in the tables of frequency was converted back into actual frequencies.
Analysis was undertaken with CGS plots (see Carrol, Green and Schaffer (1987, 1986)
which allows interpoint distances to be read directly in order to infer similarity or
dissimilarity between the clusters. The appropriateness of this approach was checked
via comparison with the actual data table and with the traditional French plot (Herman,
1991). The results are presented here as CGS plots with varimax rotation. All analyses
did a reasonable, though not outstanding, job at capturing the data in a two dimensions
(measure of fit being 0.67). Variance accounted for was X axis = 0.42 and Y axis
=0.25.
Chart 2 incorporates the variables concerning marketing strategy, with a legend
describing the figures on the map.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 45
Chart 2
1
6
GMS 5
22
9
26
20
GMS 314
11
17
27
19
GMS 4
8
5
2
3
25
15
12
7
24
10
GMS 2
13
23 4
GMS 1
(n.b. the oversize "1" in top left corner signifies its true position lies further beyond the map.
GMS 1 Generic Marketing Strategy 1 (note: read exact point at leftmost point of "G").
GMS 2 Generic Marketing Strategy 2
GMS 3 Generic Marketing Strategy 3
GMS 4 Generic Marketing Strategy 4
GMS 5 Generic Marketing Strategy 5
1 Defend
2 Steady Growth
3 Aggressive Growth
4 Expand Market
5 Win Share
6 Cost Reduction
7 Whole Market
8 Selected segments
9 Individual Customers
10 High Quality
11 Same Quality
12 Lower Quality
13 High Price
14 Same Price
15 Lower Price
16 New Growing Mkts
17 Mature Markets
18 Fluid competitive structure
19 Rapid change
20 Imitate competitors
21 Lead the market
22 No Marketing role in strat. planning
23 Major Marketing role
24 Ignore Competition
25 Take on any
26 Avoid competition
27 Moderate Risks

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 46
Chart 3 shows the Generic Marketing Strategy perceptual positions with key
discriminating variables.
Chart 3
GMS 5
GMS 3
GMS 4
GMS 2
GMS 1
Defensive, internal focus
(focus on stemming decline in growth ?)
High price & high
quality
Same (or lower ?)
quality & price;
imitate
steady growth;
selected segments
aggressive growth
Interpretation
The correspondence maps clearly show the closeness between GMS 1, high value
positioning and GMS 4, selective targeting with high quality/same prices. This pair of
clusters are situated somewhere in the middle between GMS 3 same quality same price
and GMS 2 selective targeting, premium positioning. All these strategies are distant
from the defensive, internal orientation of GMS 5.
Chart 2 shows that GMS 1 and 4 are quite similar strategies, although GMS 1
exhibits higher rates of success. This difference is possibly attributable to GMS 4
being positioned closer to “winning share” (see (5), chart 2), possibly difficult against

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 47
entrenched competitors. This contrasts with GMS 1 which appears more associated
with “expand the market” (4). The chart clarifies Hooley et al’s dual observations that
GMS 1 is oriented to new markets but that both GMS 1 and 2 emphasise growth
markets.
This raises the issue that growth as a success measure will bias results to growth
oriented firms. Researchers such as Douglas & Rhee (1989) have identified deliberate
“niche”, or small share strategies, so perhaps “growth” should not necessarily indicate
“success”. The use of a focus on growth as an independent variable when growth itself
is used as a dependent variable is an issue worthy of further debate and investigation.
Re-interpreting The relationships To The Porter Types
Bearing in mind the less than complete fitting of the data to the two dimensions, the
correspondence maps are still very different to the graph constructed from Hooley et
al’s interpretation using Porter (chart 1). Hooley et al’s description suggested that
GMS 1 (the “broad differentiation” strategy) would be quite distinct from GMS 2,3,
and 4. The correspondence map shows that GMS 1 is not as distinct as originally
suggested, and contrary to Hooley et al’s explication, GMS 2 and 4 (both supposed to
be “focus-differentiation” strategies) are somewhat less similar than was originally
suggested. The GMS 5 cluster (supposedly “focus-cost leadership”), rather than being
situated on a roughly parallel axis to GMS 2,3, and 4 as suggested by the imagined
“high differentiation/low cost” continuum, is orthogonal (at least in the two dimensions
used) to the other four generic strategy clusters; thus showing no correlation (Bendixen,
196) rather than the expected negative correlation. Indeed, rather than being a route to
competitive advantage, as would be suggested by Hooley et al’s suggestion that GMS 5
resembled a focused cost leadership strategy, GMS 5 rates poorly on most of the
performance variables used in the study (yet is not “stuck in the middle” as Porter’s
framework would suggest poor performers should be). Lastly, GMS 3 is shown to be
not “stuck in the middle” between any high differentiation and low cost/price endpoints,
but rather at the end of a high price/high quality and same price/same quality axis.
Further evidence of the lack of usefulness of the Porter types in interpreting the
results concerns the clusters’ value positioning. Hooley et al labelled the GMS 1 cluster
"high value positioners" meaning that these firms produce high quality goods but are
less likely to charge high prices, and may even charge lower prices - thus offering high
value to customers. Chart 3 shows this more clearly. GMS 1 and GMS 4 are
approximately in the "middle" between the high price/high quality strategy of GMS 2

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 48
and same quality/same price strategy of GMS 3. While the tables show GMS 4
displaying a higher incidence of superior quality than GMS 1, the number of GMS 1
firms who use higher prices pushes this strategy cluster toward the high quality/high
price endpoints (10;13 chart 2). This shows that the label of high value positioner
might be more appropriate for GMS 4 than for GMS 1. It also shows the interpretation
of GMS 1 and 4 as instances of Porter’s “differentiation” and “focused differentiation”
strategies respectively, confounded the analysis, as these strategies are meant to result
in superior profitability through higher prices (Porter, 1980, 1985). Both are shown to
include similar or even lower pricing.
A methodological point of interest highlighted by the correspondence maps is the
bipolar "quality" positions (note the question mark on chart 3 adjacent to GMS 3). It
can be seen that the spectrum ranges from "high quality" to "same quality". This might
be expected to instead be from "high" to "low". The frequency tables show very few
firms reporting they manufacture goods of a lower quality than competitors - reflecting
a problem with topic bias. Such bias is also evident in a similar study by Wong &
Saunders (1993) in which over 70% of the sample of 90 firms stated they made goods
of "superior quality" relative to competition. It suggests in some cases "superior" can
be interpreted as "parity" because by definition, the incidence of "superior" should be
less than that of "average" or "inferior".
This issue has implications for other empirical research utilising product quality as a
primary inicator of differentiation. Examples are Miller (1986) and Miller & Dess
(1993) who have operationalised differentiation primarily in terms of managerial
perceptions of relative product quality. Such an operationalisation is problematic for
several reasons. Firstly, this attribute, being a managerial rather than customer
perception, and only one of numerous factors which may or may not distinguish
products in the eyes of customers, is likely to innacurately estimate the real degree of
product heterogeneity. Secondly, if managers over-report their degree of product
quality, and it appears that this does occur on occasion, this too will result in an
inaccurate depiction of the state, or form, of product heterogeneity exhibited. Lastly, as
has been argued, these measures are not measuring the extent of differentiation but
merely product heterogeneity which is an incomplete measure.
To summarise the re-interpretation, it has been shown that the use of Porter’s
strategy types was of little use in the attempt by Hooley et al to explain the phenomena
they identified, using what has been a generally accepted, though recently criticised,
theoretical model of generic strategy. The respective proximity of the GMS clusters

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 49
bears no relationship to that suggested by superimposing Porter’s strategies on the
empirical data.
Further Discussion
This article has shown that the interpretation of observed phenomena by Hooley et al
using accepted “theory” resulted in a description which poorly captured the real
proximity of the strategy clusters to each other and their relationship to certain strategy
variables. This reanalysis has been highly supportive of Hooley et al’s (1992) overall
findings but not of the interpretation of the empirical data as being supportive of the
existence of Porter’s generic competitive strategies. In fairness to Hooley et al they
were equivocal on this point, which in itself is good justification for further analysis
and an “outside opinion” as this paper has attempted to provide.
Finally, the paper has also illustrated how correspondence analysis can be a useful
tool for marketing strategy research. Not only does it allow the presentation of large
sets of categorical data in a format which makes interpretation substantially easier but it
also due to its multivariate nature allows further insights to be drawn from the data. It
is shown to be particularly useful in interpreting empirical clusters. In this case
correspondence analysis allowed the direct comparison of the relative differences
between generic marketing strategies in terms of the variables which defined them, and
associated variables relating to market characteristics.
Further Replication Research
The work of Hooley et al identified strategy clusters which appear to explain
performance differences across firms. However, the possible problem remains, despite
precautions taken in the statistical procedures, whether such clusters do actually reflect
reality, or whether they are statistical artefacts. Additionally, empirical identification of
generic strategy types should be extended to different countries, to determine if these
broad approaches to marketing exist widely or reflect unique country characteristics or
conditions. Only replication and extension research can answer the question of just
how generic these marketing strategies really are.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 50
Cluster Validation
Hooley et al carried out useful tests for internal validity utilising chi-square and
Cramers V to test the clarity of the clusters. They also used discriminant analysis to
predict membership of one half of the sample from another half which added confidence
to the cluster solution. However, other researchers have recently (e.g. Lockshin and
Spawton, 1995) advocated additional tests for the external validity of clusters. For
example, comparing the differences of related variables (variables not included in the
cluster analysis) across clusters on the assumption that if the clusters did identify
significant differences in the variables under study, this would be reflected in at least
some differences in other related, or dependent, variables. Future work in identifying
generic strategy clusters should utilise such external validity tests. Since the whole idea
of empirical strategy research of this type is to identify strategy similarities across
industry and environment, factors are required for validation which could be expected
to alter according to generic strategy but not be sensitive to industry. Hooley et al
identified clusters created from five strategy variables (objectives, strategic focus,
targeting, quality, price). These clusters could be validated using, for example,
managerial attitudes to growth, segmentation skills, and extent of distribution (the latter
which may reflect selective targeting). Such an approach could also involve measuring
the extent of agreement to summary descriptions of the strategy clusters identified, and
sampling the same firms at a later time.
On a more specific note, one final area suggested to be fruitful for further research
concerns the GMS 5 "defender" cluster. It has been mentioned this group had a
reasonably high incidence of better profit despite poor sales and market share results.
This could be due to successful "denominator management", reducing infrastructure
and other expenditure to maintain acceptable ratios. It would be beneficial to revisit
such firms at a later date to determine if this comparatively healthy profit position vis a
vis marketplace performance has been sustained, or whether it has deteriorated further
as the long term effects of reducing expenditure are felt. Which highlights the need for
longitudinal work in strategy research.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 51
References
Bendixen, Mike (196) A Practical Guide to the Use of Correspondence Analysis in
Marketing Research, Marketing Research On-Line, Vol.1, no.1, p.16-36.
Benzecri, J. P. (1969) Statistical Analysis as a Tool to Make Patterns Emerge from
Data, in "Methodologies of Pettern Recognition", Academic Press Inc.: New York,
p.35-74.
Carroll, J. Douglas; Green, Paul E.; Schaffer, Catherine M. (1987) Comparing
Interpoint Distances in Correspondence Analysis: A Clarification, Journal of Marketing
Research, Vol.24, no.November, p.445-450.
Carroll, J. Douglas; Green, Paul E.; Schaffer, Catherine M. (1986) Interpoint
Distance Comparisons in Correspondence Analysis, Journal of Marketing Research,
Vol.23, no.August, p.271-280.
Caves, R.E.; Porter, M. (1977) From Entry Barriers to Mobility Barriers:
Conjectural Decisions and Contrived Deterrence to New Competition, Quarterly Journal
of Economics, Vol.91, p.241-330.
Chalmers, Alan Francis (1976)What is this thing called science? 1st ed. Queensland:
University of Queensland Press.
Cool, Karel O.; Schendel, Dan (1987) Strategic Group Formation and Performance:
The Case of the U.S. Pharmaceutical Industry, 1963-1982, Management Science,
Vol.33, no.9, September, p.1102-1124.
Dess, Gregory; Davis, Peter S. (1984) Porter's (1980) Generic Strategies as
Determinants of Strategic Group Membership and Organizational Performance,
Academy of Management Journal, Vol.27, no.3, p.467-488.
Douglas, Susan P.; Rhee, Dong Kee (1989) Examining Generic Competitive
Strategy Types in U.S. and European Markets, Journal of International Business
Studies, no.Fall, p.437-463.
Doyal, Len; Harris, Roger (1986)Empiricism, Explanation and Rationality. London:
Routledge.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 52
Greenacre, Michael J. (1984)Theory and Applications of Correspondence Analysis.
London: Academic Press Inc.
Hair, Joseph F.; Anderson, Rolph E.; Tatham, Ronald L.; Black, William C.
(1995)Multivariate Data Analysis. Fourth ed. New Jersey: Prentice Hall International.
Hatten, Kenneth J.; Schendel, Dan E. (1977) Heterogeneity Within an Industry:
Firm Conduct in the U.S. Brewing Industry 1952-71, The Journal of Industrial
Economics, Vol.26, no.December, p.97-113.
Hempel, Carl G. (1965) Fundamentals of Taxonomy, in "Aspects of Scientific
Explanation - And Other Essays in the Philosophy of Science", The Free Press: New
York, p.505.
Hendry, John (1990) The problem with Porter's generic strategies, European
Management Journal, Vol.8, p.443-50.
Herman, Steve (1991)MCA+. New Jersey: Bretton-Clark.
Hoffman, Donna L.; Franke, George R. (1986) Correspondence Analysis:
Graphical Representation of Categorical Data in Marketing Research, Journal of
Marketing Research, Vol.23, no.August, p.213-227.
Hooley, Graham J.; Lynch, James E.; Jobber, David (1992) Generic Marketing
Strategies, International Journal of Research in Marketing, Vol.9, p.75-89.
Lockshin, Lawrence; Spawton, Anthony (1995) Using Product, Brand and
Purchasing Involvement for Retail Segmentation, Second International Conference on
Recent Advances in Retail & Services Science, Broadbeach, Australia.
Miller, A; Dess, G.G. (1993) Assessing Porter's (1980) Model in Terms of Its
Generalizability, Accuracy and Simplicity, Journal of Management Studies, Vol.30,
no.4, p.553-585.
Miller, Danny; Friesen, Peter H. (1986) Porter's (1980) Generic Strategies and
Performance: An Empirical Examination with American Data. Part 1: Testing Porter,
Organization Studies, p.37-55.

Journal of Empirical Generalisations in Marketing Science, Volume One, 1996. Page 53
O'Shaughnessy, John (1984)Competitive marketing: a strategic approach.
Winchester, Mass: Allen & Unwin Inc.
Porter, Michael E. (1985)Competitive Advantage. New York: Free Press.
Porter, Michael E. (1980)Competitive strategy: techniques for analysing industries.
New York: Free Press.
Sharp, Byron (1995) Business Orientations and Corporate Success: A
Correspondence Analysis of Wong and Saunders' Findings, Journal of Strategic
Marketing, Vol.3, no.3, p.205-214.
Sharp, Byron; Dawes, John (1996) Is Differentiation Optional? A Critique of
Porter's Generic Strategy Typology, in "Management, Marketing and the Competitive
Process" (edited by Peter Earl), Edward Elgar: London.
Sharp, Byron M. (1991) Competitive Marketing Strategy: Porter Revisited,
Marketing Intelligence & Planning, Vol.9, no.1, p.4-10.
Speed, Richard J. (1989) Oh Mr Porter! A re-appraisal of competitive strategy,
Marketing Intelligence and Planning, Vol.7, p.8-11.
Wong, Veronica; Saunders, John (1993) Business Orientations and Corporate
Success, Journal of Strategic Marketing, Vol.1, p.20-40.