Cultural Geography
1
The Cross-Cultural Geography of Nations: Using the Spatial Autocorrelation
Coefficient to test Tobler’s First Law.
Garry A. Gelade & Paul Dobson
Cass Business School, City University, London
Abstract
This paper examines the geographical organization of some well-established
dimensions of national culture through the use of the spatial autocorrelation
coefficient. It is found that the national scores for many of these dimensions show a
significant degree of spatial patterning, such that values in nearby nations are more
similar than values in nations that are far apart. Parallel results are observed for the
states of the USA. We find substantial differences in the degree to which cultural
dimensions are organized in geographical space. A minority show no significant
organization, but some are as strongly organized as the physical climate. The
reasons for these differences are presently unclear. However, the finding that most
dimensions show a significant degree of spatial organization accords with previous
cluster analytic research which has found that nations with similar cultures are often
geographical neighbours.
Cultural Geography
2
Using the Spatial Autocorrelation Coefficient to test Tobler’s First Law
Cross-cultural psychologists have often observed that the constructs they
study are organised geographically. In an influential article, Ronen and Shenkar
(1985) reviewed eight studies that clustered countries by the similarity of work
values, and concluded, “It is apparent … that countries tend to group together
geographically.” (p. 444). A similar conclusion was reached by Schwartz (1999)
who used the co-plot technique to locate 44 national cultures on a two-dimensional
map so that nations with similar cultural profiles appeared close together, and those
with dissimilar profiles appeared far apart. The results indicated the existence of
broad cultural groupings of nations that were largely defined by geographical
proximity. Similarly, Georgas, van de Vijver and Berry (2004) noted that when
nations are clustered according to ecosocial indices, they form cultural zones
consisting of nations that share cultural similarities and that tend to be “… in
general, geographically contiguous.” (p. 78). Hofstede (2001) noted that some of
his dimensions of cultural values varied systematically with distance from the
equator, and suggested that “ … a country’s geographic position is a fundamental
fact that is bound to have a strong effect on the subjective culture of its inhabitants
…” (p. 116).
The empirical evidence therefore suggests that national cultures are
organized geographically; similar cultures are found in neighbouring countries, and
dissimilar cultures in countries that are far apart. This is an instance of Tobler’s
Cultural Geography
3
First Law of Geography, which states that everything is related to everything else,
but near things are more related than distant things. (Tobler, 1970).
To date however, the geographical distribution of the major dimensions of
cross-cultural psychology has yet to be quantified or studied systematically. Our
present understanding of how national cultures covary with physical space is mostly
qualitative, and is primarily based on the inspection of cultural maps (derived by
plotting pairs of dimensions against each other, or from multi-dimensional scaling),
and on examination of the groupings emerging from cluster analysis. While such
methods are adequate to detect the existence of an association between culture and
physical space, they cannot quantify it, and are highly susceptible to both Type I and
Type II error. Weak geographical patterns may go unnoticed, and geographical
phenomena may be inferred where none exist. It seems therefore that a quantitative
analysis of the geography of the major cultural dimensions would fill a gap in the
cross-cultural psychology literature.
In this paper geographical space is treated as an independent variable in its
own right. The aim is to develop quantitative estimates of the spatial organization of
some widely researched psychological measures of culture, with the objective of
identifying those that are highly spatially organized and those that are not. We
suggest that quantifying the spatial characteristics of different cultural dimensions
will help theorists to clarify how and why different aspects of culture develop,
interact, and spread across neighbouring regions.
Of course, it would be absurd to claim that geographical proximity alone can
explain the similarities and differences between national cultures. Indeed, as
Cultural Geography
4
pointed out by Inglehart and Baker (2000), some nations that are half-a-world apart,
for instance the UK and Australia, or Spain and Uruguay, share similar cultures
because of shared historical connections and large-scale immigration. Nevertheless,
there are good reasons to expect a higher degree of cultural similarity between social
groups that are physically close then those that are far apart. First, it has been
suggested that some cultural traits are adaptations to environmental conditions. For
example, it has been shown that warmer climates promote increased emotional
expression (Pennebaker, Rime, & Blankenship, 1996; Robbins, DeWalt, &
Pelto,1972) elevated levels of aggression and hostility (Anderson, 1989) and more
immediacy behaviours such as smiling, physical closeness and touching (Anderson,
1985). Van Vliert, Huang and Parker (2004) have also demonstrated relationships
between the physical climate and altruism and subjective well-being. Gilmore
(1990) has argued that a culture of masculinity (i.e., strong gender differentiation,
and a high regard for strength and fearlessness) is more likely to develop in harsh
climates, where survival depends on the ability to battle courageously against a
hostile environment. Such relationships between culture and the physical climate
would tend to produce geographically organized cultural zones, because countries
that have similar climates are close together. Secondly, because of the increased
opportunities for communication and migration, the processes of cultural diffusion
and borrowing are likely to occur more frequently amongst populations that are
physically close together than amongst populations that are physically far apart. This
would also tend to increase the cultural similarity between neighbouring nations.
Cultural Geography
5
The rest of this paper is organised as follows. First we introduce the spatial
autocorrelation coefficient as a measure of spatial organization, and describe its
statistical properties. Study 1 then explores how different encodings of geographical
space affect the observed autocorrelations for Hofstede’s dimensions of cultural
values. In Study 2 we calculate spatial autocorrelation coefficients for a number of
well-known cross-cultural dimensions.
Spatial Data and Spatial autocorrelation
By spatial data we mean a set of locations, {l1 ....., ln} and a set of
observations which vary by location {x (l1) ……, x(ln)}. Traditional statistics relies
on the assumption of independence between observations, but because independence
cannot be assumed for spatial data, traditional statistical techniques are not
appropriate. The appearance of Matern’s doctoral dissertation (Matern, 1960/1986)
is usually taken to herald the emergence of spatial statistics as an independent
scientific discipline, but it was in the 1970’s and 1980’s that the development of
statistical methods to handle spatial data first began to attract systematic attention
from mathematicians and statisticians (National Research Council, 1991, p. 4). In
spatial statistics, the non-independence of the data is explicitly modelled; spatially
dependent data distributions may thus be quantified, probabilistic inferences drawn,
and tests of statistical significance can be conducted.
Spatial autocorrelation is one of the most widely-used techniques within
spatial statistics, and the spatial autocorrelation coefficient measures the strength of
association between the value of an attribute and its location in space. Positive
Cultural Geography
6
spatial autocorrelation denotes a spatial structure in which locations close to each
other are associated with similar values of the attribute. A map of this type of spatial
structure would show regions of high and low values respectively; average
temperatures for the countries of Europe for example would show positive spatial
autocorrelation, because the Mediterranean countries are closer to the equator than,
and hence warmer than, the Nordic countries.
Negative spatial autocorrelation indicates that neighbouring locations are
associated with dissimilar values. A map of this type of spatial structure would look
like a checkerboard. Zero spatial autocorrelation indicates there is no spatial
structure in the data.
The most widely used measure of spatial autocorrelation (Ord & Getis,
1995) is Moran’s I (Moran, 1950). This statistic has been commended as a reliable
test for the presence of spatial pattern by Haining (1990, p. 238) provided the data
are not seriously non-normal, and provided the spatial lattice is reasonably
well-behaved (i.e. not dominated by one or two large regions.) I usually takes
values between -1 and +1 (though it can exceed those bounds under certain
circumstances); positive values of I indicate positive spatial autocorrelation,
negative values indicate the presence of negative spatial autocorrelation, and zero
indicates no spatial autocorrelation. The formula for Moran’s I can be written:
I=
N
∑ ∑ w zz
S ∑z
i
0
ij i
j
i
2
i
j
Cultural Geography
7
where N is the number of locations; wij is the element in a distance or
proximity matrix W describing the spatial relationship between locations i and j. W
is called the ‘structure’ or ‘weights’ matrix. zi and zj are observations at locations i
and j, respectively, expressed as deviations from the mean; and S0 is a normalizing
factor equal to the sum of the elements of the weights matrix, i.e., S0 = ∑i∑jwij . For
details, see Cliff and Ord (1973, 1981), and Upton and Fingleton (1985).
Moran’s I can be envisaged as the spatial analogue of the familiar Pearson
correlation coefficient; both are cross-product statistics whose numerator contains
the term zi times zj. It can easily be seen that if neighbouring regions are similar,
this term will tend to be positive, because both values will tend to be either
simultaneously above or simultaneously below the mean; while if neighbouring
regions are dissimilar, the observed values are likely to fall either side of the mean
so that the product of zi and zj will tend to be negative.
The significance of I can be assessed either analytically or by a permutation
procedure. Cliff and Ord (1973) have shown that under mild conditions, I is
asymptotically normal; for adequate samples sizes (Haining, 1990, suggests 20 or
more) the significance of I can thus be tested by reference to the standard normal
deviate. (For formulae for the expected value and variance of I, see Cliff and
Ord,1981, p. 44). Alternatively, the Quadratic Assignment Procedure (QAP) can be
used (Hubert & Schultz, 1976.) Here, rows and columns in the weights matrix are
permuted at random a large number of times, (in such a way as to preserve the
interdependencies amongst the dyads), and values of I are calculated for each
permutation. The distribution of I values is then used as a reference distribution
Cultural Geography
8
against which to compare the value of I for the un-permuted weights matrix, and to
assess its significance.
By convention the diagonal elements of W are set to 0, but the off-diagonals
may be encoded in a variety of ways depending on the data being modelled. In
many applications, wij is assigned the value 1 if i and j are within a specified
distance, d, of each other, and 0 otherwise. Alternatively, the analyst may use
α
-αd
weights proportional to 1/d , or e
, where α is a scaling factor; or if dealing with
area-based data, weights can be set proportional to the length of the common
boundary between contiguous areas. (Cliff & Ord, 1975).
Because different encodings of W lead to different numerical values for I,
the interpretation of I is not as straightforward as the interpretation of the
product-moment correlation coefficient. It is therefore only possible to directly
compare values of I for a given encoding of W.
Study1: Impact of different weights matrices
In the absence of any previous studies examining the spatial organization of
national cultural dimensions, we begin by exploring the impact of using different
weights matrices. Inconsistent results would suggest that the number of locations
used in the analyses was perhaps inadequate, or that the geographical lattice was too
irregular to give robust estimates.
Method
Weights Matrix
Cultural Geography
9
We used two types of weights matrix, both of which were constructed from
great circle distances between capital cities. In the first type, the proximity wij
between countries i and j was represented by an exponential decay function:
wij =
{
1
d <= 500 km
e-α(d-500)
d > 500 km
where d is the great circle distance (km) between the capital cities of i and j,
and α is a constant scaling factor. We investigated the effects of varying α between
.004 and .0005. As can be seen in Figure 1, smaller values of α cause proximities to
decay more slowly with distance, giving relatively greater weight to countries that
are far apart as α decreases. Because countries are irregularly spaced, those having
few near neighbours, such as Australia, receive relatively little weight in the
analysis, while countries that have many close neighbours receive more weight. The
raw wij values were therefore row-standardized, that is each element wij was divided
by
∑w
ij
.
j
-------------------- insert Figure 1 about here ------------------
The second type of matrix was a nearest neighbour matrix. Here the n
nearest neighbours of a country are coded as 1, and countries further away as zero.
This type of encoding implicitly introduces a form of standardization that partly
offsets the effects of different sizes of country and different degrees of isolation,
Cultural Geography
10
because all countries are surrounded by a fixed number of neighbours at the same
(unit) proximity. We examined values of n = 1, 2, 4, and 8.
Cultural Dimensions and Benchmarks
National measures of power distance, individualism, uncertainty avoidance
and masculinity for 50 countries were taken from Hofstede (2001, p. 500).
Temperature and precipitation for the same set of countries were included as
benchmark variables. Annual values of temperature (ºC) and precipitation (mm.)
averaged over the period 1961-1990 were taken from the Tyndall Centre dataset
TYN CY 1.1 (Mitchell, Hulme & New, 2002). The inclusion of these variables is
not meant to imply any substantive connection between culture and climate,
although such a connection is of course plausible. Rather, we present the spatial
autocorrelations for these additional variables as reference points, to which the
magnitudes of the cultural autocorrelations can be compared.
Hypothesis Testing
Standard errors and significance levels of I were assessed using 5,000
randomly generated QAP permutations. Significance was determined by a
one-tailed test, for which the null hypothesis was zero or negative autocorrelation,
and the alternative hypothesis was positive autocorrelation. An observed
autocorrelation was deemed significant with probability p if it fell at or above the
100-pth percentile of the QAP distribution.
Cultural Geography
11
Differences in I for a pair of variables were also tested for significance using
a QAP method. Here, we constructed 5,000 random QAP permutations of the
weights matrix, and for each permutation calculated I for both variables and the
absolute value of the difference between them. The observed absolute difference
between the two I values was then compared with the distribution of absolute
differences generated by the QAP permutations. An observed absolute difference
was deemed significant with probability p if it fell at or above the 100-pth percentile
of the QAP distribution. (Note that because the test statistic and the reference
distribution are represented by absolute values, p in this case represents the
two-tailed significance of a difference between the two I values.)
Results
Table 1 shows Moran’s I for the cultural and benchmark variables for four
different decay parameters and four different neighbourhood sizes.
Table 1 about here
The results for the nearest-neighbour weights matrix and the exponential
weights matrix are similar. First, spatial autocorrelation is highest when the
neighbourhood size is small or when the exponential proximity function is narrow,
and decreases as the neighbourhood size increases or the proximity function widens.
This is consistent with the idea that the cultural similarity between countries
decreases with geographical distance over the range studied; as the neighbourhoods
encoded in the weights matrix increase in size, increasing dissimilarities within the
Cultural Geography
12
neighbourhood lead to a lowering of the observed spatial autocorrelation
coefficients.
Secondly, it is clear that Hofstede’s dimensions differ substantially in their
spatial properties. Individualism consistently shows the highest levels of
autocorrelation, and masculinity consistently shows the lowest values, with power
distance and uncertainty avoidance showing values in between. These differences
are robust, and do not appear to be sensitive to variations in the way the weights
matrices are encoded, or to the size of the spatial neighbourhood.
In particular, it is of interest to see that the autocorrelations for individualism
are noticeably stronger than those for power distance. In Hofstede’s sample of
countries, these two dimensions are strongly correlated (r = -.68), and we had
therefore expected them to show similar spatial properties. In fact, they are spatially
quite distinct; national individualism is more strongly associated with physical space
than national power distance, and this difference is statistically significant for all
weights matrices. To check this counter-intuitive finding we repeated the analyses,
this time using estimates for an additional 16 nations catalogued by Hofstede (2001,
p. 502). In the combined sample of 66 countries, we once again found
individualism to be more highly regionalized than power distance. Using nearest
neighbour weights matrices with n = 1,2, 4 and 8, the I values for individualism
were .83, .70, .60 and .54 respectively, while for power distance the corresponding
values were .09, .19, .17, and .17; at each neighbourhood size, the spatial
autocorrelation for individualism was significantly greater than that for power
distance at p < .001.
Cultural Geography
13
Finally, to grasp the substantive meaning of these results, note that the
spatial autocorrelation for individualism is similar in size to that for temperature and
precipitation; in other words, the relationship between geographical space and
Hofstede’s individualism and is just about as strong as the relationship between
geographical space and the physical climate.
Study2: Spatial autocorrelation for some well-known cross-cultural dimensions
The purpose of Study 2 was to quantify the spatial organization of some
well-established measures of national culture. We also consider the case of a single
nation by analysing the geographical distribution of collectivism across the states of
the USA.
Method
National measures of Schwartz’s cultural values were taken from the
Schwartz Value Survey (Israel Social Sciences Data Centre 2005). Societal cynicism
and dynamic externality scores were taken from Bond et al. (2004), and scores for
survival vs. self-expression and traditional vs. secular-rational were taken from the
Inglehart-Welzel map of cultures (Inglehart & Welzel, 2005). Scores for Triandis’
individualism were taken from Diener, Diener and Diener (1995) and scores for the
Globe Study societal cultural practices and values were taken from House, Hanges,
Javidan, Dorfman and Gupta (2004). For the USA, the unit of analysis was the state,
and the cultural variable was a measure of state collectivism reported by Vandello
and Cohen (1999). Distances between states were calculated as the great circle
Cultural Geography
14
distances between state centroids. (State centroids rather than capital cities were
used to represent locations simply because the data was more readily available.)
As in study 1, temperature and precipitation were used as benchmark
variables. Moran’s I was computed as in Study 1, except that only one weights
matrix was used. This was a four-nearest-neighbours matrix, in which the four
nearest neighbours of a country (or state) were encoded as 1 and all other entries as
0. The QAP procedure (5,000 permutations) was used to determine the one-tailed
significance of the spatial autocorrelation coefficients, and the two-tailed
significance of the differences between coefficients.
Results
Table 2 shows the geographical characteristics of the different international
samples we examined.
Table 2 about here.
The bounds for latitude and longitude identify the geographical borders of
each sample. These consistently extend further north than south, and further east
than west, but all samples cover approximately the same area of the globe. The
median within-sample distance is the median great circle distance between all pairs
of capital cities in the sample, and the median within-neighbourhood distance is the
median distance for the cells designated as 4-nearest neighbours in the weights
matrix. These two measures indicate the degree of inter-country spacing in each
sample. It can be seen for example that the nearest neighbours in the Inglehart &
Cultural Geography
15
Welzel sample are considerably closer than in the Bond et al. sample, as would be
expected since more countries are packed into a similar space.
Table 3 shows Moran’s I for each cultural dimension, and its associated
significance level.
Table 3 about here
These results demonstrate that it is not uncommon for cultural dimensions to
be as strongly regionalized as are temperature and precipitation. Indeed, most
cultural dimensions show at least some degree of spatial organization, and it is
relatively rare for a dimension to show none.
Comparison between the different national samples is somewhat complicated
by the fact that different researchers used different sets of countries. Although all
the samples covered approximately the same area of the globe, there were some
differences in the inter-country spacing. There were also differences in regional
emphasis. Hofstede’s sample of countries for example did not include any countries
within the former Soviet Union, and societies in Africa and the Middle East were
under-represented. However, the inclusion of benchmark variables (temperature and
precipitation) allows some indication of the degree of between-sample comparability
we can assume. Differences in the spatial characteristics of the samples would be
expected to cause differences in the observed autocorrelations for the benchmark
variables. In fact, for the national samples, Moran’s I for temperature ranges
between .47 and .62, and for precipitation between .52 and .69, intervals that are not
negligible, but which imply at least some degree of comparability between the
Cultural Geography
16
samples. Nevertheless, we would suggest that direct comparisons between studies
should be treated with caution.
Relative comparisons are of course possible. In Study 1 we found the spatial
patterning of Hofstede’s individualism to be stronger than that for power distance.
This is also reflected in the Globe data. The Globe measures of institutional and
in-group collectivism (both conceptually related to individualism) show significant
spatial autocorrelation in both the practice and values domains, but the
autocorrelations for Globe power distance are significantly smaller and
indistinguishable from zero.
Discussion
This research has demonstrated that most dimensions of national culture are
geographically clustered rather than being distributed at random across the globe.
Although this general idea is not particularly new, this paper is to our knowledge the
first to offer a quantitative account of the spatial structuring of national cultural
dimensions. Some dimensions are as strongly dependent on geography as the
physical climate; many dimensions show lesser but still significant degrees of spatial
organization, and only a small minority of dimensions show no spatial organization
at all.
Overall, the results confirm that Tobler’s First Law applies to the cultural
geography of nations. Tobler’s Law is also visible on a smaller scale within the
USA, where neighbouring states have more similar levels of collectivism than do
distant states. The within-nation evidence is however limited to a single dimension
Cultural Geography
17
within a single country, and it would be of interest to know whether similar patterns
of cultural variation exist in other countries.
The findings reported here can account for much previous research which
has found that clusters of countries with similar cultures are often organized
geographically. An obvious exception is the ‘Anglo’ cluster which emerges from
several studies (e.g., Ronen & Shenkar, 1985 ). Typically including the USA,
Canada, the UK and Ireland, New Zealand and Australia, this cluster is culturally
homogeneous although its members are widely dispersed.
We suggest however that spatial autocorrelation and clustering provide
complementary perspectives on the organization of cultural phenomena. Cluster
analysis requires no a priori independent variable to be specified, but does require a
degree of subjective interpretation to account for the emergent clusters; sometimes
multiple factors (religion, language, climate, history, geography) need to be invoked
post-hoc to explain the composition of different clusters. This has the potential to
reveal previously unsuspected grouping dynamics, but can lead to
over-interpretation. Spatial autocorrelation on the other hand invokes an
independent variable (in this case geographical distance) to explain national
similarities. It provides the ability to test specific hypotheses, but it is not
exploratory, and cannot detect similarities between cultural dimensions that are
unrelated to, and uncorrelated with, the chosen independent variable.
The origins of cultural variation amongst societies have been the subject of
speculation by anthropologists and others for many years. At least three mechanisms
could account for the spatial organization of cultural beliefs and practices (see e.g.,
Cultural Geography
18
Guglielmino, Viganotti, Hewlett, and Cavalli-Sforza, 1995). First, many scholars
have argued that culture is an adaptive response to the environment. Geographically
proximate regions often share similar climates, and similar terrain, which would be
expected to produce similar cultures. Secondly, societies may impose their beliefs or
practices on one another, or adopt beliefs and practices from each other voluntarily.
In either case, spatial proximity is likely to increase the degree of interaction and
interchange between societies, producing more convergence between cultures that
are geographically close than between those that are far apart. Finally, populations
may migrate. Short-range migration would tend to increase the cultural similarity of
nearby regions (but over time, migratory chains have the ability to transport culture
over considerable distances, producing cultural similarities in groups that are a long
way apart.)
The numerous significant differences we observe between autocorrelation
coefficients show that different cross-cultural dimensions are differentially sensitive
to the mechanisms that produce spatial dispersion. Why certain cultural dimensions
should be less inclined than others to spread across national boundaries is an
intriguing question which at present remains unanswered, but which merits further
research.
The methods described in this paper can be extended in several ways. First,
weights matrices are generic proximity matrices, and are not limited to representing
distances in physical space. Any conception of distance - or indeed any measure of
interaction - between a pair of objects can be implemented as a weights matrix. It
would therefore be possible to test for example whether national cultural values are
Cultural Geography
19
related to genetic distances between populations, or to the existence of trading or
political links. A frequently raised question in the cross-cultural literature is the
relationship between national culture and physical climate (see e.g. Gupta and
Hanges, 2004 for a review.) One way to test this relationship would be to compute
autocorrelations for various dimensions of culture using a weights matrix based on
climate proximities instead of spatial proximities. Finally, cultural dimensions
themselves might also be used to construct the weights matrix, allowing the
possibility of computing autocorrelations based on cultural similarities .
Cultural Geography
20
References
Andersen, P. A. (1985). Nonverbal immediacy in interpersonal communication. In
A. W. Siegman & S. Feldstein (Eds.), Multichannel integrations of nonverbal
behavior (pp. 1-36). Hillsdale, NJ: Erlbaum.
Anderson, C. A. (1989). Temperature and aggression: Ubiquitous effects of heat on
occurrence of human violence. Psychological Bulletin, 106, 74-96
Bond, M.H. et al (2004) Culture-level Dimensions of Social Axioms and their
Correlates across 41 Cultures, Journal of Cross-Cultural Psychology, 35, 5,
548-570.
Cliff, A. K., & Ord, J. K. (1973). Spatial Autocorrelation. London: Pion.
Cliff, A. K., & Ord, J. K. (1975). Model building and the analysis of spatial pattern
in human geography. Journal of the Royal Statistical Society. Series B
(Methodological), 37, 297-348.
Cliff, A. D., & Ord, J. K. (1981). Spatial processes, models, and applications.
London: Pion.
Diener, E., Diener, M., & Diener, C. (1995). Factors Predicting the Subjective WellBeing of Nations. Journal of Personality and Social Psychology, 69, 851 -864.
Georgas, J., van de Vijver, F. J. R., & Berry, J. W. (2004). The ecocultural
framework, ecosocial indices, and psychological variables in cross-cultural
research. Journal of Cross-Cultural Psychology, 35, 74-96.
Gilmore, D. D. (1990). Manhood in the making. New Haven, CT: Yale University
Press.
Cultural Geography
21
Guglielmino, C. R, Viganotti, C., Hewlett, B., & Cavalli-Sforza, L. L. (1995).
Cultural variation in Africa: Role of mechanisms of transmission
and adaptation. Proceedings of the National. Academy of Sciences, 92, 7585-7589.
Gupta, V., & Hanges, P. J. (2004). Regional and climate clustering of societal
cultures. In R. J. House, P. J. Hanges, M. Javidan, P. W. Dorfman, & V.
Gupta (Eds.), Culture, Leadership, and Organizations. The GLOBE Study of
62 Societies (pp.178-218). Thousand Oaks, CA: Sage.
Haining, R. (1990). Spatial data analysis in the social and environmental sciences.
Cambridge, UK: Cambridge University Press.
Hofstede, G. (2001). Cultures consequences: Comparing values, behaviours,
institutions and organizations across nations (2nd ed.). Thousand Oaks, CA:
Sage.
House, R. J., Hanges, P. J., Javidan, M., Dorfman, P. W., & Gupta, V. (Eds.)
(2004). Culture, Leadership, and Organizations. The GLOBE Study of 62
Societies. Thousand Oaks, CA: Sage.
Hubert, L. & Schultz, J. (1976). Quadratic Assignment as a General Data Analysis
Strategy. British Journal of Mathematical and Statistical Psychology 29,190241.
Inglehart, R. (1997). Modernization and Postmodernization. Princeton, NJ:
Princeton University Press.
Cultural Geography
22
Inglehart, R., & Baker, W. E. (2000). Modernization, cultural change, and the
persistence of traditional values. American Sociological Review, 65, 19-51.
Inglehart, R. & Welzel, C. (2005). Modernization, Cultural Change, and
Democracy: The Human Development Sequence. New York: Cambridge
University Press.
Israel Social Sciences Data Centre. Schwartz Value Survey (SVS), Dataset 0789.
[Electronic File. Download 5th May, 2005]
Matern, B.(1960/1986) Spatial Variation. Meddelanden fran Statens
Skogsforskningsinstitut, 49. 2nd edition published as Lecture Notes in
Statistics, vol. 36. New York: Springer.
Mitchell, T. D., Hulme, M., & New, M. (2002). “Climate Data for Political Areas.”
Area 34:103-112. [Data file retrieved from
http://www.cru.uea.ac.uk/~timm/cty/obs/TYN_CY_1_1.html, 25/11/2005].
Ord, J. K., & Getis. A. (1995). Local spatial autocorrelation statistics: Distributional
issues and an application. Geographical Analysis, 27, 286-306.
Pennebaker, J. W., Rime, B., & Blankenship, V. E. (1996). Stereotypes of
Emotional Expressiveness of Northerners and Southerners: A Cross-Cultural
Test of Montesquieu's Hypotheses. Journal of Personality and Social
Psychology, 70, 372-380.
Robbins, M. C.; DeWalt, B. R.; Pelto, P. J. (1972). Climate and behavior: A
biocultural study . Journal of Cross-Cultural Psychology, 3, 331-344.
Ronen, S., & Shenkar, O. (1985). Clustering countries on attitudinal dimensions: A
review and synthesis. Academy of Management Review, 10, 435-454.
Cultural Geography
23
Schwartz, S. H. (1994). Beyond individualism/collectivism. New cultural
dimensions of values. In U. Kim, H. C. Triandis, C. Kagitcibasi, S. Choi, & G.
Yoon (Eds.) Individualism and collectivism: Theory, method, and
applications. (pp. 85-119). Thousand Oaks, CA: Sage.
Schwartz, S. H. (1999). A theory of cultural values and some implications for work.
Applied Psychology: An International Review, 48, 23-47.
Tobler, W. R. (1970). A computer movie simulating urban growth in the Detroit
region. Economic Geography, 46, 234–240.
Upton, G., & Fingleton, B. (1985). Spatial data analysis by example.
New York: John Wiley & Sons.
Van de Vliert, E., Huang, X., & Parker, P. M. (2004). Do colder and hotter climates
make richer societies more, but poorer societies less, happy and altruistic?
Journal of Environmental Psychology, 24, 17–30.
Vandello, J. A., & Cohen, D. (1999) Patterns of Individualism and Collectivism
Across the United States. Journal of Personality and Social Psychology, 77,
279-292.
Cultural Geography
24
Table 1. Spatial autocorrelation coefficients (Moran’s I) for different weights
matrices.
Exponential Decay Weights Matrix
Alpha
.004
.002
.001
.0005
Temperature
.80***
.65***
.49***
.28***
Precipitation
.79***
.64***
.47***
.26***
Individualism
.83***
.67***
.47***
.24***
Power Distance
.39***
.36***
.28***
.13***
Uncertainty Avoidance
.50***
.28***
.18**
.09**
Masculinity
.15
.02
.00
-.02
Nearest Neighbour Weights Matrix
Number of nearest neighbors
1
2
4
Temperature
.82***
.72***
.64***
.57***
Precipitation
.82***
.82***
.65***
.51***
Individualism
.86***
.69***
.58***
.54***
Power Distance
.29***
.31***
.27***
.30***
Uncertainty Avoidance
.53***
.49***
.31***
.20***
Masculinity
.08
.17
.13*
.00
Note: N=50.
*** p <= .001, ** p <= .01, * p <=.05.
8
Cultural Geography
25
Table 2. Geographic characteristics of international samples.
Latitude of capital
cities ( ° )
Sample
No. of
countries
Minimum
Longitude of capital
cities ( ° )
Maximum
Minimum
Maximum
Median Distance between
capitals (km.)
Within
sample
Within
Neighbourhood
Hofstede (2001)
50
-41.3S
60.2N
-99.2E
174.8W
8,974
1,131
Schwartz (Israel Social Sciences Data Center)
73
-41.3S
60.2N
-99.2E
178.4W
7,129
913
Bond et al. (2004)
42
-41.3S
60.2N
-77.1E
174.8W
7,116
1,160
Triandis (Diener, Diener & Diener, 1995)
53
-41.3S
64.2N
-99.2E
174.8W
7,377
1,158
Inglehart & Welzel (2005)
78
-41.3S
64.2N
-99.2E
174.8W
6,356
767
Globe (House et al., 2004)
58
-41.3S
60.2N
-99.2E
174.8W
8,262
1,109
Cultural Geography
26
Table 3. Spatial autocorrelations (Moran’s I) for different samples.
Sample
Dimension
Temperature
Precipitation
Intellectual autonomy
Egalitarian commitment
Conservatism
Hierarchy
Affiliative autonomy
Harmony
Mastery
Dynamic Externality
Societal Cynicism
Schwartz
(2005)
.59***
.53***
Bond et al
(2004)
Triandis
(Diener et al
1995)
Inglehart &
Welzel
(2005)
Globe
Practices
(House et
al 2004)
Globe
Values
(House et
al 2004)
.47***
.69***
.62***
.60***
.59***
.52***
.59***
.64***
.59***
.64***
.38***
.35***
.22**
.16*
.13
.11
.11
.09
.05
.31***
.44***
.36***
Hofstede
(2001)
.64***
.65***
Vandello &
Cohen
(1999)
.53***
.79***
.60***
.55***
.50***
.43***
.42***
.25***
.12*
.44***
.06
Individualism
Traditional vs. secular-rational
Survival vs. self-expression
Institutional Collectivism
In-Group Collectivism
Uncertainty Avoidance
Performance Orientation
Humane Orientation
Assertiveness
Gender Egalitarianism
Future Orientation
Power Distance
Individualism
Uncertainty avoidance
Power distance
Masculinity
Collectivism
*** p <= .001, ** p <= .01, * p <= .05.
.51***
.68***
.55***
.34***
-.08
.24**
.39***
.45***
.05
.58***
.31***
.27**
.13*
.37***
Cultural Geography
27
Figure 1. Weight as a function of distance.