Hofstede’s model of national culture has enjoyed enormous popularity but rests partly on faith. It has never been fully replicated and its predictive properties have been challenged. The purpose of this paper is to provide a test of the model’s coherence and utility.
Analyses of secondary data, including the World Values Survey, and a new survey across 56 countries represented by nearly 53,000 probabilistically selected respondents.
Improved operationalizations of individualism-collectivism (IDV-COLL) suggest it is a robust dimension of national culture. A modern IDV-COLL index supersedes Hofstede’s 50 year-old original one. Power distance (PD) seems to be a logical facet of IDV-COLL, rather than an independent dimension. Uncertainty avoidance (UA) lacks internal reliability. Approval of restrictive societal rules and laws is a facet of COLL and is not associated with national anxiety or neuroticism. UA is not a predictor of any of its presumed main correlates: importance of job security, preference for a safe job, trust, racism and xenophobia, subjective well-being, innovation, and economic freedom. The dimension of masculinity-femininity (MAS-FEM) lacks coherence. MAS and FEM job goals and broader values are correlated positively, not negatively, and are not related to the MAS-FEM index. MAS-FEM is not a predictor of any of its presumed main correlates: achievement and competition orientation, help and compassion, preference for a workplace with likeable people, work orientation, religiousness, gender egalitarianism, foreign aid. After a radical reconceptualization and a new operationalization, the so-called “fifth dimension” (CWD or long-term orientation) becomes more coherent and useful. The new version, called flexibility-monumentalism (FLX-MON), explains the cultural differences between East Asian Confucian societies at one extreme and Latin America plus Africa at the other, and is the best predictor of national differences in educational achievement.
Differences between subsidiaries of a multinational company, such as IBM around 1970, are not necessarily a good source of knowledge about broad cultural differences. A model of national culture must be validated across a large number of countries from all continents and its predictions should withstand various plausible controls. Much of Hofstede’s model (UA, MAS-FEM) fails this test while the remaining part (IDV-COLL, PD, LTO) needs a serious revision.
Consultancies and business schools still teach Hofstede’s model uncritically. They need to be aware of its deficiencies.
As UA and MAS-FEM are apparently misleading artifacts of Hofstede’s IBM data set, a thorough revision of Hofstede’s model is proposed, reducing it to two dimensions: IDV-COLL and FLX-MON.
Minkov, M. (2018), "A revision of Hofstede’s model of national culture: old evidence and new data from 56 countries", Cross Cultural & Strategic Management, Vol. 25 No. 2, pp. 231-256. https://doi.org/10.1108/CCSM-03-2017-0033Download as .RIS
Emerald Publishing Limited
Copyright © 2018, Emerald Publishing Limited
Geert Hofstede is the author of one of the most influential treatises on national culture (Kirkman et al., 2006), originally published in a short form (Hofstede, 1980), followed by an expanded version (Hofstede, 2001). According to Bond (2002), cross-cultural psychologists were “held in thrall” (p. 73) by Hofstede’s intellectual achievement, whereas Peterson (2003) pointed out that Hofstede’s first book shaped the basic themes, structures, and controversies of the cross-cultural field for over 20 years. Hofstede popularized the nomothetic approach to the study of culture, subsequently employed by other leading researchers (for instance Inglehart and Baker, 2000; House et al., 2004; Schwartz, 1994, 2008, etc.). Their studies have proven the utility of this approach. But how accurate is the product that it yielded in Hofstede’s research? The answer to this question is long overdue. As the issue is complex and requires a lengthy analysis, a single paper cannot provide all answers. Yet, it can outline some general conclusions.
This study starts with an analysis of secondary (published) data. Then, it analyzes primary data from a survey of nearly 53,000 respondents selected probabilistically in 56 countries. The survey was partly designed to check the structure and replicability of Hofstede’s dimensions.
I remind the readers that Hofstede designed his model at the national level of analysis, not at the individual. This means that Hofstede’s model explains patterns that are observable when the agents are whole nations, not individuals. Attempts to transpose Hofstede’s model to the individual level would be what Hofstede (2001) and others call an ecological fallacy. Unfortunately, Brewer and Venaik (2014) and Winzar (2015) found that many authors of articles in leading journals continue to project cultural patterns onto individuals or organizations. Such attempts amount to expecting the laws of classical mechanics to apply at the sub-atomic level, where very different quantum physics laws are in force. Still, if the laws are different at different levels, the logic of the discrepancy needs to be explained. In one instance, discussed further in this paper, this seems to be a problem for Hofstede’s model.
Another caveat is also important. Baumann and Winzar (2017) point out that the extent to which values drive behavior is a function of the circumstances in which individuals find themselves as well as the relative importance of competing values in particular circumstances. Minkov’s (2017) work shows that this may be especially true in the East Asian societies (all those with a Confucian heritage except Vietnam). Nevertheless, this recently highlighted challenge in the cross-cultural field will be ignored in this study and Hofstede’s model will be analyzed on the basis of the prevalent conceptualization of culture at the time of its creation, assuming that culture can be studied through snapshots of static situations rather than motion pictures.
2. Analysis of secondary data
The analysis of secondary data should shed some light on four basic questions that entail validity tests for any model in social science or psychology and are therefore relevant in this study:
Did Hofstede’s database, consisting solely of the employees of the IBM Corporation around 1970, adequately reflect the national cultures of the respondents?
Do Hofstede’s dimensions replicate?
Do Hofstede’s dimensions have internal reliability? Are their facets really correlated as the Hofstede model postulates?
Do Hofstede’s dimensions have convincing predictive properties? Are they associated with relevant external variables in accordance with Hofstede’s theoretical expectations?
2.1 Reliability of Hofstede’s IBM database as a source of knowledge about national culture
In an analysis of data from the World Values Survey (WVS) (www.worldvaluessurvey.org), Minkov et al. (2015) split all the national samples into several occupational categories. They found that national samples of respondents from a single category (e.g. only experts) do not necessarily yield the same dimensions of national culture as samples from another category (e.g. only skilled manual workers) although the items in the analysis are the same. Smith et al. (2002) explain why some seemingly matched samples may not be equivalent and, consequently, may not yield the same results: government employment is appreciated in Japan, but not in western countries. Consequently, Japanese and western government employees would probably not provide equivalent samples for comparisons of national cultures because they do not have the same status in their home countries. Likewise, while teachers enjoy respect in China, their profession is considered low status in some East European countries, and there are even derogatory words for “teacher” and “school” in those societies. Thus, it is possible that employment at IBM around 1970 did not carry the same social status across the world. As a result, the national subsidiaries of IBM may have attracted dissimilar types of job applicants. National culture also may have affected actual recruitment procedures, despite IBM’s efforts to enforce universal global standards.
The IBM database has yielded dimensions of national culture called individualism vs collectivism (IDV-COLL) and power distance (PD) that are strongly correlated with national wealth (Hofstede, 2001). This external validation means that at least some of the measured differences between the IBM subsidiaries reflect actual societal differences and are not pure artifacts of IBM’s organizational cultures, employee selection, or other local factors. But did the IBM database correctly reflect national culture in every respect? Critics, such as McSweeney (2002), were not convinced.
One of the pillars of Hofstede’s masculinity-femininity (MAS-FEM) dimension is Hofstede’s (2001) finding that distances between the values of men and women are greatest in MAS nations, whereas FEM ones have smaller distances. In other words, men and women in FEM countries, such as the Netherlands and the Scandinavian countries, have the most similar scores on values, whereas men and women in MAS countries, such as Japan, have the most dissimilar scores. By 2007, it was well established that this was demonstrably wrong. Guimond et al. (2007) summarize the literature on that subject, including Schwartz (2005) and Costa et al. (2001). The summary shows that national differences in distances between the values and personality traits of men and women are not a function of MAS-FEM but of gender emancipation, underpinned by national wealth. In other words, these gender differences are strongly associated with IDV-COLL since gender emancipation is greatest in the most IDV countries. There are diverse explanations of the larger gender differences in values and personality in the IDV societies, one of which could be that people in such societies have greater freedom to express their individuality, whereas in COLL countries there is strong pressure for men and women to be the same in terms of values and personality. Whatever the right explanation is, far from having the smallest distances between the values of men and women, as the MAS-FEM theory claims on the basis of the IBM findings, the Netherlands and the Scandinavian countries have the largest distances. The values sections in the WVS confirm this. It is noteworthy that MAS-FEM is orthogonal (unrelated) to IDV-COLL and national wealth (Hofstede, 2001). This means that, in terms of national distances between men’s and women’s values, the IBM database could hardly have been more misleading than it was, as it showed a 90-degree deviation from reality. It is highly unlikely that the structure of male-female distances across the globe has changed so drastically since 1970 that the IBM revealed a true societal pattern back then, which did not reflect any geo-economic logic, whereas today we see a 90-degree shift to something that rests on the solid logic of economic differences between nations and their societal consequences. Much more plausibly, the societal pattern was the same in 1970, yet the IBM database was contaminated with IBM-specific peculiarities that made it an unreliable source of information for extrapolations to the societal level.
Further, the IBM data set has yielded an IDV-COLL index that assigns the English-speaking countries, and particularly the USA, to the top of the ranking. Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al. (in press) reviewed all large studies of national IDV-COLL or closely related constructs, and found that the English-speaking countries do not have top scores on anything related to IDV-COLL. Of note, some of these studies are based on the nationally representative WVS or probabilistic data sets that are close in structure to the national census of each country (Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press). The leading position of the USA on Hofstede’s IDV-COLL is doubtlessly an artifact of the IBM database, reflecting its national unrepresentativeness.
2.2 Replicability of Hofstede’s dimensions
Only one peer-reviewed publication in an indexed journal so far (Merritt, 2000) describes an attempt to replicate all of Hofstede’s IBM dimensions in a single study. While IDV-COLL and PD replicated reasonably well, UA and MAS-FEM did not. Following this failure, Merritt (2000) tried constructing an UA dimension and a MAS-FEM dimension with items that were statistically correlated with those two dimensions, even though they had no face validity. The outcome was confusing. Merritt (2000) did obtain measures that were highly correlated with IBM’s UA and MAS-FEM, yet MAS-FEM was composed of classic UA items, such as a feeling of nervousness and agreement that rules should not be broken, and a PD item (employees afraid to disagree with superior). Besides, instead of being unrelated to IDV-COLL, as in the IBM study, this new MAS-FEM, just like UA, was strongly correlated with it. In short, Merritt’s (2000) work showed that UA and MAS-FEM could not be replicated, suggesting that they are problematic dimensions.
Single-dimension studies have replicated IDV-COLL successfully (Gelfand et al., 2004; Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al., in press). This implies that if PD is seen as a facet of IDV-COLL (Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press), it also replicates. Schwartz’s (1994, 2008) work demonstrates that national aggregates of values in the domain of what he calls “hierarchy” are conceptually and statistically similar to PD. Thus, Schwartz’s work indirectly confirms the replicability of PD, yet without necessarily indicating that it is independent from IDV-COLL.
There are no studies in peer-reviewed journals focusing on the replicability of MAS-FEM. Hofstede (2001) reported that MAS-FEM is highly correlated with Schwartz’s (1994) measure of “mastery vs harmony.” Yet, Schwartz (2011) stated that he sometimes regretted his 1994 publication as researchers continued to cite it despite the existence of much more refined variants of his measures. Schwartz’s latest, unpublished mastery and harmony measures (personally provided by Schwartz in 2016), recommended by him for validation purposes, are unrelated to MAS-FEM.
Project GLOBE (Sully de Luque and Javidan, 2004) attempted to replicate Hofstede’s UA. Yet, GLOBE conceptualized and operationalized UA very differently, not as a combination of anxiety and a conviction that rules and laws must be followed strictly. GLOBE’s UA is reminiscent of Hall’s (1959) high-vs-low-context concept as it measures the degree to which people perceive their societies as having clearly explained rules or wish to have such rules. The first of these two GLOBE measures (UA practices) is a variant of IDV-COLL as it creates a more or less clear contrast between economically advanced countries (more explicit rule communication) and developing countries (more implicit rule communication), which conforms to Hall’s theory. GLOBE’s UA practices measure is correlated with Hofstede’s UA at −0.66 (p<0.001, n=47), suggesting that people in countries that Hofstede defines as strongly avoiding uncertainty and ambiguity actually describe their societies as having ambiguous rule communication. This confusing result can hardly be taken as a replication of Hofstede’s UA. GLOBE’s alternative UA measure – the degree to which people wish to have clear rule communication – correlates weakly with Hofstede’s UA at 0.32 (p=0.03, n=47). Again this cannot be seen as a replication of Hofstede’s UA by any standard.
Taras et al. (2012) reported that their meta-analysis of studies devoted to Hofstede’s dimensions, most of them done in a small number of countries at a time, yielded national indices that are reasonably well correlated with all of Hofstede’s IBM measures. The authors also report that the original IBM indices for IDV-COLL and PD remain highly correlated with the corresponding meta-analytical scores from the 1980s to the 2000s, However, measures of MAS-FEM and UA from the 2000s correlate with the Hofstede’s originals at only 0.56 and 0.46, respectively, suggesting that MAS-FEM and UA are unstable dimensions whose modern variants have little to do with their counterparts decades ago.
Taras et al. (2012) conceded that they relied on studies that had used not only Hofstede’s Values Survey Module but also a variety of other tools designed by various authors to measure Hofstede’s dimensions. Without detailed information about those unknown and untested tools, it is impossible to pronounce on what they really measure and, consequently, how valid the conclusions of Taras et al. (2012) are. Of note, the Values Survey Module has never been properly tested either. There is not a single study in a peer-reviewed journal showing how it works across at least 30 countries from all continents.
2.3 Internal reliability of Hofstede’s dimensions
The issue of internal reliability is important as Hofstede’s theories are built on some key assumptions, such as the positive relationship between societal anxiety and societal restrictiveness with respect to rules and laws, underpinning the UA dimension, as well as a negative relationship between so-called MAS and FEM values, underpinning the MAS-FEM dimension. If these relationships were not confirmed, Hofstede’s model would be seriously challenged even if it were correct in other respects.
2.3.1 Internal reliability of IDV-COLL and PD
The question of whether IDV-COLL and PD, as operationalized by Hofstede, are internally reliable is probably irrelevant. Hofstede’s operationalization of IDV-COLL has not been accepted as a paradigm for major replications of that dimension simply because many of his items have been seen as lacking face validity and some have even been viewed as a mystery (Bond, 2002). Replications of IDV-COLL with entirely different items (Gelfand et al., 2004; Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press) are characterized by good internal reliability and face validity. There is no doubt that, measured in this way, IDV-COLL is a robust and important dimension of national culture.
Surprisingly, major replications of PD are simply missing in the literature. Therefore, it is impossible to pronounce on that dimension’s internal reliability. The relationship between IDV-COLL and PD has not been elucidated satisfactorily either. Hofstede’s measures of these two constructs are closely correlated statistically. Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al. (in press) argue that they are also related conceptually. If IDV-COLL reflects differences in treatment of people – as individuals or as members of particular groups – then PD is a logical facet of IDV-COLL, as it reflects differential treatment on the basis of one’s position in society. Thus, as PD is merely a conceptual facet of IDV-COLL, and not an independent dimension, the question of its internal reliability becomes irrelevant.
2.3.2 Internal reliability of uncertainty avoidance (UA)
According to UA theory, people in societies with high levels of anxiety (a facet of neuroticism in the Big Five personality model) value job security (Hofstede, 2001; Minkov and Hofstede, 2014). Minkov and Hofstede (2014) found support for this theory albeit across European countries only. Yet, Table I demonstrates that “good job security” as an important job characteristic, measured by the WVS in 2000-2004 (Item v88, subsequently discontinued) across the world, is not associated with any reported measure of national neuroticism or anxiety, including the most recent estimates of national scores on the anxiety facet of neuroticism by Allik et al. (2017).
The UA theory also postulates that societies characterized by high anxiety attempt to reduce that feeling by believing in, and in fact imposing, strict and unbendable rules and laws, so as to make life less uncertain (Hofstede, 2001; Minkov and Hofstede, 2014). However, this theory is not very convincing in view of the fact that anxiety and a rule-orientation ideology are not correlated at the individual level, which we know from Hofstede’s (2001) own findings. In other words, the anxious people and the bureaucratic-minded individuals are not necessarily the same. But if the bureaucrats do not necessarily have high anxiety levels, what makes them create and insist on unbendable rules? Are they doing it out of concern for the neurotics who need such rules? If the latter do need such rules, why do they not, too, believe in them?
Minkov and Hofstede (2014) found that, at the national level, a measure of anxiety is indeed highly correlated with a measure of the ideology that all laws must be followed strictly. Yet, this was a study across European countries only. It does not reveal whether these two facets of UA are correlated highly and positively across a wider set of countries. There are no secondary data that provide an answer to this question. Yet, the analysis of primary data below does provide an answer.
Minkov et al. (2013) extracted a dimension of national culture from WVS items measuring the strength of societal norms in the domain of creation and termination of life, such as divorce, homosexuality, prostitution, abortion, suicide, and euthanasia. They found that these items form a strong single factor, called “personal-sexual,” creating a contrast between economically advanced countries (greater permissiveness) and developing countries (greater restrictiveness). This factor is closely associated with IDV-COLL (Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press), but is not related to any published national index of neuroticism or anxiety. In fact, according to Hofstede (2001) it is MAS-FEM, rather than UA, that explains societal restrictiveness as measured by “personal-sexual” items in the WVS. Although the logic of this association is not apparent, it will be tested in the section devoted to the predictive properties of MAS-FEM.
2.3.3 Internal reliability of MAS-FEM
MAS-FEM has two main facets: MAS values (achievement, challenge, recognition, earnings, competition) and FEM values (good human relationships, including compassion). These two facets should be correlated negatively (Hofstede, 2001).
The 1995-1999 wave of the WVS allows a test of this theory. Item V73 presents the respondents with four job characteristics – “good income,” “safe job, no risk,” “working with people I like,” and “important job, feeling of accomplishment” – and asks the respondents to choose the most important one. The first item and the last two items should reflect MAS-FEM goals as they correspond to some of the items that loaded highly on MAS-FEM in Hofstede’s (2001) factor analyses of IBM items (pp. 256-257, p. 284). “Good income” corresponds to “earnings,” “working with people I like” corresponds to “friendly atmosphere,” whereas “important job, feeling of accomplishment” is a measure of the social importance of prestige and success, and corresponds to “recognition,” “challenge,” and “advancement to a higher position” in Hofstede’s analysis. This WVS item format measures relative importance and avoids response style, approximating the effect of ipsatization (also known as standardization within subject or by subject) used by Hofstede (2001) in his analysis of work values from which he extracted MAS-FEM.
Table II demonstrates that the three items are not correlated quite as the MAS-FEM theory predicts. While “good income” and “working with people I like” are indeed opposites, “important job” yields correlations that contradict the MAS-FEM theory.
Figure 1 demonstrates that if the items measuring the importance of “feeling of accomplishment” and “working with people I like” were merged into a single dimension, it would highlight a contrast between economically developed countries, in the upper right corner, and developing countries, in the lower left corner. Therefore, the observed national differences in job priorities are a function of differences in national wealth, not MAS-FEM, which is unrelated to it. Subsequent waves of the WVS confirm this finding, although the MAS item in question was fielded without the “feeling of accomplishment” component. Likewise, across the WVS waves in which these items were fielded, “good income” and “working with people I like” are negatively correlated and clearly merge into a single dimension that creates a contrast between economically developed countries (high priority of people, low priority of income) and developing countries (high priority of income, low priority of people). Evidently, these contrasts are not a function of MAS-FEM but are an outcome of differences in economic development, which is unrelated to MAS-FEM according to Hofstede (2001).
The WVS has fielded some of Schwartz’s items, including importance of success (V85) and importance of helping (V84) in the 2005-2009 WVS wave, which provides the largest set of countries that have been scored on both of these items (n=51). The two items certainly address societal MAS-FEM values. “Success” is part of Schwartz’s “mastery” domain, which according to Hofstede (2001, p. 298) contains MAS values. Helping is a major prerequisite for the maintenance of good relationships. Also, it is found across from “success” on Schwartz’s (2008) circumplex, suggesting that the two values are opposites, just as the MAS-FEM theory postulates. A reviewer of this paper helpfully provided the results of a Monte Carlo simulation analysis of these items, indicating the probability that a randomly selected male in a specific country will score higher or lower than a randomly selected female. The results are consistent with the MAS-FEM theory. For example, in 43 of the 52 countries in the WVS, men are less likely than women to value helping others and these differences are often substantial, reaching 23 percent (greater probability for males) vs 43 percent (greater probability for females). There is no doubt then that importance of helping is a FEM value, whereas importance of success is a MAS value.
Across 51 countries, importance of helping and importance of success are correlated significantly and positively: r=0.40 (p=0.004). Ipsatization at the national level reverses this correlation and makes it negative, in accordance with the MAS-FEM theory, just like in Schwartz (2008) where the two items are in opposite sections of the value circumplex, suggesting a negative correlation. This raises the question of which method is preferable: comparing raw scores or ipsatized scores? This is a complex issue beyond the topic of this study. The section on the primary data analysis explains the outcome of fielding MAS and FEM items as categorical choices without Likert scales, which makes ipsatization irrelevant.
2.3.4 Internal reliability of Confucian work dynamism or long-term orientation (LTO)
Originally called “Confucian work dynamism,” Hofstede’s fifth dimension is also known as LTO. It is an extremely important dimension as it seems to explain some of the cultural differences between the Confucian societies of East Asia at one extreme and Africa, the Middle East, and Latin America at the other. Minkov and Hofstede (2012) successfully replicated LTO with data from the WVS and confirmed its internal reliability. Still, they admitted that the dimension lacked theoretical coherence. Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al. (in press) have provided a radical reconceptualization of this dimension based on Minkov’s (2011, 2013) work, called “flexibility vs monumentalism” (FLX-MON). It is partly based on Steven Heine’s self-enhancement and self-stability theory, and reflects national differences in high vs low self-regard and self-confidence, being always the same person vs being flexible and adaptable, and liking to help people vs being reluctant to do that. Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al. (in press) do not include the concepts of persistence and thrift in FLX-MON, arguing that these facets of LTO are controversial.
2.4 Predictive properties of Hofstede’s dimensions
The predictive properties of Hofstede’s dimensions are a central topic in Culture’s Consequences, Hofstede’s (2001) main monograph. The numerous examples in that book create the impression that all the dimensions in Hofstede’s model are statistically associated with, and thus seem to predict or explain, variance in many external variables. However, given a database of diverse variables measured at the national level, it is easy to select some, or even many, that are associated at least weakly and at least across some countries. For a particular dimension of national culture to be credible and of practical utility, it should satisfy two more stringent requirements. First, it should have strong predictive properties, yielding high correlations with variables of interest across a large sample of countries from all, or most, continents, adequately representing the cultural variation across the world’s modern nations. Second, it should be a strong predictor, withstanding plausible controls. Appendix 6 at the end of Culture’s Consequences (Hofstede, 2001) shows that the great majority of Hofstede’s validity tests were performed across fewer than 30 countries, often from the economically developed part of the world, because Hofstede did not have data from other countries. This makes most of the reported associations unconvincing.
Taras et al. (2012) studied the predictive properties of Hofstede’s original indices and found reasonably high correlations, at least with respect to a small number of external variables, in the past decades. Yet, the strength of these predictive properties has been declining so much since then that, according to Taras et al. (2012), in another decade or so Hofstede’s indices may not explain adequately anything anymore. In view of the small number of dependent variables in the analysis of those authors, the lack of control variables, and the lack of information on the number and geographic location of the countries across which each correlation was calculated, their conclusions should be viewed with great caution. A more detailed dimension-by-dimension check would be beyond the scope of any article, including this one. Yet, it is noteworthy that Taras et al. (2012) found that IDV-COLL and PD yielded higher correlations with relevant external variables than did MAS-FEM and UA. This resonates with the finding of the same authors that MAS-FEM and UA are more temporarily unstable than IDV and PD, and with Merritt’s (2000) unsuccessful attempt to replicate MAS-FEM and UA, strengthening the impression that these are problematic dimensions.
2.4.1 Predictive properties of IDV-COLL and PD
Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al. (in press) show that IBM’s IDV-COLL index is still a reasonably good predictor of key variables such as rule of law, human inequality, and accident proneness. However, Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al.’s (in press) IDV-COLL index is a considerably better predictor of those key variables, suggesting that the IBM measure needs updating. Also, Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al. (in press) show that the Anglo countries, and especially the USA, do not score particularly high on IDV-COLL or any measure related to it. With the exception of Japan, the Confucian countries of East Asia score low on IDV-COLL in the work of Hofstede (2001). Yet, due to their phenomenal economic development in the past decades, they have all climbed higher on the IDV-COLL ladder, surpassing many developing countries, and currently occupy mid-range positions.
Table III compares the correlations that Hofstede’s PD and IDV-COLL yield with relevant external variables that can be expected to be associated with PD. Hofstede’s PD is the better predictor of only a few of them and in those cases the significant relationship is lowered and reduced to insignificance after controlling for a still better predictor. This seriously calls into question the utility of the PD index.
2.4.2 Predictive properties of UA
Exhibit 6 in Hofstede (2001) contains only one significant correlation between UA and a dependent variable that exceeds ±0.50 across at least 30 countries: UA predicts a country’s nurse-doctor ratio.
Hofstede and McCrae (2004) reported that UA was associated with McCrae’s national neuroticism index. Table IV shows correlations between UA and the four large studies reporting national neuroticism scores based on the NEO-PI-R or BFI questionnaires. As UA is related to some measures of neuroticism but not to others, the evidence remains inconclusive. The correlations between the available measures of neuroticism are not impressive, suggesting that some of the four studies have not measured it convincingly. Identifying those studies is beyond the scope of this paper.
Table V provides more correlations between UA and relevant societal indicators that UA could be expected to correlate with. With the exception of the “personal-sexual” factor, all selected indicators conform to Hofstede’s (2001) expectations concerning the predictive properties of UA. Namely, it should predict importance of job security, trust, subjective well-being, a focus on order, racism, corruption, slow acceptance of innovation, and a lack of economic freedom. Some correlations were calculated twice: across all available countries and then across countries in the “very high” category on the UNDP’s (2015) Human Development Index, since Hofstede (2001) reports that some of UA’s predictive properties are valid only or primarily across economically advanced countries.
The expectation that UA should be positively associated with importance of job security and negatively with interpersonal trust (Hofstede, 2001) is not confirmed by the WVS data. Both variables are closely associated with GLOBE’s UA practices. This suggests high trust and relatively low importance of job security characterize societies with detailed and properly enforced formal rules that create predictability. This has nothing to do with national anxiety.
Although Hofstede (2001) indicated that UA is not associated with risk avoidance, Table V shows that UA is a positive predictor of preference for a “safe job with no risk” as measured by the WVS. Yet, controlling for GLOBE’s UA practices reduces this correlation to statistical insignificance. The strongest preference for job safety is found in societies without detailed and strictly enforced formal rules (thus scoring low on GLOBE’s UA practices), which apparently depresses job safety while generating a high desire for it.
One of the most important practical implications of Hofstede’s UA is its presumed negative association with innovation (Hofstede, 2001). The largest study that supports this view was conducted across 68 countries more than 20 years ago (Shane, 1995). It concluded literally that uncertainty-accepting societies may be more innovative than uncertainty-avoiding societies, based on employees’ preferences for different roles within their organizations. Intrigued by these findings, Rinne et al. (2012) assessed the predictive properties of all of Hofstede’s dimensions, using data from the complex Global Innovation Index as dependent variables. They found that while IDV-COLL and PD were related to national innovation scores, UA was not related to them. The results in Table V show that UA is not related to adoption of innovative technologies.
Idiographic analyses further highlight some of the issues that plague the UA dimension and its index. According to Hofstede (2001), the South East Asian countries tend to score low on UA. This suggests that their societies should be quite liberal and allow their members to bend or ignore rules. The observed reality in those countries is precisely the opposite. They are well known for their harsh punishments, such as flogging for alcohol consumption during Ramadan in Malaysia, flogging for homosexual intercourse in Indonesia, prison sentences for graffiti writing in Singapore, and death penalty for possession of small amounts of light drugs in Malaysia, Indonesia, and Singapore. Singaporeans do not even have the right to chew gum and jokingly call their home place “a fine country:” a nice country where one can easily get fined.
2.4.3 Predictive properties of MAS-FEM
Appendix 6 in Hofstede (2001) does not list any significant correlations between MAS-FEM and a dependent variable that exceeds ±0.50 over at least 30 countries. Nevertheless, some of the presumed key predictive properties of that dimension are worth examining. Table VI provides correlations between key variables that MAS-FEM should be associated with according to Hofstede (2001). Some of the calculations were calculated also across economically developed countries as Hofstede (2001) indicates that some of the predictive properties of MAS-FEM are valid only across wealthy countries. Also, some calculations were calculated twice: using raw items and then ipsatized items, as advocated by Hofstede (2001).
Only two of the variables in Table VI – women’s share of seats in parliament and official development assistance as share of a nation’s gross domestic product – are associated, albeit weakly, with MAS-FEM, yet even these correlations become insignificant after controlling for relevant external variables, such as Welzel’s (2014) emancipative values index and GLOBE’s assertiveness, proposed by Hartog (2004) as a radical reconceptualization of, and improvement on, Hofstede’s MAS-FEM. This failure of MAS-FEM to predict female emancipation should not come as a surprise since it is well known by now that emancipation is strongly associated with variants of IDV-COLL and national wealth, and is completely unrelated to MAS-FEM. An example of such an IDV-COLL variant is Welzel’s (2014) emancipative values index. Welzel’s extensive work in the field of emancipation is fully convincing and conclusive.
Dimensions of national culture that replicate in one form or another, and have good predictive properties, produce clear geo-economic spatial configurations as shown by Dobson and Gelade (2012). This is accepted by Hofstede (Minkov et al., 2013). Yet, MAS-FEM does not yield a recognizable geo-economic configuration and neighboring countries whose populations have common ethnic and civilizational origins, such as Mexico and Guatemala, and Japan and Korea, sometimes have dramatically different scores on MAS-FEM. Naturally, an index that lacks a geo-economic structure cannot explain variables that have such a structure, such as most important national statistics and most WVS measures.
2.4.4 Predictive properties of Hofstede’s fifth dimension
Minkov and Hofstede (2012) showed that their LTO measure predicted national differences in educational achievement. However, Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al. (in press) demonstrated that FLX-MON, a radical reconceptualization of LTO, is the best-known predictor of national differences on TIMSS and PISA tests, considerably outperforming LTO measures.
3. Analysis of primary data
The analysis of primary data focuses on the internal reliability of the two problematic dimensions: UA and MAS-FEM. It uses data from a new study of national culture and personality across nearly 53,000 respondents selected probabilistically from all main geographic regions and economic sectors of 56 countries, reflecting the ethnic, linguistic, and age structure of most countries quite adequately. In economically developed countries, the samples are also close to the national census in terms of educational-level differences, whereas the samples in developing countries consist predominantly of respondents with higher education. Nevertheless, in nearly each country in the second group there are at least 100 probabilistically selected respondents without higher education, thus allowing separate cross-national analyses of samples with and without higher education. This study excluded the Dominican Republic and Puerto Rico, where the data were collected by phone, unlike all other countries. Details about the samples are freely available from Itim International (firstname.lastname@example.org), an international cross-cultural management consultancy, licensed by Geert Hofstede.
The questionnaire included 108 items, plus demographic questions, grouped in several sections. The largest section (52 items) consists of personality items and self-construals, targeting the Big Five measures of personality and Hofstede’s dimensions. Two smaller sections measure consumer behavior preferences and sources of guidance in making purchasing decisions. Analyses of these sections may have interesting implications for international business.
Comparisons of data from samples with and without higher education did not reveal any substantial differences in terms of country positions. Below, only results from comparisons of samples without higher education are reported.
In order to avoid response style associated with Likert scales, all items in this study elicit categorical responses, plus an intermediate option. Examples are provided below.
Two items target the two main facets of UA. The first is about anxiety:
1. I worry a lot and often feel nervous. 2. I am somewhere here, in between these two. 3. I am usually relaxed and do not worry much.
The second item is about the conviction that all societal rules and laws must be followed strictly, which is the societal extrapolation of the conviction that all company rules must be followed strictly (Minkov and Hofstede, 2014):
1. If I could, I would make all people in our society follow all our laws and rules very strictly. 2. I am somewhere here, in between these two. 3. If I could, I would allow people to break useless or meaningless laws and rules.
Scored on a scale from 1 to 3 and aggregated to the national level, the two items correlate positively at 0.45 (p=0.022) across the 26 European countries in the sample, supporting Minkov and Hofstede’s (2014) findings, and validating the representativeness of the database used for this analysis, as it produces the same pattern as the nationally representative European Social Survey used by Minkov and Hofstede (2014). But across 53 countries from all continents, this correlation is −0.23 (p=0.094). Figure 2 visualizes the relationship between the two variables.
Figure 2 suggests that, with the exception of Vietnam, it is mostly the economically advanced IDV societies that have the strongest tendency to give people the discretion to decide which laws and rules are not worth following. With the exception of Italy, all societies at the opposite extreme are developing countries. This supports Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al.’s (in press) assertion that this item measures IDV-COLL differences. In IDV countries, people have greater freedom and individual discretion to decide whether societal rules are meaningful or not. In COLL societies, people have to follow the rules that are imposed on them as their fellow countrymen and women do not believe that giving people discretion is a good idea. This highly meaningful and logical pattern suggests that the country scores on this item capture logical cultural differences and cannot be dismissed as a study artifact.
Table VII shows that, unlike Hofstede’s UA, the anxiety item in this study is significantly correlated with all national measures of neuroticism and anxiety obtained in large-scale studies. Besides, some of the correlations are quite high. This validates the anxiety item as a reliable measure of national anxiety. In fact, it may be the best available national measure of anxiety and neuroticism as it is the only one available that yields such (relatively) high correlations with each of the remaining measures in the other large-scale studies. In sum, we have solid evidence that this study has measured anxiety and societal restrictiveness in a very meaningful and reliable way, and that the two measures are not correlated as UA theory predicts.
Baumann and Winzar (2017) correctly point out that value prioritizations are complex processes whose outcome may depend on the values themselves and circumstances. From this perspective, asking people how much they value achievement may not be informative enough. Everybody values achievement of some sort, yet some people may value achievement of good human relationships (a FEM value or goal) more than achievement of recognition (a MAS value or goal) or vice-versa. Therefore, for the purpose of this study, the text of the achievement item was quite specific:
1. I would like to achieve fame and glory. 2. I am somewhere here, in between these two. 3. I see fame and glory as useless to me.
This conceptualization of achievement is entirely in accordance with Hofstede (2001) who indicates explicitly that high MAS stands for “achievement in terms of ego boosting, wealth, and recognition” (p. 298).
The other MAS item addresses one’s willingness to compete, based on Hofstede’s (2001) indication that MAS cultures have a “concern” for “performance and competition” (p. 313), and that “The family in masculine societies socializes especially male children toward assertiveness, ambition, and competition” (pp. 314-315):
1. I like to compete with people. 2. I am somewhere here, in between these two. 3. I hate to compete with people.
The following items capture the concept of FEM:
1. I like to help people, even if I have to do something difficult. 2. I am somewhere here, in between these two. 3. I rarely agree to do something difficult to help people.
1. I am a compassionate person. When others have problems, I feel very sorry for them. 2. I am somewhere here, in between these two. 3. If other people have problems, I am usually indifferent.
Just like helping, compassion is consistent with MAS-FEM theory since Hofstede (2001) indicates that “The mas/fem dimension affects priorities in the following areas: (1) solidarity with the weak in one’s society versus reward for the strong” (p. 317), and that “In masculine societies more people believe that the fate of the poor is the poor’s own fault” (p. 319).
Our data show that, worldwide, we have the same situation as with the “help” and “success” items in the WVS: men are more likely to adopt the supposedly MAS self-descriptions (desire for fame, competitiveness) whereas women are more likely to adopt the supposedly FEM self-descriptions (desire to help and compassion). Thus, there is no doubt that these items conform to Hofstede’s MAS-FEM theory.
Table VIII shows correlations between these items, scored on a scale from 1 to 3 and aggregated to the national level, and Hofstede’s MAS-FEM index.
Table VIII demonstrates that all MAS and FEM items are correlated positively. The use of conceptual opposites within each item instead of a Likert scale means that this pattern cannot be due to the well-known preference of some societies to agree with most statements or rate most items as important, since the respondents are not asked to agree with anything to any specific degree or rate the importance of anything. They have described themselves in terms of clear statements that they identify with.
Figure 3 visualizes the relationship between liking to compete and liking to help. The two items create a clear geographic map, with East Asia in the upper right corner, European and English-speaking countries in the middle, and Africa, Latin America, Southeast Asia, as well as the Balkans and the Middle East (Turkey, Serbia, Romania and Egypt), in the lower left corner. These two items obviously measure something very real; otherwise there would not be such a clear geographic pattern. It is evident that the East Asian Confucian cultures are least likely to socialize their members for a desire to compete and help, whereas the rest of Asia, Latin America, Africa, and the Middle East exhibit the opposite pattern. This pattern is indicative of FLX-MON differences, explained in the work of Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al. (in press).
This finding does not imply that East Asians do not engage in competitions. In fact, it is well known that they tend to be fiercely competitive in education. The results of this fierce competitiveness are also known. Children of Confucian heritage surpass those from any other societies in educational achievement, especially in mathematics. Yet, the nationally representative study TIMSS reveals that East Asian children are also those who have the most negative attitudes toward the study of mathematics (Minkov, 2011), and possibly toward the educational competitions that they have to engage in, under societal pressure. Again, we have evidence of Confucian duality and ability to adapt one’s behavior to the requirements of the situation even if this means a clash with one’s values and dispositions.
Item v10 in WVS 1999-2004 (subsequently discontinued) measured the importance of “service to others” as a personal value. The percentages of respondents who have chosen the “very important” option are highly correlated with the national scores on the reversely-scored liking-to-help item in this study: r=−0.71 (p<0.001, n=21). Considering the 15-year time difference between the two studies, this is a remarkably high correlation, strongly validating both studies: the WVS and ours. Figure 4 visualizes the relationship between the two items.
In sum, the MAS and FEM measures in this study are highly reliable and valid as measures of national culture and the positive correlations between them, refuting the MAS-FEM theory, are not due to improper measurement.
Replication and validation studies can have three possible outcomes. First, the original model may be confirmed and validated. In that case, all is just well. Second, the replication and validation attempt may produce nonsensical findings. This would not necessarily invalidate the original model. It may be the case that the original is valid whereas the replication attempt is plagued by various methodological errors. This study is an example of the third possible outcome. The original model is not replicated and is not validated but the new findings are not nonsensical at all. They are underpinned by a very solid logic, which however differs from Hofstede’s. This new logic is based on nationally representative studies, mostly the WVS and a survey of 53,000 people chosen probabilistically, reflecting the structure of the national census more or less closely in each of 56 nations, adequately representing the world’s national cultures from all continents. Which of the two logics is stronger – the old or the new – is a question that is not hard to answer.
This study documents the need for a thorough revision of Hofstede’s model of national culture. Of the four original IBM dimensions, only IDV-COLL is supported as a coherent and empirically useful dimension of national culture. Yet, the original IBM operationalization and the index that it has produced need a substantial correction. First, IBM’s IDV-COLL does not have good face validity. Second, after Hofstede’s IBM study, the USA and the other English-speaking countries have never been shown to lead the country rankings on any major dimension of national culture or any national statistics related to IDV-COLL. A much-needed correction of the IDV-COLL index is provided by Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al. (in press).
The internal reliability of PD and its independence from IDV-COLL could not be established with the data available for this study. It is however clear that IDV-COLL is a better predictor of the main variables that PD can be expected to predict, making PD empirically redundant. And since one of the main facets of IDV-COLL is differential treatment of people based on their group affiliation, PD is logically a sub-facet of IDV-COLL that need not be seen as independent from IDV-COLL.
The main pillar of UA – the assumption that societal anxiety accounts for societal preference for strict rules and laws – collapses upon scrutiny. These two presumed UA facets are correlated highly and positively only across European countries. An analysis across countries from all continents reveals quite clearly that societal preference for strict rules and laws is an aspect of COLL, and is not related to national measures of anxiety or neuroticism as UA theory predicts. This explains why, despite their low UA scores, the South and Southeast Asian countries have extremely strict rules in domains that their COLL cultures have traditionally considered important. The fact that westerners observe some lack of order in South Asia from their own perspective, such as chaotic driving, simply means that Western driving regulations are still a foreign import in South Asia that has not taken root in the local culture as it clashes with older cultural rules.
Apart from its lack of internal consistency, UA does not have any of the main predictive properties that it has been credited with. Whenever UA produces a significant zero-order correlation with a relevant external variable, that correlation is reduced to insignificance after controlling for various aspects or facets of IDV-COLL or closely related constructs, such as GLOBE’s UA or future orientation practices, or Welzel’s emancipative values index.
MAS and FEM values are correlated positively, not negatively, and are not related to the IBM MAS-FEM index. This finding, as well as the failure of the IBM’s MAS-FEM index to demonstrate the predictive properties that it is supposed to have, plus the fact that distances between the values and personality traits of men and women are not a function of MAS-FEM, discredits the MAS-FEM dimension and suggests that it is an artifact of the IBM data set without a societal equivalent.
Figure 5 is a cultural map of the world, using Hofstede’s UA and MAS-FEM as axes. It is puzzling to see the Confucian countries scattered throughout the map. It is also impossible to explain the close proximity of pairs of culturally distant countries, such as the USA and the Philippines, Canada and Indonesia, Taiwan and Brazil, Korea and Peru, Germany and Ecuador, Austria and Venezuela, and Finland and Thailand, to name just a few pairs. While the Scandinavian countries and the Netherlands have the lowest scores on MAS-FEM, the other economically advanced countries are at the other extreme on that dimension. The Latin American countries are also dispersed along the MAS-FEM axis, without any apparent logic. Some of the Confucian countries score very low on UA whereas others score very high. Taiwan is in the middle. These patterns do not have close analogues in any national statistics or other indicators.
Figures 6 and 7 present a new cultural map of the world, using the latest measures of IDV-COLL (Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press) and FLX-MON (Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al., in press).
This new map of the world is very much like the real one, drawn from a traditional European perspective, without the world’s oceans. There is one logical exception: the English-speaking countries are not scattered across the world but form a fairly compact cluster right above the center of the map. Indonesia’s proximity to the Arab world and the African countries should not seem surprising. It is supported by proximity on important national indicators, such as measures of rule of law, transparency vs corruption, accident proneness, and gender inequality (all associated with IDV-COLL), as well as educational achievement (associated with FLX-MON).
The new cultural map is also the only one available that highlights the cultural distinctiveness of the Confucian countries of East Asia. They occupy an intermediate position on IDV-COLL, yet they are leaders in terms of FLX-MON. This explains the leading position of the Confucian countries in educational achievement, followed by Finland, the Netherlands, Russia, and Kazakhstan.
This study exposes two perils in cross-cultural research. The first one is over-reliance on seemingly matched samples whose comparability is not guaranteed. The second is insufficient testing of the validity of a model of national culture. This includes use of small and globally unrepresentative samples of nations and reliance on modest zero-order correlations that have not been tested extensively by controlling for potentially better predictors. Authors who use Hofstede’s dimensions rarely test the effects of other predictors, such as Inglehart’s, Schwartz’s, and GLOBE’s measures, alongside Hofstede’s. This explains why, not only in Hofstede’s work, but also in the studies of many other authors, UA has been found to be a significant predictor of diverse variables, including some of those tested in this study. Even MAS-FEM has been reported to produce effects in some analyses, although it is obviously a fictitious dimension. Most recently, De Mooij (2017) reported several high zero-order correlations between MAS-FEM and various variables measured across European countries only. One of these, “agree with university education is more for boys” (p. 451) is reported to correlate with MAS-FEM at 0.68, without the number of countries across which the correlation was calculated (p. 451). The same variable is measured by item v52 in WVS 2011-2014 across the world. Across 30 overlapping countries, it correlates with MAS-FEM at −0.31 (p=0.09).
At the time when Hofstede developed his model, and even in 2001, when his main monograph was published, the scarcity of the available data did not allow adequate large-scale tests. Hofstede’s analysis caused the admiration of many scholars, including the author of this paper. Yet the world has changed enormously since then and the amount of information about cross-cultural differences worldwide has increased manifold. This wealth of information today reveals a picture that is very different from what Hofstede extracted from his IBM data set and, apart from the fact that the Confucian societies are now somewhat more IDV-oriented, the difference does not seem to be a result of seismic cultural restructuring across the world. It comes from the nature of Hofstede’s IBM samples: they were not good representations of the cultures from which they were drawn, whereas the samples available today are far more representative.
Might the new model, proposed here, consisting of the new measures of IDV-COLL and FLX-MON, also be refuted upon closer scrutiny? It may certainly be modified, updated, and improved, but it cannot be completely dismantled in the near future, before very significant cultural shifts have occurred across the world. IDV-COLL is a dimension that transpires in one variant or another from any large study of culture, and each variant is closely associated with differences in national wealth and a host of other national indicators. The long history of FLX-MON and its predecessors from diverse studies, including the WVS, is described by Minkov, Bond, Dutt, Schachner, Morales, Sanchez et al. (in press). The validity of that dimension is confirmed by its strikingly close and persistent association with differences in national educational achievement, and measures of self-consistency and self-esteem from a variety of reliable studies, covering many countries across the globe, including self-confidence or self-esteem measures by PISA OECD, which relies on the largest nationally representative samples in the history of cross-cultural studies. Other national indicators, such as homicide rates and adolescent fertility seem to follow the same geographic distribution, rising from Confucian East Asia toward Latin America and Africa, whereas suicide rates and tobacco consumption seem to rise in the opposite direction. These, and many other research topics, are awaiting further exploration.
Correlations between national neuroticism (N) or anxiety (A) and importance of “good job security”
|National neuroticism (N) or anxiety (A) measure||r with v88||n (countries)|
|N: McCrae (2002)||0.19||16|
|N: McCrae and Terracciano (2005)||0.34||20|
|N: Schmitt et al. (2007)||−0.07||20|
|N: Gebauer et al. (2015)||−0.05||31|
|A: Allik et al. (2017)||0.03||17|
Notes: Allik et al. (2017) provide several scores for some countries based on different studies. In those cases, I used the median score. None of the correlations are significant at 0.05
Source: World Values Survey (2000-2004, Item v88)
Correlations between three masculinity-femininity work goals
|(1) Good income||1.00||−0.69**||−0.43**|
|(2) Important job, feeling of accomplishment||1.00||0.31*|
|(3) Working with people I like||1.00|
Notes: All correlations are across 51 countries. *,**Correlations are significant at 0.05 and 0.01 levels, respectively
Source: World Values Survey (1995-1999, Item v73)
Correlation patterns of Hofstede’s power distance (PD) compared with those of Hofstede’s individualism-collectivism (IDV-COLL)
|Variable||r with PD||r with IDV-COLL||n (countries)|
|GLOBE’s power distance “practices” (Carl et al., 2004)||0.54**||0.54**||47|
|GLOBE’s power distance “values” (Carl et al., 2004)||0.07||0.11||47|
|Schwartz’s hierarchy (personally provided scores, 2016)||0.32*||−0.46**||50|
|Schwartz’s egalitarianism (personally provided scores, 2016)||−0.48**||0.56**||50|
|Schwartz’s intellectual autonomy (personally provided scores, 2016)||0.52**||0.60**||50|
|Schwartz’s affective autonomy (personally provided scores, 2016)||−0.65**||0.69**||50|
|Schwartz’s embeddedness (personally provided scores, 2016)||0.67**||0.67**||50|
|Corruption perception index 2015 (Transparency International, 2017)||−0.64**||−0.68**||68|
|Same item after controlling for GDP per person in 2014 (World Bank, 2016) and Welzel’s (2014) emancipative values index||−0.09||0.07||62|
|UNDP’s coefficient of human inequality 2015 (Jahan and Jespersen, 2015)||0.55**||0.65**||54|
|Same item after controlling for GDP per person in 2014 (World Bank, 2016) and Welzel’s (2014) emancipative values index||−0.17||−0.08||51|
|Project GLOBE’s “participative” leadership dimension (Dorfman et al., 2004)||−0.49**||0.44**||48|
|Same item after controlling for Welzel’s (2014) emancipative values index||−0.01||0.06||44|
|Use of superiors as a source of guidance for managerial decision making (Smith et al., 2002)||−0.25||0.20||45|
|Use of subordinates as a source of guidance for managerial decision making (Smith et al., 2002)||−0.37*||0.45**||45|
|Same item after controlling for Welzel’s (2014) emancipative values index||−0.13||0.19||42|
|Obedience as a desirable trait for children, WVS 2005-2009, item v21||0.53**||−0.38*||38|
|Same item, after controlling for individualism-collectivism (Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press)||0.12||0.30||30|
|Obedience as a desirable trait for children, WVS 2011-2014, item v21||0.44**||−0.35*||35|
|Same item, after controlling for individualism-collectivism (Minkov, Dutt, Varma, Schachner, Morales, Sanchez et al., in press)||−0.01||−0.29||24|
Notes: *,**Correlations are significant at 0.05 and 0.01 levels, respectively
Correlations between Hofstede’s uncertainty avoidance and measures of national neuroticism
|(1) Hofstede’s uncertainty avoidance||1.00||0.57**||0.18||0.32||43**|
|(2) Neuroticism (McRae, 2002)||1.00||40*||0.49**||0.37|
|(3) Neuroticism (McCrae and Terracciano, 2005)||1.00||0.32*||0.20|
|(4) Neuroticism (Schmitt et al., 2007)||1.00||0.23|
|(5) Neuroticism (Gebauer et al., 2015)||1.00|
Notes: *,**Significant at 0.05 and 0.01 level, respectively
Correlations between uncertainty avoidance (UA) and relevant external variables
|Variable||r with UA||n (countries)|
|Importance of job security, WVS (2000-2004, Item v88) (not fielded in the same form after 2004)||−0.11||21|
|Interpersonal trust, WVS (2005-2009, Item v23)||−0.48**||39|
|Same item after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||−0.11||28|
|Interpersonal trust, WVS (2011-2014, Item v24)||0.41*||36|
|Same item after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||0.13||24|
|Personal-sexual factor (Minkov et al., 2013), reflecting restrictive societal norms with respect to the creation and termination of life||−0.11||32|
|Preference for a “safe job with no risk,” WVS (2000-2004 and 2005-2009, average of Items v84 and v48) (not fielded after 2009)||0.47*||42|
|Same item after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||0.23||29|
|Average life satisfaction, WVS (2005-2009, Item v22)||0.01||35|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.64**||13|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for Welzel’s (2014) emancipative values index||−0.37||13|
|Average life satisfaction, WVS (2011-2014, Item v23)||0.12||35|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.25||12|
|Subjective perception of one’s own state of health, WVS (2011-2014, Item v11)||0.33||30|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||0.42 (p=0.20)||11|
|Average happiness, WVS (2011-2014, Item v10)||0.20||28|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||0.51 (p=0.11)||11|
|“Maintaining order in the nation” most important of four national goals, WVS 2005-2009 (not fielded after 2009), Item v71||0.04||38|
|Percentage respondents choosing “people of a different race” as unwanted as neighbors, WVS (2005-2009, Item v35)||0.00||48|
|Percentage respondents choosing “immigrants, foreign workers” as unwanted as neighbors, WVS (2005-2009, Item v35)||−0.04||45|
|Corruption perception index 2015 (Transparency International, 2017)||−0.23||68|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.64**||27|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||−0.32 (p=0.15)||23|
|Availability of latest technology according to the Global Competitiveness Report (2014-2015, Item 9.01) (Schwab et al., 2014)||−0.20||64|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.39*||27|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||−0.04||23|
|Percentage internet users in 2013 according to the Global Competitiveness Report (2014-2015, Item 9.01) (Schwab et al., 2014)||−0.06||64|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.50**||27|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||−0.22||23|
|Mobile phone subscriptions per 100 population in 2013 according to the Global Competitiveness Report 2014-2015, Item 2.08 (Schwab et al., 2014)||0.08||64|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.01||27|
|Index of Economic Freedom 2016 (Heritage Foundation, 2016)||−0.33*||63|
|Same item, after controlling for GLOBE’s UA practices (Sully de Luque and Javidan, 2004)||−0.07||45|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.74**||27|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for GLOBE’s future orientation practices (Ashkanasy et al., 2004) and neuroticism (McCrae and Terracciano, 2005)||−0.50 (p=0.07)||16|
Notes: *,**Significant at 0.05 and 0.01 level, respectively
Correlations between masculinity-femininity (MAS-FEM) and relevant external variables
|Variable||r with MAS-FEM||n (countries)|
|Percentage choosing “Working with people I like” as most important job characteristic, Item v73, WVS (1995-1999)||0.11||28|
|Percentage choosing “Working with people I like” as most important job characteristic, Item v48, WVS (2005-2009) (not fielded after 2009)||−0.11||36|
|Percentage choosing “Important job, feeling of accomplishment” as most important job characteristic, Item v73, WVS (1995-1999) (not fielded in this form after 1999)||−0.34||28|
|Percentage mentioning “a job in which you feel you can achieve something,” WVS (2000-2004) (not fielded after 2004)||−0.05||23|
|Importance of helping as a personal value, WVS (2005-2009, Item v84)||0.29||32|
|Same item, after ipsatization at the national level, across all 10 Schwartz items||0.18||32|
|Importance of success as a personal value, WVS 2005-2009, item v85||−0.06||32|
|Same item, after ipsatization at the national level, across all 10 Schwartz items||−0.26||32|
|“Religious faith” mentioned as an important trait for children, Item v19, WVS (2005-2009)||0.12||38|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||0.28||15|
|“Religious faith” mentioned as an important trait for children, Item v19, WVS (2011-2014)||0.20||30|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||0.07||11|
|Importance of family, item v4, WVS 2005-2009||−0.07||37|
|Same item after ipsatization by nation across five value items in that section of the WVS||−0.09||37|
|Same item after ipsatization, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.29||19|
|Importance of work, Item v8, WVS (2005-2009)||0.05||37|
|Same item after ipsatization by nation across five value items in that section of the WVS||0.05||37|
|Same item after ipsatization, only countries with a “very high” Human Development Index (UNDP, 2015)||0.08||19|
|Importance of religion, Item v9, WVS (2005-2009)||−0.05||37|
|Same item after ipsatization by nation across the five value items in that section of the WVS||−0.07||37|
|Same item after ipsatization, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.12||19|
|Average life satisfaction, WVS (2005-2009, Item v23)||0.01||35|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.16||12|
|Average life satisfaction, WVS 2011-2014, item v23||0.03||35|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.19||12|
|Subjective perception of one’s own state of health, WVS (2011-2014, Item v11)||0.09||30|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||0.21||11|
|Gender Inequality Index (UNDP, 2015)||0.08||62|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||0.20||27|
|Schwartz’s mastery (personally provided scores, 2016)||0.08||50|
|Societal restrictiveness vs permissiveness as measured by the WVS 2005-2009: “personal-sexual” factor (Minkov et al., 2013)||−0.13||32|
|Same item after controlling for Hofstede’s IDV-COLL||−0.30 (p=0.10)||32|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015)||−0.44 (p=0.11)||14|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for Hofstede’s IDV-COLL||−0.49 (p=0.09)||14|
|Women’s share of seats in parliament (UNDP, 2015)||−0.26*||61|
|Same item, after controlling for Welzel’s (2014) emancipative values index||−0.16||55|
|Same item, only countries with a very high Human Development Index (UNDP, 2015)||−0.45*||27|
|Same item, only countries with a “very high” Human Development Index (UNDP, 2015), after controlling for Welzel’s (2014) emancipative value index||−0.38||24|
|Official development assistance in 2015 (OECD, 2015)||−0.47**||27|
|Same item, after controlling for Welzel’s (2014) emancipative values index and GLOBE’s assertiveness “as should be” (Hartog, 2004)||−0.36 (p=0.14)||25|
Notes: *,**Significant at 0.05 and 0.01 level, respectively
Correlations between the reversely scored anxiety item used in this study and measures of national neuroticism and anxiety
|Neuroticism in McCrae (2002)||−0.62||31|
|Neuroticism in McCrae and Terracciano (2005)||−0.47||35|
|Neuroticism in Schmitt et al. (2007)||−0.67||37|
|Neuroticism in Gebauer et al. (2015)||−0.42||54|
|Anxiety in Allik et al. (2017)||−0.67||37|
Note: All correlations are significant at 0.01
Correlations between masculinity-femininity items (self-construals) in this study and Hofstede’s masculinity-femininity index
|(1) Desire to achieve fame||1.00||0.55**||0.32*||0.47**||−0.20|
|(2) Desire to compete||1.00||0.61**||0.59**||0.01|
|(3) Desire to help||1.00||0.77**||−0.03|
|(5) Hofstede’s masculinity-femininity||1.00|
Notes: *,**Correlations are significant at 0.05 and 0.01 levels, respectively
Vietnam exhibits an unusual pattern on this item. That country scores relatively high in terms of percentages of people who would enforce strict laws, like in a typical collectivist country. The percentage of Vietnamese who would allow others to break useless laws is small in absolute terms, yet high relative to other countries.
Allik, A., Church, T., Ortiz, T., Rossier, J., Hrebickova, M., de Fruyt, F. et al. (2017), “Mean profiles of the NEO personality inventory”, Journal of Cross-Cultural Psychology, Vol. 48, No. 3, pp. 402-420.
Ashkanasy, N.M., Gupta, V., Mayfield, M.S. and Trevor-Roberts, E. (2004), “Future orientation”, in House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (Eds), Culture, Leadership, and Organizations: The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA, pp. 282-342.
Baumann, C. and Winzar, H. (2017), “Confucianism and work ethic – introducing the ReVAMB model”, in Oh, I. and Park, G.S. (Eds), The Political Economy of Business Ethics in East Asia: A Historical and Comparative Perspective, Elsevier-Chandos, Amsterdam, pp. 33-60.
Bond, M.H. (2002), “Reclaiming the individual from Hofstede’s ecological analysis. A 20-year Odyssey: comment on Oyserman et al. (2002)”, Psychological Bulletin, Vol. 128 No. 1, pp. 73-77.
Brewer, P. and Venaik, S. (2014), “The ecological fallacy in national culture research”, Organizational Studies, Vol. 35 No. 7, pp. 1063-1086.
Carl, D., Gupta., V. and Javidan, M. (2004), “Power distance”, in House, R.J., Hanges, P.J., Mansour, J., Dorfman, P.W. and Gupta., V. (Eds), Culture, Leadership, and Organizations. The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA, pp. 513-563.
Costa, P.T., Terracciano, A. and McCrae, R.R. (2001), “Gender differences in personality traits across cultures: robust and surprising findings”, Journal of Personality and Social Psychology, Vol. 81 No. 2, pp. 322-331.
De Mooij, M. (2017), “Comparing dimensions of national culture for secondary analysis of consumer behavior data of different countries”, International Marketing Review, Vol. 34 No. 3, pp. 444-456.
Dorfman, P.W., Hanges, P.J. and Brodbeck, F.C. (2004), “Leadership and cultural variation: The identification of culturally endorsed leadership profiles”, in House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (Eds), Culture, Leadership, and Organizations: The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA, pp. 669-720.
Gebauer, J.E., Sedikides, C., Wagner, J., Bleidorn, W., Rentfrow, P.J., Potter, J. and Gosling, S.D. (2015), “Cultural norm fulfillment, interpersonal belonging, or getting ahead? A large-scale cross-cultural test of three perspectives on the function of self-esteem”, Journal of Personality and Social Psychology, Vol. 109, pp. 526-548.
Gelfand, M., Bhawuk, D., Nishii, L.H. and Bechtold, D. (2004), “Individualism and collectivism”, in House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (Eds), Culture, Leadership, and Organizations: The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA, pp. 437-512.
Guimond, S., Branscombe, N., Brunot, S., Buunk, A.P., Chatard, A., Desert, M., Garcia, D.M., Haque, S., Martinot, D. and Yzerbyt, V. (2007), “Culture, gender, and the self: variation and impact of social comparison processes”, Journal of Personality and Social Psychology, Vol. 92 No. 6, pp. 1118-1134.
Hall, E.T. (1959), The Silent Language, Anchor Books, New York, NY.
Hartog, D. (2004), “Assertiveness”, in House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (Eds), Culture, Leadership, and Organizations: The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA, pp. 395-436.
Heritage Foundation (2016), “Index of Economic Freedom”, available at: www.heritage.org/index/ranking (accessed February 15, 2017).
Hofstede, G. (1980), Culture’s Consequences: International Differences in Work-Related Values, Sage, Beverly Hills, CA.
Hofstede, G. (2001), Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations Across Nations, Sage, Thousand Oaks, CA.
Hofstede, G. and McCrae, R.R. (2004), “Personality and culture revisited: linking traits and dimensions of culture”, Cross-Cultural Research, Vol. 38 No. 1, pp. 52-88.
House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (Eds) (2004), Culture, Leadership, and Organizations. The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA.
Inglehart, R. and Baker, W.E. (2000), “Modernization, cultural change, and the persistence of traditional values”, American Sociological Review, Vol. 65 No. 1, pp. 19-51.
Jahan, S. and Jespersen, E. (2015), “Human Development Report”, United Nations Development Program, New York, NY.
Kirkman, B.L., Lowe, K.B. and Gibson, C.B. (2006), “A quarter century of Culture’s Consequences: a review of empirical research incorporating Hofstede’s cultural values framework”, Journal of International Business Studies, Vol. 37 No. 37, pp. 285-320.
McCrae, R.R. (2002), “NEO-PI-R data from 36 cultures: Further intercultural comparisons”, in McCrae, R.R. and Allik, J. (Eds), The Five-Factor Model of Personality Across Cultures, Kluwer Academic/Plenum Publishers, New York, NY, pp. 105-126.
McCrae, R.R. and Terracciano, A. (2005), “Personality profiles of cultures: aggregate personality traits”, Journal of Personality and Social Psychology, Vol. 89 No. 3, pp. 407-425.
McSweeney, B. (2002), “Hofstede’s model of national cultural differences and their consequences: a triumph of faith – a failure of analysis”, Human Relations, Vol. 55 No. 1, pp. 89-118.
Merritt, A. (2000), “Culture in the cockpit: do Hofstede’s dimensions replicate?”, Journal of Cross-Cultural Psychology, Vol. 31 No. 3, pp. 283-301.
Minkov, M. (2011), Cultural Differences in a Globalizing World, Emerald, Bingley.
Minkov, M. (2013), Cross-Cultural Analysis: The Science And Art Of Comparing The World’s Modern Societies and Their Cultures, Sage, Thousand Oaks, CA.
Minkov, M. (2017), “Middle responding: an unobtrusive measure of national cognitive ability and personality”, Personality and Individual Differences, Vol. 113, pp. 187-192.
Minkov, M. and Hofstede, G. (2012), “Hofstede’s fifth dimension: new evidence from the World Values Survey”, Journal of Cross-Cultural Psychology, Vol. 43 No. 1, pp. 3-14.
Minkov, M. and Hofstede, G. (2014), “A replication of Hofstede’s uncertainty avoidance dimension across nationally representative samples from Europe”, International Journal of Cross-Cultural Management, Vol. 14 No. 2, pp. 161-171.
Minkov, M., Blagoev, V. and Hofstede, G. (2013), “The boundaries of culture; do questions about societal norms reveal cultural differences?”, Journal of Cross-Cultural Psychology, Vol. 44 No. 7, pp. 1094-1106.
Minkov, M., Bond, M.H. and Blagoev, V. (2015), “Do different national samples yield similar dimensions of national culture?”, Cross-Cultural Management, Vol. 22 No. 2, pp. 259-277.
Minkov, M., Bond, M.H., Dutt, P., Schachner, M., Morales, O., Sanchez, C.J. et al. (in press), “A reconsideration of Hofstede’s fifth dimension: new flexibility versus monumentalism data from 54 countries”, Cross-Cultural Research.
Minkov, M., Dutt, P., Varma, T., Schachner, M., Morales, O., Sanchez, C.J. et al. (in press), “A revision of Hofstede’s individualism-collectivism dimension: a new national index from a 56-country study”, Cross-Cultural and Strategic Management.
OECD (2015), “Net ODA”, available at: https://data.oecd.org/oda/net-oda.htm (accessed February 15, 2017).
Peterson, M.F. (2003), “Review of the book Culture’s Consequences, second edition by G. Hofstede”, Administrative Science Quarterly, Vol. 48 No. 1, pp. 127-131.
Rinne, T., Steel, D.G. and Fairweather, J. (2012), “Hofstede and Shane revisited”, Cross-Cultural Research, Vol. 46 No. 2, pp. 91-108.
Schmitt, D.P., Allik, J., McCrae, R.R. and Benet-Martinez, V. (2007), “The geographic distribution of Big Five personality traits: patterns and profiles of human self-description across 56 nations”, Journal of Cross-Cultural Psychology, Vol. 38 No. 2, pp. 173-212.
Schwab, K., Sala-i-Martin, X., Eide, E.B. and Blanke, J. (2014), “Global Competitiveness Report 2014-2015”, World Economic Forum, Geneva.
Schwartz, S.H. (1994), “Beyond individualism/collectivism: new cultural dimensions of values”, in Kim, U., Kagitcibasi, C., Triandis, H.C., Choi, S.C. and Yoon, G. (Eds), Individualism and Collectivism: Theory, Method, and Application, Sage, Thousand Oaks, CA, pp. 85-119.
Schwartz, S.H. (2005), “Sex differences in value priorities: cross-cultural and multimethod studies”, Journal of Personality and Social Psychology, Vol. 89 No. 6, pp. 1010-1028.
Schwartz, S.H. (2008), Cultural Value Orientations: Nature and Implications of National Differences, State University – Higher School of Economics Press, Moscow.
Schwartz, S.H. (2011), “Studying values: personal adventure, future directions”, Journal of Cross-Cultural Psychology, Vol. 42 No. 2, pp. 307-319.
Shane, S. (1995), “Uncertainty avoidance and the preference for innovation championing roles”, Journal of International Business Studies, Vol. 26 No. 1, pp. 47-68.
Smith, P.B., Peterson, M. and Schwartz, S.H. (2002), “Cultural values, sources of guidance, and their relevance to managerial behavior: a 47-nation study”, Journal of Cross-Cultural Psychology, Vol. 33 No. 2, pp. 188-208.
Sully de Luque, M. and Javidan, M. (2004), “Uncertainty avoidance”, in House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (Eds), Culture, Leadership, and Organizations: The GLOBE Study of 62 Societies, Sage, Thousand Oaks, CA, pp. 602-653.
Taras, V., Steel, P. and Kirkman, B.L. (2012), “Improving national cultural indices using a longitudinal meta-analysis of Hofstede’s dimensions”, Journal of World Business, Vol. 47 No. 3, pp. 329-341.
Transparency International (2017), “Corruption Perception Index”, available at: www.transparency.org/cpi2015/ (accessed February 15, 2017).
UNDP (2015), “Human development report 2015”, UNDP, New York, NY.
Welzel, C. (2014), Freedom Rising; Human Empowerment and the Quest for Emancipation, Cambridge University Press, Cambridge, online appendix, available at: www.researchgate.net/profile/Christian_Welzel2/publication/269992832 (accessed February 15, 2017).
Winzar, H. (2015), “The ecological fallacy: how to spot one and tips on how to use one to your advantage”, Australasian Marketing Journal, Vol. 23 No. 1, pp. 86-92.
World Bank (2016), “GDP per capita (current USD)”, Cambridge, available at: http://data.worldbank.org/indicator/NY.GDP.PCAP.CD (accessed February 15, 2017).
The collection of primary data for this study was organized by MediaCom Ltd and the Hofstede Center at Itim International, a Dutch-Finnish cross-cultural consultancy. Financial support was provided by MediaCom. Neither of the two organizations has influenced the study design, the data analysis, the decision to write and submit this paper, or any opinion expressed in it, in any way. All main findings in this study were shared with Geert Hofstede on several occasions by February 2017.