The International Journal of Human Resource

Management

ISSN: 0958-5192 (Print) 1466-4399 (Online) Journal homepage: https://www.tandfonline.com/loi/rijh20

National culture and human resource

management: assumptions and evidence

Barry Gerhart & Meiyu Fang

To cite this article: Barry Gerhart & Meiyu Fang (2005) National culture and human resource
management: assumptions and evidence, The International Journal of Human Resource
Management, 16:6, 971-986, DOI: 10.1080/09585190500120772

To link to this article: https://doi.org/10.1080/09585190500120772

Published online: 17 Feb 2007.

Submit your article to this journal

Article views: 9649

View related articles

Citing articles: 139 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rijh20
Int. J. of Human Resource Management 16:6 June 2005 971– 986

National culture and human resource

management: assumptions and evidence

Barry Gerhart and Meiyu Fang

Abstract It is generally understood that human resource practices and strategies differ
according to contextual factors such as size, industry and so forth. However, the
globalization of business means that country differences in the environment for
organizations have received increasing attention. Although countries can differ in many
important ways (e.g. institutional and regulatory environments, labour-force character-
istics such as education), a dominant focus of the international management literature is on
national differences in culture, especially cultural values (Hofstede, 1980). We examine
the assumptions that underlie this emphasis on the importance of national culture
differences in international management and re-analyse findings from previous research,
such as Hofstede’s, to test these assumptions. We find, contrary to the interpretations
of the original authors, that these assumptions receive only weak support when country
effects are evaluated using variance explained estimates, rather than relying solely
on statistical significance tests. We conclude that, while national culture differences can
be important and must be understood, their role needs to be put in the context of other
important contextual factors, including organizational culture.

Keywords Human resource management; national culture; strategic choice.

It is widely believed that a firm’s human resource (HR) and employee relations (ER)
strategy influences its business performance (Becker and Gerhart, 1996; Boxall and
Purcell, 2003). It is also conventional wisdom that HR and ER strategy effects are
stronger or weaker as a function of a variety of contextual or contingency factors. Those
traditionally receiving the most attention include size, corporate strategy, business
strategy and operations strategy, as well as employee and occupational characteristics
(Boxall and Purcell, 2003; Dyer and Reeves, 1995; Gerhart, forthcoming; Gerhart et al.,
1996; Wright and Sherman, 1999).
However, the globalization of business and its implications for management practice
and theory have led to increased attention to another set of contingency factors, country
differences in the environment for organizations (Brewster, 1999; Dowling et al., 1999).
These include differences in the centrality of markets, institutions, regulation, collective
bargaining and labour-force characteristics. To cite a few examples: in Western Europe,
collective bargaining coverage is much higher than in countries like the United States,
Canada and Japan. Works councils (joint labour – management decision-making
institutions at the enterprise level) and codetermination (worker representation on
supervisory boards of directors of companies) are mandated by law in Western European

Barry Gerhart, School of Business, University of Wisconsin-Madison, USA (tel: þ 608 262 3985;
e-mail [email protected]). Meiyu Fang, National Central University, Taiwan (tel: þ 886 3
427 2038; e-mail [email protected]).
The International Journal of Human Resource Management
ISSN 0958-5192 print/ISSN 1466-4399 online q 2005 Taylor & Francis Group Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/09585190500120772
972 The International Journal of Human Resource Management
countries like Germany, but not in Japan or North America. In France, until recently, the
working week was limited by law to 35 hours. In China, Central/Eastern Europe and
Mexico, labour costs are significantly lower than in Western Europe, Japan, the United
States and Canada.
There is no doubt that these are critical differences between countries in the
environment for management and that management practice needs to take these and other
such differences into account (Brewster, 1999; Dunlop and Galenson, 1978). However,
what has received perhaps the most attention in international management is not these
clear-cut, readily observed and readily measured country differences in environment.
Rather, the predominant focus in international management research appears to be
something that may be less obvious at first glance, but which is nevertheless deemed
fundamental to how people think and act in different countries. This is national culture.
The purpose of our paper is to re-examine work by Hofstede (and, to a lesser
extent, related work by others) to determine whether national differences in cultural
values and related concepts are as dominant as commonly believed. Our re-examination,
focusing on effect size (i.e. variance explained), rather than on statistical significance
levels (as is done in work by Hofstede and others), suggests a smaller role for country
differences than previously claimed.
Hofstede’s work on national culture and its impact
The book that is said to be the one of the most-cited publications – ‘By now there is
hardly a cross-national paper that does not cite Hofstede’ (Hickson, 1996: 221) – in the
social sciences is Geert Hofstede’s 1980 study, Culture’s Consequences: International
Differences in Work-Related Values, a large-scale, nomothetically based approach to
quantifying average differences in values (taken to represent cultures) across countries.
Based on this logic and on his empirical findings (evaluated below) Hofstede reached
a conclusion that is now taken as a fundamental fact of international management: that
‘organizations are culture-bound’ (1980: 372). Further, Hofstede identified the following
‘consequences for policy and research’:
In the management literature there are numerous unquestioning extrapolations of organizational
solutions beyond the border of the country in which they were developed. This is especially true
for the exportation of management theories from the United States to the rest of the world, for
which the non-US importers are at least as responsible as the US exporters . . .. However, the
empirical basis for American management theories is American organizations; and we should
not assume without proof that they apply elsewhere.
(Hofstede, 1980: 373)
The impact of Hofstede’s work can be seen by looking at any international
management text. For example, Nancy Adler’s book International Dimensions of
Organizational Behavior (2002) focuses heavily on the importance and consequences of
national culture differences and uses Hofstede’s work (and that of Andre Laurent) as a
major part of her rationale. She states, for example, that:
Diversity exists both within and among cultures; however, within a single culture certain
behaviors are favored and others repressed. . .. A cultural orientation describes the attitudes of
most people most of the time, never all of the people all of the time. Accurate stereotypes reflect
societal or cultural norms (Adler, 2002: 18– 19)
It is no wonder that Hofstede’s work is seen as so important. To be able to describe ‘most
people most of the time’ simply on the basis of their nationality offers a simple, yet
powerful advance in social science and management research.
 Gerhart and Fang: National culture and HRM 973
As another example of the central role accorded to national culture in international
management, consider the leading text in international HR, International Human
Resource Management, in which Dowling et al. identify national culture as one of five
variables that ‘moderate (i.e. either diminish or accentuate) differences between
domestic and international HRM’ (1999: 10). Dowling et al. note that, ‘[b]ecause
international business involves the interaction and movement of people across national
boundaries, an appreciation of cultural differences and when these differences are
important is essential’ (1999: 12). They also go on to say that:
It is now generally recognized that culturally insensitive attitudes and behaviors stemming from
ignorance or from misguided beliefs (‘my way is best,’ or ‘what works at home will work here’)
not only are inappropriate but often cause business failure. . .. Activities such as hiring,
promoting, rewarding, and dismissal will be determined by the practices of the host country and
are often based on a value system peculiar to that country’s culture.
(Dowling et al., 1999: 14)
In summary, leading texts on international management warn that national culture
differences matter a great deal and a manager ignores them at his or her own peril. In fact,
national culture almost seems at times to be deterministic of management practice. While
we are certain that national culture differences can be critical and that insensitivity to
national culture differences can and does result in business failure (as well as failure and
career consequences for individual managers), one can still ask whether national culture,
defined in terms of values, is this important and whether its effects are this pervasive and
systematic. Does a national culture score based on the average values of individuals in that
country, in fact, ‘describe most people most of the time?’ (One can also ask whether culture
differences are best captured by values and whether these capture the most fundamental
differences between countries that will have an impact on HR practices and strategies.)

Identifying the assumptions of national culture perspectives

For national culture to play such a central role in international management theory, there
are certain assumptions that should be supported. To the degree that the following
assumptions are strongly supported, national culture’s dominant role is justified. By the
same token, to the degree support for these assumptions is not strong, then the importance
of national culture is being over-emphasized. These assumptions are:

1 Between-country differences are substantially larger than within-country differences

in culture at the individual level of analysis. Statistically, this means that in an
analysis of variance having values (Hofstede’s measure of culture) as a dependent
variable, country as an independent variable should explain a substantial amount of
variance in values as indicated by the squared correlation (where only two countries
are studied) or by omega squared or the ICC(1,1) version of the intra-class
correlation (Shrout and Fleiss, 1979) where more than two countries are studied.
(Another commonly used index of variance explained, eta2, is not recommended
because Bliese and Halverson (1998) have shown that its size is influenced by the
number of observations in each category.) If this assumption is not met, then Adler’s
claim that ‘cultural orientation describes the attitudes of most people most of the
time’ would, for example, be less convincing.
2 Country differences in culture are larger than culture differences due to other factors
such as organization. If this is not the case, then one could question the rationale for
emphasizing national culture over organizational culture (Gerhart, 2004). National
974 The International Journal of Human Resource Management
culture cannot explain differences in culture within countries. National culture
cannot explain merger and acquisition difficulties that arise between two domestic
companies that result from culture clashes. National culture cannot explain why
management practices and managers fail in domestic settings where national culture
is not an issue.
3 A misfit between national culture and management practice will reduce
effectiveness. For example, using a pay for individual performance plan in a
country having a highly collectivist culture will not work well. Statistically, this
means that there is a statistical interaction between national culture and management
practice. (A stronger version of this assumption would distinguish between the less
serious case where the pay plan has little or no impact on effectiveness versus the
case where it actually has a harmful effect.) Without this statistical interaction,
Hofstede’s (1980) claim that management theories are ‘culture-bound’ would not be
supported.
One possible complication in assessing evidence on this assumption is the
possibility of availability bias, such that problems or failures of HR practices that are
traditionally seen as a poor fit to a particular culture are more likely to be noted than
are success stories.
4 Managerial discretion is substantially limited by national culture. The adoption of
innovative employment systems and change in their nature over time to adapt to
changing competitive conditions is not possible if they conflict with national culture
averages. By contrast, note the different assumption in domestic HR and strategy
research: having an organization with a culture that is unique relative to the country
or industry can be risky, but it can also be beneficial (e.g., the resource-based view of
the firm, Barney, 1991). At the same time, there are competitive and institutional
pressures that operate across country boundaries that may serve to reduce country
differences. For example, Katz and Darbishire (2000) find increasing variation
within countries in employment systems and that this variety of employment forms
looks increasingly similar (i.e. is converging) across countries. This would seem to
suggest that national culture does not preclude using new employment systems that
have a ‘misfit’ with the current national culture.
5 Even if it is true that management practices are ‘culture-bound’ on average, for
national culture to constrain management practice significantly for each particular
organization within a nation, it must be assumed that companies attract, select
and retain employees in a random fashion (Bloom and Milkovich, 1999).
Otherwise, if, for example, a company operates in a country with a mean level
of individualism that is too low for its needs, the country would simply establish
a hiring hurdle above that mean and would be able to find enough employees
that have sufficiently high individualism to fit its management system (Bloom
and Milkovich, 1999).

It is not clear that there is sufficient evidence available at this point to evaluate all
(if any) of the assumptions. Our goal in the remaining portion of this paper is to examine
key selected studies, particularly the influential work by Hofstede and his colleagues, that
address assumptions one and two, which are among the most fundamental assumptions in
making the case for national culture having a dominant role. To reiterate, assumption one
states that between-country differences in culture are large relative to within-country
differences. Assumption two states that national differences in culture (and in HR
practices) are large relative to differences due to other factors such as occupation
or organization differences. Although similar issues have been raised in critiques
 Gerhart and Fang: National culture and HRM 975
of Hofstede’s research programme (e.g. McSweeney, 2002), these critiques have not
defined the problem in clear statistical terms nor have they re-analysed Hofstede’s own
data on these issues. We do both here.
These issues can all be framed in terms of variance ratios or variance explained, a
concept typically captured by the squared coefficient of determination (R2) in regression
analysis or by indexes such as omega2 and the intra-class correlation coefficient in
analysis of variance applications. Indeed, estimating the ‘variance explained’ by country
was one of Hofstede’s stated research goals. However, Hofstede and others have
mistakenly interpreted F statistics, p values, and other statistical significance information
as indicative of variance explained. Yet previous research demonstrates that highly
statistically significant country differences in worker views on work-related variables do
not necessarily translate into country differences explaining much variance in these
responses (Fang, 1993). Thus, fundamental questions of the following sort remain
unanswered. How much variance is there across countries? How much variance is there
within countries? What is the ratio of between-country to total (between-country þ
within-country) variance (i.e. the R2 or some other index of variance explained)? All else
being equal, the larger this ratio, the greater the variance explained by country and, thus,
the more pervasive are country differences. Variance due to organizations (and its size
relative to variance due to countries) can be approached in the same way.
As described below, upon calculating appropriate estimates of variance explained, we
find that country differences explain only a small portion of the variance in individual-
level cultural values and related concepts, based on a re-analysis of (a) Hofstede’s (1980,
2001) own results, (b) a meta-analysis of selected cultural differences (Oyserman et al.,
2002), and (c) research on meaning of work differences across countries (England and
Harpaz, 1990; Ruiz-Quintanilla and England, 1996). Thus, these re-analyses call into
question assumption 1. In addition, our re-analysis of other results (Hofstede, 1980;
Hofstede et al., 1990) also shows that Hofstede’s interpretation of his findings, as
showing that country differences in values were greater than organization differences in
values, is not supported. This calls into question assumption 2.

Assumption 1: between-country differences in cultural values and related concepts

are large relative to within-country differences
Hofstede’s (1980) study
Geert Hofstede’s 1980 book, Culture’s Consequences, used worldwide data on IBM
employees to develop the four national culture dimensions: uncertainty avoidance, mas-
culinity/femininity, power distance and individualism. (Later, Hofstede and Bond (1988)
added a fifth dimension: long-term orientation.) Hofstede described his approach as follows:
The key constructs to be used in this book for describing mental programs are values and culture.
Values are an attribute of individuals as well as of collectivities; culture presupposes a
collectivity. I define value as a ‘broad tendency to prefer certain states of affairs over others’.

(Hofstede, 1980: 19)

Hofstede et al. further state that ‘values are acquired in our early youth. . .. By the time a
child is ten, most of his or her basic values are probably programmed into his or her
mind’ (1990: 312).
To the degree these statements are correct, then culture (the values of a collectivity) is
formed early and, by implication, does not change significantly over time. To the degree
that values of individuals and of collectivities (culture) differ across nations in stable
976 The International Journal of Human Resource Management
ways and these differences influence how people respond to their environments,
including management practices, then it is easy to understand why Hofstede (1980)
concluded that organizations are ‘culture-bound’. The central role played by values in
culture also explains why Hofstede used survey responses to statements about values to
measure culture.
The key assumption made by Hofstede (see assumption 1 above) is that the values of
individuals within a nation are (substantially) more similar than the values of individuals
from different nations. As noted earlier, this hypothesis can be tested by estimating an
ANOVA using individual-level responses to value statements as the dependent variables
and country as the independent variable. From these results, the percentage of variance
explained by country can be estimated.
Hofstede (1980) reported ANOVA results on six items (and two item difference
scores) that he used to form his power distance and uncertainty avoidance national
culture scales. The survey data came from a sub-sample of 3,220 respondents in ten
countries and in five occupations. As Table 1 indicates, the F statistics from his study
show that countries do differ on the eight items he examined.
In interpreting these results, Hofstede says ‘we see that the country effect is highly
significant (beyond the .001 level) in all eight cases’ (1980: 71) and that ‘nationality of
respondents . . . affects the scores of these questions highly significantly’ (1980: 71).
Although recognizing that there is ‘a broad dispersion across individuals’ within
countries, Hofstede emphasized that these ‘relatively small shifts of group means [i.e.
country differences] . . . nevertheless, may have considerable consequences for group
behavior and institutions’ (1980: 72). It was this evidence then, the statistical significance
and size of F statistics, that Hofstede relied upon to justify the aggregation of individual
responses by country into a measure of national culture, which has ‘considerable
consequences’ for how people generally think and act in different countries.
A major problem with Hofstede’s interpretation of his analyses is his assumption that
the size and degree of statistical significance of the F statistic indicates the amount of
variance explained by a factor. It does not, because the size of the F statistic is also
influenced by sample size and by the number of categories of the independent variable.

Table 1 Country differences in culture/values questions, Hofstede (1980)

Culture/value questions F-statistica Variance explainedb
Employees afraid to disagree 7.10 .02
Preferred manager 3 3.20 .01
Preferred another manager 5.90 .02
Stress 46.70 .12
Continue less than 5 years 8.80 .02
Company rules should not be broken 6.00 .02
Personal time vs training importance 12.90 .04
Earnings vs cooperation importance 25.80 .07
Mean 14.55 .04
Median 8.0 .02

Notes
N ¼ 3,220 respondents in five occupations in ten countries.
a
Provided by Hofstede (1980: 72, Figure 2.4).
b
ICC(1,1), computed from F-statistics.
 Gerhart and Fang: National culture and HRM 977
The larger the sample, the larger the F and the larger the number of cases in a category
(e.g. country), the larger the F. So, while Hofstede emphasizes that the F statistics for
country are ‘highly significant’, these only very obliquely tell us anything about variance
explained: the key criterion for grouping by country.
As Table 1 shows, upon calculating the ICC(1,1) for country, an index of its variance
explained, we find that it is not terribly high for the eight items analysed by Hofstede.1
Our estimate is that, across the eight items, somewhere between 2 and 4 per cent of the
variance is explained by country.2 This provides something less than compelling support
for assumption 1. Interestingly, Hofstede acknowledges the low variance explained,
noting that ‘of the total variance . . . only 4.2 is accounted for’ (1980: 71). (Hofstede does
not explain how he derived the 4.2 per cent figure but it is close to our estimates.) Recall,
however, that, despite recognizing this ‘broad dispersion across individuals’, Hofstede
chose instead to emphasize that these ‘relatively small shifts of group means [i.e. country
differences] . . . nevertheless, may have considerable consequences for group behavior
and institutions’ (1980: 72). The 2 to 4 per cent figure and the important fact that it
reveals – individuals differ much more within countries than they do across countries –
received little or no further attention in Hofstede’s (1980) book.
Worse, subsequent literature has sometimes latched onto a variance explained figure,
50 per cent, to describe the importance of country differences. But, this 50 per cent figure
is not relevant to assumption 1, having been derived in a completely different analysis.
Specifically, after aggregating data by country, Hofstede performed a factor analysis on
the means of each value item using countries (not individuals) as observations (i.e. rows)
in the data matrix. He reported that ‘factor analysis showed that 50 percent of the
variance in answer patterns between countries on the value questions could be explained
[by the four national culture dimensions]’ (Hofstede, 1980: 78). It is crucial to realize that
Hofstede is referring to the amount of variance in country-level means of the value items
that is explained by his four national culture dimensions. He is not speaking of variance
explained in individual-level values. Indeed, by using country-level means, the ‘broad
dispersion across individuals’ that Hofstede recognized at one point was completely
eliminated from the data.
Thus, the following interpretation of Hofstede’s results found in a leading text on
international management should be viewed with great caution as it may give readers the
impression that nation differences explain much more variance in individual employee
values than is really the case: ‘Hofstede found striking cultural differences within a single
multinational corporation. In his study, national culture explained 50 percent of the
differences in employees’ attitudes and behaviors’ (Adler, 2002: 67). It is not correct to
say that national culture explains 50 per cent of anything at the level of the individual
employee. Rather, Hofstede’s (1980) results show that roughly 2 to 4 per cent of the
variance in individual values is explained by national differences. And there is nothing
in Hofstede’s (1980) study that pertains to individual-level ‘behaviours’.
Therefore, Hofstede’s work should not be interpreted as showing that national culture
explains 50 per cent of behaviours.

Oyserman et al.: meta-analytic results

Of the four dimensions of national culture identified by Hofstede (1980, 2001), the one
receiving the most empirical attention has been individualism –collectivism (Smith,
2002). A recent meta-analysis by Oyserman et al. (2002) provides evidence on the effect
of country on individualism and collectivism. Oyserman et al. make note of the centrality
978 The International Journal of Human Resource Management
of Hofstede’s (1980) study to this literature and, indeed, explicitly focus on the empirical
research conducted subsequent to his 1980 work.
Oyserman et al. (2002) examine research on individualism and collectivism
separately. Their measure of effect size is d, the mean difference between samples
divided by their pooled standard deviation and corrected for small sample sizes. The d
they report is the difference between an ‘American’ sample (defined by Oyserman et al.
as being from the United States or Canada) and a sample from another country.
Thus, their study does not speak directly to the overall magnitude of culture differences
across countries, but rather focuses on how much ‘Americans’ differ from citizens
of other countries. Recall, however, that Hofstede was especially concerned with
the ‘unquestioning extrapolations . . . and the exportation of management theories from
the United States to the rest of the world’ (1980: 373). In this sense, the Oyserman et al.
comparisons are very pertinent.
In the analysis of individualism, fifty-two studies comparing Americans to citizens of
twelve other countries were found. Differences between Americans and others were
statistically significant for nine of the twelve countries. The median (unweighted) d for
the twelve country comparisons was .25. Although Oyserman et al. did not report any
variance explained indexes in their study, a d ¼ .25 can be transformed and yields an
r ¼ .124 and an r 2, or variance explained, of 1.5 per cent.3 Thus, assumption 1, that
individuals vary more across countries than they vary within countries, is not well-
supported, given a variance explained by country of 1.5 per cent.
The substantial heterogeneity of responses within countries is also suggested by the
Q tests of homogeneity within countries presented by Oyserman et al. (2002).
(A statistically significant Q indicates the lack of homogeneity.) The Q tests were
statistically significant for eight of the twelve countries and for six of the nine countries
where statistically significant mean differences (from Americans) had been found.
In other words, although there were mean differences for nine of the twelve country
comparisons, six of these comparisons also showed statistically significant heterogeneity
within countries.
Oyserman et al. (2002) found seventy-two studies of American-other country
differences on collectivism. The median unweighted d across fourteen countries was
2 .43 and, as with individualism, many of the Q statistics were statistically significant,
indicating heterogeneity across studies within countries. The d of 2.43 translates into
r ¼ 2.210 and an r 2, or variance explained, of 4.4 per cent. Thus, country differences
are somewhat greater for collectivism than for individualism.
Overall, however, these results again show that there is substantial variance
unexplained by country. The fact that country explains 1.2 per cent of the variance in
individual-level individualism scores implies that 98.8 per cent of variance
in individualism is unexplained by country. Likewise, the fact that country explains
4.4 per cent of the variance in individual-level collectivism scores implies that 95.6 per
cent of the variance in individual-level collectivism is unexplained by country.

Fang (1993) and the Meaning of Work (MOW) International Research

The Meaning of Work (MOW) International Research Team (1987) has collected data on
how workers from nationally representative samples in multiple countries define and see
meaning in their work. These data are intended to help understand the factors that explain
‘why it is that people relate to their working life in different ways and to different
degrees’ (Ruiz-Quintanilla and England, 1996: 516). Among the key factors they identify
are economic (sample item: if you get money for doing it), sense of purpose or belonging
 Gerhart and Fang: National culture and HRM 979
(sample item: by doing it, you get the feeling of belonging) and normative
influences (sample item: if you have to do it). Ruiz-Quintanilla and England suggest
that information on work definition factors is potentially helpful to management in
persuading workers that there are good reasons to put forth effort on the company’s
behalf. Thus, like cultural values, work definition is seen as influencing how workers
respond to different managerial practices.
The MOW researchers have collected data from multiple countries because of their
expectation that work definition (and work meaning broadly) is likely to differ by country
(England and Harpaz, 1990; Ruiz-Quintanilla and England, 1996). England and Harpaz
developed several profiles of work definition and examined how these differed across
countries. They strongly rejected the null hypothesis of no relationship, reporting a chi-
square of 681 (degrees of freedom ¼ 25) for the relationship between country and work
definition profiles. They thus concluded: ‘There is a large and statistically significant
effect of national context on the work definition distributions’ (1990: 261).
Ruiz-Quintanilla and England (1996) also derive several profiles of work definition
(e.g. burden/control, social contribution) and then, as before, compare whether the
distribution of workers across these profiles differs by country. They report a chi-square
of 2,819.95 with 5 degrees of freedom, which is statistically significant beyond the .0001
level, rejecting the null hypothesis that different countries have similar profiles. Thus,
they conclude that ‘there are large’ country differences in work definitions.
However, as Fang (1993) first noted in critiquing the earlier England and Harpaz
(1990) study, there are two problems with the conclusion that country differences are
‘large’ based on the methodology used. First, the country differences are unadjusted for
control variables (e.g. demographics). Second, like the F statistic (see our earlier
discussion), the chi-square statistic is a function of sample size. The England and Harpaz
study had a sample size of 4,950 and the Ruiz-Quintanilla and England (1996) study had
a sample size of 38,662, quite a testament to the accomplishments of the MOW team.
However, it also means that some measure of effect size, other than the chi-square,
is necessary to establish the practical significance of the country effects. Based on
the chi-square and degrees of freedom reported in the two studies, we calculated the
Cramer’s V squared, a variance explained analogue for contingency tables, to be .028
in the England and Harpaz study and .038 in the Ruiz-Quintanilla and England study.
We note that these variance explained estimates are similar to those we computed earlier
from Hofstede’s research.
Fang (1993) used MOW data on workers from the United States, West Germany and
Japan in an effort to better isolate the net effect of country differences on work meaning.
As in the research described above, she found highly statistically significant country
differences in work definition profiles (chi-square ¼ 115, df ¼ 4, p , .0001). She,
however, also calculated the hit rate achieved by different models in correctly classifying
workers into the three different work definition profiles she derived, an especially
meaningful way of assessing the importance of an independent variable in this situation.
With three profiles, there is a one in three or 33.3 per cent base rate of classification
success. So, any model should improve on this. In her analysis, using country alone as an
independent variable resulted in a 41 per cent hit rate, a reasonable improvement over
chance. However, demographics and job category yielded a hit rate of 46 per cent,
significantly higher than using country as the independent variable. Finally, with the
demographic/job category variables in the model, she found that adding country
provided no increase in the hit rate. In other words, the modest country effect was
confounded with the demographic and job category variables and had no meaningful
predictive power once these variables were controlled.
980 The International Journal of Human Resource Management
Fang summarized her findings as showing that ‘workers in different countries define
their work in similar ways, especially when one controls for personal and situational
factors. An implication . . . may be that managerial styles and systems may need less
tailoring to individual countries than previous research might suggest’ (1993: 61).

Summary
Our re-examination of the three sets of evidence above leads to the same conclusion: the
variance explained by country in cultural values and related concepts such as the
definition and meaning of work is modest. These results suggest less support than might
be expected for assumption 1, that between-country differences in culture and related
concepts that might affect workers’ reactions to different management practices
dominate within-country differences.

Assumption 2: nation differences in culture are large relative to organization

differences in culture
An important omission from Hofstede’s (1980) book is any analysis of organization
differences in culture. Are country differences in culture greater than organization
differences? If not, why focus so much attention on national culture, rather than on
organization culture? (e.g., Barney, 1991; Kotter and Heskett, 1992; O’Reilly et al., 1991)
To compare the size of national and organization differences in cultural values,
Hofstede et al. (1990) collected new data on 1,295 respondents from twenty organizations
on ten values questions. He then compared the median F statistic, 4.0, for organization
(derived from the new data) to a median F statistic for country, 8.0, that he says is based on
the analysis from his book (1980: 72), which used 3,220 respondents from ten countries.
Based on this comparison of these median F statistics, Hofstede observed that:

The F-values shown in Table 1 are a measure of the variance explained by the criterion (country
or organization). . .. For the questions on values, country differences explain more variance than
organization differences. . .. The ANOVAs across countries explained a much larger share of
variance than the ANOVAs across organizations.

(Hofstede et al., 1990: 296– 7)

This interpretation, of course, is incorrect. The F statistic is not an index of variance
explained. The most serious problem that arises in using F statistics for comparing
relative effect sizes is that the F statistic increases as the number of categories of the
variable decreases. Thus, when Hofstede compares the F statistics for country and
organization, the ‘deck is stacked’ in favour of a larger F statistic for country because it
has ten categories, whereas organization has twenty categories.
As noted earlier, it is necessary to examine a statistic such as ICC(1,1) to assess and
compare variance explained figures. Table 2 reports the ICC(1,1) figures that correspond
to the median F statistics reported by Hofstede. It shows that Hofstede’s own findings
support an interpretation that is just the opposite of his. Specifically, organization
explains more variance (4.4 per cent) than does country (2.1 per cent) in respondents’
values.4 Indeed, organization explains twice as much as explained by country using
Hofstede et al.’s own means of comparison.5
This re-analysis calls into question the veracity of assumption 2, that country
differences in culture are larger than organization differences in culture. What is the
rationale for emphasizing national culture and de-emphasizing organizational culture
 Gerhart and Fang: National culture and HRM 981

Table 2 Country vs organization effects on cultural values, F statistics reported by Hofstede et al.
(1990) vs variance explained, ICC(1,1), computed from the F statistics
Median F-statistica Variance explainedb

Data Country Organization Country Organization

d
10 countries, 3,220 individuals 8.0 – 0.021 –d
10 values questions
2 countries, 20 ‘organizations’c –d 4.0 –d 0.044
1,295 individuals
10 values questions

Notes
a
Provided by Hofstede et al. (1990: 296, Table 1).
b
ICC (1,1). Not provided by Hofstede et al., computed from their reported F-statistics.
c
Although Hofstede et al. use the term ‘organizations’ in their table, they state earlier in the article that their unit
of analysis was ‘20 units’ taken from 10 organizations.
d
Not reported by Hofstede et al.

when, in fact, the data indicate that culture varies more between organizations than
between nations?
Indeed, in the second edition of Culture’s Consequences, Hofstede states that ‘[t]he
ANOVA results justified our research approach’ ((2001: 396) and that ‘[t]he (unpredicted)
finding that values varied more across countries than across organizations . . . led to the
distinction between national and organizational cultures pictured in Exhibit 8.3’ (2001:
396–7). Our re-analysis, however, shows that there is no basis in their ANOVA results for
this distinction. Rather, organization differences in values are similar in magnitude
and perhaps larger than those due to national differences. Recent work (Gerhart, 2004;
House et al., 2004) provides further support for this conclusion.

Using national culture averages in organization-level research

The research reviewed above indicates that country explains only a modest amount of
variance in individual-level cultural values. This means that, if Hofstede’s national
culture score averages are to be used in organizational level research (e.g. to measure
cultural differences between groups of managers from different countries), one must
document that the sample size is sufficiently large. (It must be said that Hofstede, in
his own work, focused primarily on the country-level unit of analysis where there
were more than enough individual observations to achieve highly reliable country
scores.)
Even a small ICC(1,1) may generate a large ICC(1,k), the reliability of aggregated or
averaged responses, if a sample size k of sufficient size is used (Gerhart et al., 2000;
Glick, 1985; Shrout and Fleiss, 1979). Given the ICC(1,1)s we have seen, how large must
the groups be? With an ICC(1,1) ¼ .02, culture scores having a reliability of .70 (a rule of
thumb suggested by Nunnally (1978)) can be obtained by using 114 respondents per
country and a reliability of .90 can be obtained by using 440 respondents per country.
With an ICC(1,1) ¼ .04, fifty-six respondents per country would be necessary to achieve
a reliability of .70 and 215 respondents per country to obtain a reliability of .90. The
implication for researchers doing research at the organization level of analysis is that
caution must be exercised in using average culture scores for countries to describe
anything other than a reasonably large set of respondents from each of those countries.
982 The International Journal of Human Resource Management
Conclusion
In summary, a re-analysis of Hofstede’s original research reveals two important findings
that serve to call into question the dominant role of national culture in international
management. First, in Hofstede’s data, country explains only a small share (2 to 4 per
cent) of the variance in respondents’ values. (A re-examination of similar data from the
(Oyserman et al., 2002) meta-analysis and the MOW International research on work
meaning yielded similar estimates.) Second, contrary to Hofstede’s interpretation, our re-
analysis of Hofstede’s own data shows that organization differences account for more
variance in cultural values than do country differences. Therefore, two key assumptions
of the national culture framework are called into question.
Where does this leave us? From both a research and a policy point of view, given that
there do appear to be mean differences across countries in culture (Hofstede, 1980, 2001;
Trompenaars and Hampolen-Turner, 1997), it is important to identify these in academic
work and to make managers aware of them. However, given the findings of the present
paper, it is also important to make managers aware of the fact that such differences
typically provide a poor description of any single individual or small number of
individuals from a country. Therefore, it must be emphasized that only in sufficiently
large samples will mean country differences be statistically meaningful in organization-
level research and policy decisions.
One review reported that in 79 per cent of the cross-cultural studies they examined,
‘country was a proxy for culture’ (Schaeffer and Riordan, 2003: 175). They argue that
‘researchers should minimize the use of country as a proxy for culture’ (2003: 178).
Based on our results, this seems like good advice, unless one is dealing with aggregated
data from a sufficient number of respondents.
Perhaps more important, our re-analysis of (Hofstede’s 1980; Hofstede et al., 1990)
own data calls into question his conclusion that national differences in cultural values are
larger than organization differences in cultural values, a central tenet of their theoretical
model. The fact that Hofstede et al.’s (1990) own findings (when interpreted correctly)
indicate that organization differences are larger than country differences in cultural
values suggests the need to balance the focus on national differences with a focus on
organization differences in culture. As noted earlier, many of the difficulties attributed to
lack of cultural fit (e.g. mergers, acquisitions) may, in fact, be due as much to misfit in
organizational cultures as to misfit in national cultures.
Of course, all these conclusions are ‘on average’ statements and each situation must
be evaluated on the particular circumstances and facts. For example, where two
countries have very different cultural values on average, national culture differences
will obviously warrant greater attention. In such cases, important cultural differences
between groups may operate even when small numbers are involved. Cultural distance
indexes (e.g. Kogut and Singh, 1988) are meant to capture the magnitude of such
differences. Likewise, Hofstede’s (1980: Figure 7.3, 2001: several figures) attempts to
map countries into similar and dissimilar groupings is a similar attempt to identify the
most significant differences between countries. It is also possible that, if some countries
(e.g. Japan) are more culturally homogeneous than others (e.g. the United States,
Australia), country may provide a more reliable indicator of cultural values of
individuals in those more homogeneous countries.
According to Hickson, Hofstede had ‘frail data, but robust concepts’ (1996: 222) and
‘whether [we] accept or reject his ideas does not matter. His ideas have by then done their
job. They open access to cross-cultural analysis for an increasing number of scholars and
fundamentally challenge organizational ethnocentricism’ (1996: 221). However, now
 Gerhart and Fang: National culture and HRM 983
that that job is done, it is time to move beyond being culturally aware to better
understanding the conditions under which important cultural differences can be reliably
identified (i.e. based on cultural distance and number of respondents). In addition, it is
time to focus more attention on disentangling national and organizational sources of
cultural variation rather than assuming that nation is dominant.
An ongoing dilemma with culture is, of course, whether it can be meaningfully
quantified or whether trying to do so eliminates much of what is unique and interesting.
This dilemma will not be resolved here. However, if a researcher chooses to follow a
quantitative and nomothetic approach to studying culture as Hofstede did, it is important
to work towards understanding the results as thoroughly as possible using the standards
of nomothetic research.
Earlier, we identified several assumptions that are required for national culture to play
a dominant role in international management and human resources. Here, we focused
largely on two assumptions. Future work that addresses the other assumptions would be
useful. For example, one central question is whether there is a statistical interaction
between national culture and the influence of managerial practices. This is a fundamental
question for which there is precious little evidence. That practices differ in countries with
different national cultures does not necessarily demonstrate that culture-atypical
practices are ineffective. Indeed, the finding of some degree of convergence in
employment practices across countries suggests there is room for culture- or country-
atypical practices and that being unique may be effective (Katz and Darbishire, 2000).
Another question is whether national culture differences or organization culture
differences matter most in mergers and acquisitions. Additional questions have to do
with the potential value of a ‘misfit’ between national culture and management practice
(see the resource-based view of the firm) and the possibility that successful examples of
such ‘misfit’ get less attention than failures. Attraction –selection –attrition (ASA)
processes, by which a firm may attract employees atypical of the national culture would
also be worthy of further attention. Indeed, Fang and Gerhart (2003), using white-collar
employees, find that Taiwanese companies placing greater emphasis on individual pay-
for-performance have employees that are both more extrinsically motivated and more
intrinsically interested in their work. Taiwan, of course, has a collectivist culture
according to Hofstede’s work and collectivist countries have indeed been found to rely
less on pay-for-performance (Schuler and Rogovsky, 1998). Hofstede may be correct, on
average, but this example suggests that, even within a collectivist culture, there are
workers who are not well characterized by the average that will thrive under a ‘misfit’
individual-oriented pay-for-performance policy.
Our hope is that the field will look more closely than previously at these types of
questions and be careful not to assume too large a role for national culture without more
solid evidence than we have seen previously. Management is about recognizing
constraints, including national culture, but it is also about exercising discretion and
creating value in unique ways. Any management theory that emphasizes deterministic
constraints to the exclusion of management discretion needs to be carefully scrutinized.

Notes
1 Calculation of ICC(1,1) requires knowledge of the mean square between and mean square
within from the ANOVA. If only the F statistic (the ratio of the mean square between to the
mean square within) is reported, one can fix the mean square within equal to 1.0 and then use
the F statistic itself as the estimate of the mean square between component.
984 The International Journal of Human Resource Management

2 According to Hofstede, ‘the eight questions were selected because of their ability to
discriminate between countries’ (1980: 71). This admission suggests that the F statistics (and
the related variance explained percentages) are optimistic because of capitalization on chance.
3 In the case of single degree of freedom F statistic, the following formula can be used:
r ¼ (F/(F þ error degrees of freedom))1/2.
4 Further, even though Hofstede refers to his 1990 study as examining data on ‘20
organizations’ and in the ANOVA table refers to ‘20 organizations’, a close reading of his
article indicates that he actually had data on ‘20 units from 10 different organizations’. If, as
seems likely, units from the same organization are more similar than would be completely
different organizations, then his F statistic for organizations (and the corresponding ICC)
would be biased downward.
5 The reason that country explains 2.1 per cent of the variance here versus 4 per cent of the
variance earlier, despite the fact that both analyses are based on Hofstede (1980)’s data, is
explained by the fact that Hofstede et al. (1990) reported the median F statistic for country,
whereas we used the mean F statistic earlier.

References
Adler, N.J. (2002) International Dimensions of Organizational Behavior, 4th edn. Cincinnati, OH:
South-Western.
Barney, J.B. (1986) ‘Organizational Culture: Can it Be a Source of Sustained Competitive
Advantage’, Academy of Management Review, 11: 656– 65.
Barney, J. (1991) ‘Firm Resources and Sustained Competitive Advantage’, Journal of
Management, 17: 99– 120.
Becker, B. and Gerhart, B. (1996) ‘The Impact of Human Resource Management on
Organizational Performance: Progress and Prospects’, Academy of Management Journal, 39:
779– 801.
Bliese, P.D. and Halverson, R.R. (1998) ‘Group Size and Measures of Group-Level
Properties: An Examination of Eta-Squared and ICC Values’, Journal of Management, 24:
157– 72.
Bloom, M. and Milkovich, G.T. (1999) ‘A SHRM Perspective on International Compensation and
Rewards’. In Wright, P., Dyer, L., Boudreau, J. and Milkovich, G. (eds) Research in Personnel
and HRM. Supplement 4. Greenwich, CT: JAI Press.
Boxall, P. and Purcell, J. (2003) Strategy and Human Resource Management. New York: Palgrave
Macmillan.
Brewster, C. (1999) ‘Different Paradigms in Strategic HRM: Questions Raised by Comparative
Research’. In Wright, P., Dyer, L., Boudreau, J. and Milkovich, G. (eds) Research in Personnel
and HRM. Supplement 4. Greenwich, CT: JAI Press.
Dowling, P.J., Welch, D.E. and Schuler, R.S. (1999) International Human Resource Management.
Cincinnati, OH: South-Western.
Dunlop, J.T. and Galenson, W. (1978) Labor in the Twentieth Century. New York: Academic
Press.
Dyer, L. and Reeves, T. (1995) ‘Human Resource Strategies and Firm Performance: What Do We
Know and Where Do We Need to Go?’, International Journal of Human Resource
Management, 6: 656 – 70.
England, G.W. and Harpaz, I. (1990) ‘How Working Is Defined: National Contexts and
Demographic and Organizational Role Influences’, Journal of Organizational Behavior, 11:
253– 66.
Fang, M. (1993) ‘Differences in Work Definitions: A Study of the United States, Japan, and West
Germany’, MSc thesis, Cornell University.
Fang, M. and Gerhart, B. (2003) ‘Does Pay for Performance Diminish Intrinsic Interest? A
Workplace Test Using Cognitive Evaluation Theory and the Attraction-Selection-Attrition
Model’, Working Paper, National Central University, Taiwan.
 Gerhart and Fang: National culture and HRM 985
Gerhart, B. (forthcoming) ‘Human Resource Systems’. In Ostroff and Judge, T. (eds)
Perspectives on Organizational Fit.
Gerhart, B. (2004) ‘Culture in Management Research: Nation and Industry Differences Matter (But,
What About Oraganization Differences?)’, working paper, School of Business, University of
Wisconsin-Madison.
Gerhart, B., Trevor, C. and Graham, M. (1996) ‘New Directions in Employee Compensation
Research’. In Ferris, G.R. (ed.) Research in Personnel and Human Resources Management.
Greenwich, CT: JAI Press, pp. 143 – 203.
Gerhart, B., Wright, P.M., McMahan, G.C. and Snell, S.A. (2000) ‘Measurement
Error in Research on Human Resources and Firm Performance: How Much Error Is
There and How Does It Influence Effect Size Estimates?’, Personnel Psychology, 53:
803– 34.
Glick, W.H. (1985) ‘Conceptualizing and Measuring Organizational and Psychological Climate:
Pitfalls in Multilevel Research’, Academy of Management Review, 10: 601– 16.
Hickson, D. (1996) ‘The ASQ Years Then and Now through the Eyes of a Euro-Brit’,
Administrative Science Quarterly, 41: 217 –28.
Hofstede, G. (1980) Culture’s Consequences: International Differences in Work-Related Values.
Beverly Hills, CA: Sage.
Hofstede, G. (2001) Culture’s Consequences: Comparing Values, Behaviors, Institutions, and
Organizations across Nations, 2nd edn. Thousand Oaks, CA: Sage.
Hofstede, G. and Bond, M.H. (1988) ‘Confucius and Economic Growth: New Trends in Culture’s
Consequences’, Organizational Dynamics, 16(4): 4– 21.
Hofstede, G., Neuijen, B., Ohayv, D.D. and Sanders, G. (1990) ‘Measuring Organizational
Cultures: A Qualitative and Quantitative Study across Twenty Cases’, Administrative Science
Quarterly, 35: 286 –316.
Hoose, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W. and Gupta, V. (eds) (1994) Culture,
Leadership and and Organizations: The GLOBE Study of 62 Cultures. Sage Publications.
Katz, H.C. and Darbishire, O. (2000) Converging Divergences: Worldwide Changes in
Employment Systems. Ithaca, NY: ILR Press/Cornell University Press.
Kogut, B. and Singh, H. (1988) ‘The Effect of National Culture on the Choice of Entry Mode’,
Journal of International Business Studies, 19: 411 – 32.
Kotter, J.P. and Heskett, J.P. (1992) Corporate Culture and Performance. New York: Free Press.
McSweeney, B. (2002) ‘Hofstede’s Model of National Culture Differences and
the Consequences: A Triumph of Faith—A Failure of Analysis’, Human Relations, 55:
89 –118.
MOW International Research Team (1987) The Meaning of Working. London: Academic Press.
Nunnally, J.C. (1978) Psychometric Theory, 2nd edn. New York: McGraw-Hill.
O’Reilly, C., Chatman, J. and Caldwell, D. (1991) ‘People and Organizational Culture: A Q-Sort
Approach to Assessing Person-Organization Fit’, Academy of Management Journal, 34:
487– 516.
Oyserman, D., Coon, H.M. and Kemmelmeier, M. (2002) ‘Rethinking Individualism and
Collectivism: Evaluation of Theoretical Assumptions and Meta-Analyses’, Psychological
Bulletin, 128: 3– 72.
Ruiz-Quintanilla, S.A. and England, G.W. (1996) ‘How Working Is Defined: Structure and
Stability’, Journal of Organizational Behavior, 17: 515–40.
Schaeffer, B.S. and Riordan, C.M. (2003) ‘A Review of Cross-cultural Methodologies for
Organizational Research: A Best-Practices Approach’, Organizational Research Methods, 6:
169– 215.
Schuler, R.S. and Rogovsky, N. (1998) ‘Understanding Compensation Practice Variations across
Firms: The Impact of National Culture’, Journal of International Business Studies, 29(1):
159– 77.
Shrout, P.E. and Fleiss, J.L. (1979) ‘Intraclass Correlations: Uses in Assessing Rater Reliability’,
Psychological Bulletin, 86: 420 – 8.
986 The International Journal of Human Resource Management
Smith, P.B. (2002) ‘Culture’s Consequences: Something Old and Something New’, Human
Relations, 55: 119 – 35.
Trompenaars, F. and Hampden-Turner, C. (1997) Riding the Waves of Culture: Understanding
Cultural Diversity in Business, 2nd edn. London: Nicholas Brealey.
Wright, P.M. and Sherman, W.S. (1999) ‘Failing to Find Fit in Strategic Human
Resource Management: Theoretical and Empirical Problems’. In Wright, P., Dyer, L.,
Boudreau, J. and Milkovich, G. (eds) Research in Personnel and HRM, Supplement 4.
Greenwich, CT: JAI Press.