Dr. Geert Hofstede™ Cultural Dimensions
The following Article and Editor's Note is reprinted courtesy of the European Business Forum, a joint initiative of the Community of European Management Schools and PricewaterhouseCoopers.
The EBFonline is an excellent resource for international managers, international business students, and business school teachers.
EBF’s Winter 2001/2002 edition carried a provocative article by Professors Paul Gooderham and Odd Nordhaug of the Norwegian School of Economics and Business Administration suggesting that cultural differences in Europe may be on the decline. The paper explicitly challenged the work of Professor Geert Hofstede, whose analysis of surveys conducted with IBM employees in 49 countries around 1967 and 1973 remains the benchmark for discussion of national cultures or values. The four basic dimensions to which Hofstede assigned an index value for each country - Power Distance, Uncertainty Avoidance, Individualism-Collectivism and Masculinity- Femininity - are descriptions of national norms and values which resonate among managers and students to this day. On the following pages we continue the debate with an uncompromising new critique of Hofstede’s methodology from Professor Brendan McSweeney and a personal commentary from Professor Mikael Søndergaard.
of National Cultural Differences and Their Consequences: A
triumph of faith - A failure of analysis" (abridged)
Geert Hofstede’s depiction of enduring and powerful national cultures or national cultural differences is legendary. If his findings are correct they have immense implications for management within and across countries, and for the future of nation states - including the prospects for greater European integration. However, closer examination of his research reveals that it relies, in my view, on fundamentally flawed assumptions. This article examines four crucial assumptions upon which his measurements are based. These assumptions are ‘crucial’ in the sense that each is necessary for the plausibility of his identification claims. It is argued that they are all flawed and that therefore his national cultural descriptions are invalid and misleading.
Hofstede generalises about the entire national population in each country solely on the basis of analysis of a few questionnaire responses. The respondents were simply certain categories of employees in the subsidiaries of a single company: IBM. What evidence does he have that they were nationally representative? None. He just assumes it. Sometimes he supposes that every individual in a nation shares a common national culture. At other times he claims to have found in the IBM data a "national norm" or "central tendency", or "average [national] tendency." Both claims are problematic.
A statement that every English person is violent, because on occasions some English football fans are violent, would be regarded by any rational person as absurd. Generalising about an entire national population on the basis of miniscule number of questionnaire responses - in some countries fewer than 100 - is equally absurd. If a national culture were common to all national individuals - and survey responses could identify those cultures - there would not have been significant intracountry differences in individuals’ questionnaire responses. But the IBM surveys within each country revealed radical differences in the answers to the same questions.
If somehow the "average tendency" of IBM employees in each country - constructed by statistical averaging of highly varied responses - is assumed to be nationally representative, and this is Hofstede’s assumption - then with equal plausibility, or rather equal implausibility, it must also be assumed that each Hofstedian average tendency was, and continues to be, the same as the average tendency in every other part of a country, in every company, tennis club, knitting club, political party, and massage parlour. The "average [national culture] tendency" in the New York City Young Marxist Club, for example, is (if Hofstede’s Assumption 1, above is believed) the same as in the Keep America White Cheer-Leaders Club in Smoky Hill, Kansas, USA, and amongst any other group, in any other part, of the USA. Is Hofstede’s assumption credible?
There are no evidence-based reasons for assuming that the average IBM responses reflected ‘the’ national average. Hofstede’s assumption is a mere leap of faith. It is not grounded in evidence. Furthermore, IBM subsidiaries demonstrably had many nationally atypical characteristics. These included: the company’s selective recruitment only from the ‘middle classes’; the frequent international training of employees; the technologically advanced and unusual characteristics of its products during the survey periods - which were before the development of the ‘personal computer’; the ‘frequent personal contacts’ between subsidiary and international headquarters staff; its tight, internationally centralised control; US ownership during a period in which foreign direct investment was new and controversial; and the comparatively young age of its managers. Furthermore, IBM employees diverged from the general population more in some nations than in others. For instance, during the time the survey(s) were undertaken, working for a non-family owned firm would have been much more unusual in Taiwan than in Britain. Hofstede does not demonstrate the national representativeness of what he claims to have found in each IBM subsidiary is nationally representative. He asserts it. What is said to have been identified is actually presupposed. Hofstede’s reasoning is circular - he begs the question.
Suppose that by sheer chance he did find national averages - what use would they be? As managers, employees, investors, tourists, citizens, or whatever we do not engage with statistical averages but with real local specifics - as the statistician who drowned in the river whose average depth was five centimetres unfortunately discovered.
How does Hofstede claim to have identified national culture in IBM subsidiaries? As well as supposing the existence of such cultures, and the typicality of what he could identify in that company, he had to make a number of other and equally implausible assumptions in order to be able to assert that he could describe, indeed measure, these cultures.
Three discrete cultures: Out of the potentially huge number of cultural and non-cultural influences on the questionnaire answers, Hofstede assumed that only three: organisational (OrC); occupational (OcC) and national cultures (NC), were significant. Each respondent was conceived of as exclusively carrying - as being permanently "mentally programmed" by - these three non-interacting cultures (or values). This extraordinarily reductive conception of IBM employees conveniently allowed him to argue that as there was only one IBM culture, and as he occupationally matched the respondents, the questionnaire response differences showed "national culture with unusual clarity."
Hofstede’s anorexic and mechanistic assumption can be seen from its expression below as an equation:
Convenient for processing questionnaire answers, but unrelated to reality. Was there really just one monopolistic organisational culture in IBM world-wide? Does every occupation have a single global culture?
Organisational culture: The principal problem for Hofstede’s analysis is not that he supposed that there was a single worldwide IBM organisational culture - albeit that is contestable, and not self-evident as he suggested - but he treats that culture as the only organisational culture in IBM. He ignored extensive literatures which argue for recognition of multiple, dissenting, emergent, organic, counter, plural, resisting, incomplete, contradictory, fluid, cultures in an organisation. If the assumption of a single and monopolistic IBM culture is rejected and the possibility of a host of diverse cultural and/or noncultural influences on the questionnaire responses is acknowledged, Hofstede’s underlying equation collapses.
About ten years after the initial publication of his analysis of the IBM survey data, Hofstede had begun to acknowledge that there is cultural variety within and between units of the same organisation. An acceptance that organisations have multiple cultures and not a single culture would seem to undermine a crucial, and much trumpeted, part of his analysis: that all respondents were from the same company and therefore had the same organisational culture. However, Hofstede never admits error or weaknesses in his analysis. In parallel with his acknowledgement of cultural heterogeneity in organisations, Hofstede redefined ‘organisational culture’. He stated that "national cultures and organisational cultures are phenomena of a different order." Thus he concludes that the cultural heterogeneity within IBM did not affect his cross-IBM-subsidiary comparison of values, as organisational culture does not contain/reflect values. His definitional change is problematic for many reasons (see McSweeney, 2002).
By expediently taking organisational cultures out of his definition, Hofstede’s methodology becomes even more reductive and unreal as can be seen in the revised equation which excludes organisational culture:
Occupational culture: Hofstede’s notion of uniform world-wide occupational cultures supposes early and permanently-imprinted socialisation. Occupational cultures, which he also calls "values" (and indeed national culture/values also), are, he claims, enduringly "programmed into" carriers in pre-adulthood. "Values", he states: "Are acquired in one’s early youth, mainly in the family and in the neighbourhood, and later at school. By the time a child is ten years old, most of its basic values have been programmed into its mind... for occupational values the place of socialization is the school or university, and the time is in between childhood and adulthood."
The notion that national and occupational cultures are unchanging and finalised consequences of early ‘socialization’ has few supporters. Even Talcott Parsons, the high priest of socialisation, was less rigid. And yet the continuity assumption is crucial for Hofstede’s analysis. Without it, the mere matching of respondents on an occupational basis could not plausibly be deemed to isolate national cultural values.
Hofstede’s deterministic notion of permanent programming of "occupational cultures" conveniently, but ridiculously, supposes that throughout the world members of the same occupation (plumbers, electricians, clerks, pimps, accountants, or whatever) regardless of nationally diverse entry requirements, regulations, educational backgrounds, training requirements, examinations and differences in their social status, variations in the numbers and types of trade associations or professional bodies, post-experience course requirements and content, and so forth, each share an identical and monopolistic world-wide occupational culture.
Many other factors - cultural and noncultural - could have influenced IBM employee responses. To take but one. The questionnaires were not designed to identify national cultures. They were constructed to enhance senior management’s interventions at a time of worry within the company about morale. Some years later Hofstede used these non-independent, company-administered, sometimes nonconfidential, ‘second-hand’ questionnaires for his analysis. Respondents had foreknowledge that: their managers were expected to develop strategies for corrective actions which the survey showed to be necessary. Is this not likely to have encouraged them to manipulate their answers to improve their own, and their divisions’, position, resources, remuneration, and so forth? Yet, Hofstede relies on the supposition that the answers were immune to respondents’ gaming, uninfluenced by the possible consequences of their answers. He treated the answers as the pure outcomes of their unconscious preprogrammed values.
Furthermore, it would have been remarkable if the analysis of employee responses classified on the basis of their national location had not produced response differences. Every conceivable classification of the questionnaire responses would produce differences. But what would be the representational status of those statistically conjured differences?
The IBM questionnaire answers could have been categorised, in ways which reflected possible response differences additional, or alternative to, nationality - for example - race, religion, first language, type of car, or hair colour. The problem for Hofstede’s analysis is that each of these classifications would produce response differences, and yet using the same methodology he employed, each could be deemed to have been caused by, and the means of identifying, a particular ‘culture’ or cultural difference on the basis of whatever a priori classification framed the data stratification. But Hofstede ignores this classification problem. Nationally classified data provides no evidence of either the influence or identifiability of national cultures. As Jim March observes: We can easily fool ourselves into believing that we know something simply because we have a label for it (1966:69).
Despite the criticisms above, even if it is assumed that Hofstede managed to isolate unique aspects of national cultures, it takes another non-evidence-based leap of faith to conclude that he was able to construct adequate depictions of national cultures or national cultural differences. Were the questions asked wide-ranging and deep enough? The consequences of not having comprehensively ‘identified’ national value sets is not merely incomplete descriptions, but more importantly inaccurate descriptions. Restricted questions/answers would miss influential values that might counterbalance or outweigh the values that were measured, so the resulting depictions of national cultures would be distorted. As the questionnaires were not designed to identify national cultures it is likely that the questions were not adequate for that purpose. Researchers who have asked questions different from those used by Hofstede provide evidence of this flaw. Instead of revealing the same dimensions and ranking of those dimensions, they have indeed usually produced quite different descriptions of specific national cultures. Shalom Schwartz (1994), for instance, found seven culture-level dimensions which were "quite different" from Hofstede’s. Even if it is crudely supposed that a national culture is somehow composed of separately identifiable independent dimensions, why should we accept that Hofstede successfully identified even the "dominant" dimensions? Questionnaire answers are not neutral ‘windows’ through which national cultures can be perceived.
Hofstede’s depicts his dimensions of national culture as bi-polar in the sense that each is composed of contrasting positions, for instance ‘individualism’ versus ‘collectivism’. This is also problematic - the two can coexist and are simply emphasised, more or less, depending on the situation. Every society presents a number of contradictory adages or sayings such as "look before you leap" and "he who hesitates is lost" as part of its repertoire. But Hofstede’s dimensions exclude such coexistence and are thus blind to key cultural qualities. That is not to argue that the dimensions which Hofstede used - but did not originate - or those of others, cannot usefully frame initial discussion about national particularities, to alert, or remind us, that the world is not culturally homogeneous. But Hofstede claims much more, too much more: to have been able to use those dimensions to identify and compare, indeed measure, the dominant dimensions of unique, enduring, and systematically causal cultures in numerous countries.
Such is the elusiveness of the concept of culture that in the wider literature on the subject there is no consensus about which ‘units’ or ‘dimensions’ should be used for describing it: essentially cultures are still ‘grasped’. Hofstede’s arithmetisation of some employees’ answers to survey questions is not equal to the task.
Hofstede assumes that what he identified within a workplace is situationally nonspecific. It is the same in the courtroom, on the sports field, in the bedroom, everywhere. But we all know managers who are ‘rotweilers’ at work and ‘pussycats’ at home, and the converse. We may often be individualistic within office politics, but usually act in a collectivist way on behalf of our organisations.
The IBM data was effectively restricted to the workplace. Other sections of national populations - the unemployed, full-time students, the self-employed, the retired, homeworkers, and others - were ignored. The questions were almost exclusively about workplace issues, were completed in the workplace and not replicated in other types of location.
Despite this, Hofstede claims to have identified situationally unrestricted national cultures. On what grounds does he do so? Yet again, it’s a mere assertion, not an evidencebased conclusion. That which should have been explored was conveniently, but inappropriately, presumed.
In summary, the validity of Hofstede’s national culture identification claims face two profound problems.
First, the generalisations about national level culture from an analysis of small subnational populations necessarily relies on the unproven, and unprovable, supposition that within each nation there is a uniform national culture and on a mere assertion that micro-local data from a section of IBM employees was representative of that supposed national uniformity.
Secondly, the elusiveness of culture. It was argued that what Hofstede ‘identified’ is not national culture, but an averaging of situationally specific opinions from which dimensions or aspects, of national culture are unjustifiably inferred. Even if we heroically assume that the answers to a narrow set of questions administered in constrained circumstances are ‘manifestations’ of a determining national culture, it requires an equally contestable act of faith to believe that Hofstede’s overly contrived methodology successfully identified those cultures.
Hofstede has also sought to demonstrate the validity of his findings by accounts of historical and contemporary events which he argues are explicable by, and are a consequence of, some, or all, of his dimensions of national culture. He lists a wide range of national institutions, events, and artefacts - including ‘architecture’, ‘religion’, ‘literature’, ‘industrial relations systems’, ‘family structures’, ‘religious organisations’, ‘scientific theories’, and ‘social stratification’ which he claims are ‘consequences of ’, or ‘crystallisations of ’, national cultures. Indeed he seems to suggest that the list is unlimited: "no part of our lives", he states, "is exempt". Unfortunately, the same chronic a priorism that fundamentally flawed his measurement of national cultures also invalidates his allegedly illustrative stories. They are often constructed without regard for readily available counter-evidence. The problem for Hofstede’s model is not that some, indeed many, of his stories fail when tested. Useful explanatory/predictive theories can suffer from exceptions, albeit that the greater the number of explanatory failures the less useful the theory. Much more problematic for the validity of Hofstede’s model, is the manner of the stories’ construction. Hofstede’s illustrative stories are fabricated, no doubt unwittingly, to vindicate not validate his findings.
Understanding the properties and prospects of nations requires openness to the richness and diversity of national practices and institutions. Hofstede’s reductive, closed, single-cause analysis that he has legitimated with one-sided proofs and implausible assumptions should no longer be a serious candidate for attention. Instead of seeking an explanation for assumed national uniformity from the conceptual lacuna of an essentialist notion of national culture, we need to engage with and use theories of action which can cope with change, power, variety, and multiple influences - including the nonnational, and the non-cultural. Let’s be radical, let’s be realistic in the twenty first century and try and distinguish between management theories which are genuinely developed from, and tested by, evidence from those for which ‘evidence’, even if unintentionally, merely legitimates what their advocates somehow already ‘knew’.
This article is an abridged version of B. McSweeney, 2002, Hofstede’s ‘Model of National Cultural Differences and Consequences: A Triumph of Faith - A Failure of Analysis, Human Relations’, 55.1, 89-118. Three academic specialists anonymously refereed that article. A draft copy of the Human Relations article was sent to Professor Hofstede offering him an opportunity to respond in the same edition of that journal. He declined to do so.
Hofstede’s main descriptions of his national culture research are in Culture’s Consequences, 1st edition (1980), abridged version (1984), second edition (2001) all published by Sage and in Cultures and Organizations: Software of the Mind, McGraw- Hill, (1991).
March, J., 1966, ‘The Power of Power’, in D. Easton (ed.) Varieties of Political Theory, Englewood Cliffs: Prentice Hall, 39-70.
Schwartz, S. H., 1994, ‘Beyond Individualism/Collectivism: New Cultural Dimensions of Values’., in U. Kim, H. C. Triandis, C. Kagitcibasi, S-C. Choi, and G. Yoon (Eds.) Individualism and Collectivism: Theory, Methods and Applications, London: Sage, 85-119.
Brendan McSweeney is a full professor and Director of Research at the Department of Accounting, Finance and Management, University of Essex, England.
The above Article and Editor's Note is reprinted courtesy of the European Business Forum, a joint initiative of the Community of European Management Schools and PricewaterhouseCoopers.
The EBFonline is an excellent resource for international managers, international business students, and business school teachers.
Click on country
The International Business Center recognizes the international scope of the life’s work of Dr. Stephen Guisinger, Professor at the School of Business, University of Texas at Dallas, who passed away at much too young an age.