Organization Studies
http://oss.sagepub.com/
The Ecological Fallacy in National Culture Research
Paul Brewer and Sunil Venaik
Organization Studies 2014 35: 1063 originally published online 1 May 2014
DOI: 10.1177/0170840613517602
The online version of this article can be found at:
http://oss.sagepub.com/content/35/7/1063
Published by:
http://www.sagepublications.com
On behalf of:
European Group for Organizational Studies
Additional services and information for Organization Studies can be found at:
Email Alerts: http://oss.sagepub.com/cgi/alerts
Subscriptions: http://oss.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
>> Version of Record - Jun 20, 2014
OnlineFirst Version of Record - May 1, 2014
What is This?
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
517602
OSS0010.1177/0170840613517602Organization StudiesBrewer and Venaik
research-article2014
Article
The Ecological Fallacy in National
Culture Research
Organization Studies
2014, Vol. 35(7) 1063–1086
© The Author(s) 2014
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0170840613517602
www.egosnet.org/os
Paul Brewer
University of Queensland, Australia
Sunil Venaik
University of Queensland, Australia
Abstract
This article challenges the understanding and use of the Hofstede and GLOBE national culture models
in much extant culture-related theory development. Both the Hofstede and GLOBE culture dimensions
are derived from individual-level survey data aggregated to, and analysed at, the national level. But their
culture scales that are correlated at the national (ecological) level are not correlated in the same manner
at the individual or organizational level. To presume they are is a form of ‘ecological fallacy’ that, despite
warnings, has often been overlooked by culture researchers. We analyse five research articles in top journals
in organizational behaviour, general management, international business, marketing and accounting and
show how the articles commit an ecological fallacy by projecting national-level culture characteristics onto
individuals or organizations. The implications of this ecological fallacy include the development of invalid
culture-related theory and the persistence of erroneous practitioner stereotyping. We provide the first
comprehensive explanation of the origins, effects and implications of the ecological fallacy in national culture
research and practice. A way forward for culture-related research is also suggested.
Keywords
ecological fallacy, GLOBE, Hofstede, national culture
Introduction
‘Hofstede’s dimensions of national culture were constructed at the national level. They were underpinned
by variables that correlated across nations, not across individuals or organizations. In fact, his
dimensions are meaningless as descriptors of individuals or as predictors of individual differences
because the variables that define them do not correlate meaningfully across individuals’.
(Minkov & Hofstede, 2011, p. 12, italics added)
Corresponding author:
Paul Brewer, UQ Business School, University of Queensland, Brisbane, QLD 4072, Australia.
Email: p.brewer@business.uq.edu.au
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1064
Organization Studies 35(7)
While many scholars have worked to develop theories of human culture, in the field of management research two culture models stand out: those of Hofstede (1980, 2001) and House, Hanges,
Javidan, Dorfman and Vipin (2004), known as the GLOBE project. Both models have had a profound impact on the development of culture theory (Peterson & Søndergaard, 2011), and are
described as ‘titans’ in their field (Leung, 2006, p. 881). It is claimed that these culture models
provide extremely useful country scores for common cultural dimensions which can be used to
quantitatively compare national cultures and examine the effects of national culture on organizations and their managers (Nelson & Gopalan, 2003) and they continue to be regarded as the leading
models of cross-cultural differences (Venaik & Brewer, 2010). There have also been criticisms of
the conceptualization of ‘national’ culture and of the Hofstede and GLOBE models specifically
over time. These include the difficulties of reducing the complex concept of culture to a series of
national numbers (Baskerville, 2003); probable western, male, managerial bias in survey design
(Ailon, 2008); the neglect of considerable intra-country cultural variation (McSweeney, 2009);
problems associated with measuring culture as average preferences (McSweeney, 2002); and the
failure of replication studies to support the models (Spector, Cooper & Sparks, 2001).
This article provides a detailed discussion of an additional, but more fundamental, flaw
relating to the manner in which the Hofstede and GLOBE national culture models are applied
in research and practice, which is quite different to the criticisms already explored in the literature. We show that although the Hofstede and GLOBE dimensions were developed conceptually and empirically only at the national level, they are often incorrectly applied and interpreted
at the individual and organizational levels. To assume that these national-level dimensions exist
at the levels of individuals or organizations in a society is invalid and represents a form of
‘ecological fallacy’ (hereafter mostly referred to as EF) (Hofstede, 2001, p. 16; House &
Hanges, 2004, p. 99). Our examination of key exemplar papers in top business journals that
incorporate national culture dimensions indicates that there is a lack of appreciation of this
fundamental issue.
The purpose of this article is to (1) present a comprehensive theoretical explanation of the origins of the ecological fallacy; (2) propose a typology of the different types of EF; (3) discuss the
role, pervasiveness and consequences of the EF in cross-cultural research and practical understanding; and (4) suggest a way forward to overcome the EF in culture-related research. Overall, this
article is the first to fully explain, rather than simply refer to, the ecological fallacy in national
culture research, to provide examples of its frequent occurrence in high-level research, and to provide clear implications of this for future research and practice in cross-cultural and other multilevel
research.
We begin with an explanation of the ecological fallacy phenomenon and show through a typology how it applies in differing analytical contexts. Next we provide a short overview of the
Hofstede and GLOBE culture models including the methods used to construct their national culture dimensions and scores. We explain how their methodologies result in national, not individualor organizational-level culture dimensions. We then review five articles in top business journals to
explain how the national culture characteristics are misattributed to individuals and firms via the
ecological fallacy. We explain why this problem is so widespread and its implications for culture
theory and practice. Finally, we suggest a way forward for future cross-cultural and other multilevel research.
The Ecological Fallacy Typology
Broadly, the ‘ecological fallacy’ is the error of assuming that statistical relationships at a group level
also hold for individuals in the group (King, Rosen & Tanner, 2004). It is the error of inferring
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1065
Brewer and Venaik
individual-level relationships from group aggregated data. Although the ecological fallacy has
been highlighted in the empirical literature for more than 70 years, the term has often been used
loosely and interchangeably to represent three different types of fallacies, which we label the structural ecological fallacy, the measurement ecological fallacy and the averaging fallacy. We analyse
and clarify the different types of EF in the following sections, and summarize the key differences
in Table 1.
Structural ecological fallacy
Thorndike (1939) is credited with one of the first clear expositions of the ecological fallacy. He
showed, in a constructed illustration covering 12 city districts, that the correlation between the
intelligence quotient of the individuals in these districts and the number of rooms per person (a
measure of crowdedness) for his example group was 0 within each of the districts, 0.45 for the
entire group of individuals and 0.9 for the variables’ averages over the 12 districts. Thorndike concluded that ‘if the correlation between two traits, A and B in n groups…has a certain value, K, the
correlation between A and B in the individuals or the families composing the groups need not be
K…It will usually be very much nearer zero’ (Thorndike, 1939, p. 122). Later, Robinson (1950) in
a famous study showed that correlations between two variables across different groupings of the
same people varied (he seems to be the first to use the term ‘ecological’ to describe group-level
correlations). These group correlations in turn differed from the correlation between the same variables calculated for the members of the total (all-groups) population (an individual-level correlation). Specifically, Robinson showed that in the United States the correlation between ‘percent
negro’ and ‘percent illiterate’ was 0.946 across the nine US Census Bureau geographic divisions,
0.773 across the 50 states, but only 0.203 for the United States population. He concluded, ‘There
need be no correspondence between the individual correlation and the ecological correlation’
(Robinson, 1950, p. 354). In sum, the structural EF tells us that the correlations between two variables will very likely be different at the aggregated level compared with lower levels of analysis
(individuals, families, states, geographic divisions, etc.).1
Recognition of the structural EF is important in many disciplines including epidemiology,
where, for example, a study showing a relationship across geographic areas (e.g. countries) between
two variables, say, smoking prevalence and life expectancy, may be helpful to researchers. But this
needs to be tested at the individual level before conclusions can be drawn on whether smoking
contributes to lower life expectancy. Robinson (1950) makes the point that in social science ecological correlations are generally employed for the purpose of discovering something about individuals and used on occasions when individual correlations are not available such as voting patterns
of citizens. King et al. (2004, p. 1) make the same point, ‘In these situations, where survey research
or other means of individual-level data collection are infeasible, ecological inference is the best
and often the only hope of making progress.’ Thus group-level relationships may be helpful, but
such relationships between ecological variables cannot be relied upon to exist at the individual
level; they most likely will be different (Greenland & Robins, 1994).
Measurement ecological fallacy
In the context of the Hofstede and GLOBE national culture dimensions, however, the EF problem
is of a slightly different nature. Unlike the original EF of varying structural relationships between
two variables at different levels of analysis in the Thorndike and Robinson examples, in the case of
the Hofstede and GLOBE national culture dimensions, the EF is about varying measurement relationships among the items used to measure the culture constructs at different levels of analysis.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1066
Organization Studies 35(7)
Although correlations underlie both measurement and structural models (Leung & Bond, 1989),
there is a critical difference between the two. In the classical EF examples of Thorndike and
Robinson, the interest is in the ecological versus individual-level relationship between two constructs, X and Y. However, in the case of the Hofstede and GLOBE national culture dimensions,
the EF is more fundamental as it relates to the theoretical constructs that emanate from a set of
measurement items. For example, if a set of items X1, X2, X3, X4 is used to measure a construct,
the EF question in cross-cultural research is whether the relationship among these items is isomorphic across different levels of analysis: individual, organizational and national. If the set of items
is not related in the same way across different levels of analysis, and a different set of items (say,
X5, X6, X7, X8) is needed to operationalize the namesake construct at another level of analysis,
then it is invalid to characterize the units at one level (e.g. individuals or organizations) with the
bundle of attributes used to measure the namesake construct at another level (e.g. the national culture dimensions). We provide an example to illustrate and clarify this point.
The ‘participative’ concept is found to exist both at the national level in the form of ‘national
participative systems’ and at the individual level in the form of ‘corporate participative leaders’.
However, the measures of the same concept (the degree of ‘participation’) are totally different at
these two levels. According to Freedom House (2012), the following four items are used to measure ‘national participative systems’:
1.
2.
3.
4.
Do the people have the right to organize in different political parties or other competitive
political groupings of their choice, and is the system open to the rise and fall of these competing parties or groupings?
Is there a significant opposition vote and a realistic possibility for the opposition to increase
its support or gain power through elections?
Are the people’s political choices free from domination by the military, foreign powers,
totalitarian parties, religious hierarchies, economic oligarchies, or any other powerful group?
Do cultural, ethnic, religious, or other minority groups have full political rights and electoral opportunities?
In contrast, according to the GLOBE study (see the GLOBE Phase 2 Alpha questionnaire and the
GLOBE syntax at http://business.nmsu.edu/programs-centers/globe/instruments/), ‘corporate participative leadership’ is measured with the following ten items representing two components, autocratic
leadership (six items, reversed) and non-participative leadership (four items, reversed) (codes in parentheses next to each item refer to the question numbers in the GLOBE phase 2 alpha questionnaire):
1.
2.
Autocratic leader (6 items)
a. Bossy – tells subordinates what to do in a commanding way (V2_4).
b. Autocratic – makes decisions in dictatorial way (V2_36).
c. Domineering – inclined to dominate others (V4_33).
d. Elitist – believes that a small number of people with similar backgrounds are superior
and should enjoy privileges (V4_37).
e. Ruler – is in charge and does not tolerate disagreement or questioning, gives orders
(V4_48).
f. Dictatorial – forces his/her values and opinions on others (V4_54).
Non-participative leader (4 items)
a. Non-delegator – unwilling or unable to relinquish control of projects or tasks (r4_44).
b. Micro-manager – an extremely close supervisor, one who insists on making all decisions (r4_43).
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1067
Brewer and Venaik
c.
d.
Non-egalitarian – believes that all individuals are not equal and only some should have
equal rights and privileges (r4_14).
Individually-oriented – concerned with and places high value on preserving individual
rather than group needs (r4_13).
As is clear from the above example, although the same term ‘participative’ is used to refer to
both national systems and corporate leaders, the underlying measures and characteristics are completely different at the two levels. It would therefore be erroneous to simply assume that nations
with ‘participative systems’ also have ‘participative corporate leaders’. It is this one-to-one mapping of the national culture characteristics onto individuals and firms in each country that is at the
core of the EF we discuss in this article.
Averaging ecological fallacy
Another understanding of the term ‘ecological fallacy’ has emerged in both academic manuscripts
and non-academic, common parlance which we call the ‘averaging EF’. In this interpretation, the
fallacy is committed when an average characteristic of a group is taken to necessarily apply to all
the members of that group. For example, if a particular group of people are measured to have a low
average IQ, it is an error to conclude that a randomly selected member of the group has a low IQ.
This averaging EF is also warned against in national culture research (e.g. Gerhart, 2009).
The averaging EF interpretation has, at times, been mistakenly taken to be the EF warned of by
Hofstede and GLOBE in their national culture models (e.g. Rottig, 2009). The argument is that a
national culture dimension score applies on average in a country but not to all individuals, because
people will differ on any characteristic. Consequently, within this interpretation, using Hofstede
and GLOBE scores for comparison of groups of individuals across countries (rather than just one
individual in each country) is valid because on average the scores apply, and this should be reflected
in comparisons of sizable groups. Therefore, it is argued, an (averaging) EF is avoided when analysing sample groups of people across countries using Hofstede and GLOBE national scores. And
there the EF story ends for many researchers. However, this EF interpretation is not what the EF in
Hofstede and GLOBE is about, as we have explained above. At no point do Hofstede or GLOBE
refer to this ‘averaging’ interpretation of the EF in their work. It is true that both Hofstede and
GLOBE calculated their national dimension by averaging (roughly, in the case of Hofstede) the
scores of their sample individuals’ item responses in each country. However, this calculation methodology has no meaning at the individual level, because these averages are calculated for items that
are unrelated at the individual level and which therefore do not constitute a cultural dimension at
the individual level. Therefore, while applying Hofstede and GLOBE scores in analysis comparing
groups of individuals across countries may not commit an averaging EF, it certainly does commit
a measurement EF. This is the point that the quotation at the beginning of our article makes.
To clarify these important differences, Table 1 summarizes the three types of EF and how they
are manifested in practice. We have called ‘averaging EF’ the mistake of assuming all members of
a group have the group’s average characteristics. ‘Structural EF’ is the classic Robinson case of
assuming that a relationship between variables at the group level will manifest in the same way at
the individual level. The ‘measurement EF’ of Hofstede and GLOBE occurs when a bundle of
average group items (labelled as a national culture dimension) is assumed to reflect a corresponding dimensional characteristic of the individuals in those groups (countries). It is the ‘measurement
EF’ that is commonly committed when researchers attribute the Hofstede and GLOBE national
culture dimension characteristics and scores to individuals in respective countries, as we show in
the next section.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1068
Organization Studies 35(7)
Table 1. Typology of Ecological Fallacy.
Type of EF
How does the ecological fallacy occur?
Example
From group-level attribute/ To individual-level
relationship
attribute/relationship
Averaging EF
Structural EF
Measurement EF
A group-level attribute
is calculated as the
average attribute of group
members
Two group average
variables are strongly
correlated
Individuals and sub-groups
in the group are assumed
to have the same attribute
as the group average
The same two variables
are assumed to be strongly
correlated for individuals
A national/ecological
construct emerges based
on strong correlations
among several national
average items
The same construct with
similar characteristics is
assumed to exist at the
individual level
Average group IQ is low,
therefore individuals or
sub-groups in the group
have a low IQ
Percent black and percent
illiterate are strongly
correlated across 50 states
of the USA, therefore
being black is a strong
predictor of illiteracy
National (societal)
uncertainty avoidance
dimension emerges at
national/societal level,
hence the same dimension
with similar characteristics
applies to individuals
Computation of the Hofstede and GLOBE National Culture
Scores
In order to appreciate the source and nature of the EF in national culture-related research, it is critical to understand the methodology used in the Hofstede and GLOBE studies to identify the national
culture dimensions and to compute the national culture scores. The Hofstede study was based on
responses to items in his IBM employee questionnaire covering seven occupational categories. To
identify the national culture dimensions, Hofstede aggregated the individual-level item scores
within each nation to create the national-level mean or percentage score for each item. The nationallevel aggregate item scores were factor analysed to identify the items that reflect each national culture dimension. Finally, Hofstede developed formulae to compute the national culture dimension
scores for countries from the national-level aggregate item scores, usually resulting in national
dimension scores between 0 and 100. (These formulae are shown at Hofstede, 2001, pp. 491–492.)
Hofstede’s analysis found that, while there were significant national-level correlations, that is,
between the country-level means of questionnaire items, the individual-level correlations for the
same items were mostly not significant: ‘the answers of individual respondents within countries on
these same questions do not show the same correlations’ (Hofstede, 2001, p. 65). Table 2 shows
individual-level and national-level correlations for Hofstede’s power distance and uncertainty
avoidance items (the only dimensions for which the details are available) (Hofstede, 2001, pp. 125,
184). As shown in the table, for the power distance dimension, the three pair-wise correlations
among its three measures are small and insignificant at the individual level (absolute average 0.05),
but large and strongly significant at the national level (absolute average 0.50). Similarly for the
uncertainty avoidance dimension, the three pair-wise correlations among its three measures are
small at the individual level (absolute average 0.08) in comparison to the national-level correlations (average 0.49) that are large and highly significant.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1069
Brewer and Venaik
Table 2. Individual-Level Versus Country (Ecological) Mean Correlations in Hofstede (2001) for Power
Distance and Uncertainty Avoidance.
Pairs of items
Power distance
Employees not afraid X perceived
manager autocratic/persuasive
Employees not afraid X preferred
manager consultative
Perceived manager autocratic/persuasive
X preferred manager consultative
Uncertainty avoidance
Rule orientation X Employment stability
Rule orientation X Stress
Employment stability X Stress
Correlations of individual
scores
Correlation of mean
scores across 53 countries
−.05
−.44***
.03
.53***
.07
−.53***
.14**
.00
−.11*
.58***
.46***
.44**
Note: The individual score correlations are the median correlations in five homogenous occupational groups. The country correlations are means of mean scores for seven occupational groups. *p =.05; **p = .01; ***p = .001.
The table is based on Hofstede (2001, pp. 125 and 184).
Since the relationships among the three items used to measure the culture dimensions exist only
at the national level, it follows that Hofstede’s national culture dimension scales and the characteristics underlying these dimensions also exist only at the national level and not at the individual or
organizational levels. Hofstede himself makes the point that ‘cultures are not king-size individuals:
They are wholes and their internal logic cannot be understood in the terms used for the personality
dynamics of individuals’ (Hofstede, 2001, p. 17) and ‘They [Hofstede’s dimensions] are meant to
be a test of national culture, not of individual personality; they distinguish cultural groups or populations, not individuals’ (Hofstede, 1998, p. 481). In relating national culture with individual
behaviour, Hofstede (2001, p. 65) states that ‘A culture does not produce one single type of psychological reaction; it can produce many alternative, complementary, and even conflicting types of
reaction in individuals.’ Based on his analyses, a high uncertainty avoidance score for a nation does
not mean that individuals or organizations in the nation reflect a high level of the attributes underlying the national uncertainty avoidance dimension.
The GLOBE study differed from Hofstede in that the questionnaire items were specifically
designed to measure the culture dimensions which had been previously identified through the literature (e.g. the work of Hofstede). All GLOBE items were measured on seven-point scales.
Multilevel confirmatory factor analysis (CFA) of the items ultimately confirmed ‘that the items can
be aggregated to the society level of analysis’ (Hanges & Dickson, 2006, p. 530). As shown in
Table 3, the multilevel CFA results for the GLOBE societal uncertainty avoidance practices scale
indicate that ‘the items strongly load on the single societal-level factor. The factor loadings at the
within-society level were extremely weak. Thus, this analysis supports that this scale operates at
the society level of analysis’ (Hanges & Dickson, 2006, p. 531). Based on the emergence of the
societal-level factors using national item means, GLOBE computed the national culture scores for
each dimension as the grand means of the relevant national item means. Thus, as with Hofstede,
the unit of analysis in GLOBE is the national-level mean item responses: ‘The unit of analysis for
Project GLOBE consisted of cultural-level aggregated responses of middle managers’ (House &
Hanges, 2004, p. 99).
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1070
Organization Studies 35(7)
Table 3. Multilevel CFA for GLOBE Societal Uncertainty Avoidance Practices Scale.
Standardized factor loadings
Within society
Between society
In this society, orderliness and consistency are stressed, even at
the expense of experimentation and innovation
.37
.70
In this society, most people lead highly structured lives with few
unexpected events
In this society, societal requirements and instructions are spelled
out in detail so citizens know what they are expected to do
This society has rules or laws to cover: (1) = almost all
situations, (4) = some situations, (7) = very few situations
.45
.88
.58
.93
.39
.78
The response scale for the first three items is: 1 = strongly agree, 4 = neither agree nor disagree, 7 = strongly agree.
Source: Hanges and Dickson (2006, p. 531).
Individual-level correlations between the GLOBE questionnaire items have not been reported,
since the study is not about individuals. Organization-level data has been published and is shown
at column 2 in Table 4. It is important to note that the organizational-level items have very low
internal consistency (well below the accepted cut-off of 0.7). Since the GLOBE item measures are
isomorphic between national and organizational levels, the internal consistency statistics in Table 4
confirm that the characteristics underlying the national culture dimensions are not a valid reflection
of the namesake organizational culture dimensions. Thus, as with Hofstede, the GLOBE national
culture dimensions do not exist at the individual or organizational level with the same set of items
as at the national level of analysis. Consequently, as with the Hofstede model, a society with a high
GLOBE uncertainty avoidance score does not mean that individuals and organizations in the society have a high level of the characteristics underpinning the GLOBE national uncertainty avoidance dimension.
The Ecological Nature of National Culture
Hofstede and GLOBE have themselves adopted the term ‘ecological fallacy’ to describe the mistaken application of their national culture dimension characteristics to individuals. ‘Confusion
between within-system and between-system (ecological) correlations is known as the ecological
fallacy’ (Hofstede, 2001, p. 16). ‘Inappropriately assuming that cultural-level characterizations and
relationships apply to individuals within these cultures is commonly labeled the “ecological fallacy”’ (House & Hanges, 2004, p. 99). The danger of this mistake is canvassed on occasions by the
authors of both studies, albeit in a manner that is not prominent, given the length and scope of their
monographs. Hofstede is clear in his separation of individual cultural characteristics from national
dimensions, including in articles: ‘An aggregate score no longer describes an individual, but
becomes an indicator distinguishing one social system from another’ (Hofstede, Bond & Luk,
1993, p. 488).
GLOBE (2006, p. 3), in its guidelines for use of its items, warns, ‘Caveat Emptor: It is unreasonable to expect that these scales will show the same psychometric properties at the individual
level of analysis as they did at the aggregate levels.’ In the chapter which discusses the GLOBE
scales, Hanges and Dickson (2004, p. 124) point out that ‘Clearly because the GLOBE scales were
designed to assess convergent-emergent constructs, they were not designed to assess individual
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1071
Brewer and Venaik
Table 4. Internal Consistency for Organizational and Societal Culture Scales in GLOBE.
GLOBE culture dimensions
Internal consistency
Practices (as is)
Performance orientation
Assertiveness
Future orientation
Humane orientation
Collectivism I: institutional collectivism
Collectivism II: in-group collectivism
Gender egalitarianism
Power distance
Uncertainty avoidance
Values (should be)
Performance orientation
Assertiveness
Future orientation
Humane orientation
Collectivism I: institutional collectivism
Collectivism II: in-group collectivism
Gender egalitarianism
Power distance
Uncertainty avoidance
Organizational
Societal
.68
.60
.57
.78
.44
.70
.46
.55
.67
.72
.75
.80
.88
.67
.77
.66
.80
.88
.61
.50
.52
.61
.43
.63
.65
.45
.60
.90
.53
.76
.70
.77
.66
.88
.74
.85
Extracted from House et al. (2004, p. 134).
variation within organizations or societies and should not be used for this purpose’ (Hanges &
Dickson, 2004, p. 124) and later ‘Finally, it cannot be repeated enough:…They (the scales) were
not specifically designed to measure differences within cultures or between individuals’ (Hanges &
Dickson, 2004, p. 146).
The business literature on aggregation of survey data has long recognized that individual-level
variable relationships may be different at the aggregated level (Peterson & Castro, 2006). And the
EF warnings by Hofstede and GLOBE have been repeated on occasions by others in the culture
literature. For example, the EF, in a cultural context, has been discussed by Bond (2002, p. 75),
who wrote:
As has been shown by many authors…the pattern of correlations at the national (or organizational or
group) level is not replicated at the individual level. Even though some forcefully argue that there is
pressure within the social system for an individual organization of value groupings to align itself with a
national organization of value groupings…in practice such isomorphism does not occur. Nation-level
constructs are not logically or empirically constituted the same way as individual-level constructs,
convenient as it would be.
In other words, ‘Individualism-collectivism at the level of nations is not the same as individualismcollectivism at the level of individuals, either conceptually or operationally’ (Bond, 2002, p. 76).
In a similar vein, Shweder (1973, p. 543) explained, ‘valid indicators of a theoretical variable may
be discovered across a representative sample of cultures without being discoverable within any of
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1072
Organization Studies 35(7)
them’. Schwartz (2009, p. 132) ‘reinforces the view that cultures and individuals are distinct entities and that different principles organize the normative cultural systems of societies and the motivational values systems of individuals.’ Empirical evidence of the EF at work in culture is also
readily available. For example, Ralston et al. (2009) find different relationships between their
ethics dimensions and individualism-collectivism at the societal level (using Hofstede scores)
compared with the individual level (using sampled individuals’ scores). Spector et al. (2001) found
low internal consistency at the individual level across samples in 23 nations using Hofstede’s
Values Survey Module 1994 items. These results are exactly what the EF in culture would lead us
to expect.
Thus the EF involved in attributing national-level culture characteristics to individuals and
organizations is an important issue relating to construct clarity, but, as we show in the next section,
it is often overlooked. While there is no theoretical reason for individuals to reflect the characteristics underlying their national culture dimensions, many empirical research articles (discussed in
the next section), involving the analysis of relationships between Hofstede or GLOBE scores and
other variables, commit the EF by erroneously attributing the national culture dimension characteristics to individuals or organizations.
Evidence of the Ecological Fallacy in Culture Research
This section examines evidence of the EF in culture research literature. We first present the methodology used to select our example articles and we then discuss in detail the nature of the EF in
each of the selected exemplar papers.
Methodology
To demonstrate the prevalence of the EF in cross-cultural research, we use a purposive methodology to identify exemplar cases for analysis. Our methodology involved the following steps. First,
to cover a broad range of business and management disciplines, we decided to focus on five key
areas, namely, general management, organization behaviour, international business, marketing and
accounting. Second, we selected one of the top journals (based on their citation impact factors) in
each of these disciplines for our analysis, resulting in the choice of the following five journals:
Organization Science (organizational behaviour), Academy of Management Journal (general management), Journal of International Business Studies (international business), Journal of Marketing
Research (marketing) and Accounting, Organizations and Society (accounting). Third, we used
electronic research databases to search these journals using the keywords ‘Hofstede’, ‘GLOBE’,
and ‘national culture dimensions’, to identify articles that use Hofstede and/or GLOBE national
culture dimensions and scores in their empirical analysis. We then read the most recent of these
articles in order and selected the latest paper in each journal that clearly demonstrated an EF in
their analysis and discussion, ensuring also that in making these selections a mix of Hofstede and
GLOBE dimensions were included in the studies. Five article examples across five business disciplines seems an effective number to illustrate our argument and falls within the optimal range of
cases (four to ten) recommended by Eisenhardt (1989) for multiple case studies theory
development.
As we show, these studies use national culture scores to develop theoretical propositions as if
the dimensions and associated scores represent the characteristics of individual managers in respective countries, and subsequently draw national culture dimension-related conclusions and implications at the individual level. Each of the five articles reviewed commits the EF both in the logic of
hypothesis development and in the discussion of results, thus compromising what are otherwise
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1073
Brewer and Venaik
high-quality studies. Although each article operationalizes national culture with two or more
national dimensions, and commits the EF across all the national culture dimensions used, for brevity, we focus on one or two dimensions in each article and on different dimensions in different
articles to illustrate the EF in the literature. Through the rest of this article italics shown within
quotations are our own.
Organization Science
Tosi and Greckhamer (2004) study the effect of national culture on three aspects of CEO compensation. In their article the authors apply the four original Hofstede national culture dimension
scores: power distance, uncertainty avoidance, masculinity and individualism, to explain CEO
compensation across 23 countries. Serious problems relate to the logic of their hypotheses development and in the explanation of their analysis results and theoretical implications. The authors
propose a relationship between one of their components of CEO compensation and national individualism because:
In cultures oriented toward individualism, such as the United States, the United Kingdom, and Canada,
people tend to emphasize their individual needs, concerns, and interests over those of their group or
organization. In individualistic cultures, individual initiative is encouraged (Hofstede, 1980, 2001). The
opposite is true in countries high on collectivism, such as the Asian economies such as Japan and Taiwan.
In a collectivist society, an individual is expected to interact with members of his or her group. (Tosi &
Greckhamer, 2004, p. 621)
There is no evidence for these claimed characteristics of individuals in high/low individualism
countries as measured by Hofstede. Such claims commit the EF.
An associated conclusion problem arises because while the independent variables in the regression analysis are the Hofstede national dimensions of culture, the outcomes are explained at the
individual level. In the discussion of the results the authors note that ‘In these countries (with high
Individualism), individual achievement is highly valued and rewarded. The nature of organizational practices encourages individual motivation and initiative, employees look out for their own
interests, and organizations are more impersonal.’ And ‘in a country such as the United States, that
is high on individualism, the combination of high pay and variable compensation may symbolize
the person’s achievement and attainment of higher levels of personal success’ (Tosi & Greckhamer,
2004, p. 667). These conclusions about individuals reflecting their national culture characteristics
cannot be drawn without committing the EF.
Academy of Management Journal
Steensma, Marino and Dickson (2000) study the effect of national culture on SME cooperation
with other firms and their use of equity to form alliances. In their article the authors apply three of
Hofstede’s national culture dimension scores, uncertainty avoidance, masculinity and individualism, to explain individual firm tendencies within those nations. There are two particular problems
evident in the authors’ logic relating to the specific hypotheses development and to the explanation
of the results and implications. As an example we look at the article’s proposition that higher
national masculinity results in lower interest in technological alliances. The problem is not with the
proposition per se, but rather with the manner in which the corresponding hypothesis is developed,
by attributing the ‘masculinity/femininity’ characteristics of nations to the individuals in respective
countries. For example, ‘The people from feminine countries tend to be less aggressive and more
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1074
Organization Studies 35(7)
modest’ and ‘In masculine countries, children are taught to appreciate strong and independent
heroes’ (Steensma et al., 2000, p. 956). These claims reflect the projection of a national culture
dimension onto the individuals in the relevant nation, and commit the EF.
The second EF arises in the discussion of the results. The authors use Hofstede’s national dimensions and scores in regression analysis but the rationale for the discovered relationship between
national masculinity and alliances is delivered at the individual level: ‘the win-win nature of technology alliances may be more acceptable to SMEs in feminine societies than to those in masculine
countries, where the prevalent world view is that one person’s gain is another’s loss’ (Steensma
et al., 2000, p. 966). In other words individuals in feminine societies are taken to be more amenable
to win-win situations than individuals in masculine societies who are taken to be more competitive,
even though there is no evidence of this provided by Hofstede’s masculinity-femininity which is a
national-level dimension. Overall, the development of the hypothesis and the discussion of the
results are both based on an EF due to the invalid projection of national masculine/feminine culture
characteristics onto individuals.
Journal of International Business Studies
Waldman and colleagues (2006) examine how managers value corporate social responsibility
(CSR) in different cultural and leadership contexts. The EF arises in the theorizing about the
effects of GLOBE national culture values on managerial CSR values and in the discussion of the
results. For example, in developing their hypothesis about the effect of national in-group collectivism values on CSR values, the authors state, ‘Managers in these cultures would be more
concerned with the needs of their in-group members when making decisions’ (Waldman et al.,
2006, p. 826). The rationale of the hypothesis is thus developed on the basis that the characteristics underlying the national in-group collectivism dimension are reflected in the people in these
countries, which is precisely the EF identified in GLOBE. Similarly, in developing hypothesis
H2, the authors state that ‘However, we propose that managers in societies characterized by high
power distance values are not likely to value CSR in the course of their decision making’ and
‘For example, managers in cultures with high power distance values may be less concerned with
the shareholders/owners’ needs than their own’ (Waldman et al., 2006, p. 827). Of course, no
such statement can be made about the characteristics of managers in high GLOBE power distance societies since GLOBE power distance is an ecological dimension and the attributes underlying the national culture dimension are not manifested at the individual level, as found in the
GLOBE study.
When it comes to the discussion of the results, following hierarchical regression analysis incorporating GLOBE national culture dimensions and scores, the authors again commit the EF through
the projection of national culture dimensions onto the individuals in those nations. After finding no
relationship between in-group collectivism and concern for CSR, Waldman et al. (2006, p. 834)
conclude: ‘Such constituencies (CSR recipients) may be viewed as part of the out-group, and thus
not consistently considered in the decision-making of managers in cultures stressing high in-group
values’. Similarly, on finding an inverse relationship between national power distance and CSR
values, the authors conclude that ‘When there is a strong belief in society that there should be distance among people in terms of power, relatively high-level managers who have the power…may
be more self-centered or lacking in concern for shareholders/owners’ (Waldman et al., 2006,
p. 834). Finally, in discussing the practitioner implications of their study, Waldman et al. (2006,
p. 835) state: ‘However,…managers in a subsidiary country might have weaker stakeholder CSR
values in line with that country’s weak institutional collectivism and high power distance.’ As
found in the GLOBE study, their culture dimensions exist only at the national level and not at the
individual level. Overall, this paper also commits an EF by invalidly projecting the GLOBE
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1075
Brewer and Venaik
national ecological cultural characteristics of collectivism and power distance onto the individuals
and managers in these countries.
Journal of Marketing Research
Deleersnyder, Dekimpe, Steenkamp and Leeflang (2009) assess the sensitivity of advertising
expenditure to business cycles in different countries. They propose that culture will moderate the
effect of business cycles on advertising expenditure. The analysis is conducted using Hofstede
national culture dimensions. As in our other examples, the authors’ hypotheses development is
based on an EF since the national culture characteristics are projected onto individuals. The authors
propose that advertising expenditure is more sensitive to business cycles in high collectivism countries. This proposition is made on the basis that managers in collectivistic societies follow ‘herding
behaviour (that) causes advertising to follow business cycles more closely’ (Deleersnyder et al.,
2006, p. 626). This is an EF, projecting Hofstede’s national collectivism onto individuals through
a presumed herding behaviour.
The same EF emerges in the discussion of results. Following regression analysis incorporating
Hofstede national culture dimensions and scores, the authors conclude that there is no relationship
between Hofstede’s cultural collectivism and advertising co-movement elasticity since ‘managers
in collectivistic countries are clearly tied to group support for decisions’ (Deleersnyder et al., 2009,
p. 634), and ‘internal group pressures may counteract the tendency (of managers) to engage in
herding behaviour in collectivistic societies’ (Deleersnyder et al., 2009, p. 635). Overall, in this
paper too, the authors commit the EF by invalidly projecting the national culture characteristic of
Hofstede’s collectivism onto the managers in these countries.
Accounting, Organizations and Society
Chow, Kato and Merchant (1996) compare the use and effect of a range of management controls
in multiple cost centres in Japan and the United States. The authors expect that differences in
national culture between Japanese and US managers will lead to significant differences in the
application and effect of these types of controls. While the hypothesis seems plausible, the logic
used to develop it is untenable as it is based on the EF. Invoking Hofstede’s national culture
dimensions and scores, the authors state that ‘U.S. nationals are more likely to emphasize their
individual interests over those of the firm’ (Chow et al., 1996, p. 178). But there is no evidence
about Japanese or US individuals’ traits flowing from Hofstede’s national culture dimensions.
The authors go on to claim knowledge of other characteristics of individuals in Japan and the
US, including ‘Since the Japanese are much higher in uncertainty avoidance, they can be expected
to rely more heavily on procedural control’ and ‘The higher power distance of the Japanese,
meanwhile, implies that they are far more open to receiving directives from their superiors’
(Chow et al., 1996, p. 179). These conclusions are all invalid since the Hofstede national culture
dimensions represent national-, not individual-level, characteristics. While the article cites
research which supports the contention that US people are more individualistic than Japanese
people, it provides no such evidence about individual-level differences between Japanese and
Americans for the other Hofstede national culture dimensions employed in the study, namely,
power distance and uncertainty avoidance.
Similarly, the authors again fall into the EF trap in explaining their results. Their analysis shows
significant differences between the use and effects of management controls in the Japanese and
American firms and they conclude that these differences are a result of the culture characteristics
of the individuals in these nations. For example, ‘each individual simultaneously embodies all the
dimensions of national culture’ and subsequently, ‘Because of their higher uncertainty avoidance
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1076
Organization Studies 35(7)
the Japanese can be expected to have a higher preference for this evaluation approach’ (Chow et
al., 1996, p. 189). According to the pre-eminent developers of national culture dimensions and
scores, including Hofstede on whose work the article relies, this conclusion is invalid as it is based
on an EF since the national culture dimensions characteristics cannot be projected on to the individuals or organizations in these countries.
In sum, we generalize the nature of the EF in all five examples provided above as follows. The
authors are projecting Hofstede and GLOBE national culture dimension(s) characteristics onto
individuals in respective countries. There is a presumption of isomorphism in national to individual
culture characteristics, which is invalid, and then an explanation of behaviour is provided based on
that invalid presumption.
The Spread of the Ecological Fallacy in Culture Research
While we have been critical of the exemplar articles in the last section, we readily accept that the
authors acted in good faith in applying the national culture dimensions in their work. This raises
the question as to why the EF is so prevalent in culture research. Although there are several plausible reasons for the persistence of the EF in culture research, we speculate that the problem is due,
at least in part, to two main reasons. The first results from the fact that management scholars have
for many years sought greater precision in definitions and quantitative rigour in operationalization
of concepts such as culture (Tayeb, 1994). According to Birkinshaw, Brannen and Tung (2011,
p. 573), ‘Over time, however, such (qualitative) methods have been progressively marginalized in
our field as quantitative methods have become the norm. This shift in emphasis has mirrored the
broader trend toward more positivistic empirical methods in the social sciences.’ In the context of
business-related culture research, quantitative comparisons of national culture dimensions would
be of great assistance in improving the robustness of analysis. Hofstede provided these numbers in
1980, expanded them in 2001, and GLOBE replicated and expanded the measures again in 2004.
Understandably, researchers have been enthused at the prospect of using established national culture data in models to analyse relationships of interest.
The problem is that, presented with such a precise quantification of national culture, many
scholars have simply not questioned the real nature of the dimensions and the validity of their
application (although, as noted above, some have done so). And, of course, the more the scores are
applied in journal articles the more unquestioning acceptance they generate in future studies. As
McSweeney (2009, p. 949) points out, national culture models, especially the dominant one by
Hofstede, have ‘achieved a scholarly identity, being employed in articles across a wide range of
peer-reviewed journals’. There is evidence in many journals of the view that Hofstede’s cultural
dimensions should be used at the individual level because they have been much used that way in
the past. For example, Rollinson (2005, p. 683) states, ‘Perhaps the strongest evidence of the utility
of Hofstede’s scheme lies in the large number of respected researchers…who continue to use it…’
Obviously, the EF could have been avoided if the Hofstede and GLOBE dimensions had labels
which did not semantically relate to the characteristics of individuals, but such is not the case.
Bond et al. (2004, p. 567) state:
Ecological-level and individual-level factors tap into constructs that function at different levels and must
consequently be a part of different theoretical discourses (Leung & Bond, 1989). For this reason, the
factors should always be labelled differently, with the content of the labels appropriate for the level of
analysis to which they are applied.2
This semantic isomorphism further fuels the presumption of equivalence across levels.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1077
Brewer and Venaik
The second reason is that, although the EF is cited by both Hofstede and GLOBE (specifically
Robinson, 1950), they provide no explanation about the nature of the EF problem and its role and
effects in the context of their national culture dimensions. Further, in spite of their own warnings
of the inapplicability of the national culture dimensions, scales and scores to individuals, Hofstede
and GLOBE at times do so themselves in their own published work. Earley points out that Hofstede
himself falls into the EF trap ‘in discussing the specific impact of aggregate cultural values on business practices. And it is a trap that is inevitable if one uses values measured by individual perception as an indicant of collective culture’ (Earley, 2006, p. 926). Hofstede’s dimension definitions
consistently reflect the ecological fallacy. For example, uncertainty avoidance is ‘The extent to
which the members of a culture feel threatened by uncertain or unknown situations’ (Hofstede,
2001, p. 161). (Definitions for all Hofstede dimensions are provided in the Appendix.) Hofstede
also in each of his dimension chapters projects the relevant dimension onto individuals, such as:
‘In the large power distance situation, children are expected to be obedient towards their parents.
Independent behavior on the part of the child is not encouraged’ (Hofstede, 2001, p. 99). ‘In cultures with strong uncertainty avoidance, parents are supposed to watch over their children’s proper
motivation and behavior at school – for example by signing children’s homework assignments…’
(Hofstede, 2001, p. 163). ‘In the individualist society, the relationship between employer and
employee is primarily conceived as a business transaction, a calculated relationship between actors
on a “labor market”’ (Hofstede, 2001, p. 237).
In the definitions of their dimensions, the GLOBE authors also project their national culture
dimensions onto the citizens of their sample countries. For example, in the GLOBE overview
chapter, House and Javidan (2004, p. 12) define uncertainty avoidance as ‘the extent to which
members of an organization or society strive to avoid uncertainty by relying on established social
norms, rituals and bureaucratic practices.’ (Definitions of all the GLOBE dimensions are provided
in the Appendix.) In the GLOBE chapters which discuss the specific culture dimensions, there is
evidence of the EF. For example, Javidan (2004, p. 254) comments that ‘cultures with high
Performance Orientation practices scores foster leaders who value and reward hard work, ambition, high standards and performance improvement’, thus projecting the national performance orientation characteristics onto the individuals/leaders in these countries.
All of these example quotations erroneously draw equivalence between national culture dimensions and the traits and characteristics of individuals who live in these nations, and thus commit the
EF. It is a case of ignoring the fact that ‘an aggregate variable often measures a different construct
than its name-sake at the individual level’ (Firebaugh quoted in Schwartz, 1994, p. 820). Given the
ambiguity in the usage of the national culture dimensions in the Hofstede and GLOBE studies, it
is understandable that this invalid ecological/individual equivalence is accepted and adopted by
other cross-cultural researchers.
Implications of the Ecological Fallacy in Culture Research
Implications of the ecological fallacy for culture-related theory
The nature of the Hofstede and GLOBE national culture dimensions has important implications for
the development of management theory generally. For many years anthropological culture has
been recognized as an important construct in theorizing about organizations (Allaire & Firsirotu,
1984; Nelson & Gopalan, 2003). However, the culture lenses used in this theorizing are varied
(Smircich, 1983). Some, although not all, scholars, in line with Hofstede and GLOBE, work within
the view that the collective beliefs, norms and values prevalent broadly across a society impact on
the organizations operating within that society in distinct ways (e.g. Park & Ungson, 1997). Put
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1078
Organization Studies 35(7)
another way, for researchers in this school, societal culture helps shape organizational behaviour in
recognizable ways, so that while organizations within the same country will differ across a range
of variables (effectiveness, strategy), there will also be similarities in behaviours and preferences
which are linked to the country’s culture. This approach reflects a perception of culture as an environmental condition, beyond the control of individuals and yet an important factor in individual
and organizational behaviour. The national culture models of Hofstede and GLOBE fall within this
camp; their culture dimension identification and measurement methods are etic and involve the
analysis of aggregated individual responses to culture-related questions to measure independent
but shared national culture variables (called national culture dimensions). According to these theorists, culture dimensions are common across countries, reflecting common human attitudes to life
circumstances, but there will be significant variation in the strength (or score) of the dimensions
depending on the country observed. This conceptualization of culture is common in the field of
cross-cultural management, from which our exemplar articles exhibiting the EF are drawn.
Other scholars take the view that culture is not an external variable, but rather culture should be
thought of as a phenomenon existing as the organization itself (Smircich, 1983). Thus, organizations in the same country and across countries have their own unique, or at least identifiable,
anthropological culture (not just a ‘corporate culture’), being a manifestation of the organization
itself. This perspective of culture calls for an emic or subjective approach to understanding (Bhagat
& McQuaid, 1982). Within this paradigm, an external independent national culture is not a relevant
construct in understanding organizations (nor the individuals within an organizational context). In
fact national boundaries and boundaries around other geographical groups of individuals are irrelevant; the organization is the defining boundary.
While both the culture-as-environment and the organization-as-culture perspectives can be useful in thinking about organizations and conducting research, they entail different epistemological
structures, problems and outcomes (Smircich, 1983). As we have seen, particular problems emerge
when Hofstede and GLOBE (and others), choosing to work in the external, independent or functionalist domain of culture, then attempt to objectively measure their independent national culture
through a positivist analytical approach. The theoretical basis of this process has been questioned
by several researchers, as noted early in this article (notably Ailon, 2008; Baskerville, 2003;
McSweeney, 2002, 2009), but in addition to these recognized limitations, our analysis of the implications of the ecological fallacy makes meaningful interpretation of the national culture scores
highly problematic, if not impossible. In this case, a positivist approach to culture within a functionalist framework has led many to overlook or ignore the problems associated with national
culture dimensions, resulting in a self-reinforcing literature of flawed theory development. While
the functionalist approach to understanding the interaction between culture and organizations can
be logically supported, the Hofstede and GLOBE positivist operationalization of culture within
this framework loses sight of what Smircich (1983, p. 355) describes as ‘the subjective, interpretive aspects of organizational life’ and reduces culture-related organizational behaviour to a series
of calculable, predictable, external causations. While there is not and probably never will be a
consensus on the ‘right’ approach to understanding the mix of social culture and organizations, we
should at least recognize the complexity of the interplay of the two, including the role of nonmeasurable constructs such as myths, ideology and values (Allaire & Firsirotu, 1984). A positivist,
functionalist approach such as that of Hofstede and GLOBE undervalues these aspects of culture
and, in searching for objective culture measures, in fact delivers a cultural mirage. To this extent at
least, the functionalist approach to culture and organizations should be used cautiously as it can
entice researchers towards false solutions.
While we have been critical of the ecological fallacy inherent in applying Hofstede and GLOBE
national culture scores in research, we recognize that an ecological fallacy is not an inevitable
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1079
Brewer and Venaik
consequence of applying national-level variables in multilevel analysis. For example, a large body
of literature discusses different forms of national business systems and their relationship with
organizations without committing the ecological fallacy. These include ‘diversity of modern capitalism’ (Amable, 2003), ‘varieties of capitalism’ (Hall & Soskice, 2001), ‘comparative capitalisms’
(Jackson & Deeg, 2008), ‘national business systems’ (Morgan, 2007) and ‘divergent capitalisms’
(Whitley, 1999). In this literature, the relationships across different levels of analysis are carefully
mapped out and explained. This is in sharp contrast to the national culture literature we have discussed, where the national- and individual-level characteristics are simply assumed to be isomorphic, the hypotheses developments are not supported by theoretical explanations as to how the
national culture dimensions, which do not represent individual characteristics, might shape individual behaviour and hence why they should be included in the model. Claims in the literature that
Hofstede and/or GLOBE national culture dimensions shape individual behaviour have not been
properly explained (Peterson & Søndergaard, 2011) other than by inferring some mysterious osmosis from the national environment to individuals. In other words, a key flaw in hypothesis development incorporating functionalist national culture dimensions is the lack of an explanatory theory of
causation between the individual and higher-level variables, which is essential for valid theory
development in multilevel models (Diez-Roux, 1998). As there is no clear explanation as to how
national culture characteristics affect individual- or organizational-level outcomes, other than the
invalid assumption of isomorphism, the use of national dimensions in cross-cultural analysis is at
best unhelpful and at worst misleading.
Implications of the ecological fallacy in cultural understanding for practitioners
Culture is an area where academic management research has successfully crossed over to practitioners. But this is not all good news because Hofstede and GLOBE national culture dimensions
are widely believed to reflect the characteristics of the individuals and firms in a society and this
fallacy is firmly entrenched in teaching and training of managers. Ironically, this is seen as one of
the major contributions of Hofstede’s work. Hofstede and Hofstede (2005, back cover) claim to
reveal ‘the unexamined rules by which people in different cultures think, feel, and act in business,
family, school, and political organizations’, and then proceed to base their revelations on national
culture dimensions and their relative differences. University-level international business textbooks
routinely discuss Hofstede and GLOBE national culture dimensions in terms of the characteristics
of the people who live in those societies. Browaeys and Price (2008, p. 21) advise that ‘They
[Hofstede’s dimensions] describe tendencies within a certain cultural grouping; they present orientations adopted by the majority of members of a cultural grouping in normal situations.’ According
to Deresky (2002, p. 94), ‘Singapore, Hong Kong and the United States score high on this (GLOBE
performance orientation) dimension; typically, this means that people tend to take initiative and
have a sense of urgency and confidence to get things done.’ But this guidance is built on national
cultural dimensions which do not provide any information about the characteristics of individuals
or firms in other countries that managers will meet and deal with in their international business
transactions. These descriptions create false stereotypes and, to the extent that such ideas are
accepted and adopted by international business managers, they may well be misleading and
counterproductive.
Way Forward
There are a range of arguments in the literature against the use of national culture dimensions in
cross-cultural research and explaining the implications of such use for managers, as discussed in
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1080
Organization Studies 35(7)
the introduction to this article. For these reasons and others, the national culture construct is not
recognized in disciplines such as anthropology (Breidenbach & Nyíri, 2009). In addition to the
problems raised by these debates, we have shown in this article how and why the national culture
dimension characteristics of Hofstede and GLOBE are not valid representations of the namesake
concepts at the individual level. While there is no consensus definition of ‘culture’ (Tayeb, 1994),
there are common elements across most definitions including ‘beliefs’, ‘values’, ‘norms’, ‘attitudes’ and ‘meanings’ (Dorfman & House, 2004; Hofstede, 2001; Kluckhohn & Strodtbeck, 1961)
leading to behaviours which are shared among a group of individuals (not countries per se). This
conceptual understanding is important since the Hofstede and GLOBE dimensions do not operate
at the individual level and the individuals who live in the countries covered by these two models
do not behave in the manner that is inferred by the dimension labels. Thus the dimensions are
incomprehensible to individual-level culture researchers acting within the commonly understood
contexts of culture. Given these important issues, the conclusion we must reach is that the use of
the Hofstede and GLOBE national culture dimensions should be avoided in management studies.
Another important aspect of culture that is largely ignored in national culture dimensions models
is the notion of culture as a ‘pattern’ or ‘configuration’ of characteristics that is shared by a group of
people (Kluckhohn & Strodtbeck, 1961; Tsui, Nifadkar & Ou, 2007). It is increasingly recognized
that configuration-based approaches are superior to individual variable-based approaches in understanding complex social phenomena such as organization strategy, structure, culture, etc. that involve
multiple interacting and mutually reinforcing variables (e.g. Miller, 1996). To identify configurationbased descriptive typologies of organizational strategy, structure, etc., researchers use cluster analysis
(e.g. Homburg, Furst & Kuehnl, 2012) or archetypal analysis (Midgley & Venaik, 2012). On the other
hand, for explaining configuration-based cause-and-effect relationships, qualitative comparative
analysis (QCA) or its later adaptation, fsQCA (fuzzy set QCA) are increasingly popular (e.g. Fiss,
2011; Greckhamer, 2011). QCA is a flexible and versatile methodology that combines the depth of
qualitative case studies with the rigor of quantitative methods, and works with both small and large
sample sizes (Ragin, 2008). However, researchers need to be cautious in using QCA with large numbers of variables each with multiple fuzzy states, as the number of possible configurations rise exponentially, resulting in model overfit with small sample sizes (Fiss, 2011).
The purpose of culture studies in management is to understand the differences in behaviour of
individuals belonging to different groups in society. Managers are less interested in macro-,
national-level characteristics of people. Instead, they want to know if there are segments or groups
of people in societies that have the characteristics of interest to the firm such as customers,
employees, suppliers, etc. For example, a marketing manager is interested to know if there are
market segments that can be targeted for their product or service. Similarly, a recruitment manager is interested in individuals with specific skills and knowledge, and is less concerned with the
state of these attributes at the national level. Hence, one way forward in culture research that
would also be useful for practice is to compare and contrast targeted groups of people across
countries, on specific characteristics of interest to managers for decision-making purposes, for
example, as potential customers or employees of the firm. Broad, general characterizations of
societies on national culture dimensions offer limited insights to managers about the characteristics of individuals, which is the level at which managers need and seek information about a country or society. Unfortunately, absence of such research by scholars has forced practitioners to take
the easy, albeit wrong, path of simply using the national culture dimensions to characterize individuals, thus committing an EF.
More generally, the EF issue identified in the context of national culture dimensions-related
research can potentially exist in other multilevel research including configuration-based approaches
where individual-level data is aggregated to a higher level. We recommend that researchers
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1081
Brewer and Venaik
take precautionary steps in theory building and measurement to mitigate the EF problem. First, the
concepts at the higher level need to be labelled and defined in terms of the characteristics that manifest at the level at which the construct is theorized, and not with individual-level characteristics as
done in national culture dimensions research. This is to ensure that constructs at different levels of
analysis are clear and distinct in their conceptual content and meaning. Second, when a higher-level
construct is derived by aggregating (e.g. by factor analysis) the average scores on multiple items
measured at the individual level (instead of the individual-level scores on the multiple items), the
higher-level construct needs to be cross-validated with another set of items that operate specifically
at the level at which the construct is theorized. Although seemingly redundant, this cross-validation
is critical to ensure that researchers do not commit an EF in cross-cultural and other multilevel
research by describing the higher, national-level constructs in terms of individual-level
characteristics.
Conclusion
This is the first paper that discusses the origins of the EF in the literature, proposes a typology
of EFs, examines the EF in the context of Hofstede and GLOBE national culture dimensions,
analyses cross-culture literature that commits an EF in using the Hofstede and GLOBE national
culture dimensions, and finally, recommends a way forward to avoid the EF in cross-cultural
research. Over 70 years ago Thorndike (1939, p. 122) expressed concern that ‘many readers…
often misapply correlations between features of states, counties, cities, wards, classes etc. to
their constituent units’. We show that in many culture studies his fears are still being realized, as
researchers continue to use the Hofstede and GLOBE national culture dimension characteristics
in an invalid manner, as if they apply to individuals and organizations. Specifically, in the
Hofstede and GLOBE models, the national culture dimensions are based on aggregated, nationallevel item correlations. These dimensions with their underlying characteristics do not emerge at
the individual or organizational level. To use the national dimensions to describe individual/
managerial characteristics in developing hypotheses and drawing conclusions at the individual
or firm level is committing the EF and must be avoided. More broadly, we urge researchers to be
careful about committing an EF in their studies. Theoretical conceptualization, construct measurement and data collection must all be carried out consistent with the level at which the research
problem is identified. Further, we recommend that researchers explore new, configuration-based
methodologies to gain a deeper understanding of the myriad, complex and important issues in
cultural research.
Acknowledgments
We would like to thank Neal Ashkanasy for comments on an earlier version, as well as the anonymous
reviewers and the editors for their helpful suggestions and encouragement. The authors contributed equally to
the preparation of the manuscript.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit
sectors.
Notes
1.
The error of assuming that significant individual-level relationships are explained by individual characteristics when in fact they are caused by an ecological variable is sometimes called the ‘individualistic
fallacy’ (Silver, 2000).
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1082
2.
Organization Studies 35(7)
The Schwartz (2009) culture model follows this logic. He identifies ten individual-level value types and
seven different national culture value types.
References
Ailon, G. (2008). Mirror, mirror on the wall: Culture’s consequences in a value test of its own design. Academy
of Management Review, 33, 885–904.
Allaire, Y., & Firsirotu, M. (1984). Theories of organizational culture. Organization Studies, 5, 193–226.
Amable, B. (2003). The diversity of modern capitalism. Oxford: Oxford University Press.
Baskerville, R. (2003). Hofstede never studied culture. Accounting, Organization and Society, 28, 1–14.
Bhagat, R., & McQuaid, S. (1982). Role of subjective culture in organizations: A review and directions for
future research. Journal of Applied Psychology, 67, 653–685.
Birkinshaw, J., Brannen, M. Y., & Tung, R. (2011). From a distance and generalizable to up close and
grounded: Reclaiming a place for qualitative methods in international business research. Journal of
International Business Studies, 42, 573–581.
Bond, M., Leung, K., Au, A., Tong, K., Reimel de Carrasquel, S., et al. (2004). Culture-level dimensions
of social axioms and their correlates across 41 cultures. Journal of Cross-Cultural Psychology, 35,
548–570.
Bond, M. H. (2002). Reclaiming the individual from Hofstede’s ecological analysis – a 20-year odessey: comment on Oyserman et al. (2002). Psychological Bulletin, 128, 73–77.
Breidenbach, J., & Nyíri, P. (2009). Seeing culture everywhere from genocide to consumer habits. Seattle:
University of Washington Press.
Browaeys, M.-J., & Price, R. (2008). Understanding cross-cultural management. Harlow, UK: Pearson
Education.
Chow, C., Kato, Y., & Merchant, K. (1996). The use of organizational controls and their effects on data
manipulation and management myopia: A Japan vs US comparison. Accounting, Organizations and
Society, 21, 175–192.
Deleersnyder, B., Dekimpe, M., Steenkamp, J.-B., & Leeflang, P. (2009). The role of national culture in
advertising’s sensitivity to business cycles: an investigation across continents. Journal of Marketing
Research, 46, 623–636.
Deresky, H. (2002). International management: Managing across borders and cultures. Upper Saddle River,
NJ: Prentice Hall.
Diez-Roux, A. (1998). Bringing context back into epidemiology: Variables and fallacies in multilevel analysis. American Journal of Public Health, 88, 216–222.
Dorfman, P., & House, R. (2004). Cultural influences on organizational leadership. In R. J. House, P. J.
Hanges, M. Javidan, P. W. Dorfman, & V. Gupta (Eds.), Culture, leadership and organizations: The
GLOBE study of 62 societies (pp. 51–73). Thousand Oaks, CA: SAGE Publications.
Earley, P. C. (2006). Leading cultural research in the future: A matter of paradigms and taste. Journal of
International Business Studies, 37, 922–931.
Eisenhardt, K. M. (1989). Building theories from case study research. Academy of Management Review, 14,
532–550.
Fiss, P. (2011). Building better causal theories: A fuzzy set approach to typologies in organization research.
Academy of Management Journal, 54, 393–420.
Freedom House (2012) Freedom in the world. Available at http://www.freedomhouse.org/report/freedomworld-2012/checklist-questions.
Gerhart, B. (2009). Does national culture constrain organization culture and human resource strategy? The
role of individual level mechanisms and implications for employee selection. Research in Personnel and
Human Resources Management, 28, 1–48.
GLOBE (2006). Guidelines for the use of GLOBE culture and leadership scales. Available at http://business.
nmsu.edu/wordpress/wp-content/uploads/2010/05/GLOBE-Culture-and-Leadership-Scales-GuidelinesAug-1–2006.pdf.
Greckhamer, T. (2011). Cross-cultural differences in compensation level and inequality across occupations:
A set-theoretic analysis. Organization Studies, 32, 85–115.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1083
Brewer and Venaik
Greenland, S., & Robins, J. (1994). Invited commentary: ecologic studies: biases, misconceptions, and counterexamples. American Journal of Epidemiology, 139, 747–760.
Hall, P., & Soskice, D. (2001). Varieties of capitalism: The institutional foundations of comparative advantage. Oxford: Oxford University Press.
Hanges, P. J., & Dickson, M. W. (2004). The development and validation of the GLOBE culture and leadership scales. In R. J. House, P. J. Hanges, M. Javidan, P. W. Dorfman, & V. Gupta (Eds.), Culture,
leadership and organizations: The GLOBE study of 62 societies (pp. 122–151). Thousand Oaks, CA:
SAGE Publications.
Hanges, P. J., & Dickson, M. W. (2006). Agitation over aggregation: Clarifying the development of and the
nature of the GLOBE scales. Leadership Quarterly, 17, 522–536.
Hofstede, G. (1980). Culture’s consequences: International differences in work-related values. Beverley
Hills, CA: SAGE Publications.
Hofstede, G. (1998). Attitudes, values and organizational culture: Disentangling the concepts. Organization
Studies, 19, 477–493.
Hofstede, G. (2001). Culture’s consequences: Comparing values, behaviors, institutions and organizations
across nations. Thousand Oaks, CA: SAGE Publications.
Hofstede, G., Bond, M. H., & Luk, C.-l. (1993). Individual percpetions of organization cultures: A methodological treatise on levels of analysis. Organization Studies, 14, 483–503.
Hofstede, G., & Hofstede, G. J. (2005). Culture and organizations: Software of the mind. New York:
McGraw-Hill.
Homburg, C., Furst, A., & Kuehnl, C. (2012). Ensuring international competitiveness: A configurative approach to foreign marketing subsidiaries. Journal of the Academy of Marketing Science, 40,
290–312.
House, R. J., & Hanges, P. J. (2004). Research design. In R. J. House, P. J. Hanges, M. Javidan, P. W.
Dorfman, & V. Gupta (Eds.), Culture, leadership and organizations: The GLOBE study of 62 societies
(pp. 95–101). Thousand Oaks, CA: SAGE Publications.
House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Vipin, G. (2004). Culture, leadership, and organizations: The GLOBE study of 62 societies. Thousand Oaks, CA: SAGE Publications.
House, R. J., & Javidan, M. (2004). Overview of GLOBE. In R. J. House, P. J. Hanges, M. Javidan, P. W.
Dorfman, & V. Gupta (Eds.), Culture, leadership and organizations: The GLOBE study of 62 societies
(pp. 9–28). Thousand Oaks, CA: SAGE Publications.
Jackson, G., & Deeg, R. (2008). Comparing capitalisms: Understanding institutional diversity and its implications for international business. Journal of Intentional Business Studies, 39, 540–561.
Javidan, M. (2004). Performance orientation. In R. J. House, P. J. Hanges, M. Javidan, P. W. Dorfman, & V.
Gupta (Eds), Culture, leadership and organizations: The GLOBE study of 62 societies (pp. 239–281).
Thousand Oaks, CA: SAGE Publications.
King, G., Rosen, O., & Tanner, M. (2004). Information in ecological inference: An introduction. In G.
King, O. Rosen, & M. Tanner (Eds.), Ecological inference: New methodological strategies (pp. 1–12).
Cambridge: Cambridge University Press.
Kluckhohn, F. R., & Strodtbeck, F. L. (1961). Variations in value orientations. New York: Harper Collins.
Leung, K. (2006). Editor’s introduction to the exchange between Hofstede and GLOBE. Journal of
International Business Studies, 37, 881.
Leung, K., & Bond, M. (1989). On the empirical identification of dimensions for cross-cultural comparisons.
Journal of Cross-Cultural Psychology, 20, 133–151.
McSweeney, B. (2002). Hofstede’s model of national cultural differences and their consequences: A triumph
of faith; a failure of analysis. Human Relations, 55, 89–118.
McSweeney, B. (2009). Dynamic diversity: Variety and variation within countries. Organization Studies, 30,
933–957.
Midgley, D. F., & Venaik, S. (2012). Marketing management in MNC subsidiaries: An archetypal analysis.
The 54th Annual Meeting of the Academy of International Business, Washington, DC.
Miller, D. (1996). Configurations revisited. Strategic Management Journal, 17, 505–512.
Minkov, M., & Hofstede, G. (2011). The evolution of Hofstede’s doctrine. Cross Cultural Management: an
International Journal, 13, 10–20.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1084
Organization Studies 35(7)
Morgan, G. (2007). National business systems research: Progress and prospects. Scandinavian Journal of
Management, 23, 127–145.
Nelson, R. E., & Gopalan, S. (2003). Do organizational cultures replicate national cultures? Isomorphism,
rejection and reciprocal opposition in the corporate values of three countries. Organization Studies, 24,
1115–1151.
Park, S., & Ungson, G. (1997). The effect of national culture, organizational complementarity, and economic
motivation on joint venture dissolution. Academy of Management Journal, 40, 279–307.
Peterson, M. F., & Castro, S. (2006). Measurement metrics at aggregate levels of analysis: Implications for
organization culture research and the GLOBE project. Leadership Quarterly, 17, 506–521.
Peterson, M. F., & Søndergaard, M. (2011). Traditions and transitions in quantitative societal culture research
in organization studies. Organization Studies, 32, 1539–1558.
Ragin, C. (2008). Redesigning social inquiry: Fuzzy sets and beyond. Chicago, IL: University of Chicago
Press.
Ralston, D., Egri, C., de la Garza Carranza, M. T., Ramburuth, P., Terpstra-Tong, J., et al. (2009). Ethical
preferences for influencing superiors: a 41-society study. Journal of Inernational Business Studies, 40,
1022–1045.
Robinson, W. S. (1950). Ecological correlations and the behavior of individuals. American Sociological
Review, 15, 351–357.
Rollinson, D. (2005). Organisational behaviour and analysis: An integrated approach. Harlow, UK: Prentice
Hall.
Rottig, D. (2009). Overcoming common pitfalls in cross cultural management research. International
Business: Research Teaching and Practice, 3(1), 32–51.
Schwartz, S. (1994). The fallacy of the ecological fallacy: The potential misuse of a concept and the consequences. American Journal of Public Health, 84, 819–824.
Schwartz, S. H. (2009). Culture matters: national value cultures, sources, and consequences. In R. Wyer, C.
Chiu, & Y. Hong (Eds.), Understanding culture: Theory, research and application (pp. 127–150). New
York: Taylor & Francis Group.
Shweder, R. (1973). The between and within of cross-cultural research. Ethos, 1, 531–545.
Silver, E. (2000). Race, neighborhood disadvantage, and violence among persons with mental disorders: The
importance of contextual measurement. Law and Human Behavior, 24, 449–456.
Smircich, L. (1983). Concepts of culture and organizational analysis. Administrative Science Quarterly, 28,
339–358.
Spector, P. E., Cooper, C. L., & Sparks, K. (2001). An international study of the psychometric properties of
the Hofstede Values Survey Module 1994: A comparison of individual and country/province results.
Applied Psychology: An International Review, 50, 269–281.
Steensma, H. K., Marino, L., Weaver, K. M., & Dickson, P. (2000). The influence of national culture on
the formation of technology alliances by entrepreneurial firms. Academy of Management Journal, 43,
951–973.
Tayeb, M. (1994). Organizations and national culture: Methodology considered. Organization Studies, 15,
429–446.
Thorndike, E. L. (1939). On the fallacy of imputing the correlations found for groups to the individuals or
smaller groups composing them. American Journal of Psychology, 52, 122–124.
Tosi, H., & Greckhamer, T. (2004). Culture and CEO compensation. Organization Science, 15, 657–670.
Tsui, A., Nifadkar, S., & Ou, A. (2007). Cross-national, cross-cultural organizational behavior research:
Advances, gaps, and recommendations. Journal of Management, 33, 426–478.
Venaik, S., & Brewer, P. (2010). Avoiding uncertainty in Hofstede and GLOBE. Journal of International
Business Studies, 41, 1294–1315.
Waldman, D., Sully de Luque, M., Washburn, N., House, R., Adetoun, B., et al. (2006). Cultural and leadership predictors of corporate social responsibility values of top management: A GLOBE study of 15
countries. Journal of International Business Studies, 37, 823–837.
Whitley, R. (1999). Divergent capitalisms: The social structuring and change of business systems. Oxford:
Oxford University Press.
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1085
Brewer and Venaik
Author biographies
Paul Brewer is a senior lecturer at UQ Business School, University of Queensland, Brisbane, Australia. He
obtained his PhD in management at the University of Queensland. His research interests include internationalization of the firm, psychic and cultural distance and the globalization of business.
Sunil Venaik is an associate professor at the UQ Business School in the University of Queensland, Brisbane,
Australia. He obtained his PhD in international business/marketing strategy from the Australian Graduate
School of Management at the University of New South Wales, Sydney. His research interests include national
culture, MNC strategy and management, and the impact of FDI in emerging markets.
Appendix
Hofstede and GLOBE National Culture Dimension Definitions
Hofstede.
Uncertainty Avoidance is ‘The extent to which the members of a culture feel threatened by uncertain or unknown situations.’ (Hofstede, 2001, p. 161).
‘The Power Distance between a boss B and a subordinate S in a hierarchy is the difference
between the extent to which B can determine the behavior of S and the extent to which S can determine the behavior of B.’ (Hofstede, 2001, p. 83).
‘Long Term Orientation stands for the fostering of virtues oriented towards future rewards, in
particular, perseverance and thrift. Its opposite pole, Short Term Orientation, stands for the fostering of virtues related to the past and present, in particular respect for tradition, preservation of
‘face’ and fulfilling social obligations.’ (Hofstede, 2001, p. 359).
‘Individualism stands for a society in which the ties between individuals are loose: Everyone is
expected to look after him/herself and her/his immediate family only. Collectivism stands for a
society in which people from birth onwards are integrated into strong, cohesive in-groups, which
throughout people’s lifetime continue to protect them in exchange for unquestioning loyalty.’
(Hofstede, 2001, p. 225).
‘Masculinity and Femininity…refer to the dominant gender role patterns in the vast majority of
both traditional and modern societies…statistically, men as a rule will show more “masculine” and
women more ‘feminine’ behavior.’ (Hofstede, 2001, p. 284).
GLOBE (House & Javidan, 2004, pp. 11–13).
Uncertainty Avoidance is ‘the extent to which members of an organization or society strive to avoid
uncertainty by relying on established social norms, rituals and bureaucratic practices. People in
high uncertainty avoidance cultures actively seek to decrease the probability of unpredictable
future events that could adversely affect the operation of an organization or society and remedy the
success of such adverse effects.’
‘Power Distance is the degree to which members of an organization or society expect and
agree that power should be stratified and concentrated at higher levels of an organization or
government.’
‘Future Orientation is the degree to which individuals in organizations or societies engage in
future-oriented behaviors such as planning, investing in the future, and delaying individual or collective gratification.’
‘Institutional Collectivism is the degree to which organizational and societal institutional practices encourage and reward collective distribution of resources and collective action.’
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014
1086
Organization Studies 35(7)
‘In-Group Collectivism is the degree to which individuals express pride, loyalty and cohesiveness in their organizations or families.’
‘Gender Egalitarianism is the degree to which an organization or a society minimizes gender
role differences while promoting gender equality.’
‘Assertiveness is the degree to which individuals in organizations or societies are assertive,
confrontational, and aggressive in social relationships.’
‘Performance Orientation is the degree to which an organization or society encourages and
rewards group members for performance improvement and excellence.’
‘Humane Orientation is the degree to which individuals in organizations or societies encourage
and reward individuals for being fair, altruistic, friendly, generous, caring, and kind to others.’
Downloaded from oss.sagepub.com at UQ Library on June 23, 2014