Int. J. Intercultural Rel. Vol. 20. No. 2. no. 189-198.

1996
&.

Copyright @ 1996 Elsevier Science Ltd

Pergamon Printed in Great Britain. All rights reserved
0147-1767/96 $15.00 + .OO

SO147-1767(96)00003-X

RIDING THE WAVES OF COMMERCE: A TEST OF

TROMPENAARS MODEL OF NATIONAL
CULTURE DIFFERENCES

GEERT HOFSTEDE

Institute for Research on Intercultural Cooperation,

Maastricht and Tilburg, The Netherlands

ABSTRACT. Trompenaars (1993) presents a seven-dimensional model of national

culture d@erences which he argues is particularly relevant to the conduct of inter-
national business. Data readfrom his book were treated using correlation andfactor
analysis at the country level. Results indicate that only two dimensions can be clearly
confirmed statistically: Individualism/Achievement and UniversalismlDtJiise. Both
are correlated with Hofstedes Individualism dimension. The re-analysis raised
concerns about Trompenaars conclusions and about his methodology. It is argued
that the theory in the book is not supported by the database. Suggestions are made
on how the database could be validly studied. However, the evident lack of content
validity of the instrument used will remain a major concern. Copyright @ 1996
Elsevier Science Ltd.

TROMPENAARS MODEL AND DATABASE

In Riding the Waves of Culture (London: The Economist Books, 1993)
Dutch management consultant Fons Trompenaars proposes a model of
seven fundamental dimensions of (national) culture for understanding
cultural diversity in business (p. 8). The first five factors describe relation-
ships with other people. They are: universalism versus particularism,
individualism versus collectivism, neutral versus emotional, specific versus
diffuse, and achievement versus ascription. The remaining two dimensions
are orientation in time and attitudes towards the environment.
In three opening chapters, Riding the Waves of Culture introduces the
culture concept, and relates differences in national culture to differences in
ways of organizing. Chapters 4-10 describe the seven dimensions.

The author thanks Philip Lincoln and Frank van Baren for their help in data extraction and
processing. Reprint requests should be sent to MS Ingrid Regout, IRK, University of
Limburg, P.O. Box 616, NL-6200 MD Maastricht, The Netherlands.

189
190 G. Hofstede

Chapter 11 deals with corporate cultures, and Chapter 12 offers a

perspective for international management.
In the book, Trompenaars illustrates his arguments with his own
questionnaire data. His database covers about 15,000 respondents. Some
were participants in the authors cross-cultural training programs; others
were employees in 30 companies in 50 different countries. Of the respon-
dents, 75% belong to management, while the remainder are general
administrative staff, mainly female. The book does not mention whether
the questionnaire was translated or administered in English.
The questionnaire was originally designed for the authors doctoral
dissertation (Trompenaars, 1985). The dissertation used data obtained
from a 79-item survey across 653 respondents, divided over nine
countries, two industries (oil and hosiery), and seven job categories (from
unskilled labor to managers of managers).
The survey questionnaire used in the dissertation research already
consisted of seven subscales, corresponding to the seven dimensions of
Trompenaars later work, plus demographics. For four subscales the
items were taken from existing instruments described in the (U.S.) social
science literature; for the remaining three the author designed new items.
Items included long questions describing a dilemma situation and asking
the respondent to choose a solution (several of these were taken from
Stouffer and Toby, 1951), as well as shorter questions asking the
respondent to choose between two alternative statements.
The thesis concluded that the nine countries studied could be divided
into two types: Left Brain (U.S.A., Netherlands, Sweden, Austria, and
Greece) and Right Brain (Venezuela, Spain, Italy, and Singapore).
Country scores on all subscales were correlated with this distinction, so
that Left Brain cultures were at the same time universalist, individualist,
neutral, specific, attributing status by achievement, future-oriented, and
dominating nature, while Right Brain cultures were the opposite.
The database used in the 1993 book is based on the same questionnaire,
or a revised version of it, but with a much larger group of respondents.
The analysis, however, no longer refers to Left or Right Brain cultures. In
the various chapters of the book the answers on 17 questions from the
questionnaire are shown (16 are shown in bar charts as percentages by
country, and one is shown in a country ranking). These answers are not
summarized into country scores on the seven dimensions, however, so it is
not clear where exactly a country is supposed to be positioned on a
dimension; nor are country scores on the dimensions validated against
measurements independent of the research, either country-level data (like
GNP/capita), or survey-based (like the Eurobarometer public opinion
data in the countries of the European Union).
Methodological aspects of the questionnaire survey are discussed in a
three-page Appendix (pp. 179- 18 1) by Dr Peter B. Smith of the University
 Trompenaars Model of National Culture D@erences 191

of Sussex in the U.K. Smith analysed the scores for the questionnaire items
at the country level for 47 countries with more than 50 respondents each.
The original 79-item questionnaire was reduced to 57 items to improve
reliability, measured by Cronbach alpha. The Appendix listed alpha
values for six of the seven subscales. The seventh, Time Orientation, could
not be evaluated this way, because it did not comprise a series of free-
standing items. Smith also showed a table of correlations among the seven
scales (p. 181). Correlations with Time Orientation are insignificant; the
other six scales all show significant intercorrelations. In Trompenaars
1985 dissertation the seven categories were all correlated with his Left/
Right Brain typology, also suggesting substantial intercorrelations. With
the extended database, this still appears to be the case.

CORRELATIONS AMONG
TROMPENAARS DIMENSIONS
Commenting on the significant linkages between several of the scales,
Smith (p. 180) comments:

The fact that some of these correlations are quite substantial does not
necessarily imply that separating out the different dimensions is
unnecessary. Significant country-level correlations between, for instance,
universalism and achieved status indicate only that both orientations are
high in a particular national culture, and not that they are necessarily
endorsed by the same individuals or within the same organizations.

This is a puzzling comment. Its first part is a partial truth. Significant

country-level correlations between subscales do indicate that both
orientations are high in particular national cultures, but they also indicate
that both are low in other national cultures: in fact they mean that if you
know one orientation for a country, you can fairly accurately predict the
other. The orientations may be conceptually distinct in the researchers
mind, but the empirical data show that in the real world, they usually
occur together.
The second part of the statement (orientations are not necessarily
endorsed by the same individuals etc.) is a truism, but as an argument for
separating out highly correlated dimensions at the country level it is a non
sequitur. It confuses the individual with the country level of analysis.
Individual-level, organization-level, and ecological (country-level) corre-
lations are all different things. Distinguishing between these levels is an
essential condition for cross-cultural research. Trompenaars book is
about differences between countries, not between individuals. It describes

It is not stated whether the Cronbach alpha scores presented are based on individual or on
ecological, that is country-level, scores, but in the 1985 thesis the reliability of the subscales
was assumed based on correlations of items across individual respondents.
192 G. Hofstede

(country-level) cultures, not (individual-level) personalities. Whatever one

does at the country level should be supported with arguments applicable
at that level.
Trompenaars country data have not been published before, but one
can read country scores on some of the questions from the diagrams
published in the 1993 book. Doing this, one arrives at a 17 (questions) x
39 (countries) matrix, which can be subjected to a question x question
correlation analysis. This therefore is an ecological correlation analysis.
As stated earlier, each of the seven dimensions in Trompenaars model
has a chapter in the book devoted to it. Each of these chapters (except the
one on Time Orientation) contains one or more diagrams showing the
percentage of respondents in each of the countries studied who gave a
particular answer. The questions thus serve to illustrate the meaning of
the dimensions. Three additional questions are shown in the chapter on
corporate cultures.
If the seven dimensions are both internally coherent and mutually
distinct, we should expect that questions cited within the same chapter are
strongly intercorrelated (at the country level), but questions cited within
different chapters are less strongly correlated. Table 1 tests to what extent
this is true. The actual correlation pattern does not support the attribution
of the questions to the chapters (dimensions) in the book. In all cases,
questions cited within a chapter are more strongly correlated with some
questions from other chapters than with the other questions cited within
the same chapter.
A 17 x 39 ecological correlation matrix begs some kind of simplifica-
tion by multivariate methods. I performed an ecological factor analysis
with orthogonal varimax rotation (which means that the factors to be
extracted will be statistically independent). In ecological factor analysis
the usual caution against small numbers of cases does not apply: the
reliability of the factors does not depend on the number of countries, but
on the number of individual responses integrated into the country scores,
which is nearly always more than sufficient.
The results of the factor analysis are shown in Table 2. The scree plot
shows that a maximum of six factors can be extracted (eigenvalues
over l), but that the last two factors explain hardly more than the original
variables. Not surprisingly the first factor is very strong, accounting for
40% of the total variance: this is because many of the questions are
intercorrelated. The influence of the first factor is reduced by the rotation
process which reshuffles the original factors so as to separate the variables
into groups. Successive rotation with 2, 3, and 4 factors showed the
4-factor solution to produce the best separation of variables. These four
factors together explain 72% of the variance.
The factors extracted after rotation are listed in the lower part of
Table 2. Factor 1 mainly combines items from Chapter 5 (individualism)
 Trompenaars Model of National Culture DifSerences 193

TABLE 1

Correlations among Trompenaars Questions In Relatlon to the Chapter in which they

are Clted

Correlations among items

Highest correlation
Number of with items in
Chapter Dimension items Same chapter other chapters

4 Universalism 3 .70 .77 (ch. 7)

.60 .74 (ch. 7)
.49 .73 (ch. a)
5 Individualism 3 .50 .63 (ch. 4)
.47 .51 (ch. 7)
.12 .57 (ch. a)
6 Neutral/emotional .60 (ch. 7)
7 Specific/diffuse .72 .77 (ch. 4)
.74 (ch. 4)
a Achievement .65 .73 (ch. 4)
.75 (ch. 11)
9 Time
10 Environment .02 .57 (ch. 4)
.58 (ch. 7)
.56 (ch. 8)
.64 (ch. 11)
11 Corporate cultures 3 .57 .70 (ch. 4)
.26 .66 (ch. 7)
.03 .75 (ch. a)
.64 (ch. 10)

with items from Chapter 8 (achievement); Factor 2 combines items from

Chapter 4 (universalism) with Chapter 7 (specific/diffuse) items. Factor 3
is a weak factor dealing with concepts of organization, and Factor 4 is a
residual factor dealing with the issue of controlling nature (Chapter 10);
the loading for express feelings is puzzling, but this question has only
been used in 10 out of 39 countries.
Thus the empirical analysis of Trompenaars own data provides only
limited support for his seven-dimensional model. At best, three separate
dimensions really appear, plus maybe an organization factor which was
not presented as a dimension. If we forget for a moment about Time
Orientation which was not included in the analysis, the minds of
Trompenaars respondents distinguished at maximum four separate issues
which recombined the original categories.
Now that we have empirical factors, we can also look for external
validation of these factors. I have computed approximate country scores
on the four factors (taking the two highest loading items for Factors l-3
and the single highest loading item for Factor 4), and correlated these with
194 G. Hofstede

TABLE 2

Results of an Ecologlcal Factor Analysis of 17 Questions across 39 Countrles

Wee plot
Factor Eigenvalue % of variance Cumulative

1 6.8 39.9 39.9

2 2.2 13.2 53.0
3 1.7 10.3 63.3
4 1.4 8.3 71.6
5 1.2 7.1 78.7
6 1.1 6.6 85.3

Factor loadings
Questiona Loading Issue

Factor 1
5:48 .85 Individual freedom
8:95 .79 Acting as suits you
4:39 .75 Would not tip off a friend
8:96 .71 Respect not based on family
11: 144 .69 Low hierarchical triangle
5:52 .64 Individual decisions
5154 .56 Individual responsibility

factor 2
4:37 .85 Would not write false review
4:35 .74 Would not give false testimony
7:86 .69 Company should not provide housing
11: 143 .68 Leader not seen as a father
7:80 .64 Refuse to paint boss house
10 : 128 .59 What happens to me is my own doing

Factor 3
2:18 .72 Company is system rather than social group
11: 150 .66 Function rather than personality

Factor 4
10 : 127 .83 It is worth trying to control nature
6164 - .57 (reversed) Would express feelings when upset

Chapter and page numbers as in Trompenaars (1993).

the scores on the five empirical dimensions of national cultures found in

my own research (Hofstede, 1991). Thirty-five countries are represented in
both studies, but because of missing items, scores on Factor 1 are only
available for 23 of these countries, on Factor 2 for 26 countries, on
Factor 3 for 34 countries, and on Factor 4 for 30 countries. Scores on the
Hofstede dimension of Long Term Orientation are available for 16 of the
countries only.
 Trompenaars Model of National Culture Differences 195

TABLE 3

Correlations between Four Trompenaars Factors and Five Hofstede Dlmenslons

Jrompenaars (1993)
Factor 1 Factor 2 Factor 3 Factor 4

Hofstede (1991):
Power distance - .52b -.70a
Individualism .64a .82a
Masculinity
Uncertainty avoidance - .53b
Long term orientation - .6gb -.66a

%ignificant at .Ol-level; bsignificant at .05-level; significance limits vary according to the

number of available cases.

The results are listed in Table 3. Factors 1 and 2 appear to be correlated

with both my Individualism and small Power Distance scores, which is
remarkable because Factors 1 and 2 are mutually independent.* The
correlations are strongest for Factor 2. Factor 1 is also associated with
weak Uncertainty Avoidance, and more than Factor 2 with Short Term
Orientation. There are no correlates of Masculinity/Femininity in Trom-
penaars data, nor are there correlates of his Factors 3 and 4 among my
dimensions. However, these last two factors are only based on two
questions and one question respectively, which is unlikely to give them
sufficient reliability for correlating significantly with outside measures.
To conclude, the above comparison between two independent studies
shows that Trompenaars questionnaire measures mainly what I called
Individualism; the large number of intercorrelated items dealing with this
concept can be split into two groups, one combining Trompenaars Indi-
vidualism with Achievement, and the other Universalism with Specific/
Diffuse. The other external correlations found point to possible additional
connotations in Trompenaars questions that he has not explored.

A CRITIQUE OF TROMPENAARS APPROACH

It is evident that Trompenaars confuses conceptual categories with
dimensions. Conceptual categories are present in the mind of any investi-
gator who sets out to do research. They belong to the culture of the person
or persons who designed them-in Trompenaars case, American
sociologists and anthropologists of the 1950s and 1960s.

ZThe approximated factor scores are not quite independent, both because they were
computed on two items only and because of missing data on these items. The intercorrelation
of the approximated factor scores for Factors 1 and 2 is .48, not significant. The approxi-
mated factor scores for Factor 3 show marginally significant correlations of .54 and .51 with
Factors 1 and 2, respectively.
196 G. Hofstede

The origin of the first five of Trompenaars dimensions is the

General Theory of Action by functionalist sociologist Talcott Parsons,
published with the co-authorship of Edward Shils (1951). These authors
labelled the dimensions pattern variables. Trompenaars Individual-
ism versus collectivism was called by Parsons Self-orientation versus
collectivity-orientation; it only occurs in the earlier versions of his list of
pattern variables. The term Individualism versus collectivism as a
cultural qualifier was introduced by the present author (Hofstede, 1980;
Kim, Triandis, Kagitcibasi, Choi & Yoon, 1994). Parsons theory was
speculative; it was one scholars interpretation of reality as he perceived it,
guided by a strong belief that all social phenomena should serve a
function. This kind of philosophy was rooted in American society of the
1940s. I know of no research supporting Parsons claim that these pattern
variables determine all human action, if such a claim could ever be
supported.
The other two of Trompenaars dimensions are taken from a book by
anthropologists Florence Kluckhohn and F. L. Strodtbeck (1961). Their
classification of five value orientations was inspired by a field study of
five geographically close, small ethnic or religious communities in south-
western U.S.A. The five orientations are: (1) human nature, from evil to
good; (2) relationship to the environment; (3) orientation in time; (4)
orientation toward activity, and (5) relationships among people, from
lineal to collateral. From these, Trompenaars took the second and third
only.
The purpose of research is to replace preconceived notions with
empirical findings. The choice of a research instrument is obviously
always influenced by existing theory and preconceived notions. By
exposing this instrument to the responses of a research population, data
are collected that should serve to revise the preconceived notions and
formulate a new theory.
In his dissertation, Trompenaars has gone through this process.
Unfortunately, his research samples were small and poorly matched;
also, a number of nine countries is statistically insufficient to develop a
multidimensional model. The empirical model that he derived was the
simplest possible, but probably the only one that his limited database
would allow: one single dimension, and this again simplified into two
opposite ideal types, Left Brain and Right Brain cultures. The labels are
unfortunate, and rather out of fashion now. Regardless of the labels, the
typology is not very useful for understanding cultural diversity.
Since 1985 Trompenaars has compiled much more data, which should
allow for a proper multidimensional analysis. This, however, has not yet
been done. The book has not resumed the conclusions of the dissertation;
it has gone back to the original seven preconceived dimensions. Apart
from showing some bar charts in his book-the ones I analysed -
 Trompenaars Model qf National Culture Dlferences 197

Trompenaars does not seem to have made any use of his database. Smiths
note in the books Appendix only scratches its surface. For one thing, if
six of the seven scales now have sufficient reliability,3 why does Trom-
penaars not show the scores of the countries in the data bank on these six
dimensions? That would be crucial information for readers wanting to
know about cultural diversity.
The data bank may still prove an unexplored treasure. Exploring it will
involve at least the following steps:
1. Clean the data. The data bank seems to be filled by convenience
samples, collected because they happened to be available, without
central control. Beyond errors of collection, translation, and proces-
sing, such a set contains unmatched country data that have to be
matched first.
2. Multivariate analysis at the ecological (country) level, as in the factor
analysis shown above. The particular statistical technique to be used is
a matter of taste.
3. Extract whatever sensible and robust dimensions arise from this
analysis, and reformulate the original framework accordingly.
4. Show the scores of all countries studied on these new dimensions.
5. Hypothesize what outside data, independent of this researcher, this
data bank, and this method, might correlate with the various dimen-
sions. Validate and reinterpret the dimensions according to the pattern
of correlations with outside measures found. As my quick analysis
shown above has demonstrated, Trompenaars data can produce
findings that correlate with mine, but they must be able to produce
other meaningful correlations with outside data.
A serious shortcoming of Trompenaars data bank which no profes-
sional analysis can correct is its evident lack of content validity. Content
validity is the extent to which an instrument covers the universe of
relevant aspects of the phenomenon studied, in our case national culture.
Trompenaars did not start his research with an open-ended inventory of
issues that were on the minds of his future respondents around the world;
he took his concepts, as well as most of his questions, from the American
literature of the middle of the century, which was unavoidably ethno-
centric. He did not change his concepts on the basis of his own findings

3Whether the subscales really have sufficient reliability for distinguishing between countries
remains to be proven, as it seems that Smiths reliability tests were done at the individual
level, and not at the country level.
41n my study, scores for 40 countries on four empirical dimensions were shown to correlate
significantly with nine other survey studies of narrow samples, six studies of representative
national samples, and 29 national indicators taken from economic, political, sociological,
psychological, and medical statistics (Hofstede, 1980, pp. 3266331). Since this was published,
the number of validations has continued to grow.
198 G. Hofstede

either, nor did he follow the development of the state-of-the-art in

comparative culture research since 196 1.
Trompenaars does, however, ride the waves of commerce: he tunes his
messages to what he thinks the customer likes to hear. In Parsons
functionalist scheme, the main source of Trompenaars model, there was
no place for dysfunctional and destructive elements. Therefore, in Trom-
penaars questionnaire and book, controversial issues central to cultural
conflicts, like power struggle, corruption, exploitation, aggression,
anxiety, and differing concepts of masculinity and femininity, are rarely
addressed. The result is a fast food approach to intercultural diversity and
communication.

REFERENCES
HOFSTEDE, G. (1980). Cultures consequences: International differences in work-
related values. Beverly Hills, CA: Sage.
HOFSTEDE, G. (1991). Cultures and organizations: Software of the mind.
London: McGraw-Hill.
KIM, U., TRIANDIS, H. C., KAGITCIBASI, C., CHOI, S. C., & YOON, G.
(Eds.) (1994). Individualism and collectivism: Theory, method, and application.
Thousand Oaks, CA: Sage.
KLUCKHOHN, F. R., & STRODTBECK, F. L. (1961). Variations in value
orientations. Westport, CT: Greenwood.
PARSONS, T., & SHILS, E. A. (1951). T oward a general theory of action.
Cambridge, MA: Harvard University Press.
STOUFFER, S. A., & TOBY, J. (1951). Role conflict and personality. The
American Journal of Sociology, 56, 5, 395-406.
TROMPENAARS, F. (1985). The organization of meaning and the meaning of
organization. Unpublished Ph.D. Dissertation, The Wharton School of the
University of Pennsylvania.
TROMPENAARS, F. (1993). Riding the waves of culture: understanding cultural
diversity in business. London: The Economist Books.