The Hofstede Model in Context

In 1949 U.S. psychologist Raymond Cattell published an application of the new statistical technique of factor analysis to the comparison of nations. Cattell had earlier used factor analysis for studying aspects of intelligence from test scores of individual students. This time he took a matrix of nation-level variables for a large number of countries, borrowing from geography, demographics, history, politics, economics, sociology, law, religion and medicine. The resulting factors were difficult to interpret, except for the important role of economic development. Replications of his method by others produced trivial results. More meaningful were applications to restricted facets of societies. U.S. political scientists Phillip Gregg and Arthur Banks studied aspects of political systems; U.S. economists Irma Adelman and Cynthia Taft Morris studied factors influencing the development of poor countries, and Irish psychologist Richard Lynn studied aspects of mental health. 

In the 1970s this author – more or less by accident – got access to a large survey database about values and related sentiments of people in over 50 countries around the world. These people worked in the local subsidiaries of one large multinational corporation: IBM. Most parts of the organization had been surveyed twice over a four-year interval, and the database contained more than 100,000 questionnaires. Initial analyses of the database at the level of individual respondents proved confusing, but a breakthrough occurred when the focus was directed at correlations between mean scores of survey items at the level of countries. Patterns of correlation at the country level could be strikingly different from what was found at the individual level, and needed an entirely different interpretation. One of the weaknesses of much cross-cultural research is not recognizing the difference between analysis at the societal level and at the individual level; this amounts to confusing anthropology and psychology. From 180 studies using my work reviewed by Kirkman, Lowe, and Gibson, more than half failed to distinguish between societal culture level and individual level differences, which led to numerous errors of interpretation and application. 

My hunch that the IBM data might have implications beyond this particular corporation was supported when I got the opportunity to administer a number of the same questions to nearly 400 management trainees from some 30 countries in an international program unrelated to IBM. Their mean scores by country correlated significantly with the country scores obtained from the IBM database. So it seemed that employees of this multinational enterprises – a very special kind of people – could serve for identifying differences in national value systems. The reason is that from one country to another they represented almost perfectly matched samples: they were similar in all respects except nationality, which made the effect of national differences in their answers stand out unusually clearly. 

Encouraged by the results of the country-level correlation analysis I then tried country-level factor analysis. The latter was similar to the approach used earlier by Cattell and others, except that now the variables in the matrix were not indices for the country as a whole, but mean scores and sometimes percentages of survey answers collected from individuals in those countries. Analyses of data at higher levels of aggregation are called ecological. Ecological factor analysis differs from the factor analysis of individual scores in that a usual caution no longer applies: the number of cases does not need to be (much) larger than the number of variables. The stability of the results of an ecological factor analysis does not depend on the number of cases, but on the number of individuals whose scores were aggregated into these cases. Ecological factor analysis may even be performed on matrices with fewer cases than variables. 

Factor analyzing a matrix of 32 values questions for initially 40 countries, I found these values to cluster very differently from what was found at the individual level. The new factors revealed common problems with which IBM employees in all these societies had to cope, but for which their upbringing in their country presented its own profile of solutions. These problems were: 

1. Dependence on superiors;
2. Need for rules and predictability, also associated with nervous stress; 
3. The balance between individual goals and dependence on the company; 
4. The balance between ego values (like the need for money and careers) and social values (like cooperation and a good living environment); the former were more frequently chosen by men, the latter by women, but there were also country differences. 

These empirical results were strikingly similar to the standard analytical issues described in Inkeles and Levinson's 1969 article. Dependence on superiors relates to the first, need for predictability to the third, the balance between the individual and the company to the conception of self, and the balance between ego and social values to concepts of masculinity and femininity, which were also classified under the second standard analytic issue. 

The four basic problem areas defined by Inkeles and Levinson and empirically supported in the IBM data represent dimensions of national cultures. A dimension is an aspect of a culture that can be measured relative to other cultures. The four dimensions formed the basis for my book Culture's Consequences. 

The main message of the 1980 book was that scores on the dimensions correlated significantly with conceptually related external data. Thus Power Distance scores correlated with a dimension from Gregg and Banks' analysis of political systems and also with a dimension from Adelman and Morris' study of economic development; Uncertainty Avoidance correlated with a dimension from Lynn and Hampson's study of mental health; Individualism correlated strongly with national wealth (Gross National Product per capita) and Femininity with the percentage of national income spent on development aid. The number of external validations kept expanding, and the second edition of Culture's Consequences lists more than 400 significant correlations between the IBM-based scores and results of other studies. Recent validations show no loss of validity, indicating that the country differences these dimensions describe are, indeed, basic and enduring. 

In the 1980s, on the basis of research by Canadian psychologist Michael Harris Bond centered in the Far East, a fifth dimension 'Long-Term versus Short-Term Orientation' was added. 

In the 2000s, research by Bulgarian scholar Michael Minkov using data from the World Values Survey allowed a new calculation of the fifth, and the addition of a sixth dimension. The six dimensions are labelled:

1. Power Distance, related to the different solutions to the basic problem of human inequality; 
2. Uncertainty Avoidance, related to the level of stress in a society in the face of an unknown future; 
3. Individualism versus Collectivism, related to the integration of individuals into primary groups;
4. Masculinity versus Femininity, related to the division of emotional roles between women and men; 
5. Long Term versus Short Term Orientation, related to the choice of focus for people's efforts: the future or the present and past. 
6. Indulgence versus Restraint, related to the gratification versus control of basic human desires related to enjoying life. 

Each country has been positioned relative to other countries through a score on each dimension. The dimensions are statistically distinct and do occur in all possible combinations, although some combinations are more frequent than others.

After the initial confirmation of the country differences in IBM in data from management trainees elsewhere, the Hofstede dimensions and country scores were validated through replications by others, using the same or similar questions with other cross-national populations. Between 1990 and 2002 six major replications (14 or more countries) used populations of country elites, employees and managers of other corporations and organizations, airline pilots, consumers and civil servants. 

In correlating the dimensions with other data, the influence of national wealth (Gross National Product per capita) should always be taken into account. Two of the dimensions, Individualism and small Power Distance, are significantly correlated with wealth. This means that all wealth-related phenomena tend to correlate with both these dimensions. Differences in national wealth can be considered a more parsimonious explanation of these other phenomena than differences in culture. In correlating with the culture dimensions, it is therefore advisable to always include the wealth variable. After controlling for national wealth correlations with culture usually disappear. 

Of particular interest is a link that was found between culture according to the Hofstede dimensions and personality dimensions according to the empirically based Big Five personality test. After this test had been used in over 30 countries, significant correlations were found between country norms on the five personality dimensions (Neuroticism, Extraversion, Openness to experience, Agreeableness and Conscientiousness) and national culture dimension scores. For example, 55% of country differences on Neuroticism can be explained by a combination of Uncertainty Avoidance and Masculinity, and 39% of country differences on Extraversion by Individualism alone. So culture and personality are linked but the link is statistical; there is a wide variety of individual personalities within each national culture, and national culture scores should not be used for stereotyping individuals.

Validating the dimensions is of course not only and not even mainly a quantitative issue. Equally important is the qualitative interpretation of what differences on the dimensions mean for each of the societies studied, which calls for an emic approach to each society, supporting the etic of the dimensional data.